USF Libraries
USF Digital Collections

Effect of permeate suction on the performance of spiral wound nanofiltration module

MISSING IMAGE

Material Information

Title:
Effect of permeate suction on the performance of spiral wound nanofiltration module
Physical Description:
Book
Language:
English
Creator:
El-Shamy, Awad
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Reverse osmosis
Concentration polarization
Desalination
Diffusion
Peclet number
Dissertations, Academic -- Civil Engineering -- Doctoral -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Fouling in a nanofitration membrane module is usually a result of concentration polarization. The effect of permeate suction on the slightly negatively charged spiral wound nanofiltration membrane is investigated. According to the film theory, the mass transfer coefficient is inversely proportional to concentration polarization. The effect of permeate suction destabilizes the boundary layer. This will decrease the concentration polarization layer, and consequently will increase mass transfer through the membrane's surface. To validate the hypothesis, experiments were carried out on a NF membrane that can be described by the solution-diffusion model. This model has coefficients that can be measured experimentally. Using the membrane wall concentration in this model instead of the bulk feed concentration can help estimating the mass transfer coefficient more appropriately.Two experimental studies were carried out, one with a standard high pressure pump, and another one with the added effect of suction pressure applied to the permeate collector tube. Three different concentrations of binary dilute solutions of NaCl,MgSO₄, and MgCl₂, at three different pressures (low, medium, and high) were tested. For all tested solutions, permeate suction increased the diffusive Peclet number as a function of the feed concentration (x) according to the equation Pe = a₁x²+b₁x+c₁, with R²>0.99, where x is the feed concentration in Mol/l, and a₁, b₁, and c₁ are coefficients dependent on feed pressure for every salt solution. With the increase of the Peclet number, it was observed that the concentration polarization decreased, and both the product flow and the product quality were improved.Suction had the greatest impact at the range of 100 to 110 psi feed pressure, where the concentration polarization reduced approximately 14 to 20 %. ANOVA for the concentration polarization showed that suction was significant in reducing the calculated concentration polarization layer for all tested solutions. It was concluded that permeate suction reduced concentration polarization, increased product flow rate, and improved product quality. Thus, adding permeate suction has beneficial consequences because it reduces membrane fouling and extends its useful service life.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2009.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Awad Abdel Monem El-Shamy.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 151 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002220519
oclc - 645543456
usfldc doi - E14-SFE0002868
usfldc handle - e14.2868
System ID:
SFS0027185:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200385Ka 4500
controlfield tag 001 002220519
005 20101007115846.0
007 cr cnu|||uuuuu
008 100701s2009 flua sb 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002868
035
(OCoLC)645543456
040
FHM
c FHM
049
FHMM
090
TA145 (Online)
1 100
El-Shamy, Awad.
0 245
Effect of permeate suction on the performance of spiral wound nanofiltration module
h [electronic resource] /
by Awad Abdel Monem El-Shamy.
260
[Tampa, Fla.] :
b University of South Florida,
2009.
500
Title from PDF of title page.
Document formatted into pages; contains 151 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2009.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: Fouling in a nanofitration membrane module is usually a result of concentration polarization. The effect of permeate suction on the slightly negatively charged spiral wound nanofiltration membrane is investigated. According to the film theory, the mass transfer coefficient is inversely proportional to concentration polarization. The effect of permeate suction destabilizes the boundary layer. This will decrease the concentration polarization layer, and consequently will increase mass transfer through the membrane's surface. To validate the hypothesis, experiments were carried out on a NF membrane that can be described by the solution-diffusion model. This model has coefficients that can be measured experimentally. Using the membrane wall concentration in this model instead of the bulk feed concentration can help estimating the mass transfer coefficient more appropriately.Two experimental studies were carried out, one with a standard high pressure pump, and another one with the added effect of suction pressure applied to the permeate collector tube. Three different concentrations of binary dilute solutions of NaCl,MgSO, and MgCl, at three different pressures (low, medium, and high) were tested. For all tested solutions, permeate suction increased the diffusive Peclet number as a function of the feed concentration (x) according to the equation Pe = ax+bx+c, with R>0.99, where x is the feed concentration in Mol/l, and a, b, and c are coefficients dependent on feed pressure for every salt solution. With the increase of the Peclet number, it was observed that the concentration polarization decreased, and both the product flow and the product quality were improved.Suction had the greatest impact at the range of 100 to 110 psi feed pressure, where the concentration polarization reduced approximately 14 to 20 %. ANOVA for the concentration polarization showed that suction was significant in reducing the calculated concentration polarization layer for all tested solutions. It was concluded that permeate suction reduced concentration polarization, increased product flow rate, and improved product quality. Thus, adding permeate suction has beneficial consequences because it reduces membrane fouling and extends its useful service life.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Advisor: Mahmoud Nachabe, Ph.D.
653
Reverse osmosis
Concentration polarization
Desalination
Diffusion
Peclet number
690
Dissertations, Academic
z USF
x Civil Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.2868



PAGE 1

Effect of Permeate Suction on the Performance of Spiral Wound Nanofiltration Module by Awad Abdel Monem El-Shamy A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Civil and Environmental Engineering College of Engineering University of South Florida Co-Major Professor: Mahmoud Nachabe, Ph.D. Co-Major Professor: Robert Carnahan, Ph.D. Aydin Sunol, Ph. D. Mark Ross, Ph.D. Ahmed Said, Ph.D. Date of Approval: March 12, 2009 Keywords: reverse osmosis, concentration polarization, desalina tion, diffusion, Peclet number Copyright 2009 A. El-Shamy

PAGE 2

DEDICATION To my country Egypt, the orig in of the secrets of life. To the souls of my parents, whom under thei r wings I learned how to be fascinated with knowledge and science. To my wife and my sister in law, who have always inspired me to pursue my graduate studies. To my four brother and sisters, with whom I share my love and concerns. To my four children, may God lighten their lives w ith the love of knowledge. To all of the above, I present this work.

PAGE 3

ACKNOWLEDGMENTS This work would not have be en completed without the endless contributions of two persons; my advisor Dr. Robert Carn ahan, and my wife Azzah El-Menshawi. Despite his retirement from the University of South Florida (USF) a couple of years ago, Dr. Carnahan has always been supportive for year s, armed with the real spirit of the great giving scientist, especially at the frequent critical times of my uncertainty. Azzah has been very patient with me, and ha s sacrificed long sleepless nights helping me finish this work. She encompassed me with her passion and kindness. I owe a lot to every professor in my committee. Dr. Nachabe helped me learn how to validate the hypothesis during a time when I was about to lose my way. My friend Dr. Ahmed Said taught me how to professionally present my work. Dr. Sunol was the first professor to give me guidance from the first day I entered USF with a dream of pursing my graduate studies Dr. Ross has challenged and inspired me with his deep scientific knowledge. I also want to thank the senior management of my company Crane Environmental, especially Don Borden and Dennis Greco, who took upon themselves the financial burden of this work. There are many persons that I feel grateful to because they supported this work. Some of them are Colin Leonard, Eugene Ta lly, Mario Van Sevren, Dr. Harold Fravel, Dr. Maya Trotz, and my colleague Dr. Mayssom Sallam.

PAGE 4

NOTE TO READER The original of this document contains co lor that is necessary for understanding the data. The original dissertation is on file with the USF library in Tampa, Florida.

PAGE 5

i TABLE OF CONTENTS LIST OF TABLES iv LIST OF FIGURES vii LIST OF SYMBOLS x ABSTRACT xv CHAPTER 1 INTRODUCTION TO THE RESEARCH 1 1.1 Importance of the Research Topic 1 1.2 Problem Definition 2 1.3 Research Objective 6 1.4 Research Approach 7 1.5 Dissertation Outline 9 CHAPTER 2 CONCENTRATION POLARIZATION IN RO AND NF MODLUE: LITERATURE REVIEW 11 2.1 Background 11 2.2 Preface 12 2.3 Historical Background 12 2.4 Definition of Reverse Osmosis 14 2.5 Water Treatment by Pressure Driven Membranes 15 2.6 Reverse Osmosis Membrane Properties 16 2.7 Reverse Osmosis Membrane and Module Configuration 18 2.8 Concentration Polarization 22 2.9 Previous Studies to Reduce Concentration Polarization 23 2.10 Limiting Factors for Membrane Fouling 25 2.11 The Promise of Nanofiltration Membrane 28 2.11.1 Nanofiltration for Contam inated Drinking Water 28 2.11.2 Nanofiltration for Wastewater 30 2.11.3 Nanofiltration in Hybrid Seawater Distillation 31 2.11.4 Nanofiltration in Repl acing Standard Seawater RO Membrane 32 2.12 New Generation of Nanofiltration Membrane 33 2.13 History of Using Permeate Su ction in Pressure Driven Membrane 36

PAGE 6

2.14 Effect of Increasing Suction Pr essure on the Boundary Layer 38 2.15 Reverse Osmosis Models 41 2.15.1 Irreversible Thermodynamics Models 43 2.15.2 Porous Models 43 2.15.3 Charged Membrane Models 44 2.15.4 Solution-Diffusion Models 45 2.15.4.1 Solution-Diffusion-Imperfection Model 46 2.15.4.2 Assumptions when using Solution-Diffusion Model 48 2.16 Determining Membrane Surface Concentration 49 2.17 Determining Mass Transf er Coefficient and Thickness of Concentration Polarization Layer 51 2.18 Using Sherwood Number to Determine Mass Transfer Coefficient 51 2.19 Overcoming the Disadvantages of using Sherwood Number to Determine Mass Transfer with Permeate Suction 53 2.19.1 Film Theory 53 2.19.2 Peclet Number 55 2.20 Determining Diffusion Coefficient of Strong Electrolytes 57 2.20.1 1-1 Strong Electrolyte 58 2.20.2 Non 1-1 Strong Electrolyte 59 CHAPTER 3 EXPERIMENTAL METHODOLOGY 60 3.1 Dilute Solutions Preparation 60 3.2 Reasons behind Choosing the Chemicals 61 3.3 Experimental Setup 62 3.4 Assumptions of the Experiments 70 CHAPTER 4 RESULTS AND DISCUSSION 72 4.1 Effect of Permeate Su ction on the Concentration Polarization Layer Thickness 72 4.1.1 Statistically Testing the Experimental Design for Concentration Polarization Layer Thickness 80 4.1.1.1 ANOVA for Solutions 80 2MgCl 4.1.1.2 ANOVA for Solutions 81 4MgSO 4.1.1.3 ANOVA for Solutions 82 NaCl 4.2 Effect of Permeate Suction on Mass Transfer Coefficient 83 4.3 Effect of Permeate Suction on Permeate Flow 87 4.4 Effect of Permeate Suction on Permeate Concentration 91 4.5 Effect of Permeate Suction on Concentrate Concentration 95 4.6 Effect of Permeate Suction on Membrane Wall Concentration 99 4.7 Effect of Permeate Suction on Peclet Number 104 ii

PAGE 7

CHAPTER 5 CONCLUSION AND RECOMMENDATIONS 109 5.1 Conclusion 109 5.2 Recommendations 112 5.3 Future Researches 113 LIST OF REFERENCES 114 APPENDICES 118 Appendix A: Tables of Test Results for Solutions 119 NaCl Appendix B: Tables of Test Results for Solutions 130 4MgSO Appendix C: Tabl es of Test Results for Solutions 141 2MgCl ABOUT THE AUTHOR End Page iii

PAGE 8

LIST OF TABLES Table 2-1 Comparison of new genera tion of NF membrane performance at standard operating conditions 34 Table 2-2 Several values of Sherwood number factors found in literature 52 Table 2-3 Ioni c diffusion coefficients in water at 25 degree C dilution in 58 1010 scm /2 Table 3-1 Dilute solutions concentr ation and operating pressures for the experiments 61 Table 3-2 Instrumentati on and specifications 64 Table 3-3 Geometrical parame ters for the tested FilmTec membrane Model NF270-2540 68 Table 4-1 ANOVA Table for Solutions 81 2MgCl Table 4-2 ANOVA Table for Solutions 82 4MgSO Table 4-3 ANOVA Table for Solutions 83 NaCl Table A-1 Readings for distilled water before running solutions NaCl experiments 120 Table A-2 Readings for solution at 730 mg/l and 80 psi feed pressure 121 NaCl Table A-3 Readings for solution at 730 mg/l and 110 psi feed pressure 122 NaCl Table A-4 Readings for solution at 830 mg/l and 160 psi feed pressure 123 NaCl Table A-5 Readings for solution at 1,200 mg/l and 80 psi feed pressure 124 NaCl Table A-6 Readings for solution at 1,200 mg/l and 110 psi feed pressure 125 NaCl iv

PAGE 9

Table A-7 Readings for solution at 1,200 mg/l and 160 psi feed pressure 126 NaCl Table A-8 Readings for solution at 1,750 mg/l and 80 psi feed pressure 127 NaCl Table A-9 Readings for solution at 1,750 mg/l and 110 psi feed pressure 128 NaCl Table A-10 Readings for solution at 1,750 mg/l and 160 psi feed pressure 129 NaCl Table B-1 Readings for distilled water before running solutions 4MgSO experiments 131 Table B-2 Readings for solution at 800 mg/l and 80 psi feed pressure 132 4MgSO Table B-3 Readings for solution at 800 mg/l and 100 psi feed pressure 133 4MgSO Table B-4 Readings for solution at 800 mg/l and 130 psi feed pressure 134 4MgSO Table B-5 Readings for solution at 1,250 mg/l and 80 ps i feed pressure 135 4MgSO Table B-6 Readings for solution at 1,250 mg/l and 100 psi feed pressure 136 4MgSO Table B-7 Readings for solution at 1,250 mg/l and 130 psi feed pressure 137 4MgSO Table B-8 Readings for solution at 1,750 mg/l and 80 ps i feed pressure 138 4MgSO Table B-9 Readings for solution at 1,750 mg/l and 100 psi feed pressure 139 4MgSO Table B-10 Readings for solution at 1,750 mg/l and 130 psi feed pressure 140 4MgSO Table C-1 Readings for distilled water before running solutions 2MgCl experiment 142 Table C-2 Readings for solution at 800 mg/l and 80 psi feed pressure 143 2MgCl Table C-3 Readings for solution at 800 mg/l and 100 psi feed pressure 144 2MgCl Table C-4 Readings for solution at 800 mg/l and 130 psi feed pressure 145 2MgCl Table C-5 Readings for solution at 1,250 mg/l and 80 psi feed pressure 146 2MgCl Table C-6 Readings for solution at 1,250 mg/l and 100 psi feed pressure 147 2MgCl v

PAGE 10

Table C-7 Reading for solution at 1,250 mg/l and 130 psi feed pressure 148 2MgCl Table C-8 Readings for solution at 1,750 mg/l and 80 psi feed pressure 149 2MgCl Table C-9 Readings for solution at 1,750 mg/l and 100 psi feed pressure 150 2MgCl Table C-10 Readings for solution at 1,750 mg/l and 130 ps i feed pressure 151 2MgCl vi

PAGE 11

vii LIST OF FIGURES Figure 1-1 Concentration polarization la yer near membrane surface 4 Figure 2-1 Schematic drawing of wate r and salt fluxes in direct osmosis and reverse osmosis 14 Figure 2-2 Kinds of rejected species by different pressure driven membrane types 15 Figure 2-3 Thin Film Composite (TFC) cross section view 18 Figure 2-4 Structure of thin film barri er layer of RO aromatic polyamide membrane 19 Figure 2-5 Structure of thin film barrier layer of the aromatic/aliphatic nanofiltration membrane 20 Figure 2-6 A cross section in the thin film composite (TFC) spiral wound reverse osmosis membrane showing feed channel spacer 21 Figure 2-7 Configuration of permeate sp acer (top) and feed spacer (bottom) in spiral wound RO/NF element 22 Figure 2-8 Limiting factors for RO and NF membrane fouling 27 Figure 2-9 Inorganic sca ling stages 27 Figure 2-10 Comparison of energy consum ption for seawater desalination with seawater membrane; br ackish waterseawater membrane & NF-NF membrane 33 Figure 2-11 Image of the surface of the new generation NF ESNA1-LF membrane compared to the low pressure RO ESPA3 membrane showing relative surface smoothening 35

PAGE 12

Figure 2-12 Comparison of surface charge of new generation NF ESNA1-LF membrane, typical NF LFC1 membrane, and low pressure fouling RO LFC1 membrane 36 Figure 2-13 Membrane bio-reactor using permeate suction to treat wastewater by immersed UF membrane 37 Figure 2-14 Application of suction to the membrane to prevent boundary layer separation 39 Figure 2-15 Feed side concen tration polarizatio n layer 51 Figure 3-1 Experimental equipment skid showing pressure gauges, TDS meter, and NF membrane pressure vessel 62 Figure 3-2 Variable fre quency drives for the high pressure pump, and the permeate pump 63 Figure 3-3 The assembled high pressure pump (top), and permeate pump (bottom) 63 Figure 3-4 First setup by r unning the high pressure pump only 66 Figure 3-5 Second setup by running the high pressure pump and the permeate pump 67 Figure 4-1 Concentration polari zation layer thickness versus feed concentration 74 2MgCl Figure 4-2 Concentration polariza tion layer thickness versus feed concentration 75 4MgSO Figure 4-3 Concentration polariza tion layer thickness versus feed concentration 77 NaCl Figure 4-4 Concentration polarization layer thickness versus feed concentration at 100-110 psi for the three ,, 2MgCl 4MgSO and solutions 78 NaCl Figure 4-5 Concentration polarizati on layer thickness versus net operating Pressure 79 2MgCl Figure 4-6 Mass transfer coefficien t versus feed concentration 84 2MgCl viii

PAGE 13

Figure 4-7 Mass transfer coeffici ent versus feed concentration 85 4MgSO Figure 4-8 Mass transfer coefficien t versus feed concentration 86 NaCl Figure 4-9 Permeate flow versus feed concentration 88 2MgCl Figure 4-10 Permeate flow versus feed concentration 89 4MgSO Figure 4-11 Permeate flow versus feed concentration 90 NaCl Figure 4-12 Permeate concentrati on versus feed concentration 92 2MgCl Figure 4-13 Permeate concentrati on versus feed concentration 93 4MgSO Figure 4-14 Permeate concentrati on versus feed concentration 94 NaCl Figure 4-15 Concentrate concentrat ion versus feed concentration 96 2MgCl Figure 4-16 Concentrate concentrat ion versus feed concentration 97 4MgSO Figure 4-17 Concentrate concentrat ion versus feed concentration 98 NaCl Figure 4-18 Membrane wall concentrat ion versus feed concentration 100 2MgCl Figure 4-19 Membrane wall concentrat ion versus feed concentration 101 4MgSO Figure 4-20 Membrane wall concentrat ion versus feed concentration 102 NaCl Figure 4-21 Fully developed velocity profiles and concentration profiles in the boundary layer without and with permeate suction 103 Figure 4-22 Peclet number versus feed concentration 106 2MgCl Figure 4-23 Peclet number versus feed concentration 107 4MgSO Figure 4-24 Peclet number versus feed concentration 108 NaCl ix

PAGE 14

LIST OF SYMBOLS A Pure water permeability coefficient A Angstrom 1A Wetted membrane surface area 1a ,, Coefficients in Peclet number model equations 1b 1c B Solute permeability coefficient in I.T. model BW Brackish water b Membrane length FC Bulk feed concentration ic Total dissolved salts iC Molar concentration of the solute avgmC )( Logarithmic mean solute concentration in the membrane memC Membrane wall concentration RC Reject concentration PC Product concentration Qc Dimentionless volume coefficient c Concentration of water in the membrane w D Diffusivity coefficient for solute transport through solvent x

PAGE 15

DBNPA Dibromo nitrilo propionamide 21 D Diffusivity coefficient in the ionized electrolyte DI Deionized water wD Diffusivity coefficient of water in the membrane fd Filament diameter of the feed spacer Boundary layer thickness F Concentration polarization layer feed side P Concentration polarization layer feed side M Active membrane thickness S Porous support membrane thickness DOF Degrees of freedom wD Diffusion coefficient of the solute in the membrane P Hydraulic pressure difference across the membrane Osmotic pressure difference across the membrane I. T. Irreversible thermodynamic EPA Environmental protection agency Porosity of the feed concentrate spacer wJ Permeate flux dh Hydraulic diameter of the membrane HDPE High density polyethylene HP Horse power xi

PAGE 16

Hz Hertz K Mass transfer coefficient 2k Solute permeability coefficient due to diffusion in solution diffusion model 3k Coupling coefficient in solution diffusion-imperfection model sk Distribution coefficient of the solute from the feed into the pore of the membrane F wsDk Solute transport parameter MED Multi effect distillation MF Microfiltration MSF Multi stage flash MWCO Molecular weight cut-off iMW Molecular weight of species N Number of ions formed when the solute dissociates NF Nanofiltration NPSH Net positive suction head sN Total solute flux wN Total water flux PA Synthetic polyamide aromatic eP Diffusive Peclet number Density )(FX Osmotic pressure of the feed side xii

PAGE 17

)(PX Osmotic pressure of the permeate side Ph Phase pH Negative power of hydrogen Q Quantity of fluid removed during suction L Membrane width PL Hydrodynamic permeability coefficient R Gas constant Re Reynolds number LR Reynolds number related to the length of the membrane RO Reverse Osmosis rpm Revolution per minute 2 R Proportion of variability in data Sc Schmidt number Sh Sherwood number SW Seawater T Absolute temperature Frictional shearing stress TDS Total dissolved solids TFC Thin film composite THM Trihalomethane TOC Total organic carbon u Average free stream bulk velocity at the x direction UF Ultrafiltration xiii

PAGE 18

Absolute viscosity m Micro meter s Micro siemens V Volt 0v Velocity due to suction wV Permeate velocity VC Vapor compression VFD Variable frequency drive w With suction w/o Without suction WHO World health organization FW Concentration gradient feed side PW Concentration gradient permeate side Dynamics viscosity of the solution Coefficient of coupling between salt and water in I.T. model Salt permeability coefficient X,a,b,c Coefficients in Sherwood number FX Solute mole fraction feed side pX Solute mole fraction permeate side iZ Valent number of anion or cation xiv

PAGE 19

xv Effect of Permeate Suction on the Performance of Spiral Wound Nanofiltration Module Awad Abdel Monem El-Shamy ABSTRACT Fouling in a nanofitration memb rane module is usually a result of concentration polarization. The effect of permeate sucti on on the slightly negatively charged spiral wound nanofiltration membrane is investigate d. According to the film theory, the mass transfer coefficient is inversely proportional to concentration polari zation. The effect of permeate suction destabilizes the boundary la yer. This will decrease the concentration polarization layer, and conse quently will increase mass transfer through the membranes surface. To validate the hypothesis, experiments were carried out on a NF membrane that can be described by the solution-diffusion model. This model has coefficients that can be measured experimentally. Using the membrane wall concentration in this model instead of the bulk feed concentration can help estimating the mass transfer coefficient more appropriately. Two experimental studies were carried out, one with a standard high pressure pump, and another one with the added effect of suction pressure applied to the permeate collector tube.

PAGE 20

Three different concentrations of binary dilute solutions of ,, and at three different pressures (low medium, and high) were tested. NaCl 4MgSO 2MgCl For all tested solutions, permea te suction increased the diffusive Peclet number as a function of the f eed concentration ( x ) according to the equation = + +, with eP 2 1xa xb1 `1c 2 R > 0.99, where x is the feed concentration in Mol/l, and , and are coefficients dependent on feed pressure for every salt solution. With the increase of the Peclet number, it was observed that the c oncentration polarization decreased, and both the product flow and the product quality were improved. Suction had the greatest impact at the range of 100 to110 psi feed pressure, where the concentration polarization reduced approximately 14 to 20 %. 1a 1b 1c ANOVA for the concentration pol arization showed that su ction was significant in reducing the calculated concentration polarization layer for all tested solutions. It was concluded that permeate suction reduced c oncentration polarization, increased product flow rate, and improved product quality. Thus, adding permeate suction has beneficial consequences because it reduces membrane fouling and extends its useful service life. xvi

PAGE 21

1 CHAPTER 1 INTRODUCTION TO THE RESEARCH 1.1 Importance of the Research Topic The rapid growth of the new generation of nanofiltration as an attractive membrane separation process suggests renewe d investigations of the current design methods for developing an improved design c onfiguration to reduce membrane fouling. Fouling in reverse osmosis (RO)/nanofitrat ion (NF) membrane modules is usually a result of concentration polarization. Me mbrane fouling has a serious economical implications on the water treatment plant because it causes permeate flux decline, reduces product quality, and shortens the life of the membrane. Spiral wound configuration, which is the most dominant module used in th e application of pressure driven membrane for drinking water treatment, wa s rarely investigated by resear chers as far as the permeate suction is concerned. Nanofiltration membrane, which is sometimes called loose reverse osmosis membrane, is more manageable than RO me mbrane for permeate suction because its permeability coefficient is substantially higher than RO membrane. This is despite that NF membrane systems are typically designe d like RO but with much lower driving pressures. Currently NF membranes are traditionally used to treat low salinity water, waste water, or in the proce ss industry like extracting chemicals or protein from dilute water.

PAGE 22

2 The state of the art in NF membra ne researches is taking advantages of using it in seawater desalination, either for pretreatme nt for existing thermal distillation plants; pretreatment of seawater RO plants; or to replace the traditional seawater RO membranes so that it improves production rate, or saves energy (Leon Awerbuch, 2007; Hassan, 2004;Yann Gouellec et al., 2006). This is an indication of the in spiring future that awaits the NF membrane in the pressure driven membrane technology. Despite the numerous studies that have addressed the NF membrane fouling, few researchers have addressed using permeate suct ion as a means of reducing fouling in the widely used spiral wound thin film composite membrane configuration. 1.2 Problem Definition Membrane fouling and scaling can significantly increase the cost of a membrane system as well as reduce its reliability. This lim itation is behind the great deal of research that has made significant developments in membrane science. Fouling is a term generally us ed to describe the undesirable formation of deposits on the surface of the membrane. Membra ne fouling is a complicated phenomenon because it results from a group of physical, chem ical, and biological effects that can lead to an irreversible loss of membrane perm eability (Salbani et al., 2001). Attempts to analyze the fouling phenomena have shown that its primary characteristics are adsorption of feed components, and de position of solids on the memb rane surface, accompanied by crystallization and compaction of the membra ne structure. However, the occurrence of fouling is almost always a result of con centration polarization (Jamal et al., 2004).

PAGE 23

3 Concentration polarization (Figure 1-1), may be defined as the presence of a higher concentration of rejected species at the su rface of a membrane than in the bulk solution, due to the convective transport of both solute and solvent. It is generally considered a totally reversible effect (Jamal et al, 2004) The reduction of con centration polarization layer (F ) is important for the improvement of the performance of osmotic type membranes, as it will inevitably lead to reduc tion in the fouling of the membrane (Jamal et. al, 2004). Depending on molecular weight, which will determine diffusive back-transport from the membrane, concentration pola rization is more or less distinct. Although concentration polari zation can also be found on the permeate side as indicated in Figure 1-1, it is usually neglected in pressure driven membrane since it is much less pronounced than feed side polarization (Fritz mann et al., 2007). Concentration polariza tion has several negative effects on membrane performance: (1) rejection d ecreases due to higher salt fl ux because of increased salt concentrations at the membrane surface; (2), so lubility limits can be exceeded, especially for divalent ions, leading to a precipita tion layer on the membrane surface, which negatively influences mass transfer, that dr amatically reduce permeate flux; (3) water flux is reduced due to higher osmotic pressure associated with higher salt concentration at the feed side membrane surf ace; (4) and particles are ac cumulated at the membrane which can lead to cake formation on the surface.

PAGE 24

4 Figure 1-1.Concentration polarization layer near membrane surface ( adopted from Fritzmann et al., 2007) Several approaches have been used to try to minimize the effects of fouling. In thin film composite spiral wound module, the hydraulic flow is laminar due to channels between the membrane layers. The ma ss transfer coefficient is the most widely used parameter in the design of pressure-d riven membrane separation systems such as reverse osmosis and nanofiltration. The role of suction in mass transfer through porous membranes is very important. It was identi fied by several researchers (e. g. Van den Berg et al., 1989; Gekas et al ., 1987) that the effect of pe rmeate suction enhances the mass transfer from the bulk to the membrane surface. By applying suction at the end of the collector tube of the membrane modul e, an increased rate of pressure along the stream will present. This increase in pr essure rate will destabilize the boundary layer at steady state conditions (Schlichting, 1979). The conventional way to estima te the mass transfer coefficient is to use Sherwood number relationships obtained from heat and mass transfer analogy.

PAGE 25

5 Numerous Sherwood number relationships have been proposed and extensively reviewed (C. Van de Liskdonk et al., 2000). The Graetz-Leveque correlation of Sherwood number, which is used for laminar flow when the velocity field is fully developed and the concentration boundary laye r is not fully develope d, is typically used to estimate mass transfer coefficient as: Sh = Xa(Re)bSc c hL d (1-1) where Re = Reynold's number = h wd V; Schmidt number = Sc = D ; = kinematics viscosity; wV = average cross-flow permeate velocity; hd= hydraulic diameter of the membrane element; D the diffusion coefficient for solute transport through solvent, and L is the spiral wound membrane width (Taylor et al.,1999). D in this relationship is equal to K F where K is the mass transfer coefficient, and F is the concentration polarization layer thickness. Th e terms X, a, b, and c are co efficients that have taken extremely different values by differe nt researchers (e.g. Isaacson,1979; Schocket Miquel,1979; Taylor, 1991). Further discussi ons about these terms will be included in Chapter 2. There are several lim itations upon using the above-mentioned equation: (1) the above mentioned Sherwood number re lationship is derived for flow through nonporous conduit; hence, the effect of suc tion can not be considered using these relationships; (2) the axial change in osmotic pressure at membrane surface due to the concentration polarization change is not considered in the above mentioned Sherwood number correlations; (3) and suction will change the species concentration at the

PAGE 26

6 membrane surface that will change the solutions physical properties like viscosity, density, and diffusivity, whic h are functions of the con centration. Consequently, the above-mentioned Reynolds number, and Schmidt number, will be variable along the membrane length. Those changes are not cons idered with this form of Sherwood number relationship. The diffusive Peclet number is expressed as eP= D h Vd w (1-2) where wVis the cross flow permeate velocity;D is the diffusivity coefficicent of the species; and dh is the hydraulic diameter of the spiral wound membrane. The diffusive Peclet number measures th e dimensionless ratio of convective mass transfer to the membrane to the diffusive mass transfer towards the bulk solution at the opposite direction. The Peclet number is also called the dimensionless flux. If the diffusive Peclet number is increased due to suction, while the associated concentration polarization is being reduced, this means that the suction has increased membrane production with more favorable conditions to th e membrane, as far as inorganic fouling is concerned. Therefore, the Sherwood number ca n be avoided in the calculations due to the above-mentioned limitations. 1.3 Research Objective The objective of this research is to investigate the effect of permeate suction on the mass transfer coefficient, concentr ation polarization layer, product quality, production flow rate, and membrane diffusi ve Peclet number for spiral wound NF module.

PAGE 27

7 The goal is to increase system permeate flow without subjecting the membrane to an increasing tendency for inor ganic precipitation. This was carried out by comparing the data collected from running two tests on the me mbrane: the first test will be run using the standard high pressure feed pump only, and second will be done by running the test after adding the effect of the permeate su ction pump to the original setup. 1.3 Research Approach For high rejection membra nes of the type used in reverse osmosis and nanofiltration membrane applications, the water flux can be presented by the solution diffusion model (Lonsdale et al.1965; Soltani eh and Gill, 1981), which states that the solvent flux is proportional to the effective pressure difference ( P ). The solvent flux is caused by the gradient of chemical potential which includes a concentration diffusion term, and a pressure diffusion term. For real membranes that have some imperfections like holes or microspores, the measured flux is not purely diffusive, but it contains a term contributed by convection. Recen t researches did not find pores in neither RO nor NF membranes, such that the tran sport of solvent is accomplished through the free volume between the segments of the polym ers of which the membrane is constituted (William, 2003). In diffusion controlled hyperfil tration (RO and NF) membrane process, the solution-diffusion imperfection based mode l is widely used, because most of the coefficients used in this m odel are actual operating condi tions that can be directly measured, as opposed to theoretic al models that have paramete rs difficult to be measured in reality (Williams, 2003). Furthermore, for d ilute solutions, which is the typical solution

PAGE 28

8 fed to NF membrane, this convection term is so small such that it can be neglected (Soltanieh and Gill, 1981) to further simply the real model. Mass transfer models typically assume that the bulk feed solution concentration is equal to the membrane wall solution con centration, which is not always true. This has to be related to the concentrat ion polarization expressions (Williams, 2003). Concentration polarization complicates the m odeling of membrane systems because it is very difficult to experimentally determ ine the membrane wall concentration (memc). Membrane wall concentration is necessary to be determined since it a nd not the bulk feed concentration (Fc) should be used in RO and NF transport models. In the limited feed flow rate th at is typically used for hyperfiltration (RO and NF) membrane processes, the flow in the membra ne channels is laminar, and the difference between the wall and bulk concentrations can be substantial, so calculations of the concentration at the membrane wall must be appropriately estimated. Bhattacharya et al. (1996) have developed a generalized mass-transfer relationship from first principals to obtain a theoretically modified form for the Sherwood number using the wall Peclet number to estimate th e mass-transfer coefficient using permeate suction. This correlation is debated due to the above-mentioned items that were discussed earlier. The permeate suction was teste d, and was experimentally validated for NF spiral wound TFC module in this res earch. An experimental set up for the membrane system was tested with the conventiona l operating setup in order for it to be compared with the permeate suction setup results. The research was conducted for three different dilute

PAGE 29

9 strong electrolyte binary solutions, namely:NaCl,2MgCl, 4MgSO, which are 1-1, 2-1, and 2-2 electrolytes, respectivel y, at three different pressures and three different dilute concentrations. The experiment was set at a c onstant temperature of 25 degrees Celsius in order to keep the diffusion coefficient cons tant for the different dilute solutions. 1.5 Dissertation Outline Chapter 2 of this dissertation is di vided into five parts. The first part presents general information related to RO and NF membrane properties and module conf igurations. It also includes literature review of the previous researches accomplished to reduce concentration polarization in RO and NF. The second part emphasizes the properties of NF membranes, and their benefits in recent developments in either brackish water de salination or seawater desalination. It also explores the distinguished importance of the new generation of NF membranes. The third part demonstrates the hi story of using permeate suction in pressuredriven membranes; and discusses the effect of gradually increasing suction on the boundary layer in fluid dynamics. The fourth part illustrates various mass transport models, describing the advantages and disadvantages of each model. Emphasis was placed on the solution-diffusion model which is the mass transport model used in this research. The fifth part in Chapter 2 disc usses the theories related to different equations that will be utilized to avoid using the Sher wood number relationship for mass transfer estimations. Chapter 3 presents the experi mental setup that is bein g used to validate the hypothesis. A detailed description of the a ssumptions, equipment, solutions of salts, and chemicals used in the experiments are presented.

PAGE 30

10 Chapter 4 of this disserta tion explores the results obtained from the abovementioned experiments; along with deta iled discussions for the results. Chapter 5 exhibits the c onclusion from the experi mental results and the discussions. Recommendations we re presented in this chapte r, along with the suggested future permeate suction researches that can be conducted on highe r concentrations of brackish water and seawater membranes to redu ce concentration polarization in order to elongate the membrane life.

PAGE 31

11 CHAPTER 2 CONCENTRATION POLARIZTION IN RO AND NF MODULE: LITERATURE REVIEW This chapter is divided into five main parts. The first part presents general information related to RO and NF membra ne properties and module configurations. The second part explores th e distinguished importance of the new generation of NF membranes, and the promising future of this important type of membrane. The third part demonstrates the history of using perm eate suction in pressu re-driven membranes; and discusses the effect of gradually incr easing suction on the bounda ry layer in fluid dynamics. The fourth part demonstrates va rious mass transport models, describing the advantages and disadvantages of each mode l. This includes the solution diffusion model for mass transport in the NF membrane, and the advantage of using it in the application to the new generation of NF membrane. The fifth part demonstrates the theories used to prove the hypothesis related to the chosen mass transport model in dilute solutions, which are used as a feed for NF membrane. 2.1 Background Information similar to that pr esented in this section can be found in numerous publications describing hyperfilt ration membranes and processes. References are made only occasionally and mainly when information is specific to a source.

PAGE 32

12 2.2 Preface Water shortages and lack of access to safe drinking wa ter will continue to be major global problems. At present, more th an one billion people lack access to safe drinking water, and 2.4 billion people lack acce ss to proper sanitation, nearly all of them in developing countries. At present a third of the world's population live in water-stressed countries, and by 2025, the number is expected to rise to two-thirds. Scarcity of fresh water has serious implications on human beings. It can slow down economic expansion, reduce agricultural output, hamper food independence, and degrade public health and quali ty of life. Since it was fi rst introduced in the 1950s, hyperfiltration membranes (reverse osmosis and nanofiltration) have most commonly been used for desalting seawater, and brackish water by removing salts and other impurities in order to improve the color, tast e or properties of the water for drinking and irrigation. Hyperfiltration is finding increasing uses in indus trial applications for highly pure water because of its reliability and cost-effectiveness. Membrane separation has gained considerable importance because th ey offer superior tr eatments at relatively modest capital and operating cost (Madireddi et al., 1999). However, membrane fouling will continue to be the major obstacle fo r the efficient operation of RO membrane systems (Jamal et al., 2004). 2.3 Historical Background The ancient Egyptians treate d water by siphoning water out of the huge jars after allowing the muddy Nile River to settle and se parate; the first United States water plant with filters was built in 1872 in Poughkeepsi, New York Membrane filtration represents

PAGE 33

13 the advanced ring in that hist orical development of water tr eatment, and RO is the finest level of filtration available. Th e concepts of "direct osmosis" and "reverse osmosis" have been known for many years. In fact, studies on osmosis were carried out as early as 1748 by the French scientist Nollet, and many rese archers investigated these phenomena over the next two centuries (Reid, 1966; Mason, 199 1; Williams, 2003). However, the use of reverse osmosis (RO) as a feasible separation process is a relatively young technology. In fact, only in the late 1950's did the work of Reid show that cellulose acetate RO membranes were capable of separating salt from water, even though the water fluxes obtained were too small to be practica l (Reid and Breton, 1959; Ferguson, 1980; Lonsdale, 1982; Applegate, 1984). Then, in the early 1960's, Loeb and Sourirajan developed a method for making asymmetric ce llulose acetate membranes with relatively high water fluxes and separations, thus making RO separation both possible and practical (Loeb and Sourirajan, 1962; Loeb, 1981; Sourirajan and Matsuura, 1985). Since then, the development of newer generation me mbranes such as the thin-film composite membrane that can tolerate a wider pH ra nge, higher temperatures, and harsh chemical environments, and that have highly imp roved water flux and solute separation characteristics has resulted in many RO applications. In addition to the traditional seawater and brackish water desalination pr ocesses, RO and NF membranes have found uses in wastewater treatment, production of ultrapure water, water softening, and food processing, as well as many other ap plications (Bhattacharyya, 1992).

PAGE 34

14 2.4 Definition of Reverse Osmosis Osmosis is a natural phenomenon th at occurs in all living cells in which a solvent passes through a semi-permeable barrier from th e side with lower solute concentration to the higher solute concentration. Reverse os mosis is based on a property of certain polymers called semi-permeability. While they are very permeable for water, their permeability for dissolved substances is low. By applying a pressure difference across the membrane the water contained in the feed is forced to permeate through the membrane. In order to overcome the feed side osmotic pressure, fairly high feed pressure is required. 2-1a 2-1b Figure2-1. Schematic drawing of water and salt fluxes in direct osmosis and reverse osmosis (Adopted from Ghiu et Carnahan, 2003) As shown in Figure 2-1a, solv ent flow continues until the chemical potential equilibrium of the solvent is established. At equilibrium, the pressure difference between the two sides of the membrane is equal to the osmotic pressu re of the solution. To reverse the flow of water (solvent) a pressure di fference greater than the osmotic pressure difference is applied as illustr ated in Figure 2-2b. As a resu lt, separation of water from

PAGE 35

15 the solution occurs as pure water flows from the high concentration side to the low concentration side. This phenomenon is termed reverse osmosis or hyperfiltration. The RO process is attractive because it is relatively simple in design. It consists of a feed water source, feed pretreatment, hi gh pressure pump, RO membrane modules, and in some cases post-treatment steps. 2.5 Water Treatment by Pressure-Driven Membranes The membrane processes that has the greatest immediat e application to potable water treatment are reverse osmosis (RO), nanofitration (NF), ultr afitration (UF), and microfiltartion (MF). Figure 2-2 shows the kind of rejected species by different types of pressure-driven membranes. Figure 2-2. Kinds of rejected species by different pressure-driven membrane types (Adapted from Koch membrane manufacturer, 2009) Reverse osmosis is primarily used to remove salts from brackish water or seawater, and it is also capa ble of very high rejection of synthetic organic compounds (SOCs). Nanofiltration is used to soften fres h water, and remove disinfection by-product (DBP) precursors. Ultrafiltr aion, and microfiltration are used to remove turbidity, pathogens, and particles from fresh water. A membrane, the common element of all these

PAGE 36

16 processes, could be defined as any barrier to the flow of suspended, colloidal, or dissolved species in any solvent. Contaminants larger than the maximum pore size of the membrane are removed by sieving in a diffusion-controlled process. MF and UF membranes have por es in the filtration layer of the membrane, while the active layer in RO and NF membranes is nonporous. The transport of solvent in RO and NF is accomplished through the free volum e between the segments of the polymers of which the membrane is constituted (William, 2003). Contaminants rejection by di ffusion-controlled membrane processes increases as species charge and molecular weight increa ses. Consequently, satisfactory removal of metals, total dissolved solids (TDS), radionuclides, and disinfection by-products precursors can be attained. Membranes are classified by molecular weight cutoffs, solute and solvent solubility in the membrane film, active memb rane material, active film thickness, surface charge, and smoothness of the active film surface. 2.6 Reverse Osmosis Membrane Properties Reverse osmosis membrane separation is governed by the properties of the membrane used in the process. These properties depend on the chemical nature of the membrane material as well as its physical structure. Most currently available RO membranes fall into two types of membranes: asymmetric membranes, and thin film composite (TFC). Asymmetric membrane containing one polymer, and thin film composite membranes consist of two or more polymer layers. Asymmetric RO membranes have a very thin, perm selectiv e skin layer supported on a more porous sub-

PAGE 37

17 layer of the same polymer. This membrane is used to produce the hollow fine fiber (HFF) configuration. The hollow fiber element consists of large number of fine hollow fiber membranes (with an outer diameter up to 200 m) placed in a pressure vessel; the feed flows outside the fibers and permeates through them (Allegrezza, 1988; Baker, 1990; Bhattacharyya et al., 1992). These elements have an extremely high packing density, and so can have high permeate production rates per module. However, these modules are highly prone to fouling, and thus are not feasible for many of the applications. Thin film composite membra ne is the one used in spiral wound membrane configuration. The dominant form of the synt hetic materials is TFC aromatic polyamide membrane. The development of the cross-linked fully aromatic polyamide thin film composite membrane in the 1970s represented a major advance in membrane technology. TFC membrane provides very thin ac tive film that requires much less energy to induce fluid passage than other material s, making them more economical to use on a large scale. Figure 2-3 s hows the composition of both the hydrophilic, and the hydrophobic TFC membrane. Both hydrophilic and hydrophobic films are laid in a composite film by cross-linking different poly mers. The thickness of the nonporous layer is typically less than 1 m. The widely used aromatic polyamide membranes are unfortunately susceptible to oxidation, and ar e often impacted by the side chain reaction between the disinfectant oxidi zing agent like chlorine, a nd the polyamide groups. This reaction disrupts their stable linkages, and c onformational structure. Consequently, this renders them ineffective for their intended f unction. There have been several attempts to create chlorine-resistance membranes for more than 25 years, but without success (Mukiibi, 2008).

PAGE 38

18 Membrane Surface (Polyamide Layer = 0.15 micron) Polyester Fiber = 150 microns Polysulfone Solution (Interstitial) Polysulfone Layer = 50 microns Barrier Interface Figure 2-3. Thin film composite (TFC) cross section view (Adopt ed from Hydranautics Membrane Manufacturer, 2009) 2.7 Reverse Osmosis Membrane and Module Configuration The available membrane module s using asymmetric type include plate-and-frame, tubular, and spiral-wound confi guration. Plate-and-frame modules consist of stacks of flat sheet membrane placed on supports; each membra ne and support are separated by spacers which direct the feed across each membrane and channel permeate out of the module (Allegrezza, 1988; Baker, 1990; Strathmann, 1990; Bhattacharyya et al., 1992). Tubular membrane elements consist of membrane tubes supported within perforated stainless steel t ubes; as feed flows through the tubes, the permeate passes through the membrane and the support. While the plate-and-frame module, and the tubular module are resistant to fouling, they have low membrane surface area per element. This makes them expensive and can limit their use in areas with space restrictions.

PAGE 39

19 Figure 2-4. Structure of thin film barrier layer of RO aromatic polyamide membrane (Adopted from FilmTec Membrane Manufacturer, 2009) While these elements are also fouling resistant, and are easy to clean, the modules have a low packing density, and can be expe nsive to operate because of the necessary high feed flow rates. Because of the plateand-frame, and tubular element disadvantages, these modules are used primarily for highly foul ing feeds, or in labo ratory researches. Figure 2-4, and Figure 2-5 illustrate the chemi cal structure of the RO and NF thin film composite polyamide, respectively. The spir al wound membrane configuration is the most common membrane for production of drinking, and industrial process water (Allegrezza, 1988; Bhattacharyya, 1992). This type of element has a high packing density, moderate fouling resistance, and lowe r capital and operating costs compared to plate-and-frame or tubular modules. The t ypical configuration of the spiral wound element leaves the membrane easily accessible to cleaning agents. Due to that, the spiral wound membrane can be cleaned more thor oughly, and it is less subject to fouling compared to HFF membranes (Williams et al., 1992). Spiral wound elements are manufactured using flat sheet membranes. A typical spiral wound elemen t, as shown in Figure 26, consists of envelopes (leaves) attached to a center tube th at collects the permeate stream.

PAGE 40

20 Figure 2-5. Structure of thin film barrier layer of the aromatic /aliphatic polyamide nanofiltration membrane (Adopted from FilmTec Membrane Manufacturer, 2009) The sheet itself consists of two layers forming a folded envelope. The envelope is glued along three open sides and near the fol d, completely enclosing the permeate spacer. The glue line on the fold end is a short distan ce away from the fold, because the fold end is attached to the center collec tion tube. The glue line at the fo ld end stops the flow of the feed stream, and allows the remaining pressu re in the permeate stre am to drive it through the membrane into the center collection tube. An envelope is formed by folding one flat sheet over a permeate stream spacer. Feed spacer and permeate spacer, shown in Figure 2-7, are attached to each envelope prior to establishing the fold end glue line. Several envelopes, including feed spacers, and permeate spacers are attached to the center collection tube and wrapped in a spiral around it. An epoxy shell or tape wraps are applied around the envelope, completing th e spiral wound element. The feed stream enters the end of the spiral wound element in the channel created by the feed stream spacer. The feed stream can flow either in a pa th parallel to the center collection tube or through the active membrane f ilm and membrane supports into a channel created by the permeate stream spacers.

PAGE 41

21 Figure 2-6. A cross section in th e thin film composite (TFC) spiral wound reverse osmosis membrane showing feed channel spacer ( adopted from Fritzmann et al., 2007) The permeate stream follows a sp iral path into the center collection tube, and is taken away as product for point of use. The recovery in a spiral wound element varies from approximately 5 to 20 percent. The Re ynolds number typically ranges from 100 to 1,000. The feed stream spacer creates additional turbulence and increases the Reynolds number (J. S. Taylor, 1999). The highest and lowest feed stream velocities occur at the entrance, and the exit of the el ement, respectively. The feed flow is in the laminar region, and the last element in series is the one which is most likely subjected to chemical fouling, if the species in the feed water are subjected to super saturation. Fouling from particle deposition could occur mainly in th e first element in series. Because of the importance of the membrane module used in the RO process, much research has been performed to optim ize the design of each element type. As a result, many models describing the various m odules are available, such that allowing determination of different module hydrodynamics or optimizing the membrane spacer placement and height.

PAGE 42

22 Figure 2-7. Configuration of permeate spacer (top) and feed spacer (bottom) in spiral wound RO/NF element (adopted from Hydranuatics membrane manufacturer, 2009) Reverse osmosis membrane modul es can be arranged in several configurations in the RO process (Williams et al., 1992). For a single-pass arrangement, a single high rejection membrane sufficiently removes the solute from the feed. In a double-pass configuration, the permeate of one set of membranes is used as the feed to another set of membranes in order to provide adequate overall removal of the solute. The modules can also be placed in stages in order to increas e water recoveries. In this configuration, the concentrate from one set of membranes is used as the feed for another set, and consequently high overall wate r recoveries are possible. 2.8 Concentration Polarization Concentration polarization, which is illustrated in Figure 1-1 in Chapter 1, is the term used to describe the accumulation of rejected solute at the surface of a membrane so that the solute concentration at the membrane wall is higher than that of the bulk feed

PAGE 43

23 solution. As water passes thr ough the membrane, the convectiv e flow of solute to the membrane surface is much larger than the diffusion of the solute back to the bulk feed solution. As a result, the concentr ation of the solute at the memb rane wall increases. Reviews of concentration pol arization are give n by Matthiasson and Sivik (1980), Gekas and Hallstrom (1987), Rautenbach and Albr echt (1989), and Bhattacharyya and Williams (1992). Possible negative effects of concentr ation polarization incl ude: (1) decrease in water flux due to increased osmotic pressure at the membrane wall; (2) increase in solute flux through the membrane because of incr eased concentration gradient across the membrane; (3) precipitation of the solute if the surface concentration exceeds its solubility limit, leading to sc aling or particle fouling of the membrane, and reduced water flux; (4) changes in membrane separation pr operties; (5) enhan cement of fouling by particulate or colloidal materials in the feed which block the membrane surface and reduce water flux. The extent of concentrati on polarization can be reduced by promoting good mixing of the bulk feed so lution with the solution near the membrane wall. Mixing can be enhanced through membrane module opt imization of turbulence promoters, or feed spacer geometrical configuration and he ight, or by increasing axial velocity to promote turbulent flow. 2.9 Previous Studies to Redu ce Concentration Polarization Several techniques that have the potential to reduce the concentration polarization to control the fouling have been proposed and adopted. One method is to adjust the operating parameters, e.g. using an inte rmittent mode of operation, or employing variable means to reduce concentration polarization. Both of these phenomena are

PAGE 44

24 impacting flux (Mahlab, 1978). Other tec hniques include increas ing the flow rate; assembling an intensifier for turbulent flow ; the use of impulse methods and agitating methods; the periodic depressu rization of the membrane tube flow reversal, pre-coating of the membrane surfaces; enzyme immob ilization; modification of the membranes polymeric structure; and the mechanical and ultrasonic vibration of the membranes (e.g., Mahlab, 1978; Cruver, 1973). The turbulen ce promoter acts to reduce concentration polarization and therefore fou ling is decreased by increasing the friction factor and bulk velocity. A model has been developed by Ch iolle for reverse osmosis with turbulence promoting nets for the parallel wall cha nnels module (Chiolle et al, 1978). The model developed by Drioli and Bellucci shows the effect of the interaction between concentration polarization and solute-mem brane on the pressure driven membranes, when a multi-component solution is involved (Drioli and Bellucci, 1978). The modification of the membrane s polymeric structure play s an important role in the reduction of concentrati on polarization through the flui dized bed that was developed by Van der Waal (Van der Waal, 1977). Bha ttacharyya developed a finite elements program to compute the concentration profile throughout a reverse osmosis membrane module to predict the perfor mance of the module. The finite element method allowed rapid evaluation of various membrane m odule configurations, such as tapered cell geometry and channels containing spaces (Bhattacharyya et. al, 1990; Gupta, 2005). The model developed by Van der Meer has shown that an increase of 20% in the permeate productivity of the spiral wound RO process is achievable by lowering the number of membrane modules from six per vessel to two in a pre ssure vessel (Van der Meer et al., 1988).

PAGE 45

25 2.10 Limiting Factors for Membrane Fouling Membrane processes are not only limited by increasing osmotic pressure due to concentration polarization a nd rising overall concentrati ons along the membrane, but by other factors leading to reduced performan ce and they can be differentiated by their mechanism. Various chemicals can harm the active layer of the membrane, leading to irreversible damage associated with reduced rejection capability and even destruction of the membrane. Oxidants used in pre-treatment of the reverse osmosis feed water, or as cleaning chemicals are the most important gr oup of chemicals responsible for membrane deterioration. In addition, polymeric membrane s are more or less susceptible to very low or high pH values. Therefore pH adjustment and control is necessary to ensure stable operation. During operation of a reverse osmosi s plant, care has to be taken that no dissolved, colloidal or biologic matter accumu late at the membrane surface, building a continuous layer that reduces or inhib its mass transfer across the membrane. Precipitation on the membrane is caused by super-saturation of inorganic compounds concentrated on the feed side. S uper-saturated salts can precipitate on the membrane surface building a thin layer, which hinders mass transfer through the membrane. Scaling always occurs at the me mbrane surface because of the increased salt concentration near the membrane caused by c oncentration polarizati on. Some of the most important scaling substances are 3CaCO, 4CaSO, 4BaSO, 4SrSO, 2CaF, 2) ( OH Mg, and 2SiO. Scaling can drastically reduce permeate flux, and has to be avoided by all means. Figure 2-8 illustrates the species that can foul or scale the membrane that leads to deteriorated performance. Most susceptible to scaling is the downstream part of the RO stage where concentration in the feed solution is the highest. Therefore, pre-treatment is

PAGE 46

26 used for stabilization of substa nces that could cause scali ng. By pH adjustment, and the use of antiscalants, precipitation can be inhib ited. Crystal growth is usually divided into three stages as shown in Fi gure 2-9. Antiscalants inhibit one or more of these building stages (Fritzmann et al., 2007). Membrane f ouling is caused either by convective and diffusive transport of suspended or colloid al matter, or by bio-fouling. An existing fouling layer adds to the overall resistance to mass transfer of the membrane and overall performance decreases significantly. In a ddition, membrane fouling also increases pressure loss along the membrane, while reject ion is decreased. In RO operations, fouling can never fully be prevented even with optim ized pre-treatment. Th erefore, periodical membrane cleaning has to be performed. Comp lete removal is not possible and fouling has to be tolerated up to a decrease of ma ss flux down to 75% of original flux (Fritzmann et al., 2007). Good operating practice calls fo r chemical cleaning of the membranes, either normalized permeate flow decreases by 10%, feed channel pressure loss increases by 15%, or normalized salt rejection decrease s by 10% from initial conditions during the first 48 hours of plant operation. A key phase in the membrane separation processes is the transition from concentration polarization to fou ling. This occurs at a critical flux. Song (1998) developed a mechanistic model, based on first principles, for predicting the limiting flux. He showed that fo r a given suspension there is a critical pressure below which a concentration polar ization layer will exist at the membrane surface.

PAGE 47

27 Figure 2-8. Limiting factors for RO and NF membrane fouling (adopted from Fritzmann et al., 2007) Figure 2-9. Inorganic scaling stages (Adopted from Fritzmann et al., 2007) However, a cake layer wi ll form between the polar ization layer, and the membrane surface when the applied pressure exceeds a critical pressure. The limiting or critical flux values predicted by the mechanistic model compared well with the integral model for a low concentration feed. Howe ver, it deviated at high solute feed concentrations (Salbani et al., 2001).

PAGE 48

28 2.11 The Promise of Nanofiltration Membrane NF membrane is sometimes ca lled a loose RO pressure-driven membrane process because of its relatively much higher permeab ility coefficient. NF processes operate at pressures between 50 psi, and 150 psi much lower than reverse osmosis (200 to 1,000 psi), but higher than ultrafiltration, and mi corfiltration (10 to 70 psi). The molecular weight cut-off (MWCO) is ge nerally between 300 and 1,000 Dalton. In treating brackish water, NF has been widely used due to more stringent drinking water regulations. Brackish water de salination is assumed to grow at higher rates than seawater desalinati on in the near future (Fritz mann et al., 2007). Delivery of fresh water from seawater desalination plan ts demands piping and pumping systems to transport product water from coastal regions to residential areas, which increase cost. High availability of most br ackish water in residential areas makes expensive delivery piping and pumping unnecessary. 2.11.1 Nanofiltration for Contaminated Drinking Water Nanofiltration membrane, although a relatively recent devel opment, has attracted a great deal of attention for us e in water softening, and removal of various contaminants from drinking water sources (Williams, 2003) NF membranes are usually negativelycharged, and, as a result, ion re pulsion is the major factor in determining salt rejection. For example, more highly charged ions such as 4SO, Ca, and Mg are rejected by nanofiltration membranes to a greater ex tent than monovalent ions such as Cl, or Na. NF processes can reduce or remove TDS, ha rdness, color, agricultural chemicals, and high molecular weight humic and fulvic materi als, which can form trihalomethanes when

PAGE 49

29 chlorinated. Dykes and Conlon (1989), C onlon and McClellan (1989), Watson and Hornburg (1989), and Conlon et al. (1990) have identified NF as an emerging technology for compliance with THM regulations and fo r control of TDS, TOC, color, and THM precursors. Clifford et al. (1988) discussed the use of NF70 membranes of FilmTec for contaminated groundwater treatment. Remova ls included 91% for radium-226 and 87% for TDS. Taylor et al. (1989) reported that NF70 membranes could allow control of THM formation, TOC, TDS, and produce high qua lity product water from an organic contaminated groundwater. They indicated th at the cost of a NF process would be competitive with conventional treatment proc esses which do not control THM formation. Amy et al. (1990) used NF70 membranes to remove dissolved organic matter from both groundwater, and surface water in or der to reduce THM precursors; they found that the process was effective in reducing the organics as well as conductivity in both water sources. NF membranes also reject organic compounds with molecular weights above 200 to 500. These properties have made po ssible some interesting new applications in wastewater treatment, such as selective separation, and rec overy of pollutants that have charge differences, separation of hazardous or ganics from monovalent salt solutions, and membrane softening to reduce hardness, and THM precursors in drinking water sources (Eriksson, 1988; Cadotte et al., 1988; Williams et al., 1992). Arsenic, which is the most extensive environmental poisonous chemi cal element throughout the world, can be removed by NF to meet World Health Or ganization (WHO) standards (Larry Henke, 2008).

PAGE 50

30 2.11.2 Nanofiltration for Wastewater Nanofiltration has also been used to remove both organics and inorganics in various wastewaters. Bindoff et al. (1987) re ported the use of NF membranes to remove color-causing compounds from effluent contai ning lignin, and high salt concentrations in a wood pulping process. Color removals were >9 8% at water recoveries up to 95% while the inorganic was poorly rejected, allowing th e use of low operating pressures, since the osmotic pressure of organic matters is small. Ikeda et al. (1988) i ndicated NF could give high separations of color-causing compounds su ch as lignin sulphonates in paper pulping wastewaters. Afonso et al. (1992) found NF removal (>95%) of chlorinated organic compounds from alkaline pulp and paper bleaching e ffluents with high water fluxes. Simpson et al. (1987) reported the use of NF membranes to remove hardness and organics in textile mill effluents. Gaeta and Fedele (1991) also indicated that high water recoveries (up to 90%) from textile dye house effluent could be achieved with NF membranes. Ikeda et al. (1988) and Cadotte et al. (1988) report ed the use of NF membranes in the treatment of food processing wastewaters. Some specific uses included the desalting of whey and the reduction of high BOD and nitrat e levels in potato processing waters (Anonymous 1988). Bhattacharyya et al. (1989) used NF membranes to selectively separate mixtur es of cadmium and nickel. Chu et al. (1990) detailed the use of NF in a process for treating uranium wa stewater; uranium rejections were 97% to 99.9%. Dyke and Bartels (1990) discussed the use of NF membranes to replace activated carbon filters for the removal of organics from offshore produced water containing residual oils. The produced waters contai ned ~1,000 mg/l soluble organics (mostly carboxylic acids) and high inorga nic concentrations (~15,000 mg/l Na and ~25,000

PAGE 51

31 mg/l Clas well as other dissolved ions). Orga nic rejections were suitable to meet discharge standards, while inorganics reject ions were low (<20%), allowing operation at low pressures. 2.11.3 Nanofiltration for Hybrid Seawater Distillation The use of NF membrane is th e state of the art in sea water distillation industry (Anwerbuch, 2007). The process comprises the operation of the NF selective membrane to soften the feed to distillation units. NF membrane substantially increases the water production from the mature technology of multi stage flash (MSF), muli effect distillation (MED), and vapor compression (V P) distillation techniques. The scaling in sea water disti llation systems occurs due to inverse solubility of calcium sulfate at higher temperature. In or der to increase the wa ter production of the existing distillation units, it is required to increase operation temperat ures, so that higher recovery or higher concentra tion factor are obtained. NF selective membranes were used to reject the high content of su lfate, and hardness in the sea water before it is fed to the distillation units. This allows the operators to optimize the operation of the units to run at higher temperatures than that it was designed for because of the reduction of sulfate and hardness in the re-circulated seawater after being pretr eated with NF membrane. Experiments show that the water temperature was increased from design of 105 degree C to a maximum of 117.9 degree C, whic h helped to achieve a product capacity increase of 40%, and a decrease in operati ng cost by 40% for the operation of MSF plants (Anwerbuch, 2007). Hassan (2004) from Saline Water Conversion Corporation (SWEC) of Saudi Arabia has successfully introduced a new concept to seawater desalination by

PAGE 52

32 combining the NF membrane process with one or more of the conventional seawater desalination processes in one fully integrat ed process system to form: a NF-SWRO, a NF-MSF, and a NF-SWRO reject -MSF, which we re successfully evaluated at the pilot, and demonstration plant level. The NF-SWR O hybrid has increased the productivity by 42%, and raised SWRO unit water recovery ra tio to 56% from 28%. After four years in operation, no SWRO membrane replacement, or cleaning were needed. 2.11.4 Nanofiltration Membrane in Replaci ng Standard Seawater RO Membrane The Long Beach Water Department of California (LBWD) ha s recently patented a two-pass NF membrane for seawater de salination (Le Gouellec et al., 2006). NF membranes have a sign ificantly higher permeability than seawater RO membrane, but with higher salt passage, espe cially for monovalent ions. In a two pass NF-NF system, the seawater is treated by a first pass NF system. Because of the lower salt rejection ability of the NF membrane, permeate from the first pass is further treated by a second pass NF system to produce a permeate wa ter of acceptable quality. According to the results, th e overall recovery of the system is approximately 40% to 43%. The two staged NF system is as much as 20% more energy efficient than the typical seawater desalination RO membrane for Paci fic Ocean seawater with salinity of 35,000 mg/l. The typical energy saving possible with the two configurations discussed above is illustrated in Figure (2-10). A standard RO c onfiguration is also shown for comparison. As indicated in the figure, higher recovery with lower energy consumption is possible with brackish and seawater element. The NF -NF configuration results in the lowest energy consumption at a slightly lower rec overy than standard RO seawater system.

PAGE 53

33 Power Comparison for Different Membrane Treatment for Seawater Desalination 19 195 2 205 21 215 22 225 23 235 24 Type of Pressure Driven MmembranePower Consumption (Kw/m3) SW (45%) BW & SW (55%) Two Pass NF-NF (43%) Figure 2-10. Comparison of energy consumption for seawater desalination with SW membrane; BW-SW membrane & NF-NF membrane. Recoveries are shown in parentheses (Le Gouellec et al., 2006) 2.12 New Generation of Nanofiltration Membrane New NF membranes have been recently developed which can be tailored to have a range of hardness rejection. These membranes are thee composite polyamide type, similar to the existing standard RO membranes, but are chemically treated to adjust the hardness rejection. This treatm ent also imparts fouling resistance (Wilf et al., 2007). Among the makers of the new generation of NF membranes are Hydranautics, and FilmTec membrane manufacturers. Table 2-1 illustrates the performance comparison for various types of the new generation of NF membranes. For example, the rejection characteristics of Hydranautics membrane manuf acturer new NF can be tailored to meet a variety of hardness rejection values ranging between 83% to 93 % at standard operating

PAGE 54

34 pressure of a feed of 500 mg/l of 2CaCland 75 psi feed pressure at 25 degree C, as shown in table 2-1. Table 2-1 Comparison of new generation of NF membranes performance at standard operating conditions (Adopted from FilmTec, and Hydranautics Membrane Manufacturer, 2009) Manufacturer Product Element Area Nominal Flow Rejection (2ft) (2m) (gpd) (m3/d) % Hydranautics ESNA1-LF 400 37.2 8,200 31.1 89 (1) Hydranautics ESNA1-LF2 400 37.2 10,500 39.8 86 (1) Hydranautics ESNA-LF3 400 37.2 7,200 27.3 90 (1) FilmTec NF-270-400 400 37.2 14,700 55.6 40-60 (2) 97 (3) FilmTec NF-200-400 400 37.2 8,000 30.3 50-65 (2) 97 (3) Test Conditions (1) 500 mg/l of 2CaCl, 75 psi, 25 C, 15% recovery (2) 500 mg/l of 2CaCl 70 psi, 25 C, 15% recovery (3) 2,000 mg/l of 4MgSO, 70 psi, 25 C, 15% recovery FilmTec has introduced a si milar new type of NF membrane model NF270, and model NF200. These membranes have a salt rejection of 97% at standard operating conditions of 2,000 mg/l of 4MgSO at 70 psi, and 25 degree C ., and a salt rejection of 40-60% at 500 mg/l of 2CaCl. (FilmTec, 2007). The other feature of those type s of membranes is the low fouling nature due to the smoothness of the membrane surface and th e near neutral surface charge (Wilf, 2007). Figure 2-11 indicates the re lative surface smoothness of the new NF membrane model ESNA1-LF compared to the standard low pressure RO membrane model ESPA3. Both products are manufactured by Hydranautics. The lower negative charge of the new generation NF ESNA1-LF membrane from Hydranautics can be seen in Figure 2-12. The figure shows the Zeta potential of the

PAGE 55

35 membrane surface measured as a function of the feed pH. It can be seen that the traditional NF membrane ESNA-1 has a very strong negative charge at neutral pH. In contrast, the low pressu re low fouling RO membrane, LFC1, has a slight negative charge at these pH values. Sim ilarly, the ESNA1-LF membrane has a slight charge, or near neutral surface charge. This minimal surface charge minimizes the interaction with some organic compounds. It is interesting to men tion that due to the slight negative surface charge of the new gene ration of this type of NF membranes, the membrane mass transport can be modeled using the typical Solution-Diffusion Model which is normally used for RO membrane rath er than the more sophisticated models like Donnan exclusion, and extended NernstPlanck m odels, that include electrostatic effects. These charged membrane models are disc ussed in more details later in item 2.15.3. Figure 2-11. Image of the surface of the New Generation NF ESNA1-LF membrane compared to the Low Pressure RO ESPA3 membrane Showing Relative Surface Smoothening (Adopted from Hydranautics membrane manufacturer, 2007)

PAGE 56

36 Figure 2-12 Comparison of surface charge of new generation NF ESNA1-LF membrane, typical NF LFC1 membrane, and low pressure low fouling RO LFC1 membrane (Adopted from Hydranautics membrane manufacturer, 2007) 2.13 History of Using Permeate Sucti on in Pressure-Driven Membrane Permeate suction has not been commercially used before in RO or NF. However, permeate suction was theoretically investigated by several researchers. Bhattacharya, et al. (1996) have developed a generaliz ed mass-transfer relation from first principals to obtain a theoretica lly modified form for Sherwood number using the wall Peclet number to estimate the mass-transfer coefficient using permeate suction in rectangular channel cell, tubular module, and cross flow cell. He concluded that suction through the porous membrane had a significant effect on the mass transfer coefficient, and, in turn, the permeate flux for both RO and UF. He mentioned that this should be of immense help to the process and design engineer to improve the module design. The role of suction in mass tr ansfer through porous membranes is very important. It was identified by several researchers (V an den Berg et al., 1989; Gekas et al., 1987) that the effect of permeate suction enhances the mass transfer from the bulk to the

PAGE 57

37 membrane surface. Gekas and Hallstrom ( 1987) found that suction at the membrane surface increased the mass transfer coefficient from the surface to the bulk. Sirshendu and Bhattacharya (1988) have pr oposed a modified Sherwood number relationship including the effect of property variations due to perm eate suction for laminar flow in a rectangular channel cross flow ultrafiltration for bovi ne serum and dextran (Sirshendu and Bhattacharya, 1999). Immersed UF membrane (Fig ure 2-13) is used to tr eat wastewater that has virtually no osmotic pressure, so only small suction pressure is enough to create water flow from the permeate side to penetr ate though the membrane (about 4-9 psi). In this case, suction pressure can not theoretically exceed the atmospheric pressure which is about 14.7 psi (1 atm), in order to avoid cavitation in the permeate suction pump. Figure 2-13. Membrane bio-reactor using permeate suction to treat wastewater by immersed UF membrane (Adopted from Zenon Environmental membrane manufacturer, 2006)

PAGE 58

38 2.14 Effect of Increasing Suction Pressure on the Boundary Layer: In a two dimensional laminar flow, the thickness of the boundary layer which has not separated can be estimated as follows: the inertia force per unit volume is equal to x u u And for a membrane envelop of width L, the gradient of x u is proportional to L u, where u denotes the velocity outside the boundary layer. He nce the inertia force is of the order L u2 On the other hand the friction force per unit volume is equal to y which is in the assumption of the laminar flow is equal to 2 2y u The velocity gradient y u in the direction perpendicular to the membrane is of the order u, so that the friction force per unit volume is (Schlichtings, 1979) y 2 u (2-1) From the condition of equality of the friction and inertia forces, assuming a constant viscosity, we obtain: 2 u L U2 (2-2) Solving for the boundary layer thickness at any point of the membrane length, it is found at laminar flow that: u L = u L (2-3) Hence for laminar flow region, the boundary layer thickness is: = 5 u L where L is the membrane width. (Schlichting, 1979) (2-4) In NF spiral wound membra ne, the flow is laminar, and the local Reynolds

PAGE 59

39 number ranges between 100-1,000 ( Taylor et al., 1999). By a pplying suction at the end of the collector tube of the membrane module, an increase of pressure gradient along the stream at x direction will present. This in crease in pressure will de-stabilize the boundary layer (Schlichting, 1979). This is oppisite to the case of a decreas e in the pressure gradient along x direction. In the latter case, the bounary layer will be stabilized, and will not create a suitable condition for the re duction of the concne tration polarization. Figure 2-14. Application of suction to the membrane to prevent the boundary layer separation The de-stabilzation of the boundary layer wi ll have two effects: first, it will reduces the boundary layer thickness and a thinner boundary layer is less prone to become turbulent. Secondly, since in the di lute solutions the concntration polarization layer is impedded in the boundary layer (P robstein, 1994), suction will consequently reduce concnetration polarization. In a flat surface membrane, it is assumed that the quantitie s of fluid particles in the immediate neighborhood of the membrane surface are sucked away. Flow direction Membrane surface 0 Velocity due to suction Boundary layer thickness

PAGE 60

40 This is equivalent to that the ratio of suction velocity 0v (x) to free stream velocity u is very small, say u v0 = 0.0001 to 0.01 (Schlichting, 1979). When the suction velocity is of such a small order of magnitude, it is possible to neglect the loss of mass or sink-effecton the external po tential flow. On the flat surf ace membrane, the quantity of fluid removed Q, will be expressed thr ough a dimensionless volume coefficient by putting Q = Qc1A u (2-5) where1A = wetted membrane surface area (bxL) And for the flat membrane Q = b Lv0 0[(x)] dx (2-6) Since equation (2-5) equals to equation (2-6), consequently Qc1A u = b Lv0 0[(x)] dx or Qc = Lu1 Lv0 0[(x)] dx where 1A= bL (2-7) where Q = quantity of flui d removed during suction, Qc = Dimensionless volume coefficient, u = Average bulk velocity at the x direction, and for the case of increasing suction pressure: 0vincreases over the x axis i.e. 0v(x). The suction pressure is necessary to be dependent on x and y, i.e. P (x,y) (2-8) and the continuity and Navier Stock equations that govern the laminar two dimensions flow at steady state conditi on when there is a mass transfer through

PAGE 61

41 porous membrane surface are, and using the mean flow is (Schlichting, 1979): x u + y v '= 0 (2-9) and u x u '+ 'v y u = 1 x P + 2'u (2-10) and u x v = 1 y P '+ 2'v (2-11) where 2 denotes the Laplacian operator 2 2x + 2 2y From equations 2-9, 2-10, and 2-11 the velocity gradient due to perm eate suction can be Calculated, and the bounda ry conditions are: at y =0 : u = 0, and v = 0v < 0 at x = 0, and at y = : u = U (x) 2.15 Reverse Osmosis Models Many mathematical models ha ve been proposed to describe reverse osmosis membranes. Some of these descriptions re ly on relatively simple concepts while others are far more complex and requir e sophisticated solu tion techniques. Models that adequately describe the performa nce of RO membranes are essential, since These are needed in the design of RO processe s. Reverse osmosis models can be divided Into four types: irreversible thermod ynamics (I. T.) models; nonporous homogeneous models; pore models; and charged membranes models. A fundamental difference exists between the assumptions of the homogeneous and porous membrane models.

PAGE 62

42 The homogeneous models assume that the membra ne is nonporous, that is transport takes place between the interstitial spaces of the polymer nodules, usually by diffusion. The porous models assume that transport takes pl ace through pores that run the length of the membrane barrier layer; as a result trans port can occur by both diffusion and convection through the pores. While both concepts have had some success in predicting RO separation, the question of whet her a RO membrane is nonporous or pores is still a point of debate (Williams, 2003). 2.15.1 Irreversible Thermodynamics Models Irreversible thermodynamics (I. T.) models, such as the Spiegler-Kedem model, assume that the membrane is in mechanic al equilibrium, no external force acting on the system, and flux can be described by th e phenomenological equa tions relationships. The water flux according to the Spiegler-Kedem model is given by J w = Lp (p ) (2-12) while the solute flux is expressed as J s = + (1) (Cm )avg J w (2-13) where Lp is the hydrodynamic permeability coefficient; is the coefficient of coupling between salt and water; is the difference in the osmotic pressure across the membrane;p is the operating pressure; isthesalt permeation coefficient; and (Cm ) avg is the logarithmic mean solute concentration in the membrane (Soltanieh and Gill, 1984). The Spiegler-Kedem model ha s found a wide use for the description and analysis of RO membrane separation. However a major disadvantage of the m odel is the treatment of the membrane as a black box. It does not provide insight into th e transport mechanisms

PAGE 63

43 of the membrane. I. T. models also do not include any convection effects, and considers transport of the solvent and so lute take place only by the eff ect of the chemical potential gradient, which includes concentration, and pr essure diffusion. These models assume that Onsager reciprocal relations are valid (So ltanieh et Gill, 1981). This assumption is controversial in processes far from equilibrium (Rosenbaum and Skiens, 1968). As a result, I. T. models are not very useful in optimizing separation based on membrane structure and properties. These models also do not adequately describe water flux for some solute systems, in particular some d ilute organics that have no osmotic pressure (Williams, 2003). 2.15.2 Porous Models The porous models assume th at transport takes place through pores that run the length of the membrane barrier layer; as a result, transport can occur by both diffusion and convection through the pores. The preferentia l sorption-capillary flow model (PSCP) proposed by Sourirajan (1970); S ourirajan and Matsuura (1985) states that the membrane is assumed to be microporous and the barrier la yer has chemical properties such that it has a preferential sorption for the solvent or pr eferential repulsion for the solutes of the feed solution. As a result, a layer of almost pure solvent is preferen tially sorbed on the surface and in the pores of the capillary pores under pressure. The total water flux is given by wN= A [p () (FX -) (PX ] (2-14) The total solute flux is expressed as sN= M w sD ck(FX)PX (2-15) where A is the pure water permeability constant of the membrane; p is the applied pressure difference; ) ( X represents the osmotic pressure of the feed or permeate side

PAGE 64

44 with solute mole fraction X; C is the molar concentration of salt; skis the distribution coefficient of the solute from the feed into the pore of the membrane; wDis the diffusion coefficient of the solute in the membrane; andMis the active membrane thickness. The term F w sD k, which is treated as a single para meter, is called solute transport parameter. In their experiments, Alegranti et al. (1975), suggested that the fixed pore diameter equivalent of free volume of the hollow fiber skin layer of 0.4 micron is 10 A or less. The limit of the electron microscope used in their work was 15A. This suggests that possible microspores in the skin layer could not be detect ed (Soltanieh et Gill, 1981). Sourirajan and Matsuura (1985) have utilized the above equations to analyze transport for a large number of solutes and membranes; howev er the above equations failed to describe the water flux drop, and rejection for some organics, and solutes. 2.15.3 Charged Membrane Models Charged membrane models, like the one used for st andard NF, account for electrostatic effects as well as for diffusive and/or convectiv e flow in order to describe the solute separation. Many charged membrane transport theories have been proposed. Donnan equilibrium models assume that a dynamic equilibrium is established when a charged membrane is placed in a salt solution (Bhattacharyya and Cheng, 1986; Bhattacharyya and Williams, 1992). The counter-ion of the solution, opposite in charge to the fixed membrane charge, typically car boxylic or sulfonic groups is present in the membrane at a higher concentration than that of the co-ion (same charge as the fixed membrane charge) because of electrostatic attr action and repulsion effects. This creates a

PAGE 65

45 Donnan potential which prevents the diffusive exchange of the counter-ion, and co-ion between the solution, and membrane phase. When a pressure driving force is applied to force water through the charged membrane, the effect of the Donnan potential is to repel the co-ion from the membrane; since electroneutrality must be maintained in the solution phase. The counter-ion is also rejected, resul ting in ionic solute separation. The model correctly predicted that the solute rejecti on was a function of membrane charge capacity, ion feed concentration, and ion charge. Howe ver, this model does not take into account solute diffusive and convective fluxes which are also important in charged membrane separations. Lakshminarayanaiah (1965, 1969), Dresner (1972), and Dresner and Johnson (1980) have described the use of extended Ne rnst-Planck equations for the prediction of solute ion fluxes. The model represents the solute flux due to diffusion, convection and Donnan potential. Dresner (1972) has shown that the extended Nernst-Planck model correctly predicts the trends expected for ionic solute rejection, including conditions under which a negative rejection is obtained. However, the difficulty of experimentally measuring the model parameters limits its use for solute flux, and flux prediction. 2.15.4 Solution-Diffusion Models The solution diffusion model assumes that the water transport across the membrane is only by diffusion, and so can be ex pressed by Ficks low (Soltanieh et Gill, 1981) as: wJ= wD dy dcw (2-16) where cwand Dware the concentration, and the diffusi vity coefficient of water in the membrane, respectively.

PAGE 66

46 The water flux is given by wJ= A (P ) (2-17) where A is the membrane solvent permeability co efficient and it is the property of the membrane; pis the operating pressure; is the difference in the osmotic pressure across the membrane. And the solute flux is expressed as sJ= 2k (P RC C ) (2-18) where2kis the membrane solute permeability coefficient; and RCand PC are the reject and product concentration, respectively. 2.15.4.1 Solution-Diffusion-Imperfection Model Sherwood et al. (1967) have extended the solution-diff usion model by including additional terms due to pore flow in additi on to diffusion of solv ent and solute through the membrane as the mechanism of transporta tion. This modified model recognizes that there may be small imperfections or def ects (pores) on the surface of the membrane through which transport can occur. The total water flux,wN, and the total salt flux, sN, are given by: wN = wJ+ memPC k3= A ) ( P+ memPC k 3 (2-19) sN= sJ+ RPC k3 = ) (2 R PC C k + RPC k 3 (2-20) where memC is the water concentration on th e upstream side of the membrane. The coefficient 3kcan be viewed as a coupling coefficient.

PAGE 67

47 If we divide both sides of equation (2-19) bymemC, the left hand side will be equal to the water permeation velocity expressed as wV= mem wC N, which is very close to the total permeation velocity. For real membranes, which have some imperfections, the measured flux (wN) is not purely diffusive (wJ), but it contains a term contribu ted by convection. It is necessary to distinguish between the tw o, although in the literatur e, they are usually used interchangeably (Soltanieh et Gill, 1981). The solute flux is equal to the permeation velocity multiplied by the product concentration, i.e. sN= wVPC Equations (2-21) & (2-22) then can be written in terms of the permeation velocity: wV= P k P k 3 1) ( (2-21) wV= R R PPC k C C k 3 2) ( (2-22) where 1k= memC A It is interesting to compare the relative contribution of diffusive and pore flow Fluxes based on the calculated values of 2 1&k kand 3k from Applegate, and Antoson (1972). Applegate, and Antoson used the a bove equations to analyze the rejection pressure drop data for asymmetric aromatic polyamide membranes, and cellulose acetate membrane. The values of 3k from all membranes, and concen trations were at least two orders of magnitude smaller than those of1k. Since in the solvent flux equation (2-21), P and) ( P, are of the same order of magnitude, the second term P k 3(the pore flow) is negligible as compared to

PAGE 68

48 the first term) (1 P k (the diffusive flow). This is true for dilute solutions. In the solute fl ux equation (2-22), we have to compare P k 3with 2k since RC is of the same order as) (P RC C. Calculations of Soltanieh and Gill (1981) illustrated that the calculations of Applegate and Antonson ex periments for dilute solutions (< 0.05 M), showed that the contribution of the pore flow to the solute fl ux is very small (about 2% of total flux). At higher concentra tions, say about 0.1 M, the co ntribution of pore flow is about 8%, and for a 0.5 M feed, the pore flow contributes to about 2540% of the solute flux for polyamide membrane. The standard NF membrane used to be described by the Donnan equilibrium model (Battacharyya and Cheng, 1986; Battacharyya and Williams, 1992), or by the extended Nernst-Planck m odel (Lakshminaraiah, 1969; Dresner and Jonson, 1980) to predict the solvent and solute flux, because it is negatively charged. The new generation of NF membranes has a slight negative charge at the membrane surface, and this charge is approximately close to neutral as shown in Fi gure 2-12 (Wilf et al, 2006), so these sophisticated models are not re quired any more to describe the solvent, and solute flux in this type of membranes. Instead, the above-mentioned simpler solutiondiffusion model can describe the membrane, when the pore flow terms are neglected. 2.15.4.2 Assumptions when us ing Solution-Diffusion Model Based on the above equations for the solution diffusion-im perfection model, the following assumptions are going to be considered (Fritzmann et al, 2006): (1) the active membrane layer is a de nse membrane without pores. Permeating components dissolve in the membrane phase ; (2) at all times there is chemical equilibrium at the phase inte rface between membrane and feed /permeate side; (3) salt and water flux are independent of each other. Sa lt flux results solely from concentration

PAGE 69

49 gradient, but not from pressure; (4) due to membrane swelling, wa ter concentration and water diffusion coefficient across the membrane are constant; (5) the driving force for permeation of each component can be split into two terms, the concentration or activity difference, and the pressure difference between the feed and the permeate sides; (6) at relatively low salt concentrations, the pressure driving force for permeating salt components is negligible; (7) due to the assu mption of constant wa ter concentration in the membrane, solely the applied pressure difference p causes water flux across the membrane; (8) the measured fluxwN, and the purely diffusive flux wJ are assumed to be equal. 2.16 Determining Membrane Surface Concentration RO membrane transport models typically assume that the bulk feed solution concentration is equal to the membrane wall solution concentration, which is not always true. This has to be related to the concentration polarization expressions (Williams, 2003). Concentration polariza tion complicates the modeli ng of membrane systems because it is very difficult to experiment ally determine the solute membrane wall concentration (memC). The membrane wall concentration is necessary to be determined since it is not equal to th e bulk feed concentration (FC). In the limited feed flow rate that is typically used for hyperfiltration (RO and NF) membrane processes, the flow in the membrane channels is laminar, and th e difference between the membrane wall concentration and bulk concentrations can be substantial. So, calculating the membrane wall concentration must be appropriately estimated. For dilute solutions, and from equation (221) above, the water flux will be equal to

PAGE 70

50wJ = A (P ) (2-23) where A is the permeability coefficient of the membrane, and it is a function of the membrane construction. The membrane permeability coefficient A can be determined from distilled water where in equation (2-23) in this cas e will be approximately equal to zero. The term P is the hydraulic pressure difference across the membrane, and is equal to the applied membrane pressure minus the permeate pressure, while wJis the permeate flux and equal to the permeate flow rate divided by the membrane cross flow area. The term is equal to the difference between the osmotic pressure at the membrane surface minus the permeate osmotic pressure. Osmotic pressure is a property of the solution and does not in any way depe nd on the membrane properties (Probstein, 1994). For dilute solutions the os motic pressure is independent of the solute species, and is given by Vant Hoff equation: i = inRT n i iC1 (2-24) Where in= number of ions formed wh en the solute dissociates. And iC= molar concentration of the solute = ic/ 1000 iMW and ic= total dissolved solids as mg/l ; iMW= molecular weight of the dilute solution; R= gas constant; and T= absolute temperature. From equations (2-23) and (2-24) above, membrane wall concentration Cmem can be calculated.

PAGE 71

51 2.17 Determining Mass Transfer Co efficient and Thicknesses of the Concentration Polarization Layer Under the condition of mass tr ansfer-limited permeate fl ux shown in Figure 2-15, the accumulation of materials near the membra ne can be envisioned as a balance between advection of materials towards the membrane due to permeation, and back diffusion that occurs as a concentration gradient builds up n ear the membrane (Letterman and Taylor, et al. 1999). Figure 2-15 illustrates the flow directions of solv ent and solute near the membrane surface. (y) direction Flow Direction (x) direction MembraneF y c D memcbulkc wJ Figure 2-15. Feed side concentration polarization layer 2.18 Using Sherwood Number to Dete rmine Mass Transfer Coefficient The conventional way to estimate the mass transfer coeffici ent is to use Sherwood number relationships obtained from the h eat and mass transfer analogy. Numerous Sherwood number relationships have been proposed and extensively reviewed. The Graetz-Leveque correlation of Sherwood number, which is used for laminar flow when the velocity field is fully devel oped and the concentration boundary layer is not fully developed, is typically used to estimate mass transfer coefficient as: Sh = Xa(Re)bSc c hL d (2-25)

PAGE 72

52 where Re = Reynold's number = h wd V; Schmidt number = Sc = D ; = kinematics viscosity; wV = average cross-flow permeate velocity; hd= hydraulic diameter of the membrane element; D the diffusion coefficient fo r solute transport through solvent, and L is the spiral wound membrane wi dth. (Taylor et al.,1999). D in this relationship is equal to K F where K is the mass transfer coefficient, and F is the concentration polarization layer thickness.The terms X, a, b, and c are coefficients that have taken extremely different values by different re searchers (Isaacson, 1 976; Schock & Miquel, 1987; Xuesong, 1987; Taylor, 1991; Van de Lisdonk et.al, 2001). Table 2-2 demonstrates the different values of coefficients found in the literature by some of those researchers to calculate Sherwood number at different operatin g conditions. It is clear from the table that the different coefficients vary considerably. Table 2-2. Several values of Sherwood number coefficients found in literature Literature Year X a b c Isaacson 1976 0.20.27 0.5 0.33 0 Schock & Miquel 1987 0.065 0.875 0.25 0 Wang Xuesong 1987 1.66 0.36 0.34 0.42 Taylor 1991 1.86 0.33 0.33 0.33 Van de Lisdonk 2001 0.265 0.33 0.52 0 There are several limitations in using the above-mentione d equation: (1) the above mentioned Sherwood number relationship is derived for flow through non-porous conduit; hence, the effect of suction can not be considered using these relationships; (2) the axial change in osmotic pressure at membrane surface due to the concentration polarization change is not considered in the above mentioned Sherwood number correlations; (3) suction will alter the veloci ty gradient in the bulk stream through the

PAGE 73

53 boundary layer (Schlichting, 1979) that will impact Reynolds number, which is not considered in calculating the Graetz-Lev eque Sherwood number; (4) and suction will change the species concentration at the me mbrane surface that will change the solution physical properties like visc osity, density, and diffusivity, which are function of the concentration. Consequently, the above mentioned Reynolds number, and Schmidt number, will be variable along the membrane length. These changes are not considered with this form of Sherwood number relations hips. It is concluded that in the case of suction, calculating the concentration polarizat ion layer using the traditional way of using Sherwood number correlations will lead to e rroneous results due to the change in the solution properties that are not considered in the above mentioned relationship. 2.19 Overcoming the Disadvantages of us ing Sherwood Number to Determine Mass Transfer Coefficient with Permeate Suction Of great interest is the case of dilute so lution where Sc >> 1. In this case the diffusion concentration polarization layer is imbedded in the viscous boundary layer, and the velocity it sees is that cl ose to the wall (Probstein, 1994). 2.19.1 Film Theory According to the film theory, mass transfer coe fficient is invers ely proportional to the concentration polarization layer (Taylo r et al., 1999; Williams, 2003), so if the concentration polarization layer decreases, the mass transfer coefficient will increase, and the permeate flow and quality will improve; an d consequently membrane fouling can be reduced. Referring to Figure 2-15 at item 2.17 above, the Navier-Stockes diffusion-

PAGE 74

54 convection equation for flow over a flat shee t membrane, gives the concentration profile by (Bhattacharyya and Williams, 1992) u x C + wV y C D ( x C2 2 + y C2 2 ) = 0 (2-26) With boundary conditions: For x direction C (0,y) = FC And for y direction y x C ) 0 (= 0 Assuming a constant permeation rate, and a concentration polarization layer with axial distance (Figure 2-14), mass balance on the concentration polarization layer yields C Jw = D y c (2-27) where wJ is the permeate water flux; C is the c oncentration of the sp ecies subject to concentration-polarization, D is the diffusivity coefficient for solute tran sport through solvent (calculated from Cussler, 1984), and y is the distance with the boundary layer such that C = memC at y = 0; and C = FC at y = F If the boundary layer is assu med to be stagnant over the channel length, the equation will be: wJ y C = D y C2 2 (2-28) Integrating this expression over the thickness of the stagnant concen tration-polarization layer F results in the following expression: P F P memC C C C = exp D JF w (2-29)

PAGE 75

55 This is the widely applied film theory developed by Brian (Brian 1966; Bhattaryya and Williams, 1992). The ratio of the diffusion coefficient for solute transport through solvent to the concentration polarization layer thickness in this film thoery model defines a mass transfer coefficient K: K = FD (2-30) where K is the mass transfer coefficent. Using equation (2-30), the con centration polarization layerF can be calculated, if the diffusivity coeficient of the speices D, and the mass transfer coeficient K are determined, without having to use Sh erwood number corelations. 2.19.2 Peclet Number Peclet number is defined as the dimentionless ratio of the rate of mass transported by convection to the membrane, and the rate of mass transported by diffusion back to the bulck solution; in other words the diffusive membrane Peclet number is expressed as: eP= 2 1 D h Vd w (2-31) where wVis the permeate velocity which is equa l to the permeate flux per one sheet of membranes;2 1D is the diffusivity coefficicent of the ionized electrolye; and dh is the hydraulic diameter of the spiral wound memb rane. The diffusive Peclet number is a measure of how permeate goes through the membane, so if eP is less than unity, this is an indication of no cocentration polarization, while a large Peclet number means that there is a concentration polarization at the membra ne surface. Peclet nu mber is called the

PAGE 76

56 dimensionless flux. If the diffusive Peclet num ber is increased due to suction, while the associated concentration polarization is be ing reduced, this means that suction has increased membrane production with more favor able conditions to the membrane, as far as inorganic fouling is concer ned. Therefore, we using Sher wood number can be avoid in the calculations due to the above-mentione d limitations. For the 2.5 inch diameter membrane, which is composed of two envel opes (leaves) of membranes, where every envelope is composed of two sheets, the hydr aulic diameter is the ratio of the cross section of the flow channel to the wetted circ umference and can be calculated with the dimensions of the feed-concentrat e spacer according to (Schock and Miquel, 1987): dh= f fd d 4 ). 1 ( 1 4 (2-32) where is the porosity of the f eed-concentrate spacer; and fdis the filment diameter of the feed spacer, which is equal to half th e feed spacer height. As in the case of NF membrane, it is observed that for dilute solutions the Schmidt number is very large, as a consequence of which the diffusive membrane Peclet number is genera lly large. This is true even at a moderate Reynol ds number (Probstein,1994). Probstein (1994) also showed that at typical conditi on of operting the mebrane system without permeate suction, as eP number increases, the c oncentration polarization boundary layer increases, and consequelty the mass transfer along with the product permeate decreases. If permeate suction increases theeP, and the concentration polarization layer is decreased, this will in dicate that permeate suction will help in reducing the membrane fouling, wh ile product flow is increased.

PAGE 77

57 2.20 Determinig Diffusion Coeffi cient for Strong Elecrolytes A transport of mass or diffu sion of mass will take place in a fluid mixture of two or more species whenever there is a sap tial gradient in the properties of the mixure, that is, a concentration gradient (Pr obstein, 1994). Diffusion causes convection. Convection flow can have ma ny causes. For example, it can occur because of pressure gradient or through differences in temperature. However, even in isothermal and isobaric systems, diffusion will al ways produce convection (Cussler, 1984). This combination of diffusion and co nvention could complicate our analysis. Multi-ions salt solutions calculations of diffusi on coefficient are more complicated because more than one cation can be accompanied by one anion or vice versa, depending on the ion valence. For exam ple, for a ternary system, there would be two concentration gradients, and the diffusive flux of each species could be affected by both concentration gradients. One instance where this is not so is th e infinitely dilute solutions for which each component is unaffect ed by the presence of the other (Probstein, 1994). The diffusivity for dilute liquid solu tions, like the case of the feed to NF membrane, may be estimated theoretically from simple hydrodynamic consideration (Probstein, 1994). In solute-s olute interaction dilute so lutions, like the case of NF membrane, the convection caused by diffusion is vanishingly small, and dependent on the solute concentration, and thus on temperature. This is the frame work of this research. It is worthy to mention that estimates of diffu sion for concentrated solutions are far more difficult (Probstein, 1994).

PAGE 78

58 Salts are ionized when they are dissolved in water. For example, a Soduim Chloride solution in water diffuses as a singl e ion, and not as a single molecule; instead sodium ions and chloride ions move freely through the solutio n (Cussler, 1984). Table 2-3 shows values of ioni c diffusion coefficients in water at 25 degree C at infinite dilution. 2.20.1 1-1 Strong Electrolyte Referring to the below menti oned Table 2-3, and in th e above example of Soduim Chloride solution, the diffusion of Na is slower than that of Cl, and the diffusion of both ions in a dilute soltion of NaCl is going to be dominated by the larger ion because the two ions are tied together electrostatically. Table 2-3. Ionic diffusion coefficients in water at 25 degree C at infinite dilution in 510 scm/2. (Calculated from data of Robinson and Stocks (adopted from Cussler, 1984) When describing the ion fl uxes of a single strong 11 electrolyte, such an electrolyte ionizes completlety, and it w ill be producing equal numbers of cations and anions. Although the concentration of anions and cations may vary through the solution, the concentration gradient of these species are equal everywhere because of the

PAGE 79

59 electroneutrality (Cussler,1984). and 2 1D= 2 11 1 2 D D (2-33) Where 2 1D, 1Dand2D are the diffusivity coefficicent of the ionized electrolye, the anion, and cation respectively. 2.19.2 Non 1-1 Strong Elecrolyte The non 1-1 electrolyes like4MgSO, and 2MgCl are parallel to the above mentioned 1-1 electrolyte, but more co mplicated algebraica lly (Cussler,1984). The diffusivity coefficicent of the ionized elect rolyte in this case is going to be expressed as: 2 1D= 1 2 2 1 2 1D Z D Z Z Z (2-34) where 2 1D,1D, and 2D are the diffusivity coefficicent of the ionized electrolye, the anion, and cation respectively, while 1Zand2Z are the absloute valent numbers of the the anion, and the cation respectively.

PAGE 80

60 CHAPTER 3 EXPERIMENTAL METHODOLOGY This chapter describes th e technical procedure ut ilized in performing the experimental part of this research. Th e dilute solutions preparation, and the instrumentation used in the experiments are described first, followed by the experimental procedures. For the two types of experiment s, a diagram of the equipment set-up is shown with each component briefly descri bed. The description of each type of experiment contains a summary table of the salts solutions used as feed water, their concentration and the system operating para meters. The sampling and the measurement protocols, as well as the procedure for the re plicate runs are presented for each type of experiment. 3.1 Dilute Solutions Preparation The feed water for all the expe riments is prepared using de-ionized (DI) water and analytical grade salts. DI water is produced from a feed of tap water using RO unit. The conductivity of the product is 16 s Only simple binary solutions ofNaCl, 2MgCl, and 4MgSOwhich are 1-1, 2-1, and 2-2, resp ectively, strong electrolyte dilute solutions are considered in this research The diffusion coefficients for the binary solutions were calculated using th e equations from Cussler (1984). Salt concentration was prepared to have c oncentration of the th ree different binary

PAGE 81

61 solutions according to Table 3-1. The experi ments were run at thr ee concentrations for low, medium, and high values of the applied feed pressures. 3.2 Reasons behind Choosing the Chemicals It is required to study the effect of both monovalent and divalent ions on the performance of the new generation of NF membrane. Na, and Cl ions are monovalent ions. The solubility of chloride salts and s odium salts are high and do not create a RO scaling problems, that may allow running th e experiments at a relatively high recovery. Sodium and Chloride, in brackish and seaw ater, are the prevalent ions. Magnesium ( Mg) is a divalent cation, an d accounts for about a third of the hardness in brackish water, leaving about the two third to Calcium ( C) cations. Table 3-1. Dilute solutions concentration and operating pressures for the experiments Dilute Solution Feed Concentration (mg/l) Feed Concentration (Mol/l) Feed Pressure (psi) 750 0.0249 80, 110,160 Sodium Chloride 1200 0.0411 80, 110,160 1750 0.0599 80, 110,160 820 0.0136 80, 100,130 Magnesium Sulfate 1235 0.0205 80, 100,130 1770 0.0606 80, 100,130 840 0.0176 80, 100,130 Magnesium Chloride 1260 0.0265 80, 100,130 1750 0.0368 80, 100,130 The solubility of magnesium salts is high, and they typical ly do not cause a scaling problem in membrane systems. Sulfate ( 4SO) is a divalent anion, and 4MgSOdoes not cause scaling on membranes. This is unlik e calcium, barium, and strontium sulfate,

PAGE 82

62 which have low solubility limits, and can cau se scaling problem in the back-end of the system. 3.3 Experimental Setup The apparatus used in the ex periment are assembled on the equipment skid shown in the pictures of figures (3-1, 3-2 and 3-3). Figure 3-1. Experiential equipment skid showing pressure gauges, TDS meters, flowmeters, and NF membrane pressure vessel The instruments used are pr esented in Table 3-2 together with calibration requirements, manufacturer, readings range, and accuracy. The sensors are connected to the control panel of the equipment skid for data collection and analysis. Samples were taken during the runs to test the feed water pH and the biocide cont ent. Samples were also taken for feed flow, product flow, and c oncentrate flow to test the conductivity panel readings against manually held TDS meter. Wh enever samples were taken for test, they were poured back to the feed water tank to keep feed water conductivity constant.

PAGE 83

63 Figure 3-2. Variable frequency drives for the high pressure pump, and the permeate pump Figure 3-3. The assembled high pressure pump (top), and the permeate pump (bottom)

PAGE 84

64Table 3-2. Instrumentation and specifications No. Instrument Manuf. Model Calib. Req. Range Accuracy 1 Electrical Top loading Balance Ohaus Adventure No 0 -3,100 g 0.1 mg 2 Conductivity Meter & Probe Hanna BL983318 Yes 0 -10,000 mg/l 1 mg/l 3 Control Box R&D Specialties CE2-IPC Yes 0 999 mg/l 1 mg/l 4 Permeate Conductivity Cell R&D Specialties 80TDS150R1 Yes 0 999 mg/l 1 mg/l 5 Flowmeter Blue-White F440 No 0 1 gpm 0.05 gpm 6 Flowmeter Blue-White F44375 No 0 5 gpm 0.05 gpm 7 2.5" Pressure Gauge Wika 316SS tube & Connection Yes 0 30 psi 0.1 psi 8 2.5" Pressure Gauge Wika 316SS tube & Connection Yes 0 60 psi 0.1 psi 9 2.5 Pressure Gauge Wika 316SS tube & Connection Yes 0 300 psi 0.5 psi 10 0.5" NC Solenoid Valve GC-Valves H211YF02J7DG4 No On/Off N/A 11 Volumetric Flask Kimax 20024 No 250 ml 1.4% 12 Volumetric Flask Pyrex 3024 No 100 ml 1.0% 13 Hand Held Conductivity Meter Hanna TDS-3 Yes 0-19,999 10 mg/l 14 Hand Held pH Meter Hanna HI 98107 Yes 0-14 0.1 15 Volumetric Pipette Pyrex 7101 No 0 50 ml 0.2% 16 Beaker Kimax 14000 No 0 600 ml 5.0% 17 Beaker Pyrex 1000 No 0 1,000 ml 5.0% 18 Low Pressure Switch Barksdale E1H-H90 No 040 psi 1 psi As shown on the schematic fl ow diagram Figures 3-4 and 3-5, solutions were pumped from a HDPE 35 gallons feed tank by a booster pump to the high pressure pump through 5 m cartridge filter to protect the pump, and the membrane from any suspended solids that may be available in the solution tank. The booster pump is a

PAGE 85

65 centrifugal type; model number 594-154; manuf actured by Surflow Company, USA, and rated at 3.3 gpm at 45 psi. The membrane was pre-compacted at 120 psi using recirculated DI water, and the re-circulated so lution is kept disinfected using 0.2 0.3 mg/l of biocide. The used biocide is Model number RoCide DB-20 manufactured by Avista Company, USA, is formulated to keep the membrane sanitized and non-oxidized. RoCide DB-20 is approved by the EPA to be used in RO systems as a fast acting; non-oxidizing biocide based on a 20% solution of the act ive ingredient DBNPA (Dibromo nitrilo propionamide), (Avista Company, 2008). The same concentration of biocide was kept for all the solutions during the en tire experiment runs. The bi ocide content was periodically checked using oxidant reagent test kit. At the beginning of the experiments, the pure water permeability coefficient (A) was calculated using equa tion (2-23), where wJ = A (P ), since A is the permeability coefficien t of the membrane, and it is a function of the membrane chemical structure. The membrane permeability coefficient A was determined from distilled water where in equation (2-23) in this case is approximately equal to zero. The term P is the hydraulic pressure difference across the membrane, and is equal to the applied membrane pressure minus the permeate pressure, while wJis the permeate flux, and is equal to the permeate flow rate divided by the membrane cross flow area. The capacity of the high pressure pump was determined according to the membrane software program of the manufacturer follo wing the design guide lines at the laboratory temperature. The pump is a rotary vane t ype manufactured by Procon company, USA; constructed from 304 SS, and rated at 207 gph at 200 psi and 1725 rpm.

PAGE 86

66 The pump Model number is 105E265F31BA215 The motor is 1 HP at 1725 rpm; inverted duty non washdown manufactured by Bal dor with electrical specifications of 230/460 V/60 Hz/ 3 Ph; and Model number ID NM3581T bolt on. The variable frequency drive (VFD) on the high pressure pump along w ith the concentrate control valve allows unlimited control of the membrane feed pressure according to the feed water concentration. Concentrate return flow Concentrate Flowmeter Feed Tank Conc Control Valve Permeate Flowmeter Membrane Module Permeate return flow High Pressure Pump Press. In Press. Out Press. In TDS2 TDS3 Cooler VFD Booster Pump Cartridge Filter Press. In Press. Out TDS1 Figure 3-4. First setup by running the high pressure pump only The setup shows the spiral wound NF TFC membrane used in the experiment. Table (3-3) demonstrates the geometrical di mensions of the tested new generation NF membrane, which is a standard size commercial 2.5 inch nominal diameter, and a 40 inch long spiral wound aromatic po lyamide thin-film composite membrane Model NF2702540, manufactured by Dow-Film Tec Inc., Minneapolis, MN. The effective surface area is 282ft(2.62m). This membrane was chosen for this research

PAGE 87

67 as a representative of a class of the ne w generation membranes, which is used increasingly in water treatment applications. The2CaClrejection is about 40-60%, and the magnesium sulfate rejection is 93%, as reported by the manufactur er (test conditions: feed TDS is 500 mg/l for 2CaCl, and 2,000 mg/ l for 4MgSO, 70 psi, 10% recovery, and 25C. The feed spacer height is 28 mil (0.711 mm) with a porosity ( ) of 0.89, computed as described by Schock and Miquel (1987). Concentrate return flow Concentrate Flowmeter Feed Tank Suction Pump Conc Control Valve Permeate Flowmeter Membrane Module Permeate return flow High Pressure Pump Press. In Press. Out Press. Out Press. In TDS1 TDS2 VFD Chiller VFD Booster Pump Cartridge Filter Figure 3-5. Second setup by running the high pressure pump and the permeate suction pump The membrane zeta potential is close to neutral, so the membrane charge was assumed to be zero, to eliminate the charge effect. The membrane pressure vessel is made of reinforced fiberglass and has a rated maximum operating pressure of 300 psi. The pressure vessel is produced by Crane Environmental. The experiments were divided into two setups: the first setup was conducted by running the experiment using th e high pressure pump only.

PAGE 88

68Table3-3. Geometrical Parameters for the tested FilmTec Membrane Model NF270-2540 Parameter Specifications Membrane Model No. 270NF-2540 Active Area 28 2ft(2.6 2m) Active Length 35.4 in.(90.4 mm) Feed Spacer Thickness 28 mil (0.71 mm) No. of envelops (leaves) 2 Feed Spacer Porosity 0.89 Hydraulic Diameter 28 mil (0.877 mm) Filament Diameter 14 mil (0.355 mm) In the second setup of the e xperiments, the permeate suction pump was run along with the high pressure pump, to apply suction on the membrane permeate side. The capacity of the permeate suction pump was determined according to the projected permeate flow rate given by using the manu facturer software program. The pump is a positive displacement rotary vane type manu factured from bronze by Procon Company, USA, and rated at 37 gph at pressure of 50 psi and 1725 rpm. The Model number is 102E125F31BA250. The motor is 0.5 HP at 1725 rpm; inverted duty non washdown. It is manufactured by Baldor comp any with electrical specifi cations of 230/460 V/60 Hz / 3 Ph, and Model number IDNM3538. The VFD which is used to control the suction pressure is rated at 0.5 HP manufactre rd by Woods company with electrical specifications of 115 V/ 1 Phase/ 60 Hz, and Model number SE1C1S005D01. It is worthy to mention that the suction pressure of the permeate suction pump can not be lower than the pump net positive suction head (NPSH), in order to avoid cavitations in the pump. All the interc onnecting piping was made of anti-corrosion Stainless Steel tubes, or flexible breaded Stainless Steel. Permeate tubing was made of polypropylene tubing. The permeate flow, c oncentrate flow, pressures, and total dissolved solids for the feed, permeate, and concentrate side were measured on the

PAGE 89

69 control panel. The permeate flow, and concentr ate flow were returned to the feed tank where the temperature was kept constant at 25 1 degree C by a cooler. The cooler is a drop in chille r manufactured by Current company, USA, rated at 3,926 BTU, and model No. CD-22308HP. Keeping the re-circulated flow at a constant temperature eliminates the effect of the change of the diffusivity coefficient of the species due to temperature change, since diffusivity coefficient is a function of the solvent temperature and solute concentration. The salt diffusivity coefficien t for the 1-1 strong ionized solution is calculated from equation (2-33), and (2-34) in Chapter 2 by: 2 1 D= 2 11 1 2D D where 2 1 D, 1Dand2D are the diffusivity coefficicent of the ionized electrolye, the anion, and cation respectively, and for non 1-1 strong ionized solution is calculated from: 2 1 D= 12 2 1 2 1D Z D Z Z Z where 2 1 D,1D, and 2D are the diffusivity coefficicent of the ionized electrolye, the anion, and cation respectively, while 1Zand2Z are the absloute valent numbers of the anion and cation respectively. The run for each salt solution is finished when the permeate water conductivity and flow rate are at equilibrium. At the end of the runs for every salt solution, the pure water permeability coefficient (A) is re-calculated using DI water. The value of the new permeability coeffi cient was used for the calculations of the

PAGE 90

70 next salt solution. This value is typically reduced due the membrane compaction with aging. After the completion of each solution run, the system is flushed with DI water before running the next solution experiment When applying pressure on the dilute solutions, the term is equal to the difference between the osmotic pressure at the membrane surface minus the permeate osmotic pr essure, as indicated in Chapter 2 above from equation number (2-24): where i = inRT n i iC1 where in= number of ions formed when the solute dissociates. And iC= molar concentration of the solute = ic/1000 iMW; and ic= Total Dissolved Solids as mg/l; iMW= Molecular Weight of the dilute solution; R= gas constant; and T= absolute temperature. From the equations (2-23), and (2-24) in Chapter 2 above, membrane wall concentration Cmem was calculated. From equations (2-29), and (230) in Chapter 2, the mass tran sfer coeficient K, and the concentration polarization layerF were calculated. Peclet number was calculated as per equations (2-31), and (2-32) in Chapter 2. 3.4 Assumptions of the Experiments Based on the above discussi on, the following assumptions in the experimental setup were made: (1) the membrane is neutral because the zeta potential is close to zero; (2) the diffusion coefficient of the binary solution species is only a function of temperature and solution concentration, hen ce the diffusivity is constant because the experiments are carried out at a constant temperature and constant feed concentration;

PAGE 91

71 (3) the mass transfer coefficien t of the whole membrane is constant, and is not a function of the permeate flux or concentrate flow ve locity; (4) the concentrate flow is equally distributed on the membrane envelope and the feed and permeate spacers; (5) the type of membrane material and chemical structure do es not influence concen tration polarization; (6) there are no dead zones where salt can ac cumulate on the membrane; (7) the binary solutions used in the experiments are complete ly ionized in the water, and the concentrate solutions have not reached the saturation limits so that there is no species precipitations on the membrane; (8) the concentration polarizat ion layer is stagnant and constant along the run of the membrane surface.

PAGE 92

72 CHAPTER 4 RESULTS AND DISCUSSION This chapter presents the re sults obtained from the tw o setups of experiments described in Chapter 3. It also discusses the results in details. The results are presented in tables and graphs. The graphs are arranged such that they can be compared with each others for each dilute solution, at different pressures. The details of the replicate runs are shown in the tables of Appendices A, B, and C. Pressures were identified as low, medium and high; while solute dilutions were also determined as low, medium, and high, as pe r table 3-2 presented pr eviously in Chapter 3. For an easier illustration, the medium pressure 100 to110 psi is indicated on the graphs by the solid line for both operating set ups, i.e without permeate suction, and with permeate suction, while the other two pre ssures, i.e. low, and high pressures, are indicated by different dotted lines. 4.1 Effect of Permeate Suction on the C oncentration Polarization Layer Thickness According to the equations (233), and (2-34) in Ch apter 2, the diffusion coefficient for the dilute solutions of4MgSO, and2MgCl, and NaClare 0.85 x910, 1.243 x910, and 1.612 x910s m/2, respectively. The molecular weights of the three binary salt solutions are 120.3, 95.3, and 58. 5, respectively. This indicates that the diffusion coefficient is a function of the mol ecular weight; and at constant temperature,

PAGE 93

73 the back diffusion of the salt is proportio nal to the diffusion coefficient and solute concentration. Hence, the diffusivity is assu med to be constant for all the experiments, since a cooler was used to keep the temper ature at 25 degrees C. Figures 4-1, and 4-2 illustrate that the concentration polarizat ion layer thickness, in general, is decreased with permeate suction at all the tested pressu res and feed concentrations. According to the film theory, the concentration polariza tion layer is inversel y proportional to the mass transfer coefficient. This suggests that the effect of permeate suction enhances the mass transfer from the bulk to the membrane surface, and it de-stabilizes the laminar flow condition in the conduit due to the gradual increase in the positive pr essure on the bulk solution due to the permeate suc tion at the permeate collector tube. The above mentioned two figures also show that the c oncentration polarization layer thickness increases with the increase of the TDS of the feed. A more detailed study of Figure 4-1 and 4-2 shows that th e greatest impact on the concentration polarization layer was achiev ed at the medium feed pr essure, which is 100 psi. Also, the above two figures along with Figure 4-3 show that the thickness of the concentration polarization layer is proportional to the rejected ions in the binary dilute solution, since the solute flux in the membrane is propositional to concentration difference across the membrane.

PAGE 94

74 Concentration Polarization Layer Thickness Versus Feed Concentration MgCl2 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 0.0170.0220.0270.0320.037 Feed Concentration (Mol/l) ConcentrationPolarizationLayerThicknessx10 -5 (m) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Linear (HPP + Permeate Pump 80 psi) Linear (HPP + Permeate Pump 100 psi) Linear (HPP Only 80 psi) Linear (HPP Only 100 psi) Linear (HPP Only 130 psi) Linear (HPP + Permeate Pump 130 psi) Figure 4-1 Concentration polarization layer thickness versus feed concentration -2MgCl

PAGE 95

75 Concentration Polarization Layer Thickness Versus Feed Concentration MgSO4 3 4 5 6 7 8 0.0130.0150.0170.0190.0210.0230.0250.0270.029 Feed Concnetration (Mol/l) ConcentrationPolarizationLayerThickness10 5 (m) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Poly. (HPP + Permeate Pump 130 psi) Poly. (HPP Only 130 psi) Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump 80 psi) Linear (HPP + Permeate Pump 100 psi) Linear (HPP Only 100 psi) Figure 4-2. Concentration polarization layer thickness versus feed concentration -4MgSO

PAGE 96

76 It is also illustrated from Figure 4-3 that NaCl solutions are showing the same trend as2MgCl, and 4MgSOsolutions, but with one higher order of magnitude. This is due to the much lower rejection of the NF membrane for both monovalent ions in theNaClsolutions. According to the solution-diffusion transport model, the solute transport in the membrane is a function in the difference between the membrane wall concentration, and the permeate concentrati on, regardless of the operating pressure. Since the permeate con centration is relatively high, the concentration polarization layer thickness was relatively high. Figure 4-3 al so shows that the greatest reduction in concentration polari zation layer thickness in the NaCl solution is achievable at the medium pressure which is 110 psi. As it was mentioned above, in the case of NaCl solutions the concentration polarizati on layer thickness is of one order of magnitude higher than that of the case of2MgCl, and 4MgSOsolutions. Part of this higher magnitude is due to th e higher tested pressure for NaClsolutions (110 psi), as compared to the other two solutions of 2MgCland 4MgSO(100 psi). Figure 4-4 summarizes the conclu sion from the above three figures. It shows the relative effect of the permeate suction on the three binary solutions at the medium range of pressure (100 to110 psi) in which the pe rmeate suction had the greatest impact. It clearly shows that the concentration pol arization layer thickness is reduced with suction for all the tested solutions. Detail ed study of the above-mentioned figure also shows that the concentration polarization la yer thickness is a function of the diffusion coefficient of the solute. As the diffusion coefficient increases from 4MgSO to

PAGE 97

772MgCltoNaCl, the concentration polarization layer increases if the feed solution concentration and the temperature are kept constant. Concentration Polarization Layer Thickness Versus Feed Concentration NaCl0.7 0.9 1.1 1.3 1.5 1.7 0.02350.02850.03350.03850.04350.04850.05350.0585 Feed Concentration (Mol/l) ConcentrationPolarizationLayerThicknessx10 4 (m) HPP Only -80 psi HPP + Permeate Pump HPP Only 110 psi HPP + Permeate Pump 110 psi HPP Only 160 psi hpp + Permeate Pump -160 psi Linear (HPP Only -80 psi) Linear (HPP + Permeate Pump) Linear (HPP Only 110 psi) Linear (HPP + Permeate Pump 110 psi) Linear (HPP Only 160 psi) Linear (hpp + Permeate Pump -160 psi) Figure 4-3. Concentration polarization layer thickness versus feed concentration NaCl

PAGE 98

78 Concentration Polarization Layer Thickness Vresus Feed Concentration 100 to 110 psi for the Three NaCl, MgCl2, and MgSO4 Solutions 3.0000E+00 5.0000E+00 7.0000E+00 9.0000E+00 1.1000E+01 1.3000E+01 7009001100130015001700 Feed Concentration (mg/l) ConcentrationPolarizationLayerThicknessx10 -5 (m) HPP Only NaCl HPP + Permeate Pump NaCl HPP Only MgSO4 HPP + Permeate Pump MgSO4 HPP Only MgCl2 HPP + Permeate Pump MgCl2 Linear (HPP Only NaCl) Linear (HPP + Permeate Pump NaCl) Linear (HPP Only MgCl2) Linear (HPP + Permeate Pump MgCl2) Linear (HPP Only MgSO4) Linear (HPP + Permeate Pump MgSO4) Figure 4-4. Concentration polarization layer thickness versus feed concentration at 100 to 110 psi for the three NaCl, 2MgCl, and 4MgSOsolutions

PAGE 99

79 Figure 4-5 is plotting the tr end in concentration polar ization layer thickness against the net operating pressure (P ) for2MgClsolutions. The solid line is showing the trend with permeate suction, while the dotted line is presenting the trend without permeate suction. For example, at a net operating pr essure of 5.45 atm, the average concentration polarization layer th ickness without permeate suction was 7.9 x 510 m. This was reduced to about 6.7 x 510m with permeate suction, resulting in a reduction of about 15%. Concentration Polarization Layer Thickness Versus Net Operating Pressure -MgCl2 5.4 5.9 6.4 6.9 7.4 7.9 8.4 8.9 9.4 3.954.454.955.455.956.456.957.45 Net Operating Pressure (atm)Concentration Polarization La y er Thickness x 10 5 ( m ) HPP Only HPP + Permeate Pump Linear (HPP Only) Linear (HPP + Permeate Pump) Figure 4-5. Concentration polarization layer thickness versus net operating pressure 2MgCl

PAGE 100

80 4.1.1 Statistically Testing the Experiment al Design for Concentration Polarization Layer Thickness The measured concentration polarization layer thicknesses were in the order of magnitude of 510to 410m. To validate the experiments, and to eliminate experimental errors, the Analysis of Variance (ANOVA) was tested. ANOVA were carried out on the two different treatments (without permeate suction, and with permeate suction) to compar e the mean value of the two tests, and to check if the permeate suction has made a sign ificant change from the case of not having permeate suction. Two statisti cal hypotheses were tested: 0H: 1 = 2 and 1H: 1 2 where 1 is the mean concentration layer thickne ss in the case of r unning the test with out using permeate suction in m multiplied by 510, and 2 is the mean concentration layer thickness in the case of using permeate suction in m multiplied by 510. The following ANOVA tables for th ree tested salts are showing the results of the experiment analysis. 4.1.1.1 ANOVA for2MgCl Solutions Referring to Table 4-1 below, and from Table IV of Design and Analysis of Experiments book (Montgomery D., 2001): Critical value of )1)(1),(1,(050 baaF= 8,1,050F= 5.32 Since oF = 288.552 >8,1,050F= 5.32

PAGE 101

81 so 0H: 1 = 2 is rejected, and 1H: 1 2 It was concluded that there was a difference between the two treatments. Hence, the permeate suction was significantly different from the case of no suction. Table 4-1. ANOVA Table for 2MgCl solutions Source of Variation Sum of Square DOF Mean Square oF Treatment (w/o & w suction) 6.8672 1 6.8672 288.5552 Blocks 17.9010 8 2.2376 Error 0.1903 8 0.0238 Total 26.1177 17 4.1.1.2 ANOVA for4MgSO Solutions Referring to Table 4-2 below, and from table IV of Design and Analysis of Experiments text book (Montgomery D., 2001): Critical value of )1)(1),(1,(050 baaF= 8,1,050F= 5.32 Since oF = 267.2989 >8,1,050F= 5.32 so 0H: 1 = 2 is rejected, and 1H: 1 2 It was concluded that there was a di fference between the two treatments. Hence, the permeate suction was significantly different from the case of no suction.

PAGE 102

82Table 4-2. ANOVA Table for 4MgSO solutions Source of Variation Sum of Square DOF Mean Square Fo Treatment (w/o & w suction) 2.2493 1 2.2493 267.2989 Blocks 19.9345 8 2.4918 Error 0.0673 8 0.0084 Total 22.2512 17 4.1.1.3 ANOVA forNaCl Solutions Referring to Table 4-3 below, and from table IV of Design and Analysis of Experiments text book (Montgomery D., 2001): Critical value of )1)(1),(1,(050 baaF= 8,1,050F= 5.32 Since oF = 10.36771 >8,1,050F= 5.32 so 0H: 1 = 2 is rejected and 1H: 1 2 Table 4-3. ANOVA Table for NaCl solutions Source of Variation Sum of Square DOF Mean Square Fo Treatment (w/o & w suction) 0.5894 1 0.5894 10.3677 Blocks 104.1044 8 13.0130 Error 0.4548 8 0.0568 Total 105.1486 17

PAGE 103

83 4.2 Effect of Permeate Sucti on on Mass Transfer Coefficient As was mentioned above, the mass transfer coefficient is inversely proportional to the concentration polarization layer thickness. Figures 4-6, 4-7, a nd 4-8 show that the mass transfer coefficient for all dilute solu tions increased with permeate suction, if compared with the case of no permeate suction. Again, it is deduced that th e permeate suction destabili zes the boundary layer in the laminar flow condition that reduces c oncentration polarizati on and enhances the mass transfer coefficient. The above-menti oned figures show that the greatest mass transfer coefficients were achievable when the operating conditions we re in the range of 100 to 110 psi. However, the mass transfer coefficient rates for4MgSOsolutions were reduced faster at the higher feed concentr ation greater than 0.0225 Mol/l, due to the higher rate of the increase of concen tration polarization layer at the higher concentrations.

PAGE 104

84 Mass Transfer Coefficient Versus Feed Concentration MgCl2 1.25 1.45 1.65 1.85 2.05 2.25 0.0170.0220.0270.0320.037 Feed Concnetration (Mol/l) MassTransferCoefficientx10 -5 (m/s) HPP Only 80 psi HPP + Permeate Pump80 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Linear (HPP Only 100 psi) Linear (HPP + Permeate Pump 100 psi) Linear (HPP + Permeate Pump80 psi) Linear (HPP Only 80 psi) Linear (HPP Only 130 psi) Linear (HPP + Permeate Pump 130 psi) Figure 4-6. Mass transfer coefficient versus feed concentration 2MgCl

PAGE 105

85 Mass Transfer Coefficient Versus Feed Concentration -MgSO4 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 0.0130.0150.0170.0190.0210.0230.0250.0270.029 Feed Concentration (Mol/l) MassTransferCoefficientx10 5 (m/s) HPP Only 130 psi HPP + Permeate Pump -130 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 80 psi HPP + Permeate Pump 80 psi Poly. (HPP Only 80 psi) Linear (HPP Only 130 psi) Linear (HPP + Permeate Pump 100 psi) Linear (HPP Only 100 psi) Linear (HPP + Permeate Pump -130 psi) Linear (HPP + Permeate Pump 80 psi) Figure 4-7. Mass transfer coefficient versus feed concentration 4MgSO

PAGE 106

86 Mass Transfer Coefficient Versus Feed Concentration NaCl0.8 1 1.2 1.4 1.6 1.8 2 2.2 0.02350.02850.03350.03850.04350.04850.05350.0585 Feed Concentration (Mol/l) MassTransferCoefficientx10 -5 (m/s) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 110 psi HPP + Permeate Pump 110 psi HPP + Permeate Pump 160 psi HPP Only 160 psi Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump 80 psi) Linear (HPP Only 110 psi) Linear (HPP + Permeate Pump 110 psi) Linear (HPP + Permeate Pump 160 psi) Linear (HPP Only 160 psi) Figure 4-8. Mass transfer coefficient versus feed concentration -NaCl

PAGE 107

87 4.3 Effect of Permeate Suction on Permeate Flow Figures number 4-9, 4-10, and 4-11 show the relationship between the feed concentration, and the product flow for the different dilute solu tions at the three pressures. It is evident from the figures that the product flow has increased due to permeate suction for all salt so lutions, and under the three te sted pressures. As it was expected, the permeate flow rate, and cons equently the permeate flux was increased with the higher feed pressure. In general, th e product flow rate was reduced as the feed concentration increased. However, for4MgSOand NaCl solutions the greatest rate of increase due to permeate suction was achieve d at the medium operating feed pressure range (100 to110 psi).

PAGE 108

88 Permeate Flow Versus Feed Concentration MgCl2 2.3 2.8 3.3 3.8 4.3 0.0170.0220.0270.0320.037 Feed Concentration (Mol/l) PermeateFlowx10 5(m3/s) HPP Only 80 psi HPP + Permeate Pump HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump) Linear (HPP Only 100 psi) Linear (HPP + Permeate Pump 100 psi) Linear (HPP Only 130 psi) Linear (HPP + Permeate Pump 130 psi) Figure 4-9. Permeate flow versus feed concentration 2MgCl

PAGE 109

89 Permeate Flow Versus Feed Concentration MgSO4 2.3 2.8 3.3 3.8 4.3 4.8 0.0130.0150.0170.0190.02 10.0230.0250.0270.029 Feed Concnetration (Mol/l) PermeateFlowx10 -5 (m3/s) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP + Permete Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump 80 psi) Linear (HPP Only 100 psi) Linear (HPP + Permete Pump 100 psi) Linear (HPP Only 130 psi) Linear (HPP + Permeate Pump 130 psi) Figure 4-10. Permeate flow versus feed concentration 4MgSO

PAGE 110

90 Permeate Flow Versus Feed Concentration NaCl 2.5 3 3.5 4 4.5 5 5.5 6 0.0240.0290.0340.0390.0440.0490.0540.059 Feed Concentration (Mol/l) PermeateFlowx10 5 (m 3 /s) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 110 psi HPP + Permeate Pump 110 psi HHP Only 160 psi HPP + Permeate Pump 160 psi Linear (HPP + Permeate Pump 80 psi) Linear (HPP Only 80 psi) Linear (HPP Only 110 psi) Linear (HPP + Permeate Pump 110 psi) Linear (HPP + Permeate Pump 160 psi) Linear (HHP Only 160 psi) Figure 4-11. Permeate flow versus feed concentration NaCl

PAGE 111

91 4.4 Effect of Permeate Suction on Permeate Concentration Figures 4-12, 4-13, and 4-14 i llustrate the change in permeate concentration versus feed concentration with permeate sucti on at various pressures. It is clear from the three figures that permeate suction ha s improved the quality of the permeate concentration at all pressures. It is intere sting to notice on the first two figures that the greatest improvement was achieved at the medium feed pressure (100 psi). It is worthy to me ntion that the permeate concentration of 2MgClsolutions was much higher than the pe rmeate concentration of 4MgSOsolutions under the same operating conditions. This as indicated in item 4.1 above, is due to th e low rejection of the Clmonovalent anion by the NF membrane, while the rest of ions in the two solutions, namely, Mg, and 4SOare highly rejected because they are divalent ions. The ionic mobility of ions has also contributed to the rate of their rejection. Robinson and Stockes (1965) have indicated th at the radius of chloride ion is 1.81 A, versus a greater radius for sulfate compound ion. The fact that the sulfate compound anion has a radius greater than that of the chlo ride element anion has helped the latter in its greater passage rate through the membrane. In all cases, the use of permeate suction has resulted in a better permeate quality. This was more pronounced in the solutions of the divalent ions than th e monovalent ion solutions.

PAGE 112

92 Permeate Concentration Versus Feed Concentration MgCl2 17 19 21 23 25 27 29 31 33 35 0.0170.0220.0270.0320.037 Feed Concentration (Mol/l) PermeateConcentrationx10 5(Mol/l) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump 80 psi) Linear (HPP Only 100 psi) Linear (HPP + Permeate Pump 100 psi) Linear (HPP Only 130 psi) Linear (HPP + Permeate Pump 130 psi) Figure 4-12. Permeate concentration versus feed concentration 2MgCl

PAGE 113

93 Permeate Concentration Versus Feed Concentration MgSO4 3.3 4.3 5.3 6.3 7.3 8.3 9.3 0.0130.0150.0170.0190.0210.0230.0250.0270.029 Feed Concentration (Mol/l) PermeateConcentrationx10 -5 (Mol/l) HPP Only 130 psi HPP + Permeate Pump 130 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 80 psi HPP + Permeate Pump 80 psi Linear (HPP + Permeate Pump 80 psi) Linear (HPP Only 80 psi) Linear (HPP Only 100 psi) Linear (HPP Only 130 psi) Linear (HPP + Permeate Pump 130 psi) Linear (HPP + Permeate Pump 100 psi) Figure 4-13. Permeate concentration versus feed concentration 4MgSO

PAGE 114

94 Permeate Concentration Versus Feed Concentration NaCl 1 1.5 2 2.5 3 3.5 4 4.5 0.02350.02850.03350.03850.04350.04850.05350.0585 Feed Concentration (Mol/l) PermeateConcentrationx10 4 (Mol/l) HPP only 80 psi HPP + Permeate Pump 80 psi HPP Only 110 psi HPP + Permeate Pump 110 psi HPP Only 160 psi HPP + Permeate Pump 160 psi Linear (HPP + Permeate Pump 110 psi) Linear (HPP Only 110 psi) Linear (HPP + Permeate Pump 80 psi) Linear (HPP only 80 psi) Linear (HPP + Permeate Pump 160 psi) Linear (HPP Only 160 psi) Figure 4-14. Permeate concentration versus Feed concentration NaCl

PAGE 115

95 4.5 Effect of Permeate Suction on Concentrate Concentration Concentrate concentration is generally increased due to permeate suction for all dilute solutions at different concentrations and pressures as shown in Figures (4-15, 4-16, and 4-17). The increase is due to the reduction in the membrane wall concentration as it will be explained later in item 4.6. Especially at the medium feed pressure range (100 to 110 psi), the rate of increase of the concen trate concentration due to suction in 4MgSOsolutions was higher. This was due to the higher rejection of its two divalent ions Mg, and 4SO, as opposed to the lo wer rejection of the monovalent ion Cl that is available in the2MgClsolutions, as was indicated in item 4.1 above.

PAGE 116

96 Concentrate Concentration Versus Feed Concentration MgCl2 2.6 3.1 3.6 4.1 4.6 5.1 5.6 6.1 0.0170.0220.0270.0320.037 Feed Concentration (Mol/l) ConcentrateConcentrationX10 -2 (Mol/l) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump 130 psi) Linear (HPP Only 100 psi) Linear (HPP Only 130 psi) Linear (HPP + Permeate Pump 100 psi) Linear (HPP + Permeate Pump 80 psi) Figure 4-15. Concentrate concentration versus feed concentration -2MgCl

PAGE 117

97 Concentrate Concentration Versus Feed Concentration MgSO41.9 2.4 2.9 3.4 3.9 4.4 4.9 0.01250.01450.01650.01850.02050.02250.02450.02650.0285 Feed Concentration (Mol/l) ConcentrateConcentrationx10 -2 (Mol/l) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Poly. (HPP Only 100 psi) Poly. (HPP + Permeate Pump 100 psi) Linear (HPP Only 130 psi) Linear (HPP + Permeate Pump 130 psi) Linear (HPP + Permeate Pump 80 psi) Linear (HPP Only 80 psi) Figure 4-16. Concentrate concentration versus feed concentration -4MgSO

PAGE 118

98 Concentrate Concentration Versus Feed Concentration -NaCl 36 46 56 66 76 86 96 0023500285003350038500435004850053500585 Feed Concentration (Mol/l) ConcentrateConcentrationx10 2 (Mol/l) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP + Permeate Pump 110 psi HPP Only 110 psi HPP Only 160 psi HPP + Permeate Pump -160 psi Linear (HPP + Permeate Pump 110 psi) Linear (HPP Only 110 psi) Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump 80 psi) Linear (HPP Only 160 psi) Linear (HPP + Permeate Pump -160 psi) Figure 4-17. Concentrate concentration versus feed concentration -NaCl

PAGE 119

99 4.6 Effect of Permeate Suction on Membrane Wall Concentration Membrane wall concentration can not be experimentally measured. Therefore it was calculated using the solution diffusion mode l, as it was discussed in Chapter 2. Figures 4-18, 4-19, and 4-20 show that th e calculated membrane wall concentration is lower with permeate suction if compared to the case of running the high pressure pump only at all pressures for all th e tested solutions. A more detailed analysis of the abovementioned figures can be illustrated if they are compared to the corresponding three figures of the concentrate concentration in it em 4.5, namely Figures (4-15, 4-16, and 417). It is deduced from that comparison th at the membrane wall concentration is a function of the effect of c oncentration polarization. The higher the feed pressure, the more pronounced the difference in concentr ation between no suction, and permeate suction. The difference between the ionic sp ecies radii of chlori de element ion, and sulfate compound ions as it was discussed in item 4.1 and 4.4 above has contributed to the distinction. For example, at the medium pressure range (100 to 110 psi), where the permeate suction had the greatest impact, the fact that the sulfate compound anion has a radius greater than that of th e chloride element anion has helped the latter in its higher passage rate through the membrane in bothNaCl, and 2MgClsolutions such that the membrane wall concentration in those cases is greater than the case of 4MgSOsolutions.

PAGE 120

100 Membrane Wall Concentration Versus Feed Concentration MgCl2 2.3 3.3 4.3 5.3 6.3 7.3 8.3 9.3 10.3 0.0170.0220.0270.0320.037 Feed Concentration (Mol/l) MembraneWallConcentrationx10 2(Mol/l) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Linear (HPP + Permeate Pump 80 psi) Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump 100 psi) Linear (HPP Only 100 psi) Linear (HPP + Permeate Pump 130 psi) Linear (HPP Only 130 psi) Figure 4-18. Membrane wall concentration versus feed concentration -2MgCl

PAGE 121

101 Membrane Wall Concentration Versus Feed Concentration MgSO4 3.3 4.3 5.3 6.3 7.3 8.3 9.3 0.0130.0150.0170.01900210023002500270029 Feed Concentration (Mol/l) MembraneWallConcentrationx10 2 (Mol/l) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi HPP + Permeate Pump 100 psi Linear (HPP Only 130 psi) Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump 80 psi) Linear (HPP + Permeate Pump 100 psi) Linear (HPP Only 100 psi) Linear (HPP + Permeate Pump 130 psi) Figure 4-19. Membrane wall concentration versus feed concentration -4MgSO

PAGE 122

102 Membrane Wall Concentration Versus Feed Concentration NaCl3.5 7 10.5 0.0240.0290.0340.0390.0440.0490.0540.059 Feed Concentration (Mol/l) MembraneWallConcentrationx10 -2 (Mol/l) HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 110 psi HPP + Permeate Pump 110 psi HPP Only 160 psi HPP + Permeate Pump 160 psi Linear (HPP Only 80 psi) Linear (HPP + Permeate Pump 80 psi) Linear (HPP + Permeate Pump 110 psi) Linear (HPP Only 110 psi) Linear (HPP + Permeate Pump 160 psi) Linear (HPP Only 160 psi) Figure 4-20. Membrane wall concen tration versus feed concentration -NaCl

PAGE 124

alter the velocity profile from profile number (1) to profile number (2) due to the destabilization of the boundary layer. In additi on, let us take a cross section perpendicular to the membrane at steady state condition. Al so, let us consider a random point at the boundary layer velocity profile, and its correspondent point at the concentration profile. Suction will increase velocity U1 to veloci ty U2, which will increase the local flow at that point. Since the corresponding point at th e concentration profile without suction is C1, the concentration at the same point after applying suction will be decreased to C2 because the corresponding velocity and flow rates have increased. When applying the mass balance equation for the whole system at steady state conditions, the average concentrate concentration at the bulk solu tion at the case of applying suction will be increased, if compared to the average concentrate concentration of the bulk solution before suction. This is due to the increased flow near the membrane that will partially wash away the accumulated species on the membrane. This was the observation of the experiment for the concentrate concentrati on as indicated in Fi gures 4-15, 4-16, and 417 in item 4.5 above for all th e solutions at all pressures. 4.7 Effect of Permeate Suction on Peclet Number Peclet number is defined as the dimentionless ratio of the rate of mass transported by convection to the membrane, to the rate of mass transported by diffusion back to the bulk solution. In other words, th e diffusive membrane Peclet number which is expressed as per equation (2-31) in Chapter 2 is: eP = 21 D hVdw where is the permeate velocity which is equal to the permeate flux per one sheet of wV 104

PAGE 125

membranes; is the diffusivity coefficicent of the ionized electrolye; and is the 21 D dh hydraulic diameter of the spiral wound memb rane. The diffusive Peclet number is a measure of how permeate goes through the memb rane. It is observed that for the dilute solutions the Schmidt number is very large, as a consequence of which the diffusion Peclet number is large. This is tr ue even at moderate Reynolds number (Probstien,1994). In the case of using the standard hi gh pressure pump only, both the permeate concentration and the concentration polari zation layer increase as the Peclet number increase (Probstein, 1994). However Figures 4-22, 4-23, and 4-24 show a remarkable result. They show that the diffusive Peclet number for the binary dilute solutions has increased with the permeate suction at al l pressures, although the associated permeate concentration, and concentration polarizati on layer thicknesses have decreased as was discussed in item 4.1 and 4.3 above. This is a proof that the permeate suction has stabilized the flow conditions, and has e nhanced the mass transfer coefficient. From the above mentioned figures, at permeat e suction, and at different feed operating conditions, the diffusive Peclet number can be expressed according to the equation : eP = + + with 2 1xa xb1 `1c 2 R > 0.99, where x is the feed concentration in Mol/l, and , and are coefficients dependent on feed pr essure for every binary salt solution. 1a 1b 1c 105

PAGE 126

Peclet Number Versus Feed Concentration MgCl2 y = -307.37x2 + 1.6963x + 2.0213 R2 = 0.9998 y = -30.336x + 3.7124 R2 = 0.9995 y = -1758.8x2 + 73.385x + 1.5321 R2 = 0.9934 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 0.017 0.022 0.027 0.032 0.037 Feed Concentration (Mol/l) P ecl et N u m b er HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP + Permeate Pump 130 psi Poly. (HPP Only 80 psi) Poly. (HPP + Permeate Pump 80 psi) Linear (HPP + Permeate Pump 130 psi) Poly. (HPP Only 100 psi) Poly. (HPP + Permeate Pump 100 psi) Poly. (HPP Only 130 psi) Figure 4-22. Diffusive Peclet number versus feed concentration 2MgCl 106

PAGE 127

Peclet Number Versus Feed Concentration MgSO4 y = -2111.4x2 + 58.393x + 2.5756 R2 = 0.9993 y = -1797.1x2 + 40.006x + 3.6133 R2 = 0.9947 y = 1686.8x2 99.483x + 5.7313 R2 = 0.9994 2.3 2.8 3.3 3.8 4.3 4.8 0.0130.0150.0170.0190.0210.0230.0250.0270.029 Feed Concentration (Mol/l) P e cl et N u m b er HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 100 psi HPP + Permeate Pump 100 psi HPP Only 130 psi HPP Only + Permeate Pump 130 psi Poly. (HPP Only 80 psi) Poly. (HPP + Permeate Pump 80 psi) Poly. (HPP Only 100 psi) Poly. (HPP + Permeate Pump 100 psi) Poly. (HPP Only + Permeate Pump 130 psi) Poly. (HPP Only 130 psi) Figure 4-23. Diffusive Peclet number versus feed concentration 4MgSO 107

PAGE 128

Peclet Number Versus Feed Concentration NaCly = -28.205x2 9.538x + 2.4562 R2 = 0.9969 y = -450.34x2 + 29.473x + 2.6508 R2 = 0.996 y = 240.83x2 26.213x + 2.1404 R2 = 09992 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 0.024 0.029 0.034 0.039 0.044 0.049 0.054 0.059 Feed Concentration (Mol/l) P e c l et N u m b er HPP Only 80 psi HPP + Permeate Pump 80 psi HPP Only 110 psi HPP + Permeate Pump 110 psi HPP + Permeate Pump 160 psi HPP Only 160 psi Poly. (HPP Only 110 psi) Poly. (HPP + Permeate Pump 110 psi) Poly. (HPP + Permeate Pump 110 psi) Poly. (HPP Only 160 psi) Poly. (HPP + Permeate Pump 160 psi) Poly. (HPP + Permeate Pump 80 psi) Poly. (HPP Only 80 psi) Figure 4-24. Diffusive Peclet number versus feed concentration NaCl 108

PAGE 129

109 CHAPTER 5 CONCLUSION AND RECOMMENDATIONS This chapter consists of three parts. The first part conc ludes the results of the experiments. The second part emphasizes the importance of the findings on the design improvements for NF plants. The third part discusses the suggested and recommended future researches that can be carried out as a further step to this research. 5.1 Conclusion The goal of this research was to make use of the effect of the de-stabilization of the laminar flow that exists at the memb rane surface due to the gradually increased permeate suction, in an attempt to reduce th e concentration polarization layer in the module. Previous researches showed that the concentration polarization is almost always the reason behind membrane fouling. The technique of using permeate suction is not practically used in the NF of RO membrane industry so far, despite that it has been theoretically investigated at the la boratory scale by a few researchers. (1) This research showed that when using binary dilute so lutions, the permeate suction reduced the concentr ation polarization at the f eed side of the industry scale NF membrane surface that helped to increase mass transfer coefficient, and increased the product flux without subj ecting the membrane to less favorable conditions.

PAGE 130

(2) The research also showed that permeate suction had the greate st impact at the medium range (100 to110 psi) feed pressure, resulting in a reduction of concentration polarization layer thickne ss with an average between 14 to 20%. (3) The magnitude of the measured concentr ation polarization layer thickness in the experiments was very small in the order of 3 x to 8 x meters. To eliminate experimental errors, the analys is of the variances of the experiments (ANOVA) were tested in both experiment treatments (without suction, and with suction) to investigate the significanc e of the permeate suction. ANOVA Tables for the three tested binary dilute salt solutions showed that the permeate suction was significant, and applying the permeate suction was statistically different from the case of not applying suction. 510 410 (4) This research showed that calculating the concentra tion polarization layer using the traditional way of using Sherw ood number correlation would lead to erroneous results due to the changes in the solution properties because of suction, that are not considered in th is relationship. In addition, it is even believed that the use of the correlation of Sherwood number in the literature for the traditional high pressure pump desi gn only might be inadequate. This is shown by the different values of the coe fficients that were deduced by different researchers in calculating the relationships in that correlation. (5) The use of Peclet number instead, which does not primarily depend on most of the changed physical properties of the so lutions would eliminate the need to use the Sherwood number. The values used in calculating the diffusive Peclet number could be easily and appropriate ly calculated. Although the Peclet 110

PAGE 131

number is a function of the solute diffu sion coefficient, keeping the temperature constant through all the experiments has el iminated the change that could occur in the solute diffusivity back to the bulk solution. The velocity term in the Peclet number was easily calculated from the experimentally measurable permeate flux. The hydraulic diameter coefficient in Peclet number is dependent on the membrane structure, which was calculated from the geometrical dimensions of the tested membrane. (6) The diffusive Peclet number increased with the permeate suction at all the experimental testes. The Peclet number followed a pattern of = + + with eP 2 1xa xb1 `1c 2 R > 0.99, where x is the feed concentration in Mol/l. The terms , and are coefficients dependent on feed pressure for every binary salt solution. The results showed that the perm eate flow increased at the same time, while the concentration polarization was reduced. This was opposite to the traditional case where permeate suction is not used. In the later case, the increase of the Peclet number increases the concentration polarization layer. 1a 1b 1c (7) Although some researchers used the concentration polar ization layer thickness, and boundary layer thickness interchang eably (e.g. Lisdonk, C., et al. 2001; Sablani S., et al, 2001) they are actually different. At the dilute solutions where Schmidt number is >> 1, the concentrati on polarization layer is imbedded in the viscous boundary layer, and its velocity is that the one close to the wall. Schlichting (1979) estimat ed the boundary layer thickness at the porous wall suction as a function of Reynolds number The calculations for this estimation 111

PAGE 132

112 are based on an assumption of a consta nt viscosity along the membrane width, which may not be accurate, especially for highly concentrated solutions. (8) It is worthy to mention that the equa tion of transport in the membrane is complex for multi component solutions with more than one anion and one cation because more than one cation can be accompanied by one anion or vice versa, depending on the ion valence. It has been shown that the concentration polarization for individual salts changes substantially with the presence of other salts (Srinivasan and Tein, 1970). The solu tion-diffusion model is valid only for binary solution systems, and can not be us ed for a mixture of salts. To estimate the transport in the membrane for a multi component solution, one alternative is the use of the Nernst-Plank model, whic h is for a mixture of n ions (3n+2) equations are required (Ghiu and Carnahan, 2003). 5.2 Recommendations (1) The NF membrane makers traditionally us e the standard testing pressure of 70 to75 psi range. It is suggested that th e membrane manufacturers might have to change their standard test conditions to 100 to110 ps i, since this research showed that the greatest impact on the NF membranes was achieved at the medium feed pressure which was 100 to110 psi. (2) The applications of NF membrane in th e water treatment industry are numerous, and based on the above-mentioned results using the permeate suction can help reduce the concentration polarization la yer in NF modules, and increase mass transfer coefficient, that will el ongate the useful membrane life.

PAGE 133

113 (3) One interesting recent application is using two staged NF-NF to desalinate seawater is very promising. In this case using the permeate suction at the end of both stages may be a way to reduce concen tration polarization in this application for seawater desalination, which woul d lead to a prolonged life of the membrane, in addition to increa sing the NF plant productivity. 5.3 Future Researches (1) It is thought that an econo mical study of adding the cost of the permeate suction pump to the existing traditional module design should be addressed in the upcoming researches, and the total water cost of a system should be evaluated based on that addition. (2) Another proposed future study is dete rmining the minimum suction pressure required to reduce the concentration pol arization layer, so that the plant modification from the current standa rd system design can be optimized. (3) It is also believed that extending the application of the perm eate suction to the higher concentration of brackish water RO modules, or seawater RO plants can critically be investigated, despite the complexity of predicting the mass transfer model for highly concentrated mixed salt solutions.

PAGE 134

LIST OF REFERENCES 1144 2 42 Afonso, Maria et al, 2000, Transport Of, And Across an Amphoteric Nanofiltration Membrane, Journa l of Membrane Science, 179-137-154. MgSO MgCl SONa Al-Bastaki N., 2001, Use of Fl uid Instabilities To Enhance Membrane Performance: A Review, Desalination 136 255-262. Amjad, Z., 1993, Reverse Osmosis Membrane Technology, Van Nostrand Reinhold, New York. Bartels, Craig, 2007, Novel New Fouling Na nofiltration Membranes, Hydranautics Membrane Manufacturer. Bhattacharyya et al, 1999, Separation Of Organic Pollutants By Reverse Osmosis and Nanofiltration Membranes: Mathematical Models and Experimental Verification, Department of Chemical and Materials Engi neering, University of Kentucky, Lexington, Kentucky. Bhattacharyya, D, S.L. Back and R.I. Kermode, 1990, Prediction of Concentration Polarization and Flux Behavior in Reverse Osmosis by Numerical Analysis, Membrane Science, 48 231. Chiolle, A., G Gianotti, M. Gramondo a nd G Parrini, 1978, Mathematical Model of Reverse Osmosis in Parallel Wall Channe ls with Turbulence Promoting Nets, Desalination, 26 3-16. Cruver, J. E., 1978, 1JS Department of In terior, Office of Saline Water Resource, Development Progress Report, NTIS: PB 223 181, No. 882, Desalination, 34 297-303. Cussler, E. L., 1984, Diffusion Mass Transfer in Fluid Systems, Press Syndicate of the University of Cambridge. Drioli, E. and F. Bellucci, 1978, Concentr ation Polarization and Solute-Membrane Interactions Affecting Pressure Driven Memb rane Processes, Desalination, 26 1736. FilmTec membrane handbook, 2009, FilmTec Membrane Manufacturer.

PAGE 135

115 Flemming, Hans Curt, 1993, Mechanistic As pects of Reverse Osmosis Membrane Biofouling and Prevention in Reverse Os mosis Membrane Technology, edited by Zahid Amjad, Van Nostrand Reinhold, New York. Flugzeugbau, DG, 1990, Performance Increase Possibilities of Gliders, Internet Wikipedia Encyclopedia, Fritzmann, C., 2007, State-of-Art of Reverse Osmosis Desalination, Desalination 216-176. Gekas and Hallsrom, 1987, Mass Transfer in the Membrane Concentration Polarization Layer under Turbulent Cross Flow, Geraldes, V., V. Semiao, 2003, Hydrodynamics and Concentration Polarization in NF/RO Spiral Wound Modules with Ladder-T ype Spacers, Desalination 157 395-402. Ghiu, S. and R. Carnahan, 2003, Mass Transfer of Ionic Species in Direct and Reverse Osmosis Processes, Dissertation, University of South Florida. Gill and Soltanieh, 1981, Review of Reverse Osmosis Membranes and Transport Models, 1981, Gordon and Breach Science Publisher, Inc, USA. Gupta, Sharad K. et al, 2005, Modeling of SpiralWound Module and Estimation of Model Parameters using Numerical Techniques, Desalination 173 269-286. Hassan, A., 2004, Development of a Novel NF-Seawater Desalination Process and Review of Application From Pilot Plant to Commercial Production Plant Stages, Beirut Conference. Hydranautics Membrane Manufacturer, 2009. Jamal, J., M.A. Khan, M. Kamil, 2004, Mathematical Modeling Of Reverse Osmosis Systems, Desalination 160 29-42. Le Gouelle, Yann et al, 2006, A Novel Appro ach to Seawater Desa lination using DualStaged Nanofiltration, American Water Works Association. Lide, D.R., 1995, Handbook of Chemistry and Phsics,76 th Edition, CRC Press, Inc. Lisdonk, C. A. et al, 2001,The Influence of Concentration Polarizat ion on the Risk of Scaling in Spiral Wound Membrane Sy stem, Kiua Research and Consultancy, Netherlands.

PAGE 136

116 Mahlab, D., N. Ben Yosef and G. Belfort, 1998, Concentration Polarization Profile for Dissolved Species in Unstirred Batch Hyperfiltration (Reverse Osmosis), Desalination,11957 Mohammadi, Toraj, 2005, Mathematical Mode ling of Flux Decline in Ultrafiltration, Desalination 184-367-373. Montgomery, Douglas C., 2004, Design and Analys is of Experiments, Wiley Interscience Publication, John Wiley & Sons, Inc., New York. Mukiibi, Mo, 2008, Reverse Osmosis Opportuni ties and Challenges, Water Conditioning & Purification Magazine, USA. Probstein, Ronald F., 1994, Physicochemical Hydrodynamics, Wiley Interscience Publication, John Wiley & Sons,Inc., New York. Robinson, R. and Stokes, R, 1965, Electrolyte Solutions, Second Edition, London Butterworth. Rogers, Laminar Flow Analysis, 19 92, Cambridge University Press. Sablani, S. S., M.F. A. Goosen, R.Al-Bel ushi, M.Wilf, 2001, Concen tration Polarization in Ultrafitration and RO: A Critic al Review, Desalination 141, 269-289. Sablani, Shyam et al, Influence of Spacer Thickness on Permeate Flux in Spiral-Wound Seawater Reverse Osmosis Systems, Desalination 146 (2002) 225-230. Sangyoup, Lee, Jaeweon Cho, Menachem Elimelec h, Influence of Collo idal Fouling and Feed Water Recovery on Sa lt Rejection of RO and NF Membranes, Desalination 160(2004) 1-12. Schaep, Johan, et al, 2001, Modeling the Rete ntion of Ionic Components for Different Nanofiltration Membranes, Separation and Purification Technology, 22-23-169-179. Schlichting, Boundary Layer Theory, Mc Graw-Hill, Inc., 1979, pages 26, 379-385. Schock, G., and Miquel, A., 1987, Mass Transf er and Pressure Loss in Spiral Wound Modules, Desalination 64-339-352. Singh, R., Hybrid Membrane Systems for Water Purification: Technology, Systems Design and Operation, Elsevier Science Publishers, UK. Sirshendu, De, & Bhattarya, P. K., 1999, Mass Transfer Coefficient with Suction Including Property Variations in Applications of Cross-Flow Ultrafiltraion, Separation Purification Technology, 16-61-73.

PAGE 137

117 Song L., 1998, A New Model For The Calculati on of The Limiting Flux in Ultrafiltration, Journal of Membrane Science, 173-185. Srinivasan, S. and Tien, C., 1970, A Si mplified method for the Prediction of Concentration Polarization in Reverse Osmo sis Operation for Multi Component Systems, Desalination, 7: 133-145. Taylor, Water Quality and Treatment, edit ed by Letterman, 1999, Chapter 11, McGraw Hill inc. Tepper, F. et al., 2006, Emerging Technology, Performance Testing of a New Electropositive Filter, Ultra Pure Water Periodical. Vicevic Glenn, 2003, RO Pretreatment with Immersed Hollow Fiber Ultrafiltration, Zenon Environmental, Intern ational Water Conference. Waal, M.J. van der, P.M. van der Veide n, J. Koning, C.A. Smolders and W.P.M. van Swaay, Use of Fluidized Beds as Tur bulence Promoters in Tubular Membrane, Desalination 22 (1977) 465 -483. Williams M., 2003, A Brief Review of Reverse Osmosis in Membrane Technology, Williams Engineering Services Company, Inc. Williams M., 2003, A Review of Reverse Osmosis Theory, EET Corporation and Williams Engineering Services Company, Inc. Yu Zhao, Yu, 2004, Modeling of Membrane Solute Mass Transfer in NF/RO Membrane Systems, Dissertation, University of Florida.

PAGE 138

118 APPENDICES

PAGE 139

APPENDIX A: Tables of Test Results for Solutions NaCl 119

PAGE 150

APPENDIX B: Tables of Test Results for Solutions 4MgSO 130

PAGE 161

APPENDIX C: Tables of Test Results for Solutions 2MgCl 141

PAGE 172

ABOUT THE AUTHOR Awad A. El-Shamy is currently a senior design engineer at Crane Environmental Company, Florida, USA. His du ties include working closely w ith engineering design and technical proposals for water tr eatment systems. He has exte nsive experience working in several technical positions for over 20 years, with American and Saudi companies, in the field of design, operation, and maintenance of seawater and brackis h water desalination plants. He submitted two papers at the International Desalination Association (IDA) conferences, and published a number of technica l articles in the specialized engineering magazines in the United States and Egypt. He received his B.Sc. degree in Mech anical Engineering, and M. Sc. in Operations Research from Cairo University, Egypt. His research interests are in the fi eld of pressure-driven membranes.