UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A study of heavy oil ultrafiltration using ceramic membranes Duong, Anna Manhoa 1996

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

Item Metadata

Download

Media
831-ubc_1996-0164.pdf [ 6.09MB ]
Metadata
JSON: 831-1.0058522.json
JSON-LD: 831-1.0058522-ld.json
RDF/XML (Pretty): 831-1.0058522-rdf.xml
RDF/JSON: 831-1.0058522-rdf.json
Turtle: 831-1.0058522-turtle.txt
N-Triples: 831-1.0058522-rdf-ntriples.txt
Original Record: 831-1.0058522-source.json
Full Text
831-1.0058522-fulltext.txt
Citation
831-1.0058522.ris

Full Text

A STUDY OF H E A V Y OIL ULTRAFILTRATION USING CERAMIC MEMBRANES By ANNA MANHO A DUONG BA.Sc. The University of British Columbia, 1993  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES Department of Chemical Engineering  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA March 1996 © Anna Duong, 1996  In  presenting this  degree at the  thesis in  University of  partial  fulfilment  of  of  department  this thesis for or  by  his  or  requirements  British Columbia, I agree that the  freely available for reference and study. I further copying  the  representatives.  an advanced  Library shall make it  agree that permission for extensive  scholarly purposes may be granted her  for  It  is  by the  understood  that  head of copying  my or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department The University of British Columbia Vancouver, Canada  DE-6 (2/88)  ABSTRACT  Heavy oil refining is complicated by the presence of relatively high concentrations of contaminants that include sulfur, nitrogen, heavy crude components (asphaltenes) and heavy metals such as nickel and vanadium. The heavy crude components lead to coke formation and the metals deactivate catalysts irreversibly during hydroprocessing. In the present study, the feasibility of heavy oil ultrafiltration using ceramic membranes to remove the problematic asphaltenic components has been examined.  The study was divided into two parts: an  experimental study and a modeling study of membrane fouling.  A number of experiments were conducted to examine the effect of membrane pore size, crossflow velocity and membrane regeneration on the permeate flux and asphaltene reduction. Results from the experiments showed that both the flux and the asphaltene concentration of the permeate declined with time due to membrane pore restriction. For example, with a O.lfim nominal pore size membrane operating at 110°C, 600kPa and a cross-flow velocity of 6.9m/s, a flux of 55 kg/m /day and an asphaltene reduction of 80wt% were obtained after 9 2  hours on stream. Permeate flux increased with increased cross-flow velocity and membrane pore size whereas asphaltene reduction increased with decreasing membrane pore size. A number of correlations were obtained for the viscosity, density, vanadium and nickel content versus asphaltene content in heavy oil.  Ill  A mechanistic model with onefittedparameter was developed to predict the permeate flux decline with time due to fouling of adsorbed asphaltene molecules inside the membrane pore. Good agreement was found between the measured and calculated permeate flux when the average molecular size of asphaltenes was assumed large (i.e. r,=153A), whereas at smaller asphaltene size (i.e. r,=26A) good agreement was obtained between the measured and calculated asphaltene content in the permeate.  iv  TABLE OF CONTENTS  ABSTRACT  u  TABLE OF CONTENTS  iv  LIST OF TABLES  vii  LIST OF FIGURES  viii  ACKNOWLEDGMENTS  xii  CHAPTER 1: INTRODUCTION  01  1.1 Heavy Oil Resources and Motivation of the Study  01  1.2 Heavy Oil Upgrading Processes  03  1.3 Membrane Separation Technology  05  1.3.1 Characteristics of Membrane Separation Processes  05  1.3.2 Factors Limiting the Applications of Membrane Processes  10  1.4 Heavy Oil Upgrading by Ultrafiltration  10  1.5 Research Obj ectives  12  CHAPTER 2: LITERATURE REVIEW  13  2.1 Oil Upgrading Using Membrane Separation Technology  13  2.1.1 Polymeric Membranes  13  2.1.2 Ceramic Membranes  16  2.2 Polymeric Versus Ceramic Membranes  18  2.2.1 The Advantages of Ceramic Membranes  18  2.2.2 The Disadvantages of Ceramic Membranes  19  2.3 Characteristics of Ceramic Membranes  20  2.4 Membrane Fouling Characteristics  21  2.4.1 Flux Decline Due to Gel Layer Formation on Ultrafiltration Membranes  22  2.4.2 Flux Decline Due to Pore Plugging or Pore Narrowing  27  V  2.4.3 Semi-Empirical Mathematical Models  CHAPTER 3: EXPERIMENTAL STUDIES 3.1 Heavy Oil Properties  31  33 33  3.1.1 Feed Properties  36  3.1.2 Asphaltenes  39  3.2 Experimental Set-Up and Procedure  46  3.2.1 Experimental Apparatus  48  3.2.2 A Summary of the Experimental Procedure  54  3.3 Analytical Methods  CHAPTER 4: DEVELOPMENT OF A FOULING MODEL  55  57  4.1 Model Assumptions  58  4.2 Model Development  63  CHAPTER 5: RESULTS AND DISCUSSION 5.1 Experimental Studies  72 72  5.1.1 Effect of Cross-Flow Velocity  74  5.1.2 Asphaltene Reduction with Time  75  5.1.3 Correlations of Metals and Asphaltene Concentration  77  5.1.4 Viscosity, Density and Asphaltenes Correlations  79  5.1.5 Flux Decline with Time Due to Membrane Pore Restriction  82  5.1.6 Effect of Pore Size  90  5.1.7 Effect of Membrane Regeneration  95  5.2 Modeling Studies  98  5.2.1 Initial Permeate Flux Determination  98  5.2.2 Other Parameters Used in the Pore Restriction Model  105  5.2.3 Effect of Asphaltene Size  107  5.2.4 Effect of Cross-Flow Velocity  118  5.2.5 Effect of Membrane Pore Diameter  118  vi  CHAPTER 6: CONCLUSIONS AND RECOMMENDATIONS  134  6.1 Conclusions  134  6.2 Recommendations  136  NOMENCLATURE  138  REFERENCES  140  APPENDIX A  144  APPENDIX B  148  APPENDIX C  150  APPENDIX D  158  APPENDIX E  170  vii  LIST OF TABLES Table 1.1 : Principle Characteristics of Commercialized Membrane Separation Processes Using Pressure Driving Force  07  Table 2.1 : Mathematical Models of the Fouling Process  32  Table 3.1 : Composition Ranges for the Bulk Fraction in Crude Oils  35  Table 3.2 : Comparison of Cold Lake Heavy Oil and Conventional Crude Oil Properties  37  Table 3.3 : Asphaltene Fraction and Molecular Radius in Tetrahydrofuran  43  Table 5.1 : Heavy Oil Ultrafiltration at Different Cross-Flow Velocities Using Membranes with Average Pore Diameter of 0.1 urn 73 Table 5.2 : Summary of the Repeatability of Experimental Data for Heavy Oil Ultrafiltration Using the 0.1 um Ceramic Membranes at High Cross-Flow Velocities  86  Table 5.3 : Effect of Pore Size on Membrane Heavy Oil Separation  92  Table 5.4: Summary of Operating Conditions Used in Heavy Oil Ultrafiltration with New and Regenerated Membranes of 0.1 um Pore Size  96  Table 5.5 : A Comparison of Experimental and Predicted Viscosity of Cold Lake Heavy Oil  103  Table 5.6 : A Comparison of Experimental and Predicted Density of Cold Lake Heavy Oil  103  Table 5.7 : A Comparison of the Calculated and the Experimental Initial Permeate Flux  104  Table 5.8: Summary of the Minimum SSQ Obtained in Determining the Best Fitted K* for the 0.1 urn Pore Diameter Membrane at Different Asphaltene Radii  115  Table 5.9: Summary of the Minimum SSQ Obtained in Determining the Best Fitted K* for the 0.05pm Pore Diameter Membrane at Different Asphaltene Radii  120  Table 5.10: Summary of the Minimum SSQ Obtained in Determining the Best Fitted K* for the 0.02p.m Pore Diameter Membrane at Different Asphaltene Radii  126  VII1  LIST OF FIGURES  Figure 1.1 : A Simplified Schematic Diagram of Cross-Flow Membrane Filtration System 09 Figure 3.1 : Classification of Crude Oils by Density -Gravity  34  Figure 3.2 : Simulated Distillation of Cold Lake Heavy Oil  38  Figure 3.3 : Hypothetical Structures for Asphaltenes From (A) Venezuelan Crude Oil; (B) Californian Crude Oil  40  Figure 3.4 : Representation of an Asphaltene ClusterfromX-Ray Analysis  41  Figure 3.5 : GPC Elution Curve of Asphaltene Fraction in Tetrahydrofuran at 25°C  43  Figure 3.6(a) : A Flow Diagram of the Experimental Set-Up  47  Figure 3.6(b) : Batch Ultrafiltration with Partial Recycle of Retentate  48  Figure 3.7 : Membrane Housing Unit  49  Figure 4.1 : A Pore Restriction Schematic Diagram  62  Figure 4.2 : The Schematic Flowsheet of the Pore Restriction Model  71  Figure 5.1 : Asphaltene Reduction Versus Time Using 0.1 urn Pore Diameter Membranes at Different Cross-Flow Velocities  76  Figure 5.2 : Nickel and Vanadium Versus Asphaltene Content in Heavy Oil  78  Figure 5.3(a): Viscosity Measured at 40°C vs. Asphaltene Content in Heavy Oil  80  Figure 5.3(b): Viscosity Measured at 110°C vs. Asphaltene Content in Heavy Oil  80  Figure 5.4 : Density Measured at 15.6°C vs. Asphaltene Content in Heavy Oil  81  Figure 5.5 : Average Permeate Flux vs. Time Using the 0.1 urn Membranes Operated at Different Cross-Flow Velocities Figure 5.6 : Average Permeate Flux vs. Time Operated at Uc=6.9m/s Using the 0. lum Membranes  83  Figure 5.7 : Average Permeate Flux vs. Time Operated at Uc=8.8m/s Using the 0. lum Membranes  87 88  ix  Figure 5.8 : Asphaltene Reduction vs. Time Using the 0.1 pm Membranes at High Cross-Flow Velocities  89  Figure 5.9 : Permeate Flux vs. Time Using Different Membrane Pore Sizes at Uc=6.9m/s 93 Figure 5.10 : Asphaltene Reduction vs. Time Using Different Membrane Pore Sizes  94  Figure 5.11(a) : Comparing Average Permeate Flux vs. Time of New and Regenerated 0.1 pm Membrane  97  Figure 5.11(b): Comparing Asphaltene Content vs. Time of New and Regenerated 0.1pm Membrane  99  Figure 5.12 : Experimental and Theoretical Membrane Resistances of Membralox® During Filtration of Water  100  Figure 5.13(a): Experimental and Calculated Flux vs. Time Using the 0.1pm Membrane at Uc=6.9m/s, r.=0.0026um. K*=1.75xl0" m/s  108  u  Figure 5.13(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.1 um Membrane at Uc=6.9m/s, r =0.0026um. K*=l .75xl0" m/s 108 u  8  Figure 5.14(a): Experimental and Calculated Flux vs. Time Using the 0.1pm Membrane at Uc=6.9m/s, r =0.0032um. K*=1.75xl0" m/s n  8  109  Figure 5.14(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the O.lumMembrane at Uc=6.9m/s, r=0.0032um. K*=1.75xl0" m/s 109 u  s  Figure 5.15(a): Experimental and Calculated Flux vs. Time Using the 0.1 urn Membrane at Uc=6.9m/s, r=0.0047um. K*=1.75xl0- m/s n  s  110  Figure 5.15(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.lpm Membrane at Uc=6.9m/s, r„=0.0047um. K*=1.75xl0" m/s 110 u  Figure 5.16(a): Experimental and Calculated Flux vs. Time Using the 0.1 um Membrane atUc=6.9m/s, r =0.0079um. K*=1.5xl0" m/s n  8  111  Figure 5.16(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0. lpm Membrane at Uc=6.9m/s, r =0.0079um. K*=1.5xl0" m/s 111 n  8  Figure 5.17(a): Experimental and Calculated Flux vs. Time Using the 0.1 urn Membrane atUc=6.9m/s, r =0.0153pm. K*=1.5xl0""m/s s  112  Figure 5.17(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the O.lum Membrane at Uc=6.9m/s, r,=0.0153um. K*=1.5xl0' m/s 112 u  Figure 5.18(a): Effect of r, on Calculated Permeate Flux vs. Time Using the 0.lum Membrane at Uc=6.9m/s. K*=1.75xl0' m/s  117  Figure 5.18(b): Effect of r on Calculated Permeate Asphaltene Content Flux vs. Time Using the O.lum Membrane at Uc=6.9m/s. K*=1.75xl0' m/s  117  11  8  u  Figure 5.19(a): Experimental and Calculated Flux vs. Time Using the 0.05um Membrane at Uc=6.9m/s, r=0.0026um. K*=1.75xl0" m/s 121 n  s  Figure 5.19(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.05um Membrane at Uc=6.9m/s, r =0.0026um. K*=1.75xl0' m/s 121 n  8  Figure 5.20(a): Experimental and Calculated Flux vs. Time Using the 0.05um Membrane at Uc=6.9m/s, r=0.0032um. K*=1.75xl0" m/s 122 n  s  Figure 5.20(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.05um Membrane at Uc=6.9m/s, r„=0.0032um. K*=1.75xl0" m/s 122 n  Figure 5.21(a): Experimental and Calculated Flux vs. Time Using the 0.05um Membrane at Uc=6.9m/s, r =0.0047um. K*=1.5xl0" m/s 123 n  9  Figure 5.21(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.05um Membrane at Uc=6.9m/s, r„=0.0047um. K*=l.5xl0" m/s 123 n  Figure 5.22(a): Experimental and Calculated Flux vs. Time Using the 0.05um Membrane at Uc=6.9m/s, r=0.0079um. K*=1.75xl0" m/s 124 n  s  Figure 5.22(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.05um Membrane at Uc=6.9m/s, r,=0.0079um. K*=l .75xl0" m/s 124 u  Figure 5.23(a): Experimental and Calculated Flux vs. Time Using the 0.05um Membrane atUc=6.9m/s, r=0.0153um. K*=1.75xl0" m/s 125 n  s  Figure 5.23(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.05um Membrane at Uc=6.9m/s, r =0.0153um. K*=1.75xl0" m/s 125 u  8  Figure 5.24(a): Experimental and Calculated Flux vs. Time Using the 0.02um Membrane atUc=6.9m/s, r,=0.0026um. K*=1.0xl0" m/s 127 n  Figure 5.24(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.02um Membrane at Uc=6.9m/s, r=0.0026um. K*=l.0xl0' m/s 127 n  s  xi  Figure 5.25(a): Experimental and Calculated Flux vs. Time Using the 0.02pm Membrane at Uc=6.9m/s, r,=0.0032um. K*=1.0xl0" m/s 128 u  Figure 5.25(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.02pm Membrane at Uc=6.9m/s, r =0.0032pm. K*=l .0xl0" m/s 128 u  8  Figure 5.26(a): Experimental and Calculated Flux vs. Time Using the 0.02pm Membrane at Uc=6.9m/s, r =0.0047pm. K*=1.25xl0- m/s 129 n  8  Figure 5.26(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.02pm Membrane at Uc=6.9m/s, r,=0.0047pm. K*=1.25xl0" m/s 129 u  Figure 5.27(a): Experimental and Calculated Flux vs. Time Using the 0.02pm Membrane at Uc=6.9m/s, r =0.0079um. K*=1.25xl0" m/s 130 u  8  Figure 5.27(b): Experimental and Calculated Asphaltene Concentration vs. Time Using the 0.02pm Membrane at Uc=6.9m/s, r =0.0079pm. K*=1.25xl0' m/s 130 n  8  Figure 5.28(a): Effect of K* on Calculated Permeate Flux vs. Time Using the 0.05pm Membrane at Uc=6.9m/s, r =0.0153pm.  133  Figure 5.28(b): Effect of K* on Calculated Permeate Asphaltene Content Flux vs. Time Using the 0.05pm Membrane at Uc=6.9m/s, r =0.0153um.  133  8  8  Xll  ACKNOWLEDGMENTS  I would like to thank Dr. Smith for his continual guidance, support and encouragement throughout this project. His valuable advices, ideas and suggestions are greatly appreciated. I would also like to thank Dr. W. Kwok, Dr. Chattopadhyaya, the staff at the Chemical Engineering workshop and stores for their assistance. Thefinancialsupport of Discovery West Corporation is also gratefully acknowledged.  Finally, I would like to thank the  professors of the Chemical Engineering Department, who made graduate study so useful and memorable.  1  CHAPTER I: INTRODUCTION  1.1 HEAVY OIL RESOURCES AND MOTIVATION FOR THE STUDY  Today, supplies of light oil appear to be stable, abundant and currently satisfy more than 90 percent of the world demand for petroleum. However, they account for less than 25 percent of the remaining world petroleum reserves (AOSTRA, 1990). Reinsch and Considine (1991) reported that the industrialized countries including Canada, the United States, Western Europe, and Australasia (Japan, Australia and New Zealand) consumed around 57 percent of the world oil production but produced only 24 percent of the world oil supply over the four year period from 1987-1990.  With the continued increase in fuel consumption by  industrialized nations and decline in known crude oil reserves, the potential for a severe and costly imbalance between petroleum supply and demand exists.  Past experiences revealed that oil alone could cause economic shock waves across the world. The energy crisis in the 1970s and 1980s proved extraordinarily costly to the world economy. The most recent oil shock occurred in 1990-1991 during the Iraq-Kuwait War. The sudden interruption of exportsfromthese major oil suppliers caused the oil price to increase from $16 a barrel in July to about $34 in September 1990 (Fried and Trezise, 1993). Average oil prices between September and December 1990 were about $10 per barrel higher than had been expected before the outbreak of the war. The Organization of Petroleum Exporting Countries  2 (OPEC) also forecasted oil prices to rise from $18.40 in 1991 to $33.40 in 2010 or an average of 3.2% a year based on 1990 dollars (Fried and Trezise, 1993).  Many alternative energy sources are being investigated in order to replace petroleum products as well as to reduce pollution and conserve energy. However, until these technologies are commercialized, petroleum based products will remain the major energy source of industrialized nations.  Due to the high demand for oil products and rapid decline of  conventional crude oil reserves, the processing of heavier crude oil (heavy oil) has become 1  increasingly important since the 'energy crisis' of the 1970s.  There are over three trillion barrels of heavy oil around the world. In Canada, most of the heavy oil deposits are located in the province of Alberta which has approximately 8 billion barrels of heavy oil in-place (AOSTRA, 1990). The four major deposits are Athabasca, Wabasca, Peace River, and Cold Lake. Many smaller deposits (<10,000 barrels/day) are located across Canada. The marketability of heavy oil is limited by the presence of a high proportion of contaminants such as sulfur, nitrogen, nickel, vanadium and heavy crude oil components known as asphaltenes . The sulfur concentration in heavy crude oil can be four 2  times that in conventional crude. Besides the environmental problems associated with sulfur and N O emissionsfromfuels' combustion, sulfur and nitrogen compounds from heavy oils x  poison reforming catalysts.  In addition, asphaltenes can lead to coke formation during  'Heavy Crude oil or heavy oil has higher viscosity and contains more high molar mass components than conventional crude oil (see Chapter 3). Asphaltenes are classified as the fraction of heavy oil that is insoluble in n-pentane but soluble in an aromatic solvent such as toluene (see Chapter 3).  2  3  hydroprocessing while the heavy metals such as nickel and vanadium tend to deactivate hydroprocessing catalysts rapidly and irreversibly. A detailed discussion of heavy oil and asphaltenes characteristics can be found in Chapter 3 of this thesis.  Due to its distinguishing characteristics, refining heavy oil has proven technically difficult and financially costly. In fact, most conventional refineries can only process limited quantities of heavy oil. Thus upgrading of heavy oil to yield products that could be transported by pipeline and processed as conventional crude oil would have a significant impact on the economic viability of such heavy oil deposits.  1.2 H E A V Y OIL UPGRADING PROCESSES  Many conventional heavy oil upgrading processes have been developed throughout the years. Catalytic upgrading processes have typically been the most effective means of converting heavier oils to lighter and more valuable products (Booz, 1980). However, because of the presence of high levels of asphaltenes and contaminants in heavy oil, conventional heavy oil upgrading processes are noncatalytic. They include solvent extraction,fluidcoking and partial oxidation processes.  Solvent extraction, also known as solvent deasphalting, was originally developed to extract heavy lubricating oilsfromresidual oil. The solvent deasphalting process is basically a liquidliquid extraction process in which addition of hydrocarbon solvent such as propane, butane or  4  n-pentane is used to extract the heavier componentsfromheavy oil. The main disadvantages with the process are the large remaining volume of very low quality residue (524+ °C) and also the large amount of steam required to extract the solvent (Booz, 1980).  Fluid coking is a noncatalytic thermal cracking process. As with any other thermal cracking process, this process requires intensive coke-handling facilities to remove the coke from the reactor. Although it is a reliable process, it produces relatively more light ends and gaseous fuels, and less of the valuable liquid products. In addition, it requires a gas desulfurization unit to remove the sulfurfromthe products.  Partial oxidation is a continuous non-catalytic process in which oxygen is added with the heavy oil to produce a synthesis gas primarily of hydrogen and carbon monoxide.  This  process yields no liquid hydrocarbon products. Thus the partial oxidation process would only be attractive if a large market for synthesis gas existed. This process requires an oxygen plant as well as an H S and C 0 removal system. 2  2  Flowsheets and summary tables of the advantages and disadvantages of the above processes can be found in Appendix A. There are many factors that affect the selection of an upgrading process including the product site, product market, process costs and construction lead time. The conventional heavy oil upgrading processes often need a large plant and many processing steps that imply a very high capital cost. Continuing effort has been devoted to find "a more effective,  efficient, and environmentally acceptable bitumen and heavy oil upgrading  5 technology" (AOSTRA, 1979). Due to the presence of many small heavy oil deposits, the primary objective of the upgrading is to produce a transportable crude that can be upgraded further in existing or modified refineries (Booz, 1980). Alternative upgrading methods are sought to find uses and add value to the available heavy oil resources. To minimize the loss of valuable hydrocarbon, a highly selective process would be required where only the heaviest components of the heavy oil are removed.  1.3 M E M B R A N E SEPARATION TECHNOLOGY  Membrane technology has been widely developed and applied in gas, aqueous and protein separation processes. With the development of new membrane materials and commercial units, membrane technology has recently been applied to the separation of organic liquids. Examples of this concept include deasphalting a hydrocarbon oil (Kutowy et al., 1985; Kutowy et al., 1989) and removal of metals including nickel and vanadium from a hydrocarbon feed stock (Kulkarni et al, 1988; Osterhuber and Phillipeburg, 1989; Sparks et al, 1990; Bitter etal, 1992).  1.3.1 Characteristics Of Membrane Separation Processes  In all membrane processes, the feed is separated into a stream that passes through the membrane, called the permeate, and afractionof feed that does not go through the membrane, called the retentate.  Table 2.1 shows the principle characteristics of commercialized  6  membrane separation processes for liquids using pressure difference as a driving force (Ho and Sirkar, 1992). As seenfromTable 2.1, both ultrafiltration (UF) and microfiltration (MF) membranes separate solute molecules by a molecular sieving mechanism, while reverse osmosis (RO) membranes separate solute molecules by molecular interactions between solutes, solvents and membrane materials.  The difference between UF and MF membranes is that UF membranes separate macrosolutes (high molecular weight solutes and colloidal dispersed substances such as clays, pigments and latex particles) from microsolutes or their solvents (usually water or lower molecular weight media) whereas MF membranes separate particles or microsolutes (approximate molecular diameter of 0.02-10um). Depending on the nature of the species transported through and retained by the membrane and the mechanism for transport, a preliminary selection of the membrane process relevant to the separation goal can be identified.  In most membrane separation processes, a sufficient pressure driving force is required to establish a transmembrane pressure differential, P , for permeate to flow through the T  membrane (Cheryan, 1986). In the case of a highly viscous feed material such as heavy oil, heating is necessary to reduce the feed viscosity to make it aflowablefluid.Theoretically, the transmembrane pressure, PT, can be expressed as follows: P = AP - An  (l.l.a)  AP = P - P  (l.l.b)  T  with,  T  T  F  P  w fc. 0  60  e  1 Q  C/3  ON  4) la  9 1  s cn 09 (U ha  PH OX)  e  cn  . o B « _  60  2  cn cn <U O  CO 4>  O  u  I  co  "8 a.  PM  e  e s.s  5  03  <u <u  cn  o cn  1  CO  V3  e «  5  XI  >, o  U-, H  E  g °  CU  g  +3  •o v .a  3  C/D  1  "« o u  s H o  U  C  <U  <-i  on  1 O  co  2  1^  o •a  c o  O  o -a  >  o  CU  o  «  CN  3  •2  co o co 2 .2  2 8  CO  ^  o  CN  'p  °. a o  cu o «  •1-1  08  x: U "3 c  s a, u  C/3  8  I  g  O  3  S  S  co  g  1 8 1 8 o o o o '3  o  I  C/5  co  co  O  en  C  C/3  Pi  0  1  <N  >  oi  .3 .3 <fa B .3 '2 CO 5 CO o s Q o |  p b- .2 o  to,  0  o  O  o  CO  I  •2, 2 S g °  ( <  O.  0  p .2  leg  1  C3 co O «  ••§ .2  o  1 o  g  and,  An = 7CF - n  (1.1.c)  P  where, PF  = the applied pressure (Pa).  P  = the back pressure on the permeate side (Pa).  p  7T.F  = the osmotic pressure of the feed solution (Pa)  7t  = the osmotic pressure of the permeate solution (Pa)  P  In most ultrafiltration applications, due to the high molecular weights of the retained solutes (macromolecules and colloidal particles), the osmotic pressure of the solutes is small compared to the applied pressure. Thus the An term is neglected and the transmembrane pressure is equal to AP . In a cross-flowfiltrationsystem (see Figure 2.1), the applied X  pressure ( P F ) is the average of the membrane inlet and outlet pressures. Therefore, the net driving force for permeateflowin a cross-flowfiltrationsystem can be written as:  P  T  = AP  T  = (Pi:+ Po)/2 - P  p  where, Pi  = the membrane inlet pressure (Pa)  P  = the membrane outlet pressure (Pa)  0  (1.2)  9  Permeate  Cross-flow Membrane Module  Feed  Retentate  Recycle  Figure 1.1. A Simplified schematic Diagram of A Cross-Flow Membrane Filtration System  10 1.3.2 Factors Limiting the Applications of Membrane Processes  There are many barriers to the development of commercial membrane separation processes. These include concerns regarding membrane process reliability, membrane life and membrane cost as well as the limited choice of membrane materials and configurations. In addition, the complexity of studying membrane performance characteristics is also a factor corifining the application of membranefiltrationin many separation processes.  Flux loss due to fouling is a major drawback in many membrane separation processes and this is expected to be particularly severe in the ultrafiltration of heavy oil. Studies have been carried out to understand and control membrane fouling. Due to the complexity of the fouling process, it is difficult to develop general rules or theories to characterize the membrane fouling mechanism. Many studies aimed at developing new membrane materials, determining optimal operating conditions and techniques to reduce fouling and increase flux have been reported in the literature.  1.4 H E A V Y OIL UPGRADING BY ULTRAFILTRATION  Recent developments in membrane materials and commercial units have encouraged many studies in the membrane separation domain.  Due to the macro-molecular structure of  asphaltene molecules in heavy oil, the potential to apply membrane ultrafiltration technology to heavy oil upgrading exists.  In addition, membrane processing only requires a sufficient  11 energy for pumping and heating in the case of a highly viscous feed material. Membrane separation units are compact and can be built on a small site. For small heavy oil deposits, there may be some advantage in building a small, low capital cost upgrading process on site, that could remove some of the contaminants and reduce the density and viscosity of heavy oil to make it transportable by pipeline to existing refineries.  Ideally in membrane separation processes, large molecules are retained by the membrane while small molecules permeate through the membrane pores more or less freely.  In heavy oil  ultrafiltration, the heavy oil is upgraded by removal of the most problematic components, asphaltenes. Asphaltenes have a macromolecular size range from 10 to 1000A (Baltus and Anderson, 1983). They consist of complex hydrocarbon rings to which sulfur, nitrogen, oxygen and metals are bound. Selective removal of asphaltenes will reduce the viscosity, density and the metal content of the heavy crude oil. The properties of asphaltenes will be discussed further in Chapter 4.  Ceramic membranes are presently available as single tubes and in commercial-size monolith units, and have pore sizes as low as 50A suitable for ultrafiltration. In addition, the ceramic membranes are chemically and thermally stable. Hence, ultrafiltration of heavy oil at elevated temperature using these membranes seems feasible.  12 1.5 RESEARCH OBJECTIVES  Many studies have been reported on the ultrafiltration of diluted waste oils, residua and bitumens (heavy, viscous, black crude oils) at low temperature using polymeric membranes (see Chapter 2). Limited information on heavy oil ultrafiltration using ceramic membranes is available. The present study has been concerned with ultrafiltration of undiluted heavy oil at elevated temperature.  The goal of the ultrafiltration process is to reduce the density and  viscosity of the heavy oil, and to remove at least some of the most problematic components, like asphaltenes, so that transportation and processing are less difficult. The primary objective of this study was to experimentally investigate heavy oil ultrafiltration using ceramic membranes. The effect of cross-flow velocity and membrane pore size on permeate flux and separation were examined.  Membrane fouling is a major draw-back in ultrafiltration and this is expected to be particularly severe with heavy oil feed. No studies on membrane fouling during heavy oil ultrafiltration have been reported in the literature. The second objective of the present work was to identify the fouling characteristics of the ceramic membrane during heavy oil ultrafiltration and to develop a mechanistic model of the fouling process.  13  CHAPTER 2; LITERATURE REVIEW  2.1 OIL UPGRADING USING MEMBRANE SEPARATION TECHNOLOGY  2.1.1 Polymeric Membranes  Upgrading heavy oil using polymeric membranes has recently been reported by several investigators.  Kutowy and colleagues (1985) studied the performance of polysulfone  membranes for the separation of a variety of feedstocks including spent diesel and diluted light and heavy oil. They reported a significant reduction in viscosity and metals content following ultrafiltration. For example, a Cold Lake heavy oil diluted with 34% naphtha was separated using polysulfone membranes operated at 45°C and a transmembrane pressure of 2.5MPa with a feed flowrate of 0.085 kg/s. The results obtained showed a significant reduction in viscosity (from 89.6 to 1.72 mPa.s) while the vanadium and nickel were reduced from 124.5 to 2.8 ppm and 41 to <0.6 ppm, respectively. sulphur and nitrogen contents.  Similarly, 59% and 86% reductions were obtained for  Corresponding to the reductions in the contaminants, the  average molecular weight found in the permeate was 412 g/mol compared to the feed of 879.5 g/mol. In addition, they obtained an average flux of 27.2 kg/m /day after several hours of 2  operation. The exact operating time was not reported. The membrane used in this particular case was Udel polysulfone membrane, GR81, which had an 88% separation for PEG 6,000 (polyethylene glycol, molecular weight 6,000) in water.  14  More recent work by Sparks et al. (1990) using the same polymeric membranes to upgrade Athabasca bitumen with 37% naphtha dilution showed very similar results. They claimed that the permeate contained lighter fractions than that of the feed. In addition, they also found that the product molecular weight distribution shifted slightly toward higher values as the pore size of the membrane increased. However, removal of contaminants such as the sulfur, nitrogen and heavy metals content of the products was found to be very similar for all of the membranes used. Typically, the sulfur, nitrogen and heavy metals were reduced by 50, 70 and 95%, respectively.  On the other hand, Sparks et al. (1990) reported that membranes with a more open structure had higher fluxes but still similar values for contaminant reduction. For example, the Udel polysulfone membrane, GR81, displayed a lower flux of 63 kg/m /day compared to a 2  membrane with a more open structure (such as Victrex membrane, 20VP) which had a flux of 113 kg/m /day when operated at 35°C, a transmembrane pressure of 2.25MPa and a flowrate 2  of 0.108 kg/s.  Both membranes gave very similar separation results. The sulfur, nitrogen,  nickel and vanadium reductions using GR81 were 61, 86, 97 and 98%, respectively while using the 20VP membrane, the reductions were 55, 83, 96, and 94%, respectively.  In  addition, the viscosity and average molecular weight reductions were very similar for both membranes. Using the GR81 membrane, a 95% reduction in viscosity and a 50% reduction in the average molecular weight were obtained whereas using the 20VP membrane, 94 and 42% reductions were measured, respectively.  15  Recent patents have also described the use of membranes for the separation of heavy oil/solvent mixtures at low temperatures.  Osterhuber and Phillipeburg (1989) patented a  process wherein heavy oil residue was dissolved in an organic solvent such as toluene and the resulting mixture passed through a polymeric ultrafiltration membrane. In one example, heavy Arab vacuum residual oil was upgraded using a Nuclepore type F ultrafiltration membrane having a nominal molecular weight cutoff of 5,000. The vacuum residual oil was dissolved in toluene with a weight ratio of 1:1. The separation was done in a stirred batch ultrafiltration cell operated at 1 MPa and 27°C. The resulting product contained 11 wt% of Conradson Carbon Residue compared to 22 wt% in the feed.  The permeate also had much lower  vanadium and nickel contents, 45 and 10 ppm, respectively compared to 180 and 43 ppm in the feed, respectively. Parallel to the above reductions, the molecular weight in the permeate was 794 compared to the feed of 1098 g/mol. The permeate flux was not reported in this patent.  In another disclosure, Kulkarni and coworkers (1988) used polymeric membranes to upgrade heavy oil which was diluted with cycloparaffinic solvents such as cyclohexane. They claimed that aromatic solvents such as xylene, benzene and toluene would deteriorate the membrane and therefore would not be preferred solvents. Preferred membranes were polysulfones with pores of about 5 to 500 A diameter. The produced permeates had lower molecular weights and metal contents than that of the feeds.  16  In one example, Kulkarni and coworkers (1988) reported that the permeate of a 28% Boscan heavy crude oil dissolved in cyclohexane operated at 25°C and 0.2MPa gauge was found to have a molecular weight of 840g/mol, nickel and vanadium contents of 14 ppm and 115 ppm respectively, compared to the feed which had a molecular weight of 2932g/mol and nickel and vanadium contents of 66 ppm and 733 ppm, respectively. The membrane pore size and the flux were not reported.  2.1.2 Ceramic Membranes  Trambouze and coworkers (1989) used porous inorganic membranes having pore size in the range of 2-15 nm to upgrade deasphalted vacuum residue. For example, a Safaniya vacuum residue was deasphalted by addition of n-pentane. The asphaltic phase was discharged. The oil-solvent phase, which contained about 23 wt% of solvent, was then separated using an aluminum oxide membrane of 4.5 nm pore radius operated at a temperature of 180°C, a transmembrane pressure of 3.6 MPa and a cross-flow velocity of 3.5m/s.  The resulting  deasphalted oil contained a lower Conradson carbon, sulphur and aliphatic C 5 and C7 hydrocarbon concentration compared to the prefiltered oil-solvent feed. They reported that C5 and C7 hydrocarbons were reducedfrom24 to 0.3 wt% and 13 to 0.05 wt% respectively. A 51% reduction in the Conradson Carbon and a 15% reduction in the sulfur content were also reported. The permeatefluxwas not reported in this patent.  17  In an earlier disclosure, Arod et a/.(1983) showed that a ceramic or metallic support membrane, coated with a layer of, for example, titanium dioxide, magnesium oxide, aluminum oxide, MgAl 04, or silica could significantly reduce the asphaltene and metals content of an 2  undiluted waste oil. The system was operated at elevated temperature (>100°C) to reduce the oil viscosity. In one example, ultrafiltration of a vacuum distillation residue of a crude oil originating from Kirkuk, Iraq, was carried out on a module comprising of three membrane tubes of 80cm in length and 1.5cm inside diameter each coated with a layer of titanium dioxide and aluminum oxide. The average membrane pore size was 100A. The experiment was operated at 330°C with a relative pressure of 0.5MPa and a cross-flow velocity of 5.6 m/s. The rate of ultrafiltration was 667 L/m /day. Asphaltenes were reducedfrom6.30 in the 2  feed to 4.14 wt% in the permeate, while vanadium was reduced from 128ppm to 90ppm, and viscosity from 0.66 kg/m s to 0.184 kg/m s.  Available data for heavy oil upgrading using membrane technology are very limited and most data provided by the patents discussed above are incomplete.  Furthermore, due to the  variation in membrane materials and the dilution of the heavy oil feedstocks, permeate flux and separation data are not available for comparison. In addition, no data are available on the decline in permeatefluxwith time for heavy oil using either polymeric or ceramic membranes.  The present thesis provides a preliminary study of the permeate flux decline with time for heavy oil ultrafiltration as well as an assessment of the potential for heavy oil upgrading using ceramic membranes.  The following section briefly describes some of the advantages and  18 disadvantages of polymeric and ceramic membranes followed by an explanation of why ceramic membranes were chosen for the present study.  2.2 POLYMERIC VERSUS CERAMIC MEMBRANES  The direct application of various polymeric membranes to diluted heavy oil upgrading has been demonstrated by many researchers.  However, operational limitations exist with these  organic membranes since they are chemically and thermally unstable.  The presence of  solvents can deform the membrane structure. Hazlett et al.(1986) investigated the effect of various solvents on the pore structure of polysulfone membrane. They reported that the presence of alcohol such as ethanol, n-butanol, n-hexanol can enlarge the pore structure whereas paraffinic hydrocarbons such as n-hexane, n-decane, n-tetradecane, n-dodecane contract the pore structure.  In addition, high operating temperature (>100°C) can also  degrade and damage the membranes (Goldsmith, 1988; Hazlett et al., 1985).  2.2.1 The Advantages of Ceramic Membranes  Porous inorganic (ceramic) membranes are thermally and chemically stable (Goldsmith, 1988). They permit long term operation at elevated temperatures.  This advantage is of special  interest in applications in which highly viscous feed materials are used. Operating at high temperature can reduce the feed viscosity and thus increase membrane flux. Many ceramic  19  membranes can operate at a temperature in excess of 1000°C depending on the other components of the system.  Another advantage of ceramic membranes is their ability to withstand prolonged exposure to aqueous media with extreme pH as well as non-aqueous media including strong solvents such as toluene and benzene.  Hence, ceramic membranes are more readily cleaned than many  polymeric membranes. This advantage is of special interest in membrane regeneration. With their tubular configuration, ceramic membranes are less susceptible to membrane fouling. The tubular configuration minimizes deposits within the membrane module.  Membrane  regeneration by back flushing with a strong solvent or by heating at high temperature can be achieved without damage to the membrane.  Commercial ceramic membranes are now  available with average pore diameter as low as 40 A (Hsieh et al, 1988).  2.2.2 The Disadvantages of Ceramic Membranes  The major disadvantage of ceramic membranes is their high price per unit area relative to polymeric membranes. They are approximately ten times more expensive than polymeric membranes.  This results in high capital cost of any installation.  In addition, ceramic  membranes are quite fragile due to the brittleness associated with the ceramic support material and this is especially severe for single tubular elements and small diameter multiplepassageway monoliths. However, the problem is relatively unimportant for larger diameter multiple-passageway monolithic devices.  Finally, ceramic UF membranes have not been  20  reliably produced with retention properties comparable to those of the lowest 'molecular weight cut-off' polymeric membranes (Goldsmith, 1988) but this problem is less significant as advances in their preparation continue to be made.  Ceramic membranes are used in this study due to their many advantages over polymeric membranes as discussed above.  By using ceramic membranes, heavy oil viscosity can be  reduced by operation at high temperature rather than by the addition of solvent.  2.3 CHARACTERISTICS OF CERAMIC MEMBRANES  Ceramic membranes are asymmetric layered composites. They are usually composed of aluminum oxide. The membrane consists of three major layers. The dense layer, also called the active layer, has the smallest pores and a thickness of about 5 to 10 p.m. The active layer is usually the inner most layer of the membrane and is responsible for the separation. Typical porosity ranges from 35% to 50%. The second intermediate layer is approximately 30 to 45 pm thick with a porosity of about 40% and is used to protect and support the dense layer. Finally, the outer-most layer provides the membrane with mechanical strength and is approximately 1.5 to 2 mm thick with a porosity of 40%-45%.  Appendix B shows a SEM (Scanning-Electron Microscope) micrograph of the cross-section of a commercial Membralox® ceramic membrane.  In the present study, single-pass,  cylindrical Membralox® ceramic membranes with 0.02, 0.05 and 0.1 p.m norninal pore  21  diameters were used.  They had a height of 0.2m and an inside and outside cross-sectional  diameter of 0.008m and 0.01m, respectively. Due to their small pore sizes, the dense layer of these membranes were made of zirconium oxide, Zr0 . All of their support layers were made 2  of aluminum oxide,  AI2O3.  2.4 M E M B R A N E FOULING CHARACTERISTICS  The main industrial applications of membrane technology have been in aqueous and gas phase separation processes. processes.  Membrane technology is a recent application in non-aqueous  Thus, fouling studies related to non-aqueous ultrafiltration are limited. The  following section gives a brief review of many studies on membrane fouling characteristics of aqueous and protein ultrafiltration (UF) processes.  Generally,fluxdecline due to membrane fouling is a result of specific interactions between the membrane and solutes in the feed stream. Marshall et al. (1993) defined fouling as the 'coupling" of deposited material to the membrane through the intermediate step of concentration polarisation.  Concentration polarisation refers to the development of a  concentration gradient of the retained components (solutes) near the membrane wall that causes an increase in concentration at the membrane surface. Concentration polarisation is a function of the hydrodynamic conditions in the membrane system and is independent of the physical properties of the membrane. In general, membrane fouling can be divided into two  22  categories: fouling due to the deposition/absorption of material on the membrane surface or within the membrane pores. In both cases, the result is a change in membrane behaviour.  2.4.1 Flux Decline Due to Gel Layer Formation on Ultrafiltration Membranes  Howell and Velicangil (1980) found that the initial flux decline in UF of cheese whey or other proteins is due to local convective deposition of protein molecules close to or in the pores. This process was completed within a few seconds of the establishment of a quasi-steady-state concentration polarization layer. Then flux continued to drop due to protein adsorption of a monolayer at the membrane surface which occurred within one to ten minutes. Thefinalflux decay over a period of hours is attributed to the reversible polymerization of protein to gel.  The concentration polarization mechanism was derived from the conservation equation for solute across the boundary layer, ignoring flow and gradients in the plane of the membrane (Howell and Velicangil, 1980): 6C ct  = D £C _ dx?  USC dx.  (2.1)  with the initial and boundary conditions, at t = 0,  C=C  b  at x = 0,  C=C  b  at x = L ,  dC =UCandC = C dx  w  where, D Cb  = the diffusivity of proteins in whey (m /s) 2  = the solute concentration in the bulk solution (kg/m ) 3  23  Cw  = the solute concentration at the membrane wall (kg/m )  U  = the permeatefluxacross the membrane (m/s)  L  = the boundary layer thickness (m).  3  They estimated that the maximum time constant for the full effect of concentration polarization to occur could be around 5 seconds or even faster.  To determine the gel  concentration at the membrane surface, the steady-state velocity, Us, was evaluated from the steady-state solution of Equation 2.1 with a wall concentration (Cw) equal to the gel concentration (Cg): (2.2)  U =_D InQ, L C s  b  The permeate flux decline across the membrane due to the gel layer growth can also be expressed as: Us =  AP  (2.3)  where, u,  =fluidviscosity (kg/m.s)  AP  = the transmembrane pressure (Pa)  Rm  = the membrane resistivity (m' )  I  = the gel layer thickness (m)  P  1  = the gel layer permeability (m ) 2  g  24  The gel layer permeability, P , can be determinedfromthe Carman-Kozeny equation for flow g  through porous media, P = ?  (f 180  (2.4)  E/ (l-£g)  2  where, d  = the average pore diameter (m)  Eg  = the gel layer voidfraction(dimensionless)  By analyzing variousfluxdata, they found a second order relationship between the rate of gel layer growth (d£/dt) and the wall concentration (C): d#dt = k, Cj  (2.5)  where k, is the rate constant Combining equations 2.2 and 2.3 with equation 2.5 yields: d&dt = k C exp [2AP/k„ fi (R + #Pg)J 2  r  b  m  (2.6)  where, km  = D/ L = the mass transfer coefficient (m/s) x  Thus the gel layer thickness, I, can be calculatedfromEquation 2.6 provided the values for the physical parameters such as the diffusivity, D, the average pore diameter, d, the bulk solute concentration, Cb, the gel concentration, C , the transmembrane pressure, AP, the gel g  permeability, P , and the membrane resistivity, Rm are known. By substituting the calculated g  gel layer thickness, £, into Equation 2.3, the proteinfluxacross the UF membrane of protein could be determined.  They found good agreement between the experimental and the  25  predicted data for the ultrafiltration of 0.5% bovine albumin through an ordinary PM-10 membrane.  Aimar et al. (1988) studied the flux decline during ultrafiltration of cheese whey with Carbosep M4 inorganic membranes. They also found that flux decline occurs in three steps. In the first step, concentration polarization is established within thefirstminute of operation. This reduces the applied pressure driving force by an equivalent osmotic pressure down to an effective pressure which then remains constant until the end of the run. In the second step, a sharp decrease in flux was observed during thefirsthour of operation. Membrane fouling in this stage is due to either protein adsorption or particle deposition. Finally, a slow decrease of flux occurs until the end of run (over 3 hours) due to convective deposition of particles.  In addition, Aimar et al. (1988) also found that the rejection of proteins is a function of time and operating conditions. The observed rejection coefficient was calculated as follows, Rob, = l-  (2.7)  C/C  b  where, Robs  = the observed rejection coefficient (dimensionless)  C  = the permeate concentration (kg/m ) 3  p  Cb  = the bulk concentration (kg/m ) 3  They found that a-lactalbumin proteins were completely rejected by the membrane at the beginning of the experiment. This was due to the adsorption of proteins onto the membrane or into the pores and hence no proteins were found in the permeate. After a progressive saturation of proteins on the membrane material, the proteins were less and less retained by  26 adsorption and hence were found in increasing amounts in the permeate until the cake layer of proteins built-up on the membrane. This cake layer acted as a retentive membrane which retained the proteins progressively with time. Aimar et al. (1988) reported that the rejection coefficient of a-lactalbumin generally decreased to a minimum during thefirst90 minutes of the operating period and then increased. The level of this minimum as well as the final retention was dependent upon the operating conditions. Increasing the applied pressure and decreasing the cross-flow velocity raised the level of the minimum and increased the final value of the rejection coefficient. Furthermore, they found that the rejection of larger protein molecules (i.e. P-lactoglobulin, M.W.=36,000 daltons) was always greater than that of smaller molecules (i.e. a-lactalbumin, M.W.=14,000 daltons).  In another article, HallstrOm et a/.(1989) described fouling by proteins as a three-step process. Initial flux decline is caused by a rapid increase in the membrane resistance due to rapid deposition of proteins on the membrane surface and at the entrances to the pores. Further deposition then occurs on top of the first deposited layer, causing a slower rate of increase in the membrane resistance than during the initial deposition. Then, eventual bridging of the pore entrances occurs and a continuous surface layer results. At this stage the properties of the fouling layer dominate the system behavior. These three stages correspond to a transfer from membrane-controlled separation to fouling-layer-controlled separation.  Marshall et al. (1993) summarized the permeate flux decline due to protein fouling into three separate phases. Thefirstphase occurred in thefirstminute and is due primarily to  27 concentration polarization which causes an initial rapid drop in flux. In the second phase, the flux continued to decline, but not as rapidly, for one hour due to protein deposition. This deposition is initially a monolayer adsorption and then a complete surface layer builds up. The third phase is a quasi-steady-state period where the flux declines slowly due to further deposition of particles or to consolidation of the fouling layer.  In addition to the decline in  flux, in case of UF, it was found that the retention of protein generally increased with time.  2.4.2 Flux Decline Due to Pore Plugging or Pore Narrowing  Hanemaaijer et al. (1989) studied the effect of fouling during UF of whey using polysulfone and regenerated cellulose membranes.  They claimed that the fouling occurred due to the  interaction of solutes and membrane pores.  The primary cause of this effect was the  narrowing and blocking of pores due to adsorption of proteins and possibly precipitation of poorly soluble salts taking place mainly inside the membrane pores.  Evidence for location of  precipitation and adsorption inside the membrane was obtained by measuring the amounts of salts precipitated and protein adsorbed, and by comparing the protein adsorption with the amounts on model surfaces having membrane-like characteristics. They also showed that membrane structure and surface properties appear to have a considerable influence on the amount and type of protein adsorbed, suggesting a potential for controlling membrane fouling by modification of membrane properties.  28 Zeman (1983) studied the effect of adsorption of polymers and bovine serum albumin on a flat -sheet UF membrane which had a average MWCO of 10,000.  He proposed that the flux  decline during ultrafiltration of polyethylene oxides and dextrans was due to a pore restriction. The reduction in the pore size was a result of adsorption of solutes on pore walls. Zeman modified the Ferry equation to reflect the changes in membrane pore dimension due to solute adsorption. The original Ferry equation (Ferry, 1936) states, R = 1-C/C =[X(X-2)f  (2.8.a)  X = r,/r  (2.8.b)  f  and,  p  where, R  = the rejection coefficient (dimensionless)  r,  = the radius of the solutes (m)  r  = the radius of the membrane pore (m)  p  C  = the solute concentration of the permeate (kg/m ) 3  p  Q  = the solute concentration of thefiltrate(kg/m ) 3  In the case of adsorbing solutes, Equation 2.8.a became, (2.9.a)  R = [X'(X -2)f (  and, where, Ar  (2.9.b)  X' = r,/ (r - Ar ) p  p  p  = the pore adsorbed fouling layer thickness (m) = 2r, (for the case of adsorbing rigid sphere)  On the other hand, the pore adsorbed layer thickness, Ar , can be calculated using the p  experimental measurements.  By substituting the clean and the fouled water flux  29  measurements into the Hagen-Poiseuille equation for laminar liquid flow in cylindrical pores, Ar can be obtainedfromthe following equation: p  Ar r  = / - /' J °"  (2.10)  0 2 5  P  where, J'  = the waterfluxafter solute adsorption (m /m /s)  J  = the waterfluxprior to solute adsorption (m /m /s)  3  2  3  0  2  They found good agreement between the experimental and the predicted (from Equation 2.10) data for polyethylene oxides (Carbowax) using an Abcor HFM-100 membrane.  Permeatefluxdecline in ultrafiltration membranes can be characterized by a number of fouling phenomena. Due to the complex fouling mechanism, most of thefluxmodels have to include several empirical parameters to replace the unknown properties such as membrane structure, fouling layer thickness, and gel or wall concentration, to name a few.  In general, two  principal model types are used to describe the flux loss, the 'serial resistance" model and "interacting" models (Dejmek and Nilsson, 1989).  In the 'serial resistance" model, an additional resistance much larger than that of the clean membrane is formed due to fouling. This dynamic membrane resistance controls membrane behaviour while the intrinsic properties of the membrane are not changed (Marshall et al, 1993). Thus, the total membrane resistance can be expressed as the sum of all the resistances, R otal = Rm t  +R  (2.11)  a  where, Rm  = the clean membrane resistance (m") 1  30  Ra  = the additional resistance due to fouling (m' ) 1  The model can be visualized as consisting of a layer of fouling material on top of the membrane, as in cake filtration (Dejmek and Nilsson, 1989). By assuming that the dynamic membrane behaves like afiltercake, the Carmen-Kozeny equation can be used to calculate its resistance,  R= a  K (l-«fS 'JL r  (2-12)  r  where, Kozeny's constant (a function of porosity and particle shape)  e  c  S 6  the void fraction of a randomly packed cake (dimensionless) the surface area per unit volume of solids in the cake (m /m ) 2  c  C  3  the cake thickness (m)  In this model, the membrane and the fouling layer are treated as independent of each other (Dejmek and Nilsson, 1989).  Fouling within the membrane structure is usually modeled by the 'interacting" model where the change of permeability depends on clean membrane structure (Dejmek and Nilsson, 1989). Hence, the total membrane resistance is a function of the change in the apparent pore size, pore size distribution and the pore structure. It can be expressed as follows, Rtotai = function (membrane structure, fouling)  (213)  The 'interacting" model can be described by the narrowing of the equivalent pore structure. This internal membrane fouling can be modeled using the Hagen-Poiseuille equation to calculate an apparent reduction in pore size,  31  // = jcAP(r -Ar ) /8fjL  (2.14)  4  p  p  where, Ji Ar  = the fouled membraneflux(m /m /day) 3  p  2  = the equivalent thickness of internal fouling layer (m)  For this model, it can be easily shown that the membrane resistance due to the internal membrane fouling can be expressed as, R. = R {[r /(r -Ar )] -l}  (2.15.a)  R  (2.15.b)  4  m  and,  total  p  = R[ m  p  p  r /(r -Ar )]  4  p  p  p  Hence, for a uniform, cylindrical pore structure, the total membrane resistance,  Rtoui,  is simply  a function of the change in membrane pore size to the fourth power.  2.4.3 Semi-Empirical Mathematical Models  There are many simple, semi-empirical mathematical models developed to relate the flux to the time and/or volume permeated.  Table 2.1 shows some of these models which will  adequately fit almost any fouling data (flux vs. time) reasonably well (Cheryan, 1986; Fane and Fell, 1987). However, some of the models in Table 2.1, such as Equations 2.16 and 2.17, predict that the flux will be zero as time approaches infinity. This may not actually occur in practice. In addition, since these models are semi-empirical in nature, they do not help to understand the fouling mechanism.  32  Table 2.1. Mathematical Models of the Fouling Process (Cheryan, 1986; Fane et al, 1987)  2.16. J, = J f  (n<0)  0  2.17. J =  Je  2.18. J, =  JV  nt  t  0  0  2.19. J, = A + Be"'  2.20. 7 = e T 2.21. J = AP/[p(R + R + R )J m  d  bl  where, Rd = ajMj dM / dt = r d  or  dM /dt =  i.e.  M  or  -r  d  K [M *-Md]  d  d  e  d  d  =M *[l-exp(-K t)] d  = K, J  d  0  c  b  -K  2  where J is initialfluxat zero time or at 1 minute; J isfluxat any time t; V is volume permeated; R„, is clean 0  t  membrane resistance; R<| is deposit resistance; Rti is boundary layer resistance; Md is mass per unit area of deposit; Mj* is the maximum amount of mass deposited per unit area for a given time interval; aj is specific resistance of deposit; r is rate of deposition; r is rate of removal; IQ, Ki, or K is deposition rate constant; Cb d  is bulk concentration.  e  2  33  CHAPTER 3: EXPERIMENTAL STUDIES  3.1 H E A V Y OIL PROPERTIES  The generic term 'heavy oil' is often applied to a petroleum that has an API gravity of less 1  than 20° and usually, but not always, a sulfur content higher than 2% by weight (Speight, 1991). Heavy oil has a much higher viscosity and density or lower API gravity than conventional crude oil or petroleum. Figure 3.1 classifies the different crude oils based on the API gravity and viscosity.  Heavy oil is a complex mixture of high molecular weight hydrocarbons and contains relatively high concentrations of contaminants such as sulfur, nitrogen, oxygen, nickel, vanadium and asphaltenes.  Heavy oil contains a higher proportion of high boiling components such as  asphaltic components and residuum than conventional crude oil.  Table 3.1 tabulates the  fraction of the asphaltenes, resins and oils found in different crude oils. The non-asphaltic 2  fraction, which comprises oils and resins are categorized as maltenes.  The asphaltenes  fraction is also known as the solid precipitate fraction (Reerink, 1973).  gravity =  141.5  -131.5 ; SG is fluid's specific gravity.  (SG)i5. °C 6  Resins is an extremely viscous oil. It can be defined as the difference in the amount of precipitate obtained by adding propane in the crude oil and n-pentane (Mehrotra et al, 1994). 2  34  Characteristics  Type of crude 1. Conventional or "light" crude oil  Density-gravity range less than 934 kg/gm (>20° API) Density-gravity range from 1000 kg/m to more than 934 kg/m (10° API to < 20° API) Maximum viscosity of 10,000 mPa.s (cp) Density-gravity greater than 1000 kg/m (< 10° API) Maximum viscosity of 10,000 mPa.s (cp) Viscosity greater than 10,000 mPa.s (cp) Density-gravity greater than 1000 kg/m (< 10° API) 3  2. "Heavy" crude oil  3. "Extra-heavy" crude oil; may also include atmospheric residua (b.p. > 340°C; > 650°F) 4. Tar sand bitumen or natural asphalt; may also include vacuum residua (b.p. > 510°C; > 950°F)  3  3  3  3  Figure 3.1. Classification Of Crude Oils By Density-Gravity (Speight, 1991)  35  T a b l e 3.1.  Composition Ranges For The Bulk Fractions In Crude Oils (Speight, 1991)  RANGE OF COMPOSITION Crude Oils  Asphaltenes  (wt/wt%)  Resins  Oils  Conventional  <0.1 - 12  3-22  67-97  Heavy  11 -45  14-39  24-64  Residua  11-29  29-39  0-49  36 3.1.1 Feed Properties  The heavy oil used in this study was obtained from Imperial Oil Limited, Cold Lake, Alberta. Table 3.2 and Figure 3.2 summarize historical analytical data for this Cold Lake heavy oil. Typical conventional crude oil property data are also reported in Table 3.2 for comparison. As seen in Table 3.2, Cold Lake heavy oil viscosity at room temperature is over 100,000 cP which is more than 1000 times that of conventional crude oil. In addition, it is found that asphaltene and nitrogen contents of the Cold Lake heavy oil is two times higher than that of conventional crude oil while sulfur content is 20 times higher. Furthermore, as seen in Figure 3.2, only 50% by volume of Cold Lake heavy oil is distillable. However, Speight (1991) reported that more than 70% by volume of conventional crude oil is distillable.  37  Table 3.2. Comparison of Cold Lake Heavy Oil and Conventional Crude Oil Properties  HISTORICAL D A T A  C O L D L A K E H E A V Y OIL*  CONVENTIONAL  Maximum  Minimum  11.3  9.1  DENSITY:  0.991  1.006  VISCOSITY(cP):  410,000 176,000 30,200 1,050  96,800 22,600 1,050 170  <30(38°C)  C A R B O N R E S I D U E (wt%):  12.1  10.4  1-2  A S H (wt%):  0.13  0.03  0  C A R B O N (wt%):  84  83  86  10.7  10.2  14  N I T R O G E N (wt%)  0.9  0.3  0.2  S U L F U R (wt%):  4.7  4.1  0.1 -0.2  2.4  0.9  V A N A D I U M (ppm wt.):  189  146  N I C K E L (ppm wt.):  85  64  A S P H A L T E N E S (wt%):  18  15  I^LGRJ^TTY15°C 25°C 60°C 100°C RAMSBOTTOM  HYDROGEN  OXYGEN  (wt%):  (wt%):  •Source: Esso Resources Canada Ltd. " S o u r c e : Speight (1978)  C R U D E OIL ** 25-37 0.839-0.914  < 10  Figure  3.2. Simulated Distillation Of Cold Lake Heavy Oil  Cut Temperature (C)  39  3.1.2 Asphaltenes  Asphaltenes are the most problematic class of materials present in heavy petroleum feedstocks. They are dark brown to black friable solids that have no definite melting point, and when heated, usually swell, then decompose leaving a carbonaceous residue (Speight, 1981). Asphaltenes are usually defined as the fraction of heavy crude oil that is insoluble in non-polar solvents such as n-pentane and n-heptane, but soluble in polar solvents such as toluene and benzene.  Asphaltenes are believed to be made of mixtures of micelles which resultfromthe association of smaller units, called clusters (Speight, 1991). These clusters are condensed aromatic nuclei that carry alkyl and alicyclic substituents with heteroatoms (i.e. nitrogen, oxygen, and sulfur) scattered throughout in various locations. Metals such as vanadium and nickel are embedded within these clusters and are thought to help in the formation of micelles. The micelles are believed to be held together primarily by weak hydrogen bonds and van der Waals forces (Sane et ai, 1992). Studies have found that heteroatoms and metal contents increases with increasing asphaltenes present in the oil. Figure 3.3 shows two hypothetical structures of asphaltenes derivedfromstudying nuclear magnetic resonance (NMR) and spectroscopic data of asphaltenes from various petroleums (Speight, 1991). A diagram of an asphaltene cluster, as deducedfromx-ray diffraction can be found in Figure 3.4.  Figure 3.3. Hypothetical Structures For Asphaltenes From (A)Venezuelan Crude Oil; (B) Californian Crude Oil (Speight, 1991)  41  ^sr  \ wvvw  A r o m a t i c sheets  vwvJ — I 3.6-3.8 A (C/2) — V«/WW^  10-16 A (L )  'WwVWAA  6-15 A  N  c  (L )  =L / C / 2 = 3 - 5 c  Where N is the number of lamellae; L is the height of the unit cell; c/2 is the interlamellar distance; L is the layer diameter c  c  a  Figure 3 4 Representation Of An Asphaltene Cluster From X-Ray Analysis (Speight, 1991)  42  The molecular nature of asphaltenes is currently a topic of much study. Determining the molecular weight (MW) of asphaltenes is a difficult problem because of their low solubility in the liquids which are used for the determination (Speight, 1981). In addition, asphaltenes are dynamic aggregates that are constantly changing, either delaminating to form smaller components or associating to form larger micelles (Sane et al, 1992). The delamination rates depend on temperature, concentration and the solvent.  Speight (1991) has summarized the asphaltene molecular weights reported in the literature and determined by different analytical methods. For example, molecular weight investigations using an ultracentrifuge gave values up to 300,000 g/mol. However, asphaltene molecular weights determined by the ebullioscopic method were lower, ranging from 2500 to 4000 g/mol and by vapor pressure osmometry from 1000 to 5000 g/mol. Markhasin et al (1969) reported a range of 1000-4000 g/mol by light absorption coefficients.  Baltus and Anderson (1983) studied the transport of asphaltenes through porous membranes. They used a Wicke-Kallenbach type diffusion cell to examine the relationship between diffusion coefficient of asphaltenes and membrane pore size using track-etched mica membranes.  They found a good linear relationship between the molecular weight of  monodisperse polystyrenes and asphaltenes in tetrahydrofuran (THF) at 25°C.  Using  polystyrene-equivalent molecular weight, they divided the asphaltene MW into five ranges and assigned each to a hypothetical asphaltene fraction.  43  I0  5  I0 2xl0 Toluene Polystyrene Molecular Weight (MW*) 4  5  Figure 3.5. GPC Elution Curve Of Asphaltene Fraction In Tetrahydrofuran At 25°C (Baltus and Anderson, 1983)  Table 3.3. Asphaltene Fraction And Molecular Radius In Tetrahydrofuran At 25°C (Baltus and Anderson, 1983)  Asphaltene Fraction  MW* xlO"  3  Doo x 10 (cm /sec) 7  Molecular Radius (A)  2  Stokes-Einstein A B C D E  2481632-  4 8 16 32 64  16.1 13.1 8.85 5.24 2.71  26 32 47 79 153  Experimental** 20 32 44 69 152  MW* is the molecular weight of polystyrene which was eluted from the GPC column at the same solvent throughput as the asphaltene component. **The equivalent polystyrene molecular radius.  44  Figure 3.5 shows the exclusion chromatography (GPC) elution curve of the asphaltene fractions based on an equivalent polystyrene molecular weight (MW*). Baltus and Anderson (1983) measured the diffusion rates of asphaltene fractions through porous membranes with pore diameter in the range 80-2200 A. By assuming that asphaltene components are spherical, the molecular radii were determined from the experimental diffusivity data and the StokesEinstein equation: r, =K T/(6jnjD ) B  (3.1)  0O  where, r,  = the solute radius (i.e. asphaltene) (m)  KB  = the Boltzmann constant = R/N  A  R  = the gas constant (m .Pa/mol.K)  NA  = the avogadro's number (atom/mol)  T  = the absolute temperature (K)  Do,  = the bulk diffusion coefficient (m /s)  r|  = the fluid viscosity (Pa.s)  3  2  Table 3.3 lists the values of Doo and molecular radius of each of the hypothetical asphaltene fractions reported by Baltus and Anderson (1983).  Storm et a/.(1995) reported that the density of dry asphaltene (i.e. precipitatedfromn-pentane or n-heptane) was l.lg/cm and the density of asphaltene in heavy oil was believed to be less. 3  They suggested that asphaltenes behave as a solvated spheres dispersed in the nonasphaltenic  45  fraction (oil) and so asphaltenes have a larger hydrodynamic volume than would be expected based on their dry density. In addition, Mehrotra et al. (1994) reported that the solubility of asphaltenes in a crude oil is dependent on the relatively lighter components, whose solubility is in turn dependent on the still lighter components. Thus by changing the solubility of the component species, asphaltenes precipitation might be expected.  46  3.2 EXPERIMENTAL SET-UP AND PROCEDURE  3.2.1 Experimental Apparatus  A schematic description of the experimental set-up used in the present study is shown in Figure 3.6(a). The key components of the system include a single-tube membrane housing unit, two feed tanks, two positive displacement pumps, a recycle flowmeter, a permeate flowmeter, a pressure relief valve, an on-linefilter,and a 486DX2/33MHz computer for data acquisition and automatic temperature control. This system operates as a batch ultrafiltration unit with partial recycle of retentate, sometimes referred to as a "batch with recirculating loop, topped off' (Cheryan, 1986). In this system, the feed oilflowsat a high pressure into the tube side and the permeate is collected on the shell side of the housing. Figure 3.6(b) shows a simple schematic diagram of the operation used in this thesis. The system was designed and constructed by researchers in Dr. Smith's group in the Chemical Engineering Department of the University of British Columbia. Trouble shooting and reconstruction were accomplished in the present study.  Membrane Housing Unit The membrane housing unit, shown in Figure 3.7 was made of 316 stainless steel with an outside diameter of 5.08 cm (2 inches) and an inside diameter of 1.91 cm (3/4 inches). A single tube ceramic element was placed in the housing and sealed by two O-rings on the  Return Line  Feed Pump  Recycle Pump  Figure 3.6(b). Batch Ultrafiltration with Partial Recycle of Retentate  1/4" NPT (FEMALE)  1/4" NPT (FEMALE)  S.S. HOUSING (2" O.D. & 3/4" I.D.)"  CERAMIC ELEMENT FERRULE O-RING  O-RING  1/4" NPT (FEMALE)  6 HOLES FOR 1/4" ALLEN  Figure 3.7. Membrane Housing Unit  50  ferrules located at both ends of the ceramic element. Two openings located close to both ends of the housing unit was used for permeate collection.  Feed Tank Two 3.25L 316 stainless steel feed tanks were used: one for heavy oil feed (Tl in Figure 3.6(a)) and the other (T2 in Figure 3.6(a)) for the solvent used to clean the system after each experiment.  The heavy oil feed tank included a stirrer for complete mixing of the fluid, a  thermocouple located at the bottom of the tank to detect the feed tank temperature, and two heating bands to heat the heavy oil. In addition, air was used to purge the stirrer motor to prevent the buildup of explosive vapours near the motor.  Feed Pump The Feed pump (PI in Figure 3.6) was a Viking C432 gear pump installed with a mechanical seal made of carbon graphite that could withstand temperatures to 150°C and strong solvents such as toluene or benzene. It was driven by a 3/4 horsepower DC motor with a maximum capacity 3.15xlO" m /s (0.5GPM) at 100% DC power input. 5  3  The pump was controlled  manually through a DC controller. It could operate at a maximum recommended discharge pressure of 1.03MPa.  Recycle Pump The Viking HI 15 rotor gear or gear-in-gear pump (P2 in Figure 3.6) was used to create a high cross flow velocity in the membrane by recirculating a fraction of the retentate.  A 1V  2  51  horsepower DC motor was used to drive this pump. It was also manually controlled by a DC controller but with both forward and reverse mode that made cleaning up the system convenient. This pump could operate at a maximum differential pressure of 1.03MPa. The maximum capacity of the pump was 6.31X10" m /s running at 0.69MPa with a feed viscosity 4  3  of 165mPa.s. To allow such a high flow capacity, 3/4" (1.91cm) diameter flow lines were used for the recirculation loop and the inlet to the recycle pump. The outlet tube diameter was 1/2" (1.27cm). The pump has a cavity of 34cm . 3  Recycle Flow Meter An MK 525 Metalex flow sensor (FM2 in Figure 3.6) was used to detect the recycle flow. The meter was made of stainless steel and could operate under high temperature and pressure conditions (temperatures up to 150°C and pressures up to 1500psi or 10.3MPa). The flow meter was designed to operate at aflow-throughvelocity of 0.46-6.10 m/s.  Permeate Flowmeter A Max Series 210flowmeter(FM1 in Figure 3.6) was used to measure the permeate flow. The meter could detect a maximum flow of 70 ml/min, and the in line pressure could not exceed 15 psi (103kPa). Theflowmeterwas connected directly to the computer to record the permeate flow. However, thisflowmeterwas not used during the experimental runs reported in this thesis (see Chapter 5) due to the effect of rapid membrane fouling. The initial permeatefluxcould not be measured with this meter.  52  Pressure Relief Valve The system pressure was regulated by a RL3 pressure relief valve (V19 in Figure 3.6), which had a nominal cracking pressure of 0-225 psig (1.55 MPa) at 21°C. With a standard viton seal, the valve could withstand temperatures to 204°C and strong solvents such as toluene and benzene. Trapped air could be released by fully opening this valve or by opening the needle valve (V19) located on top of the membrane unit.  Pressure Transducer Three pressure transducers were used to monitor the inlet and outlet pressure of the membrane unit, and the pressure in the permeate line. They were all configured on Labtech for data acquisition. The pressure transducers had a full scale output of 6 Volts and a minimum output of 1 Volt. They could operate at high temperature up to 125°C and pressure up to 3.45 MPa.  Pressure gauge A pressure gauge (PG in Figure 3.6) was located after the on-linefilterand between the feed pump and the recycle pump. It was used to ensure the presence of oil in the line as well as to observe the system pressure manually in case the pressure transducers failed.  Heating Tape Four Silicon Rubber Thermolyne heating tapes (0.5" width x 6' length or 1.27cm x 183cm each) were used to heat the system flow lines and maintain a constant temperature. The  53  heating tape had a maximum tape exposure temperature of 260°C.  Each one required a  120VAC and had a load capacity of 156W. The heating tapes were controlled by on/off closed loops and were monitored continuously by computer.  Band Heater Two electrical band heaters (3" width x 5" length or 7.6cm x 12.7cm each) wrapped around the feed tank (Tl) were used to heat the feed material. They were the major heat source of the system. They required 240VAC input voltage. Each band heater had a load capacity of 850W and a watt density of 20W/in  2  They were monitored and automatically controlled  using Labtech.  On-line Filter A 'TF"series tee type removablefilter(Fl in Figure 3.6) was installed after the feed pump. A 140pmfilterelement was used tofilterout large particles, allowing cleanfluidto pass through the membrane system. The element was constructed of 316 stainless steel wire mesh.  System Piping Due to the viscous properties of heavy oil, all tubes and valves selected were no less than 3/8" (0.95 cm) in diameter. The total volume of all piping in the system was 480 cm . 3  54  Data Acquisition Labtech Notebook computer software was used to perform the data acquisition and control functions of the membrane test unit. During the experiments, temperature and pressure measurements were obtained from thermocouples and pressure transducers located at the membrane inlet and outlet as well as at the permeate sample line. The permeate and the recycle flows were also measured and reported.  The data were recorded every minute for up to eight hours mnning time. The computer needed to reset and restart if the run time was longer than eight hours. The data reported for each of the measurements was the arithmetic average of data measured every three seconds up to a duration of one minute. All data were recorded in a newfileevery one hour.  Detailed description of both computer software and hardware can be found in Appendix C.  3.2.2 A Summary of the Experimental Procedure  Prior to the experiment, the feed oil was heated in the feed tank to approximately 100°C for about two hours to remove water from the heavy oil. At the beginning of the experimental run, heavy oil was pre-heated in the feed tank to between 90 and 120°C. It was then fed to the membrane unit at high pressure using the feed pump (PI).  It took approximately 30  seconds to fill the whole system. The recycle pump (P2) was turned on after the system had filled. The high pressure feed stream entered the tube side and permeate was collected on the  55  shell side of the housing unit. The major portion of the retentate was then combined with the injected feed and recirculated by the recycle pump (P2) in the unit at a high flow rate. A small amount of the retentate liquid was recycled back to the feed tank and mixed with the feed stock.  Permeate samples were collected from the sample line periodically during the  experiment.  Detailed procedures for operating and shutting down the membrane unit are described in Appendix C of this thesis. In addition, a brief discussion of the troubling shooting experienced in this work can also be found in Appendix C.  3.3. ANALYTICAL METHODS  Asphaltene Analysis Asphaltene content in the heavy oil feed and products was measured as thefractioninsoluble in n-pentane determined according to the Syncrude Analytical Method 5.1 (Liu and Gunning, 1991). A sample of heavy oil, approximately 2 to 3 gram was dissolved in an equal volume of toluene or benzene. When fully dissolved, 40 ml of n-pentane was slowly introduced for each volume of toluene previously added. The resulting precipitate wasfilteredusing a medium pore (10-15um) fritted glass filter under slight vacuum. It was then dried at 105°C and weighed as Asphaltenes. Hence, asphaltene content in the heavy oil was recorded as weight percent.  56  Metals Analysis Nickel and vanadium content was measured by the Alberta Research Council using Inductively Coupled Argon Plasma Atomic Emission Spectrometer (ICAP-AES).  A 2 to 10  gram sample of the hydrocarbon was diluted to 50 ml with xylene. This solution was then directly aspirated into the ICAP-AES. The emission intensities of the nickel and vanadium were then compared with those of standard solutions of certified concentrations, previously analyzed. Results were reported in terms of mg/kg (ppm) of sample for each element.  Density Fluid density was measured by the standard method for Specific Gravity and Density of SemiSolid Bituminous Materials according to the ASTM procedure D70. The sample was slightly heated (> 56°C and <111°C) and poured into a clean, dry, and warmed pycnometer with a weight of 'A' grams to about 3/4 of its capacity.  The sample was cooled to ambient  temperature (25°C or 15.6°C) and weighed with the stopper, this weight is designated as ' C . The pycnometer was weighed again with the addition of water, this weight is designated as 'D'. Another pycnometer wasfilledwith water and weighed as 'B'. Density of the bitumen was calculated based on the following equation, Density = SG x WT and  SG = (C- A)/f(B  -A)-(D-Q]  WT is the density of water at test temperature in desired units (g/cm ). 3  57  CHAPTER 4; DEVELOPMENT OF A FOULING MODEL  The present study is concerned with heavy oil upgrading by ultrafiltration using ceramic membranes. Progressive flux decline with time due to fouling is a major drawback in many ultrafiltration processes. No study on the fouling characteristics of heavy oil using either polymeric or ceramic membranes has been reported. Thus, this study is needed to develop a better understanding of the fouling mechanism and to identify methods to limit the effect of fouling during heavy oil ultrafiltration.  A review of fouling, both on the membrane surface and within the membrane pores during ultrafiltration processes, was presented in Chapter 2. In the present study, permeate flux decline during heavy oil ultrafiltration with ceramic membranes was determined for membranes of various pore sizes and a range of cross-flow velocities.  The results of these  experiments described in detail in Chapter 5 suggested that at high cross-flow velocity (Uc > 6.9m/s) the flux decline was due to a reduction in membrane pore diameter. Hence a fouling model, based on pore restriction due to deposition of asphaltenes within the membrane pore has been developed to predict the permeate flux decline and asphaltene concentration during heavy oil ultrafiltration.  58  4.1 M O D E L ASSUMPTIONS  As discussed in Chapter 3, heavy oil is a complex mixture of organic molecules, the asphaltene fraction of which is not well defined on a molecular level. In addition, the membrane pore structure is determined by the packing of small solid particles so that there is significant tortuosity and pore size distribution associated with the pores. The complexity of this system makes the development of a detailed mathematical model to describe the fouling a difficult task. The model developed in this preliminary study is based on a series of approximations and simplifications that are discussed below.  Hence tractable solutions to the equations  derived on the basis of the phenomena thought to cause fouling, are possible. Figure 4.1 shows a schematic representation of an ideal pore and the fouling by adsorption/deposition of asphaltenes.  The assumptions made in the pore restriction model of heavy oil ultrafiltration using a ceramic membrane are as follows:  I. The solute molecules of the system are assumed to be made up of asphaltenes only. This assumption was made since the removal of the asphaltenes was a major goal of the ultrafiltration process and as is shown in Chapter 5, reduction in asphaltenes correlates with reduction in metals and viscosity of the heavy oil.  59  n. The asphaltenes are assumed to be a unique molecular species of rigid spherical shape. Asphaltenes are complex mixtures of long chain hydrocarbons and heteroatoms. Their size and shape are not well defined and their molecular structure is the subject of continuing research (Storm et al, 1995). The assumption of spherical molecular species allows the use of simplified hydrodynamic equations in describing solute flow.  m. The solute (asphaltene) concentration in the feed is assumed to be constant over the ultrafiltration period. The experimental system was configured with partial recycle of the retentate.  As the  lighter material permeates through the membrane, the asphaltene concentration in the feed increases. However, since the permeate flow was very small compared to thefluidflow through the membrane, i.e. the recoveries were < 1%, the change in feed concentration during an experiment was neglected. The validity of this assumption was confirmed by analysis of the feed concentration before and after each experiment.  IV. The active layer of the membrane is assumed to be aggregates of straight cylindrical pores, oriented perpendicular to the membrane surface. For mathematical simplicity, the pores in the membranes are assumed to be uniform cylinders. The effect of pore size distribution, although important, increases the model complexity and was considered beyond the scope of this preliminary study.  60  V. The diffusional flow of solutes through the membrane pores is neglected. Due to the large and complex structure of asphaltene molecules, the effect of asphaltene diffusion through the membrane pores is very small and was therefore neglected.  VI. The membrane structure is assumed to be very stable. The inorganic membranes used in this work are made of aluminum oxide which is chemically and thermally stable. Thus, compression of the membrane structure due to operating conditions such as high temperature and pressure would not occur.  VJI.Osmotic pressure is neglected. Since asphaltenes are macromolecules, the force of molecular attraction between them and heavy oil molecules is assumed to be very small compared to the applied system pressure.  V m . Concentration polarization is assumed to be established immediately after the run started. In membrane ultrafiltration, large molecules are retained by the membrane while small molecules permeate through the membrane pores more or less freely. The retained solute accumulates at the membrane surface where its concentration will gradually increase. As the concentration of solute increases, it will cause back diffusion of the solutefromthe membrane surface back into the bulk liquid due to the concentration gradient. This is called concentration polarization.  61  As discussed in Chapter 2, studies found that concentration polarization for membrane ultrafiltration of macro-molecules was established within a few seconds of the start of the experiment. In the present work, the above assumption is valid, since asphaltene has a macro-molecular structure and the permeate flux decline was modeled over hours of operation.  Legend Cb : Bulk (feed) concentration. C»: Membrane wall concentration C : Permeate concentration J : Permeate flux Tpo : Initial pore radius L : Membrane active layer thickness p  Figure 4.1. A Pore Restriction Schematic Diagra  63  4. 2 MODEL DEVELOPMENT  As discussed in the previous section, the active layer of the ceramic membrane is assumed to be composed of straight cylindrical and uniform pores. Based on the pore restriction model, permeate flux decline is due to adsorption/deposition of asphaltenes within the membrane pore that causes the pore to restrict or become smaller in size. According to the Hagen-Poiseuille equation for straight cylindrical and uniform pores, the average permeate flux (volumetric flowrate per total membrane area) decline with time t can be expressed as follows,  J(t)_r;(t)[r -Ar (t)r po  p  where, J  = the initial average permeate flux (m /m /day) 3  0  2  J(t)  = the average permeatefluxat time t (m /m /day)  rpo  = the initial average pore radius (m)  r (t)  = the average pore radius at time t (m)  p  3  2  Ar (t) = the solute adsorption/deposition layer thickness at time t (m) p  A detailed derivation of Equation 4.1 can be found in Appendix D. Based on Equation 4.1, the change in membrane pore diameter needs to be modeled in order to predict the flux decline with time.  Since using ceramic membranes for heavy oil upgrading is a new development, very few studies are available that describe the adsorption or deposition of asphaltenes in porous membranes. Shimura et a/.(1986) studied the effect of metal sulfide and coke deposits within  64  catalyst pores during hydrodemetalation of asphaltenes in catalytic hydrotreating.  They  modeled the changes of the catalyst pore structure due to these deposits using the following volume loss equation: + M r e A t) t ' M r e A A t )  er(r ^,t+At) =9 c  C  (4 2)  where, 0x(r ,z,t+At) c  = the fraction of total volume loss in the pore at time t+At (dimensioniess)  0  = the fraction of volume loss in the pore due to coke deposit  C  after the early stage of the run (dimensioniess) 0M(r ,z, t),  = thefractionof volume loss in the pore due to particulate  c  At)  QM(T ,Z, C  r  depositions at time t and At, respectively (dimensioniess) = the radial distancefromthe center of the catalyst (m)  c  t  = the experimental time (s)  z  = the axial distance in the catalyst bed (m).  For an increment of timefromt to t+At, the volume loss in the pore due to the deposition of metal sulfides was expressed as follows,  e (r ,z,At) = k Sa (r„z,t)Cv(r,..z,tV' At pv, Pv (r„z,0) M  c  h  (4.3)  r  c  where, kh  = the intrinsic reaction rate constant per unit surface area for hydrodemetalation  Sac(r ,z,t)  = the pore surface area per unit weight of catalyst (m /kg)  Cv(r ,z,t)  = the vanadium concentration in the pore (kg/m )  n  = the intrinsic reaction order  pv,  = the density of vanadium deposits (kg/m )  c  c  2  3  3  65  Pv (r ,z,0) c  = the volume of pore per unit weight of catalyst at time t = 0  c  (i.e. clean catalyst) (m /kg) 3  Shimura et al. (1986) found that the catalyst pore volume loss due to vanadium sulfide deposits by the hydrodemetalation of asphaltenes could be determined using Equation 4.3 with the intrinsic reaction order, n equal to 2. By using 8T , the change in catalyst pore structure due to the deposits was calculated using the following equation,  Pd(r„z,t+At) = Pd(r„z,0) [ 1 - er(r ,z,t+At)]  (4.4)  1/2  c  where, Pd(r ,z,t+At) = the pore diameter at time t+At (m). c  Based on this volume loss model for the change in catalyst pore structure due to fouling, a model of the change in membrane pore size was developed to determine the amount of adsorption/deposition of asphaltenes inside the membrane pore in heavy oil ultrafiltration. Thus, Equation 4.3 was modified to model the membrane pore volume loss as a function of time-on-stream. For an increment of time At, the fraction of volume loss in the membrane pore, 9, can be expressed as follows,  0(At)= K* Sa(Y) C. (t) At P.PV  (4.5)  0  where, K*  = the rate of adsorption/deposition of solutes (m/s)  Sa (t) = the surface area of the pore at time t (m ) 2  = 7T.d (t)L p  a  (for a cylindrical pore)  d (t)  = the diameter of the pore at time t (m)  L,  = the thickness of the active membrane layer (m)  C  = the solute (asphaltene) concentration inside the pore (kg/m )  p  3  s  66  p  = the solute (asphaltene) density (kg/m ) 3  a  Pv  = the initial or clean pore volume (m ) 3  0  = 7tdpo L /4 2  a  dpo  (for a cylindrical pore)  = the initial diameter of the pore (m)  The volume loss at time t +At is the volume loss at time t plus the volume loss at an increment of time At, 9 (t+At) = 0 (t) + 0 (At)  (4.6)  By using the above fraction of volume loss in the pore due to adsorption/deposition of asphaltenes, the change in diameter of a cylindrical pore at time t + At can be expressed as follows, d (t+At) = d „ [ l - 0(t+At)] p  (4.7)  1/2  p  or in terms of pore radius, r (t+At) = r [ l - 0(t+Af)]  (4.8)  1/2  p  p o  Assuming the membrane has uniform pore size, by substituting Equation 4.8 into Equation 4.1, the permeate flux decline with time due to the pore restriction can be determined as a function of the change in pore size. For example, the permeatefluxat time t - ti +At can be 2  expressed as follows, [r(t,+At)] J ( t ) = J„ " < 2  4  (4.9)  67  Therefore, in order to predict the permeate flux at t, the pore volume loss due to adsorption/deposition of solute needs to be determined. Hence, all of the parameters in Equation 4.5 need to be defined in order to calculate the initial pore volume loss at time t given the initialfluxand initial clean membrane pore size. The simplest way to detennine the solute concentration inside the pore, C,, is to assume that C, equals the solute concentration in the permeate, Cp.  In the ultrafiltration process, the permeate solute concentration, C , at any time t can be p  determined from the wall solute concentration, C , and the true rejection coefficient, R, as follows: C (f) = C (t)[l -R(t)] p  (4.10)  w  where, C (t) p  = the permeate concentration at time t (kg/m ) 3  Cw(t) = the wall concentration at time t (kg/m ) 3  R(t)  = the membrane true rejection coefficient (dimensionless)  Assuming that solute transport through the membrane pores is based only onfiltrationflow, the Ferry Equation (1936) can be used to determine the true rejection coefficient, R. For laminarflowthrough a membrane tube with uniform and cylindrical pores and circular crosssection, the true rejection coefficient as a function of time is as follows: R(t) = 1 - [2(1-^) - (l-X) ] 2  or,  =  and  X = rjr (t)  (4.11)  4  [X(X-2)]  2  p  \<X<0  (4.12)  68  where r  = the solute radius (m)  s  However, the above equation was derived for the steric hindrance effect at the entrance to the pores during ultrafiltration. The friction between a molecule moving within a pore and its wall had not been taken into consideration by Ferry. Renkin (1955) reported that an additional factor to the Ferry Equation, needed to account for the frictional interaction between the solute and the membrane pore wall, could be obtained using the Faxen Equation.  This  fictional effect also depends on the relative sizes of the solute, r„ and the pore, r , and was p  shown to be given by: F = 1-2.104X +2.09X - 0.95X 3  (4.13)  5  By combining these two factors, the steric hindrance at the entrance to the pores and the frictional effect on solute molecules within the pores, the total restriction to molecular diffusion (Ferry-Faxen Equation) is given as follows: F  = [2(1-X) - (1-A.) ] [1-2.104*. + 2.09X - 0.951 ] 2  T  4  3  s  (4.14)  and the rejection coefficient in Equation 4.11 can be written as:  R(t) = 1 - F  T  (4.15)  69  The solute concentration at the membrane wall surface, Cw, can be obtained based on the concentration polarization equation as follows: C (t) = exp{J (t)/ k}*[C - C (t)] + C (t) w  L  b  p  (4.16)  p  where, k  = the mass transfer coefficient (m/s)  Cb  = the bulk concentration (kg/m )  C (t)  = the permeate concentration at time t (kg/m )  JiXt)  = the local permeate flux (volumetric flow per pores area)  3  p  3  at time t (m /m /day) 3  2  For a membrane with a uniform pore size, the local permeate flux, J , is equal the overall L  permeate flux, J , divided by the clean membrane porosity, e , or J = Jo/e . 0  0  L  0  Since the  membrane pore structure restricts with time, the membrane porosity will change with time. Hence the local permeate flux as a function of time can be derived as follows, J L ( I ) =M * SM  (4.17)  A detail derivation of Equation 4.17 can be found in Appendix D. Substituting Equation 4.10 into Equation 4.16 and rearranging, we obtain: C (t) = w  Ch explJr (O/kl R(t)+ [l-R(t)]exp[J (t)/k]  (4.18)  L  The mass transfer coefficient for laminar flow through thin channels can be calculated from the Leveque Equation (Porter, 1972): k = 1.62 [UtD^L-J d Lt h  1/3  (4.19)  70  where, Ub  = the bulk axial velocity (m/s)  Doo  = the bulk diffusivity coefficient (m /s)  dh  = the equivalent hydraulic diameter which is equal to d for circular  2  tube (m). Lt  = the channel length or membrane tube length (m)  Most of the parameters in this model can be obtained either from correlations or theoretical mathematical relationships. The rate of asphaltenes adsorption/deposition, K* is the only unknown parameter in the model, determined byfittingthe experimental data. All of the equations used in the model are simple algebraic equations. Thus they can be easily calculated using any spreadsheet or programming language.  A schematic flowsheet of the pore  restriction model can be found in Figure 4.2. Sample calculations of the model and the calculated data can be found in Appendix D.  ' po  R(t) [Eg. 4.15]  C (t) [Eq. 4.18] w  /C^CpCt)'! [Eq. 4.10]  JL(0  ©(At) [Eq. 4.5]  [Eq. 4.17]  Legend Tpo : Initial pore radius r,: Solute radius r (t): Pore radius at time t J : Initial permeate Flux J (t): Permeate Flux at time t Cp(t): Solute concentration in the permeate at time t C,(t): Solute concentration inside the pore at time t R(t): Rejection coefficient at time t 0 (t): Pore volume loss at time t p  0  Figure 4.2. T h e Schematic Flowsheet of the Pore Restriction M o d e l  72  CHAPTER 5: RESULTS AND DISCUSSION  5.1 Experimental Studies  Experiments were performed to examine the separation of Cold Lake heavy oil over a range of operating conditions using porous ceramic membranes.  Since heavy oil is extremely  viscous (u>100,000 mPa.s at room temperature, see Chapter 3), the experiments were done at elevated temperatures (90°C and 120°C) to reduce viscosity to a suitable level for ultrafiltration. The transmembrane pressure in all experiments was between 414 kPa to 689 kPa.  For each experiment, the permeate flux was measured as a function of time and the  asphaltene, nickel and vanadium concentration in the permeate were also measured.  Since fouling is expected to reduce the permeate flux, one way to minimize this effect is to operate at high cross flow velocity. This will minimize the effect of fouling by deposition on the membrane surface as well as the effect of concentration polarization. To examine the effect of cross-flow velocity an initial series of experiments was performed at different crossflow velocities using ceramic membranes with an average pore diameter of 0.1p.m. Table 5.1 presents the experimental operating conditions, the average permeate flux and the feed asphaltene concentration at the end of run for these experiments.  Note that the Reynolds number calculated in Table 5.1 indicates that the systemfluidflowwas laminar.  This is due to the small diameter of the available commercial single-pass  73  Table 5.1. Heavy Oil Ultrafiltration at Different Cross-flow Velocities Using Membranes with Average Pore Diameter of O.lum Run 1  Run 2  Run 3  Run 4  Run 5  2.3 100 421 407 17.1 335  5.1 98 669 614 16.3 466  6.9 110 662 600 17.3 571  8.8 118 669 538 18.3 586  Reynolds Number: 99  148  297  675  1154  Average Flux in 5 hrs (kg/m /day):  37  85  99  98  16.9  17.4  19.1  18.5  Operating Conditions: U (m/s): T (°C): Pi (kPa): AP (kPa): C (wt%): Duration (min.): c  T  FS  2  1.5 100 510 503 15.4  340  41  C (wt%): 16.0 (at the End of Run) FS  where, U is the cross-flow velocity; T is the system operating temperature; AP is the transmembrane pressure; Pi is the membrane inlet pressure; and C is the asphaltene concentration in the heavy oil feed. c  FS  74  membrane tube used in this work, d = 0.008m and the high heavy oil viscosity (u>100mPa.s t  at 100°C as reported in Chapter 3 and in Section 5.2 of this Chapter).  It was quite difficult to have exactly the same operating temperature for each experiment. Heating tapes used to heat the system flow were held in place by bands of fiber glass insulation. These bands had to be removed after each experiment. Since these bands were not replaced in an identical location for each experiment, the overall system heat transfer rate likely variedfromeach experiment which affected the system temperature.  5.1.1 Effect of Cross-flow Velocity  As seenfromTable 5.1, the average permeate flux after 5 hours of operation increased with increasing cross-flow velocity.  Increasing cross-flow velocity increases the mass transfer  coefficient across the boundary layer. In addition, increased cross-flow velocity likely also increased the mixing and agitation near the rough membrane surface reducing the accumulation of solute on the membrane surface.  However, the effect of cross-flow velocity on permeatefluxis not as important as the effect of transmembrane pressure. Permeatefluxis proportional to the transmembrane pressure, AP , T  according to Darcy's law, whereas the mass transfer coefficient, k, is proportional to U c  1/3  according to the Leveque correlation for laminarflowthrough a cylindrical circular tube (see Chapter 4). Hence according to the different APx and Uc in Run 1 and Run 2, permeate flux  75  of Run 1 should be (503/407)*(1.5/2.3) = 1.1 times higher than that of Run 2. As seen from 1/3  Table 5.1, the average experimental permeate flux obtained after 5 hours of operation from Run 1 was 10% or 1.1 times higher than Run 2 (41 and 37 kg/m /day, respectively). This is 2  exactly what is expected theoretically. Similar calculations were also applied to Run 3. The average permeate flux for this run after 5 hours of operation was predicted to be equal to 74kg/m /day whereas the measured value was 85kg/m /day. The slight difference between the 2  2  calculated and measured flux could be due to the experimental errors and/or the effect of cross-flow velocity on the fouling mechanism as discussed above. Furthermore, the results obtained in Run 1 to 3 indicate that the experiments were conducted within the pressure control region in which permeate flux increases with increasing pressure. However, in this thesis, the effect of pressure was not examined in detail.  5.1.2 Asphaltene Reduction with Time  Figure 5.1 shows the permeate asphaltene reduction at different cross-flow velocities as a function of time-on-stream. As seen from Figure 5.1, significant asphaltene reduction was obtained by passing the heavy oil through the ceramic membranes with a nominal pore diameter of 0.1pm. It was found that asphaltene reduction was a strong function of time-onstream.  Asphaltene reduction increased almost linearly with time up to about 200 minutes. A gradual increase in the asphaltene reduction then followed as the run proceeded. From Figure 5.1, approximately  60%  reduction in asphaltene  concentration  was  obtained  in the  76  Figure 5.1. Asphaltene Reduction vs. Time Using 0.1 u,m Pore Diameter Membranes at Different Cross-flow Velocities 90 80  x  o  70  o  60  '•43 O  3 50  "O  <B DC  <2 40 ra Q.  W <  x Run 1 o Run 2  30  A Run 3  20  o Run 4 • Run 5  10  100  200  300  Time (min)  400  500  600  77  first 200 minutes of operation and this only increased to about 80% asphaltene reduction in the next 300 minutes.  The maximum asphaltene reduction obtained at the end of each  experiment ranged from 70 to 81%. Moreover, the highest asphaltene reduction was accomplished at the longest run time. As seen in Run 5, 81% asphaltene reduction was obtained at the end of 586 minutes of the time-on-stream. These results suggest a fouling mechanism in which membrane pore size is being reduced with time-on-stream. The reduced pore size increases the asphaltene rejection.  5.1.3 Correlations of Metals and Asphaltene Concentration  As discussed in Section 3.1.2, metal contaminants such as nickel and vanadium are believed to be embedded within the asphaltene structure. Thus removing asphaltenes from the heavy oil should also result in removal of these heavy metal contaminants.  Indeed, the experimental results of the present study showed that the metals content of the permeate correlated with the asphaltene content. Figure 5.2 shows the results of the nickel and vanadium concentrations versus different asphaltene concentration in the permeates and the feed heavy oil. Data for the heavy oil feed are also included in this Figure. The concentration of nickel and vanadium was linearly dependent on the asphaltene concentration. A linear correlation was determined using a least-squares estimation: Ym = 3.15*X  asphaL  Yva = 8.12 where,  + 20.44 + 46.84  (5.1) (5.2)  YM  = the concentration of Nickel in the oil (ppm)  Yv»  = the concentration of Vanadium in the oil (ppm)  X^hai.  = the concentration of asphaltenes in the heavy oil (wt%)  Figure 5.2. Nickel and Vanadium vs. Asphaltene Content in Heavy Oil 200  T  0  5  10  15  Asphaltene Content (wt%) •  Nickel  * Vanadium  y1(V)  y2(Ni)  20  79  5.1.4 Viscosity, Density and Asphaltenes Correlations  As reviewed in Chapter 3, asphaltenes have a complex, macro-molecular structure, made up of hydrocarbons embedded with heteroatoms and metals. Thus, the removal of asphaltene molecules from the heavy oil will induce a reduction in oil molecular weight, viscosity and density (as reviewed in Chapter 2). In the present study, it was found that the amount of asphaltenes present in the heavy oil significantly affected the fluid viscosity and density as shown in Figure 5.3 and 5.4. The oil with lower asphaltene content was lighter and less viscous than that with higher asphaltene content. An exponential correlation between the heavy oil viscosity and asphaltene content was found byfittingthe measured data: Y^o = 295.96 exp (0.166X^0,) Y^uo = 16.38 exp (0.073X  (5.3.a)  J  (5.3.b)  ttspha  where, Yn-40  = is the heavy oil viscosity at 40°C (mPa.s)  Y^-no = is the heavy oil viscosity at 110°C (mPa.s) Xasphai.  = the heavy oil asphaltene content (wt%)  On the other hand, the effect of asphaltenes on heavy oil density was correlated by a linear relationship between the oil density and the asphaltene content (see Figure 5.4): Y = 1.97+XaspH.i + 965.3  (5.4)  p  where, Y  = the heavy oil density at 15.6°C (Kg/m ) 3  p  The heavy oil viscosity and density can be reduced significantly by removing the asphaltenes from the oil.  Figure 5.3(a). Viscosity Measured at 40C vs. Asphaltene Content in Heavy Oil 7000  T  0  5  10  15  20  Asphaltene Content (wt%)  Figure 5.3(b). Viscosity Measured at 110C vs. Asphaltene Content in Heavy Oil 70  x  60 M CO  50  0.0731X  y = 16.376e' R = 0.8815 2  E 40 + 30 cn  > 20 10 0  • Measured — Expon. (y) —t—  15 10 Asphaltene Content (wt%)  20  81  Figure 5.4 Density Measured at 15.6C vs. Asphaltene Content in Heavy Oil 1000 y = 1.9702x + 965.32 995 4-  R = 0.9761 2  990  J=  985  "55  980  O)  c  a>  Q  975 Measured  970  Linear (y) 965 10  Asphaltene Content (wt%)  15  20  82  5.1.5 Flux Decline with Time Due to Membrane Pore Restriction  Permeate flux decline with time due to fouling is as significant in heavy oil ultrafiltration with ceramic membranes as with other membrane separation processes. Figure 5.5 shows the average permeate flux versus time at different cross-flow velocities using ceramic membranes with a O.lum nominal pore diameter. The data correspond to the experiments of Table 5.1. As seenfromFigure 5.5, an initial rapid drop in the permeate flux occurred within the first 2 to 3 hours of operation, followed by a gradual decline with the permeate flux leveling off at the end of the experiment. Furthermore, Figure 5.5 shows that the permeate flux increased with increasing cross-flow velocities up to 6.9m/s. Further increases in velocity had little or no effect on the permeate flux.  At low cross-flow velocity (Uc < 5.1 m/s), the permeate flux decreased continuously up to approximately 8 hours of operation (Run 1, 2 and 3 of Figure 5.5). However, asphaltene reduction followed a similar trend as observed at high cross-flow velocity (see Figure 5.1). These observations suggest that at low cross-flow velocity, fouling was dominated by the formation of a 'gel' layer. The gel layer thickness increases with time-on-stream thereby decreasing the permeate flux. However, the gel layer thickness does not affect the membrane rejection.  83  o o  CO  E T-  ^—  CD  3 a:  © JS  X  o> o x: o  CM co C c 3 3 o  •«*  c 3 OH o  •4  m c 3  o o  VO  •  tL © ]?>  o o  3J  .i 18  s  M  O  o o  > X  £  LL  «  0  CO  E  ©  E * V.  ^  °- « O)  ©  ^  c co  + § CM  2o  k_  3  O O  E ©  LL  o to  CM  O LO  O O CM  (ABP/Z UJ/6)J) V  O O  O  to  xn|j a6BJ3AV  E <»  E  84 At high cross velocity (Uc > 5.1m/s), Run 4 and 5 in Figure 5.5, initial flux decline occurred relatively quickly. It is important to examine the trend of this initialfluxdecline. Experiments 4 and 5 were repeated to estimate the repeatability of the experimental data as well as to monitor the initial flux decline.  Table 5.2 tabulates the operating conditions for these  experiments The average permeatefluxwith time for Run 4 and the repeated Run 6 can be found in Figure 5.6, whereas the average permeate flux versus time for Run 5 and the repeated Run 7 are reported in Figure 5.7. The repeatability of these experiments was very good considering the complexity of the experimental set-up. The standard deviation of the permeatefluxat the end of these experiments was less than 8%.  Figure 5.6 and 5.7 show a sharp drop in the initial permeate flux in thefirsttwo hours, from about 660 kg/m /day to approximately 60 kg/m /day followed by a very slow decrease that 2  2  appeared to be almost stable at around 55 kg/m /day for the remainder of the experiment. 2  Figure 5.8 shows the asphaltene reduction with time-on-stream of all four experiments at high cross-flow velocities (i.e. Run 4, 5, 6 and 7). As shown in Figure 5.8, asphaltene reduction increased continuously with time-on-stream up to about 3 hours of operating time and gradually leveled off until the end of the experiment. As seen from Figure 5.8, asphaltene reduction can be divided into three stages. Thefirststage occurred in thefirsthour. In this stage, asphaltene reduction increased gradually with time. The high asphaltene content of the permeate during this initial stage suggested that the membrane pore size was too large and provided essentially little separation. As the experiment proceeded, fouling of the membrane pore surfaces is thought to have occurred which resulted in a decrease in the effective pore  85  size. Hence, the asphaltene concentration in the permeate decreased significantly with time. This second stage occurred within the next three hours of operation. Finally, the asphaltene reduction increased gradually and was almost constant as the experiment proceeded in the third stage. This was due to the membrane pore being restricted to a certain size through which only very small asphaltene molecules were able to pass but large asphaltene molecules were rejected completely.  As the membrane pore size decreased due to the adsorption/deposition of solute molecules in the pore, the membrane resistance increased and hence the permeate flux decreased. Initially, due to the large pore opening, it allowed a large amount offluidpassed through the pore. As a result, a large amount of asphaltene molecules adsorbed/deposited in the membrane pore and fouled the pore quickly. Hence, the initial flux dropped rapidly. As the membrane pore restricted to a certain size, only a very small amount of fluid and small asphaltene molecules could pass through. Thus, the pore fouled much more slowly and the permeatefluxremained almost constant.  The experiment data at high cross-flow velocity were therefore consistent with fouling by pore restriction. This mechanism explains the change in asphaltene reduction with time and the almost constant permeate flux once the pore had restricted to a certain size. Based on the experimental results at high cross-flow velocity, a pore restriction fouling model was developed (see Section 5.2) to describe the fouling mechanism by heavy oil ultrafiltration using ceramic membranes.  86  Table 5.2. S u m m a r y of the Repeatability of Experimental D a t a for H e a v y O i l Ultrafiltration Using the 0.1pm C e r a m i c Membranes at H i g h Cross-Flow Velocities Run  4  Run 6  Run 5  Run 7  6.9  8.8  8.8  T (°C):  6.9 110  105  P, (kPa):  662  662  118 669  111 662  A P (kPa):  600 17.3  586 19.0  538 18.3  538 19.2  571  472  586  483  Operating Conditions: U  c  (m/s):  T  C (wt%): Duration (min.): F S  where, U is the cross-flow velocity; T is the system operating temperature; AP is the transmembrane pressure; Pi is the membrane inlet pressure; and C s is the asphaltene concentration in the heavy oil feed. c  F  Figure 5.6. Average Permeate Flux vs. Time Operated at Uc=6.9m/s Using the 0.1 um Membranes 700 600  o- - • Run 4 • Run 6  j? 500 CN  <  1  400  X  E 300 CD O) CO  I  200  -f  100 •o-  . o -  -f-  100  200  300  Time (min)  400  500  Figure 5.7. Average Permeate Flux vs. Time Operated at Uc=8.8m/s Using the O.lum Membranes 700  600 +  • • • - - Run 5 --a— Run7  CO  Si < E O)  500  400  X C  0) co  > <  300 200  100 '-a -  -- - - % n  1  100  200  ~~ ' " ~ ~ n -  •  Q  m  1  300 Time (min)  r—  400  500  600  89  Figure 5.8. Asphaltene Reduction vs. Time Using the 0.1 um Membranes at High Cross-Flow Velocities 90 80 +  o •  70 c o •o cu ct  CD  60 + 50 40  c  CD  Q_ W <  o Run4  30  • Run5 • Run6  20  • Run7 10 1  100  200  300 Time (min)  400  500  600  90  5.1.6 Effect of Pore Size  The effect of membrane pore size on permeate flux and asphaltene reduction was examined using ceramic membranes with an average pore diameter of 0.02um, 0.05um and 0.lum. The experiments were all operated at the same high cross-flow velocity of 6.9 m/s. Table 5.3 summarizes the operating conditions.  Figures 5.9 and 5.10 show the permeate flux and  asphaltene reduction with time-on-stream for these experiments. As seen in Table 5.3, the average permeate flux in 7 hours of operation obtained with the O.lum membrane was significantly greater than that obtained with the 0.05um and 0.02um membrane.  After  approximately 4 hours of operation, however, the permeate flux obtained from the O.lum membrane was only slightly higher than thatfrom0.05um membrane (53 kg/m /day compared 2  to 45 kg/m /day, respectively). 2  In the case of the 0.02um membrane, the permeate flux  dropped to approximately 36 kg/m /day at the end of 4 hours operation. 2  On the other hand, higher asphaltene reduction was obtained using the smaller membrane pore size. As seen in Figure 5.10, asphaltene reduction for the 0.02um membrane size increased significantly in the first 100 minutes of operation to approximately 80% reduction and slowly increased to 90% at the end of the experiment. The initial asphaltene reduction with time-onstream for the 0.05um membrane was significantly higher than for the 0.1 um membrane. However, as the experiments proceeded, the differences in asphaltene reduction between the two membranes was negligible (see Figure 5.10).  91 The results of these experiments indicate that membrane pore size has a significant effect on heavy oil separation. The smaller the membrane pore size the greater the degree of reduction of asphaltenes in the heavy oil.  In addition, membranes with larger pores have lower  membrane resistance and therefore higher initial permeate flux. Furthermore, because of the higher permeate flux, the larger membrane pores foul faster than the smaller pores. As a result, a sharp drop in permeate flux occurred in the initial period with the 0.lum and 0.05u,m membrane (see Figure 5.9). After several hours of operation, the pores of all the membranes were reduced to a certain size. Hence, the permeate flux after several hours of operation was independent of the original membrane pore size.  92  Table 5.3. Effect of Pore Size on Membrane Heavy Oil Separation  Run 6  Run 8  Run 9  6.9 0.1 105 662 586 19.0 472  6.9 0.05 108 669 586 18.0 477  6.9 0.02 107 662 600 16.4 570  110  52  45  Operating Conditions: U (m/s): dp. (um): T (°C): P, (kPa): AP (kPa): C (wt%): Duration (min.): c  T  F S  Average Flux in 7 hours (kg/m /day): 2  where, Uc is the cross-flow velocity; dpo is the average membrane pore diameter; T is the system operating temperature; AP is the transmembrane pressure; P is the membrane inlet pressure; and C s is the asphaltene concentration in the heavy oil feed. T  ;  F  93  Figure 5.9. Permeate Flux vs. Time Using Different Membrane Pore Sizes at Uc=6.9m/s 700  600  500  • Run 6  E O) 400  xRun 8  x  + Run 9  <  E cv  300  a  |  200 100  4-  *x, * —I  100  x  •+  x*x  +  x  1  1  200  300  #  x+ x * * f  Time (min)  400  «  + 500  600  94  Figure 5.10. Asphaltene Reduction vs. Time Using Different Membrane Pore Sizes  100 90 80  r 70 u  60  •o a> (£  50 4-  3  CO CD c  40  e Run 6  30  x  20 10  + Run 9  CO 0) o.  CO  <  0  •  0  •  Run 8  1—  100  200  300 Time (min)  400  500  600  95  5.1.7 Effect of Membrane Regeneration  As shown in the previous sections, permeate flux decline with time due to fouling is a significant problem in heavy oil ultrafiltration. The experimental data suggest that fouling was due to adsorption/deposition of asphaltenes in the membrane pores. To regain the high initial membrane permeability and control fouling, the absorbed/deposited asphaltene molecules must be removed. Asphaltenes are completely dissolved in polar solvents such as toluene and benzene. Hence, in this study, toluene was used to remove the asphaltenes from the ceramic membranes. Due to the thermal and chemical stability of the ceramic membranes, strong solvents like toluene would not affect the membrane structure.  The membrane regeneration was examined in the following way.  At the end of Run 5,  without removing the membrane from the test unit, toluene was pumped in a back-flush mode through the system for about 15 minutes at ambient temperature, a flowrate of 80 ml/min, and a pressure of 30 to 50 kPa gauge. Then the membrane was soaked in toluene overnight (about 24 hours).  Following this procedure the ultrafiltration experiment was repeated. A  comparison of results obtained from a new and regenerated membrane are presented in Figure 5.11. Table 5.4 summarizes the operating conditions used in both runs. As shown in Figure 5.11, the permeate flux and the asphaltene reduction of the regenerated membrane followed exactly the same trend as that obtained with the new membrane. Hence, the membrane regeneration by back-flushing with toluene seemed to be relatively effective in removing the asphaltenes from the membrane pores.  96  Table 5.4 Summary of Operating Conditions Used in Heavy Oil Ultrafiltration with New and Regenerated Membranes of 0.1 pjn Pore Diameter Run 5 (New)  Run 10 (Regenerated)  g.g  Operating Conditions: U (m/s): T (°C): P, (kPa): AP (kPa): C (wt%):  118 669 538 18.3  8.8 116 662 524 19.5  Duration (min.):  586  483  c  T  FS  where, U is the cross-flow velocity; T is the system operating temperature; AP is the transmembrane pressure; P is the membrane inlet pressure; and C is the asphaltene concentration in the heavy oil feed. c  T  ;  FS  97  Figure 5.11(a). Comparing Average Permeate Flux vs. Time of New and Regenerated O.lum Membrane 250  x  • New (Run 5) oLIsed (Run 10)  -+-  100  200  300  400  500  600  Time (min)  Figure 5.11(b). Comparing Asphaltene Content vs. Time of New and Regenerated 0.1um Membrane  • New (Run 5) oUsed (Run 10)  100  200  300 Time (min)  400  500  600  98  5.2 MODELING STUDIES  Based on the experimental studies discussed in Section S.l, the pore restriction model described in Chapter 4 was applied to the flux data. The model assumes that flux decline is due to a reduction in the membrane pore diameter due to adsorption/deposition of asphaltenes inside the pore. As the membrane pore diameter decreased, it prevented larger asphaltene molecules from passing through the membrane thereby increasing the asphaltene rejection. The model has one fitted parameter, K*, the asphaltenes adsorption/deposition rate constant.  5.2.1 Initial Permeate Flux Determination  In order to calculate the permeate flux loss with time, the initial permeate flux and the change in pore radius with time need to be determined (see Equation 4.1). Based on Darcy's law, the initial permeate rate can be determined knowing the transmembrane pressure, the viscosity of the feed used and the clean membrane resistance: (5.5)  AP m  where, AP  the driving pressure (Pa) the fluid dynamic viscosity (Pa.s)  Rm  the clean membrane resistance (m' )  Jo  the initial permeate flux (volumetric flow rate per total membrane  1  area) (m /m /s) 3  2  99  Membrane Resistance, R  m  The membrane resistance, Rm, can be determined theoretically or experimentally and is a function of membrane properties such as membrane structure, orientation, material, porosity, and tortuosity factor. The resistance of the ceramic membranes used in the present study (Membralox® membranes) has been examined by Elmaleh and Naceur (1992) both experimentally and theoretically.  Figure 5.12 shows the measured and calculated clean membrane resistance as a function of the membrane average pore size reported by Elmaleh and Naceur (1992). They found major discrepancies between the predicted and the measured membrane resistance due to large differences in the granule size that exists in the transition zones of the various layers of the membrane. As a result, the experimental values of the membrane resistance reported by Elmaleh and Naceur were used in the present study. For Membralox® membrane with a 0.05u.m nominal pore diameter, the clean membrane resistance was 1.41xl0 m' . The clean 12  1  membrane resistances for 0.1 and 0.02um membranes was calculated by applying linear interpolation to the reported experimental values of the 0;05um and 0.2um membranes. As a result, for the 0.1 urn and 0.02um nominal pore diameter membrane, the clean membrane resistances used in this study were l.OlxlO m" and 1.65xl0 m", respectively. 12  1  12  1  100  Figure 5.12. Experimental and Theoretical Membrane Resistances of Membralox During Filtration of Water (Elmaleh and Naceur, 1992) 1.6  T  0  0.1  0.2  0.3  0.4  0.5  0.6  Membrane Mean Pore Size (nm)  0.7  0.8  101  Viscosity and Density Correlations As presented in Chapter 4, good estimates of fluid viscosity and density at the conditions of ultrafiltration are needed to apply the pore restriction model. Fluid viscosity and density are strongly dependent on temperature, especially for viscous fluids like heavy oil whereas pressure has a relative minor influence on viscosity (AOSTRA, 1989). Furthermore, heavy oil properties also vary markedly depending on their source.  Mehrotra and Svrcek (1987) modeled the Cold Lake heavy oil viscosity data over the temperature range of 37°C to 115°C and the pressure range of 0 to 10 MPa gauge. The asphaltene content of their Cold Lake samples was 17.0 wt%. Mehrotra and Svrcek reported the Cold Lake oil viscosity as a function of temperature and pressure according to the following correlation: Infln(/i)J= 22.1325 - 3.47381 HnfTg) + 0.004288*(P)  (5.6)  where, p  = the heavy oil viscosity (mPa.s)  TK  = the system temperature ( K )  P  = the system pressure (MPa)  Table 5.5 shows a comparison between the Cold Lake heavy oil viscosity measured in the present work and calculated from Equation 5.6.  The average absolute relative deviation  between the measured and calculated viscosity reported in Table 5.5 was 16%. On the other hand, the standard deviation of the measured values was about 10%.  Despite the slight  102 difference in asphaltene content of the Cold Lake samples, the agreement between the measured and calculated viscosities was relatively good.  In another article, Mehrotra and Svrcek (1988) claimed that the density of Cold Lake heavy oil decreased linearly with temperature between 25.3°C to 119.9°C at atmospheric pressure. The linear relationship was as follows:  (5.7)  p = -0.7822 *T + 1038 e  where, p T  = the heavy oil density (kg/m ) 3  c  = the experiment temperature (°C)  A comparison between the density calculated from the correlation and the density measured in the present work is shown in Table 5.6. The calculated density from Equation 5.7 is in good agreement with the measured density as indicated by the average absolute deviation between the measured and calculated density of 2.4%.  Given the operating conditions, the initial permeate rate was determined by substituting the value of R m reported by Elmaleh and Naceur (1992) as well as the Cold Lake heavy oil viscosity, calculated by the correlation of Mehrotra and Svrcek (1987), into Equation 5.5.  103  Table 5.5. A Comparison of Experimental and Predicted Viscosity of Cold Lake Heavy Oil  Experimental*  Correlation (Eq. 5.6)  Temperature  Viscosity  Viscosity  (°C)  (mPa.s)  (mPa.s)  (%)  40  5825  6342  9  80  425  319  -25  110  68  77  13  Deviation  * Measured viscosity of heavy oil used i n the present study  Table 5.6. A Comparison of Experimental and Predicted Density of Cold Lake Heavy Oil  1  Experimental*  Correlation (Eq. 5.7)  Temperature  Density  Density  (°C)  (kg/m )  (kg/m )  (%)  15.6  999.4  1025.8  2.6  18.0  1047  1023.9  -2.2  3  3  * Measured density of heavy oil used in the present study  Deviation  |  1  104  The heavy oil density obtainedfromEquation 5.7 was used to convert the calculated permeate flux from m /m /day to kg/m /day for comparison with the permeate flux measured in 3  2  2  kg/m /day. 2  As seenfromthe experimental data of both Run 6 and Run 7, thefirstpermeate product was observed after about 7.3 minutes using 0.1pm membranes operated at high cross-flow velocity.  Knowing the permeate flow line dead volume, the initial permeate flux was  determined from this time.  Comparing the initial flux measured to the initial clean flux  calculated from Equation 5.5 shows that the calculated values were within 5% of the experimental values. Considering the errors associated with measuring the initialfluxfrom the time at which permeatefirstappears at the sampling point, the agreement is excellent.  Table 5.7. A Comparison of the Calculated and the Experimental Initial Permeate Flux  Permeate Flux  Operating Condition Temperature  Pressure  Calculated  Experimental (at 7.3 mins)  (°C)  (kPa)  (kg/m /day)  (kg/m /day)  6  110  586  687  664  7  111  538  655  664  Run  2  2  105  5.2.1 Other Parameters Used in the Pore Restriction Model  Besides the calculated initial permeate flux, the pore restriction model also requires an estimate of the solute molecular diameter because the solute rejection is deterrnined by the ratio of solute diameter to pore diameter (see Section 4.2). Since asphaltenes are not a welldefined molecular species, it is difficult to determine their average molecular diameter. Various attempts at estimating the size of asphaltene molecules have been reported in the literature (Baltus and Anderson, 1983; Sane et al., 1992). In the present study the effect of asphaltene diameter on the model calculation has been examined based on the asphaltene diameters reported by Baltus and Anderson (1983).  Another parameter required to determine the mass transfer coefficient, k, using Equation 4.19 is the asphaltene bulk diffusivity. Asphaltenes are a mixture of micelles that are constantly delaminating to smaller molecules or agglomerating to larger molecules (as discussed in Section 3.1.2). The delaminating or agglomerating rate are dependent upon temperature, concentration and solvent (Fane et ai, 1992; Mehrotra et ai, 1994; Storm et al, 1995). Therefore, it is difficult to measure the asphaltene diffusion through porous membranes directly and accurately (Fane et al, 1992). As reported in Section 3.1.2 of this thesis, Baltus and Anderson (1983) were the first to study the effect of the asphaltene diffusivity and their data were used in the model calculation.  106  To calculate the average local permeate flux, J , the porosity of the clean membrane active L  layer also needs to be determined. As mentioned in Section 2.3, ceramic membrane porosity can range from 35% to 50% depending on the pore diameter. Elmaleh and Naceur (1992) reported that the porosity of the active layer of Membralox® membranes with nominal pore diameters of 0.05 to 0.8 pm rangedfrom0.38 to 0.43. In this model, an average value of the reported porosities was used, e = 0.4. Thus substituting the calculated initial permeate flux and the initial membrane porosity, the average permeate flux decline with time can be determined as the pore restricts using Equation 4.9.  In addition, the model has one fitted parameter, K*, which is the asphaltenes adsorption/deposition rate. K* was estimated by deterrnining the value that minimized the residual sum of squares (SSQ) between the measured and calculated permeate flux and asphaltene concentration in the permeate, i.e,  SSQ= £  ffJ." - J ^ l / f N i - l )  1=1  + £  [(Cp," - C ) ]/(N -1) 2  pi  C  (5.8)  '=i  Ji where, J"  = the measuredfluxat point i in time (kg/m /day)  J;  = the calculatedfluxat point i in time (kg/m /day)  Cpi"  = the measured asphaltene concentration at point i in time (wt%)  Cpi  = the calculated asphaltene concentration at point i in time (wt%)  Nj  = the number of measuredfluxdata points  N  = the number of measured asphaltene concentration data points  2  2  c  107  5.2.3 Effect of Asphaltene Size  Figures 5.13(a) to 5.17(a) show the permeate flux calculated using the pore restriction model at different asphaltene radii: r, = 0.0026, 0.0032, 0.0047, 0.0079 and 0.0153 urn compared to the experimental data of Run 6. For each asphaltene size, the fitted parameter, K*, was obtained by determining the smallest SSQ using Equation 5.8. Good agreement between the calculated and experimental permeate flux was obtained when a large asphaltene radius was assumed (i.e. r, = 0.0153 um, see Figure 5.17(a)), but the agreement was less successful for smaller asphaltene radius (i.e. r, = 0.0026u.m, see Figure 5.13(a)).  However, for the  calculation of the permeate asphaltene concentration, a better match was found with smaller asphaltene radius than larger radius (see Figures 5.13(b) to 5.17(b)).  As seen in Figures 5.13 to 5.17, the calculated asphaltene content in the permeate decreased with time as the permeate flux declined.  Very good agreement was found between the  calculated and the measured permeate flux and asphaltene concentration in the first hour of the operating period for all of the assumed asphaltene sizes. As seen from the experimental data (Run 6), the initial asphaltene removal was very low (i.e. approximately <2% asphaltene reduction obtained in the first hour of operation). According to the pore restriction model, the initial pore openings were much larger than the asphaltenes and hence the asphaltenes could still flow through the membrane more or less freely. Indeed, the membrane used in this case had a nominal pore diameter of 0.1 um which is 19 times larger than the smallest asphaltene  diameter,  d,-00052Lim  and  3.3  times  larger  than  the  largest  Figure 5.13(a). Experimental and Calculated Flux vs. Time Using the 0.1 um Membrane at Uc=6.9m/s rs=0.0026um. K* =1.75 E-11 m/s  500 Time (min)  Figure 5.13(b). Experimental and Calculated Asphaltene Concentration Using the 0.1 um Membrane at Uc=6.9m/s.  100  200  300  Time (min)  400  Figure 5.14(a). Experimental and Calculated Flux vs. Time Using the 0.1 um Membrane at Uc=6.9m/s rs=0.0032 um. K* =1.75 E-11 m/s  100  200  300  400  Time (min)  Figure 5.14(b). Experimental and Calculated Asphaltene Concentration Using the 0.1 um Membrane at Uc=6.9m/s.  100  200  300  Time (min)  400  500  Figure 5.15(a). Experimental and Calculated Flux vs. Time Using the 0.1 urn Membrane at Uc=6.9m/s rs=0.0047 pm. K* =1.75 E-11 m/s  200  300  400  Time (min)  Figure 5.15(b). Experimental and Calculated Asphaltene Concentration Using the 0.1 um Membrane at Uc=6.9m/s.  100  200  300  Time (min)  400  500  Figure 5.16(a). Experimental and Calculated Flux vs. Time Using the 0.1 um Membrane at Uc=6.9m/s rs=0.0079 um. K* =1.5 E-11 m/s 700  T  100  200  300  400  500  Time (min)  Figure 5.16(b). Experimental and Calculated Asphaltene Concentration Using the 0.1 urn Membrane at Uc=6.9m/s. rs=0.0079 um. K*=1.5 E-11 m/s  o  4  1  1  1  1  1  0  100  200  300  400  500  Time (min)  Figure 5.17(a). Experimental and Calculated Flux vs. Time Using the 0.1 um Membrane at Uc=6.9m/s rs=0.0153 um. K* =1.5 E-11 m/s  100  200  300  400  Time (min)  Figure 5.17(b). Experimental and Calculated Asphaltene Concentration Using the 0.1 um Membrane at Uc=6.9m/s.  100  200  300  Time (min)  400  500  113  asphaltene diameter, ds= 0.0306um, used in this model study. Figures 5.13(b) to 5.17(b) show that the calculated asphaltene concentration decreased very slowly with time in the first hour of the operating period. For example, the calculated asphaltene reduction in this period was 2.2% with the assumed smallest asphaltene size, r=0.0026um (see Figure 5.13(b)). The s  agreement between the calculated and measured asphaltene concentration in this initial operating period was excellent.  The average relative standard deviation between the  calculated and measured asphaltene concentration at this initial stage was < 1% for all of the assumed asphaltene sizes.  On the other hand, the initial calculated permeate flux decreased significantly with time. The reason is that the permeate flux calculated by the model varies as r and hence a small 4  decrease in the pore size will significantly affect the permeate flux. Due to the initial high asphaltene concentration passing through the pores, the pores were fouled very quickly. As a result, a sharp drop in the permeate flux was obtained initially. The experimental data (Run 6) also show that, at least qualitatively, the initial permeate flux dropped significantly in the first hour of the operating period.  As the asphaltenes passed through the membrane, they continuously adsorbed or deposited in the membrane pores and hence decreased the pore openings and increased the membrane rejection coefficient.  The result was a significant increase in the asphaltene separation (see  Figures 5.13(b) to 5.17(b)). Due to the decrease in the asphaltene concentration as the pore restricted, the pore volume loss for an interval of time, 9(At), also decreased with time-on-  114 stream. Thus, the pores fouled much more slowly in the later period than in the initial operating period. As a result, the permeatefluxdecreased gradually with time-on-stream after the initial stage.  As shown in Figures 5.13 to 5.17, the measured permeate flux and  asphaltene concentration followed the same trends as what were predicted.  Table 5.8 summarizes the rninimum sum of squares, SSQ, obtained in determining the best fitted K* at different asphaltene radii. As seen in Table 5.8, the best agreement between the calculated and measured asphaltene concentration was obtained at r,=0.0079pm whereas the best agreement between the calculated and measured permeatefluxwas at r,=0.0153pm. The best fitted K* obtained for smaller asphaltene sizes (i.e. r,= 0.0026pm, 0.0032pm and 0.0047pm) was 1.75xl0" m/s whereas for larger asphaltene sizes (i.e. r,= 0.0079pm and u  0.0153pm) it was 1.5xl0" m/s. However, for each assumed asphaltene size, the difference n  between the SSQ obtained from these two K* values was very small.  Hence the  adsorption/deposition rate, K*, was not a strong function of asphaltene size.  Figures 5.18(a) and 5.18(b) show the effect of asphaltene size on the permeate flux and asphaltene  concentration with K*=1.75xl0" m/s. n  For large asphaltene  size (i.e.  r„=0.0153pm), a sharp drop in the permeate asphaltene concentration was found after the first hour offiltrationperiod. This is due to the membrane sieving effect. Larger molecules are more readily rejected by the membrane than smaller molecules.  Hence, better separation  115  Table 5.8. Summary of the Minimum SSQ Obtained in Determining the Best Fitted K * for the O.lum Pore Diameter Membrane at Different Asphaltene Radii  K*=1.5xlO m/s  K * =1.75xlO m/s  n  n  Asphaltene  SSQ,  SSQc  2 SSQ  SSQ,  SSQcp  SSSQ  0.0026  0.531  0.089  0.620*  0.524  0.243  0.767  0.0032  0.523  0.062  0.585*  0.515  0.169  0.684  0.0047  0.498  0.034  0.532*  0.492  0.068  0.560  0.0079  0.427  0.055  0.482  0.426  0.024  0.450*  0.0153  0.241  0.158  0.399  0.272  0.095  0.367*  P  Radius, r, (um)  where SSQj and SSQcp is the residual sum of squares of the permeatefluxand asphaltene concentration data, respectively. * The minimum residual sum of squares in determining the bestfittedK* reported in Figures 5.13 to 5.17.  116  was found for larger asphaltene size than smaller asphaltene size.  However, due to the  molecular sieving effect, the smaller asphaltene molecules pass through the membrane more readily than the larger molecules. With a high asphaltene concentration passing through the pore, the pore restricted at a much faster rate and hence lower permeate flux was obtained. As seen from Figure 5.18(a), the calculated permeate flux for larger asphaltene size (r,=0.0153pm) was higher than that for smaller asphaltene size (r„=0.0026pm).  117  Figure 5.18(a). Effect of rs on Calculated Permeate Flux vs. Time Using the 0.1 um Membrane at Uc=6.9m/s. K*=1.75 E-11 m/s  100  200  300  400  500  Time (min)  Figure 5.18(b). Effect of rs on Calculated Permeate Asphaltene Content vs. Time Using the 0.1 urn Membrane at Uc=6.9m/s. K*=1.75 E-11 m/s  0  100  200  300 .  Time (min)  400  500  118  5.2.4 Effect of Cross-flow Velocity  According to the pore restriction model, since the mass transfer coefficient, k, varies with the cross-flow velocity to the one-third power, U  1/3 c  , the effect of U on the permeate flux and c  asphaltene concentration will be small. Indeed, as seen in Figures 5.6 and 5.7, at cross-flow velocities of 6.9m/s and 8.8m/s, the permeate fluxes were almost the same with a slightly lower flux found at the lower cross-flow velocity (i.e. 6.9m/s).  The small difference in  permeate flux and asphaltene concentration between these experiment is due in part to the experimental error associated with the experiments.  5.2.5. Effect of Membrane Pore Diameter  As discussed in section 5.1, smaller membrane pore diameter gives better separation and lower permeate flux. Figures 5.19 to 5.23 show the predicted permeate flux and asphaltene concentration with time at different asphaltene sizes compared with the experimental data for the 0.05pm membrane (Run 8). Table 5.9 summarizes the minimum sum of squares, SSQ, obtained in determining the best fitted K* reported in the above Figures. As shown from Figures 5.19 to 5.23, good agreement between the calculated and measured asphaltene concentration was obtained when smaller asphaltene sizes were assumed (i.e. r,=0.0026pm, 0.0032pm and 0.0047pm) and the best agreement was found at r=0.0047pm. s  The best  agreement between the calculated and measured permeatefluxwas found at r,=0.0079pm.  119  Figures 5.24 to 5.27 show the calculated and measured permeate flux and asphaltene concentration for the 0.02um membrane at different asphaltene sizes. The minimum sum of squares, SSQ, found for these figures are summarized in Table 5.10.  The best agreement  between the calculated and measured values was obtained at r =0.0032um. g  For the assumed  largest asphaltene size, r,=0.0153um or d,=0.0306u,m, the model assumed that asphaltene molecules of this size were totally rejected by the 0.02um membrane due to the membrane sieving effect.  Similar trends were obtained between the calculated and the measured permeate flux and the asphaltene concentration using the 0.05um and 0.02um membrane compared with the O.lum membrane discussed in Section 5.2.2.  It was found that better separation and higher  permeatefluxwere obtained at larger asphaltene sizes than smaller sizes.  Furthermore, it was found that the adsorption/deposition rate, K*, seemed to change with the membrane pore size.  For example, the minimum SSQ for the 0.1 urn membrane with  r=0.0047um (see Table 5.8) occurred at K*=1.75xl0* m/s whereas for the 0.05um and the u  s  0.02um membranes, the value was 1.5xl0' m/s and 1.25xl0" m/s, respectively. u  n  120  Table 5.9. Summary of the Minimum SSQ Obtained in Determining the Best Fitted K* for the 0.05um Pore Diameter Membrane at Different Asphaltene Radii  K*'=1.5xl0' m/s  K* =1.75xlO m/s  n  n  Asphaltene  SSQ,  SSQcp  I SSQ  SSQ,  SSQcp  SSSQ  0.0026  0.816  0.481  1.297*  0.887  0.433  1.320  0.0032  0.788  0.456  1.244*  0.858  0.394  1.252  0.0047  0.670  0.461  1.131  0.742  0.381  1.123*  0.0079  0.379  0.560  0.939*  0.496  0.481  0.977  0.0153**  2.219  0.769  2.988*  2.644  0.717  3.361  Radius, r„ (urn)  where SSQj and SSQcp is the residual sum of squares of the permeate flux and asphaltene concentration data, respectively. * The minimum residual sum of squares in determining the best fitted K* reported in Figures 5.19 to 5.23. **Can not obtain the minimum SSQj due to the large error in fitting the permeate flux.  Figure 5.19(a). Experimental and Calculated Flux vs. Time Using the 0.05 um Membrane at Uc=6.9m/s rs=0.0026 um. K*=1.75 E-11 m/s.  100  200  300  400  500  Time (min)  Figure 5.19(b). Experimental and Calculated Asphaltene Concentration Using the 0.05 um Membrane at Uc=6.9m/s. rs=0.0026 um. K*=1.75 E-11 m/s  100  200  300  Time (min)  400  500  Figure 5.20(a). Experimental and Calculated Flux vs. Time Using the 0.05 um Membrane at Uc=6.9m/s rs=0.0032 um. K*=1.75 E-11 m/s  100  200  300  400  500  Time (min)  Figure 5.20(b). Experimental and Calculated Asphaltene Concentration Using the 0.05 um Membrane at Uc=6.9m/s.  0  100  200  300  Time (min)  400  500  Figure 5.21(a). Experimental and Calculated Flux vs. Time Using the 0.05 um Membrane at Uc=6.9m/s rs=0.0047 urn. K* =1.5 E-11 m/s 500  T  450  -r  100  200  300  400  500  Time (min)  Figure 5.21(b). Experimental and Calculated Asphaltene Concentration Using the 0.05 um Membrane at Uc=6.9m/s.  0  100  200  300  Time (min)  400  500  Figure 5.22(a). Experimental and Calculated Flux vs. Time Using the 0.05 um Membrane at Uc=6.9m/s rs=0.0079 um. K* = 1.75 E-11 m/s  —i  1  200  300  1-  400  Time (min)  Figure 5.22(b). Experimental and Calculated Asphaltene Concentration Using the 0.05 um Membrane at Uc=6.9m/s. rs=0.0079 um. K*= 1.75 E-11 m/s 18 t  100  200  300  Time (min)  400  500  Figure 5.23(a). Experimental and Calculated Flux vs. Time Using the 0.05 um Membrane at Uc=6.9m/s rs=0.0153 um. K* = 1.75 E-11 m/s  •Experimental (Run 8) -Calculated  100  200  300  400  500  Time (min)  Figure 5.23(b). Experimental and Calculated Asphaltene Concentration Using the 0.05 um Membrane at Uc=6.9m/s. rs=0.0153 nm. K*=1.75 E-11 m/s 18 f  —I  100  200  300  Time (min)  400  500  126  Table 5.10. Summary of the Minimum SSQ Obtained in Determining the Best Fitted K* for the 0.02pm Pore Diameter Membrane at Different Asphaltene Radii  K*=1.0xl0 m/s  K* =1.25x10 m/s  n  n  Asphaltene  SSQ,  SSQcp  ESSQ  SSQ,  SSQcp  I SSQ  0.0026  0.316  0.370  0.686  0.258  0.341  0.599*  0.0032  0.126  0.371  0.497  0.135  0.299  0.434*  0.0047  1.028  0.496  1.524*  1.604  0.398  2.002  0.0079**  21.647  0.964  22.611*  23.982  0.904  24.886  Radius, r (pm) s  where SSQj and SSQcp is the residual sum of squares of the permeate flux and asphaltene concentration data, respectively. * The minimum residual sum of squares in determining the best fitted K* reported in Figures 5.24 to 5.27. **Can not obtain the minimum SSQ, due to the large error in fitting the permeate flux.  Figure 5.24(a). Experimental and Calculated Flux vs. Time Using the 0.02 um Membrane at Uc=6.9m/s rs=0.0026 um. K* =1 E-11 m/s  100  200  300  400  500  600  Time (min)  Figure 5.24(b). Experimental and Calculated Asphaltene Concentration Using the 0.02 um Membrane at Uc=6.9m/s. , rs=0.0026 um. K*=1 E-11 m/s 16 t  100  200  300  Time (min)  400  500  Figure 5.25(a). Experimental and Calculated Flux vs. Time Using the 0.02 um Membrane at Uc=6.9m/s rs=0.0032 um. K* =1 E-11 m/s  100  200  300  400  500  600  Time (min)  Figure 5.25(b). Experimental and Calculated Asphaltene Concentration Using the 0.02 um Membrane at Uc=6.9m/s. rs=0.0032 um. K*=1 E-11 m/s.  •Experimental (Run 9) • Calculated  100  200  300  Time (min)  400  500  Figure 5.26(a). Experimental and Calculated Flux vs. Time Using the the 0.02 um Membrane at Uc=6.9m/s rs=0.0047 um. K* =1.25 E 11 m/s :  400  T  100  200  300  400  500  600  Time (min)  Figure 5.26(b). Experimental and Calculated Asphaltene Concentration Using the 0.02 urn Membrane at Uc=6.9m/s. rs=0.0047 nm. K*=1.25 E-11 m/s  18  16 %  H—Experimental (Run 9)  14 +  ~ 12 +  — Calculated  CD  I  10 +  in  CD c C D < -* cs  n a.  <  -+-  100  200 300 Time (min)  400  500  Fiaure 5.27(a). Experimental and Calculated Flux vs. Time Using the 0.02 um Membrane at Uc=6.9m/s rs=0.0079 um. K* =1.25 E-11 m/s 400 -r •Experimental (Run 9) - Calculated  400  300  200  100  500  600  Time (min)  Figure 5.27(b). Experimental and Calculated Asphaltene Concentration Using the 0.02 um Membrane at Uc=6.9m/s. rs=0.0079 um. K*=1.25 E-11 m/s  18  16 • Experimental (Run 9)  14  c  1 2  §  1 0  o CO  D  a)  o  c  a>  5 j=  6  3  4  a.  -Calculated  I  100  1—  200  300  Time (min)  400  500  131  The estimate of K* was based on the least-squares fit in which the permeate flux and asphaltene concentration data were weighted equally. However, the magnitude of K* affects the permeate and asphaltene concentration prediction in different ways as shown in Figure 5.28. In future work, attention should be paid to estimating K* using a weighted least squares method. In this way, differences in the measurement errors of the data, particularly the initial flux data, will be accounted for.  The best fitted permeate flux and asphaltene concentration at various asphaltene diameters show some deviations from the experimental values although agreement between the measured and calculated values was qualitatively good.  The differences likely reflect the  approximation and simplifying assumptions made in this preliminary modeling study. The model is based on the assumption of first order adsorption/deposition of asphaltenes in the pores of the membrane. Although the used membranes contain carbonaceous deposits within the pores, there is no independent evidence that they deposit according to a first order process.  Furthermore, and more importantly, the model assumes an average pore size and average asphaltene size. Since flux varies as r , the presence of even a small number of large pores 4  will significantly alter thefluxcalculation (Fane and Fell, 1987). According to the model, the membrane will allow more fluid to pass through the large pores and the result is a high permeatefluxand a low asphaltene separation. As the permeateflowsthrough the pores, the asphaltenes will continuously adsorb or deposit inside the pore. Hence, the large membrane  132  pores will restrict much faster due to the large amount of deposits. Depending on the fraction of small or large pores presence, the permeate flux and asphaltenes concentration could decrease more slowly or faster than that calculated by the model.  In addition, the effect of asphaltene size on the model suggests that the asphaltene size distribution ignored in the present work, will also affect the membrane rejection coefficient. Better separation and higher permeate flux could be obtained at larger asphaltene sizes than smaller sizes. Consequently, with the asphaltene size distribution, the calculated asphaltene concentration and the pore volume loss or degree of pore restriction will be different than with the assumption of uniform asphaltene size. Hence, the amount of permeate flux through the membrane would also be different to that calculated in the present work.  Figure 5.28(a). Effect of K* on Calculated Permeate Flux vs. Time Using the 0.05um Membrane at Uc=6.9m/s. rs=0.0153um. 500  T  450 +  100  200  300  400  500  Time (min) Figure 5.28(b). Effect of K* on Calculated Asphaltene Concentration vs. Time Using the 0.05um Membrane at Uc=6.9m/s. rs=0.0153um. 20 j 18 »  0  100  200  300  Time (min)  400  500  134  CHAPTER 6: CONCLUSIONS AND RECOMMENDATIONS  6.1 CONCLUSIONS  Results from the present study indicate that undiluted heavy oil ultrafiltration using ceramic membranes is a viable and effective process to reduce the asphaltene, nickel and vanadium contents of heavy oil.  Significant asphaltene reduction was found using  membranes with nominal pore diameters of 0.1, 0.05 and 0.02um operated at high temperature (around 100-120 °C) and pressure (around 400-700 kPa). By reducing the asphaltene concentration in the heavy oil, lighter and less viscous oil was obtained. Nickel and vanadium content of the permeate was found to be linearly proportional to the asphaltene content. Permeate asphaltene content was a strong function of time-on-stream, the maximum reduction in asphaltene content occurring at the end of an experimental run. In addition, better separation was found using the smallest membrane pore diameter. For example, the 0.02um membrane gave an asphaltene reduction of 90% when operated at T=107°C, APr=600kPa, U =6.9m/s, whereas the O.lum gave a reduction of 80% at the c  same conditions.  Permeate flux dropped significantly with time in thefirsttwo to three hours of operation, followed by a gradual decrease until the flux reached a constant level after four hours. The permeate flux increased with increasing cross-flow velocity up to a cross-flow velocity of 6.9m/s. Further increases in Uc had a very small effect on the permeate flux.  135  From these data it was concluded that membrane fouling was dominated by a pore restriction mechanism at high cross-flow velocities (i.e. Uc ^ 6.9m/s). At low cross-flow velocities, fouling by the formation of a 'gel' layer likely also occurred. In addition, it was found that the permeate flux increased with increasing membrane pore diameter. For example, the permeate flux at the end of the run using the 0.02, 0.05 and 0.1pm membranes was found to be approximately 40, 48 and 55 kg/m /day, respectively. Back2  flushing the membrane with toluene has been shown to be effective for membrane regeneration.  A pore restriction model was developed to explain the permeate flux decline and asphaltene content with time and is based on fouling caused by adsorption/deposition of asphaltenes inside the membrane pores. As the pores become smaller, they prevent larger solute molecules from passing through them. The model has onefittedparameter, K*, the adsorption/deposition rate of asphaltene on the membrane pore surface. The best fitted K* was obtained by determining the minimum residual sum of squares between the calculated and measured permeate flux and asphaltene content.  Literature values of  asphaltene diameter were used to examine the effect of asphaltene size on the permeate flux and asphaltene separation.  Good agreement between the calculated and experimental asphaltene concentration was found using smaller rather than larger values of the asphaltene radius. On the other hand, a better match was found between the calculated and measured permeate flux using the  136  larger asphaltene radius than the smaller radius. For example, for the 0.1 um membrane, the best agreement between the calculated and measured asphaltene concentration was found at the assumed asphaltene size, r,=0.0047um whereas between the calculated and measured permeate flux, better agreement was found at r,=0.0153um.  In addition, K*  was also found to be a function of membrane pore diameter, smaller pores yielding a smaller K* value. Moreover, from the model prediction, the presence of larger asphaltene size would give better separation and higher permeate flux rate than smaller asphaltene size.  Two important assumptions made in the model calculations were that the asphaltene and membrane pore had a uniform size distribution. These assumptions are known not to be correct and this is thought to be the main reason for the observed differences between measured and calculated values of asphaltene content of the permeate.  6.2 RECOMMENDATIONS  Permeate rates dropped significantly at the beginning of the experiments indicating that the primary focus of future work should be on controlling or reducing membrane fouling. From the model study, increasing permeate flux and asphaltene separation can be obtained with large asphaltene size. Storm et al. (1995) reported that asphaltene molecules could agglomerate at high temperature (>150°C).  In addition, increasing the operating  temperature can significantly reduce the heavy oil viscosity.  Thus studies at high  137  temperature (>150°C) should be performed to increase both the permeability and selectively of the ultrafiltration process. Another factor that has been shown to have a significant effect on permeation is the transmembrane pressure. Increasing the pressure will increase the initial permeate flux according to Darcy's law. Thus future work should be focused on operating the system at high temperature and pressure.  Future work on the modeling study should aim to include effects of asphaltene and pore size distribution. In addition, a better datafittingtechnique using weighted least squares that accounts for the larger error associated with the initial flux data, should be investigated.  138  NOMENCLATURE  Cb  bulk concentration (kg/m )  Cp(t)  permeate concentration at time t (kg/m )  Cpi  calculated asphaltene concentration at point i in time (wt%)  Cpi"  measured asphaltene concentration at point i in time (wt%)  C  solute gel concentration on the membrane wall (kg/m )  3  3  3  g  C,(t)  solute (asphaltene) concentration inside the pore at time t (kg/m )  Cv(rc, z, t)  vanadium deposits concentration in the catalyst pore (kg/m )  Cw(t)  solute (asphaltene) concentration on the membrane wall at time t (kg/m )  D„  bulk diffusivity coefficient (m /s)  dh  equivalent hydraulic diameter which is equal to d for circular tube (m)  d (t)  membrane pore diameter at time t (m)  J (t)  average permeation flux at time t (m /m "membrane area'Vs)  JL(0  local permeate flux at time t (m /m "pores area'Vs)  Ji  calculated flux at point i in time (kg/m /day)  J"  measured flux at point i in time (kg/m /day)  k  mass transfer coefficient (m/s)  K*  rate of adsorption/deposition of solutes (asphaltenes) (m/s)  kh  intrinsic reaction rate constant per unit surface area of catalyst for  3  3  3  2  p  3  3  2  2  2  2  hydrometalation in Equation 4.3 L,  thickness of the active membrane layer (m)  n  reaction order in Equation 4.3  Nj  number of measured flux data points  Nc  number of measured asphaltene concentration data points  Pv  initial or clean pore volume (m ) 3  0  Pv (r , z, 0) c  c  pore volume per unit weight of catalyst at time t = 0 (i.e. clean catalyst) (m /kg) 3  Pd(r , z, t+At) pore diameter at time t+At (m) c  APT  driving force or the transmembrane pressure (Pa)  139  r,  solute (asphaltene) radius (m)  r  radial distancefromthe center of the catalyst (m)  c  R(t)  solute rejection coefficient at time t (dimensioniess)  R„  clean membrane resistance (m*) 1  Sac(r , z,t)  catalyst pore surface area per unit weight of catalyst (m /kg)  Sa(t)  membrane pore surface area at time t (m )  TK  system temperature in (°K)  T  experiment temperature (°C)  c  2  2  C  UbOrUc  bulk axial velocity or cross-flow velocity (m/s)  Xasphal.  concentration of asphaltenes in heavy oil (wt%)  Y>«  concentration of Nickel in heavy oil (ppm)  Y .  concentration of Vanadium in heavy oil (ppm)  Y^o  heavy oil viscosity at 40°C (mPa.s)  Y^no  heavy oil viscosity at 110°C (mPa.s)  Y  heavy oil density at 15.6°C (kg/m )  v  3  P  axial distance in the catalyst bed (m)  z  Greek Letters Pvs  density of vanadium deposits (kg/m )  p  solute (asphaltene) density (kg/m )  3  3  a  e  porosity of the membrane active layer (pores area/membrane area)  p  fluid  dynamic viscosity (Pa.s)  0(f)  fraction  of volume loss due to adsorption/deposition of solutes (asphaltenes)  in the membrane pore (dimensioniess)  Subscripts o  initial value at t=0  s  solutes (asphaltenes)  140  REFERENCES  "A 15 Year Portfolio Of Achievement," Alberta Oil Sands Technology and Research Authority (AOSTRA), 1990. "Fourth Annual Report, " Alberta Oil Sands Technology and Research Authority (AOSTRA),  March 31, 1979. Aimar, P., C. Taddei, J.P. Lafaille and V. Sanchez, "Mass Transfer Limitations During Ultrafiltration of Cheese Whey with Inorganic Membranes, " Journal of Membrane Science, 38, p.203-221, 1988.  Arod, J., B. Bartoli, P. Bergez, J. Biedermann, P. Caminade, J.M. Martinet, J. Maurin, and J. Rossarie, "Process for the Treatment of a Hydrocarbon Charge by High Temperature  Ultrafiltration, " U.S. Patent 4,411,790, 1983. Baltus, R.E. and J.L. Anderson, "Hindered Diffusion of Asphaltenes Through Microporous  Membranes, " Chemical Engineering Science, 38, p.1959-1969, 1983. Bitter, J.G.A., J.P. Haan, and H.C. Rijkens, "Solvent Recovery Using Membranes in the Lube  Oil Dewaxing Process, " The American Institute of Chemical Engineers Symposium, Series No. 272, 85, p.98-99, 1989. Bitter, J.G.A., R.H. Clark, J.L.W.C. Den Boestert and J.B. Rajani, "Process for Reducing the Metal Content of a Hydrocarbon Mixture, "U.S. Patent 5,133,851, 1992. Black, L.E., P.G. Miasek and  G. Adriaens,  "Aromatic Solvent Upgrading Using  Membranes, " U.S. Patent 4,532,029, 1985.  Booz, Allen and Hamilton, "An Assessment of International Tar Sands Recovery and Upgrading Processes," Booz-Allen & Hamilton Inc., December 1980. Chen, T.J. and J.R. Sweet, "Integrated Solvent Extraction/Membrane Extraction With Retentate Recycle for Improved Raffmate Yield " U.S. Patent 4,810,366, 1989.  Cheryan, M . , Ultrafiltration Handbook. Technomic Publishing Company, Inc., Lancaster, 1986. Dejmek, P. and J.L. Nilsson, "Flux-Based Measured of Adsorption to Ultrafiltration  Membranes," Journal of Membrane Science, 40, p. 189-197, 1989. Elmaleh, S. and W. Naceur, "Transport of Water Through an Inorganic Composite  Membrane, " Journal of Membrane Science, 66, p.227-234, 1992.  141 Fane, A G . and C.J.D. Fell, "A Review of Fouling and Fouling Control in Ultrafiltration, "  Desalination, 62, p. 117-136, 1987. Ferry, J.D., "Statistical Evaluation of Sieve Constants in Ultrafiltration," J. Gen. Physiol.,  20, p.95-104, 1936. Fried, E.R. and P.H. Trezise, Oil Security: Institution, Washington, D.C., 1993.  Retrospect and Prospect The Brookings  Goldsmith, R.L., "Special Issue on Ceramic Membranes, " Journal of Membrane Science, 39, p. 197-201, 1988. HallstrOm, B., G. Trag&rdh and J.L. Nilsson, in: W.E.L. Spiess and H. Schubert (Eds.), Engineering and Food. Vol. 3 - Advanced Processes, Elsevier Applied Science, London, p. 194-208, 1989. Hanemaajer, J.H., T.Robbertsen, Th. van den Boomgaard and J.W. Gnjruiink, "Fouling of Ultrafiltration Membranes. The Role of Protein Adsorption and Salt Precipitation," Journal  of Membrane Science, 40, p. 199-217, 1989. Hazlett, J., O. Kutowy and T.A. Tweddle, "Processing of Crude Oils With Polymeric  Ultrafiltration Membranes," AIChE Symp., Series No. 272, 85, p.101-107, 1985. Hazlett, J , O. Kutowy, T.A. Tweddle, M.D. Guiver and T.W. McCracken, "Polysulfone Membranes In Non-Aqueous Applications, " Proc. Int. Membrane Conf, NRC Ottawa, 1986. Ho, W.S.W. and K.K. Sirkar, Membrane Handbook. Van Nostrand Reinhold, New York, 1992. Howell, J. A., and 0 Velicangil, "Protein Ultrafiltration: Theory of Membrane Fouling and  Its Treatment with Immobilized Proteases, " in: A.R. Cooper (Ed.), Ultrafiltration Membranes and Applications. Plenum Press, New York, p.217-229, 1980. Hsieh, H.P., R.R. Bhave and H.L. Fleming, "Microporous Alumina Membranes," J. Membrane Sci., 39, p.221-241, 1988. Kulkarni, S.S., Y.A. Chang, J.G. Gatsis and E.E. Funk, "Membrane Separation of Hydrocarbons Using Cycloparaffinic Solvents, " U.S. Patent 4,750,990, 1988.  Kutowy, O , P. Guerin, T.A. Tweddle and J. Woods, "Use of Membranes for Oil Upgrading," in Proceeding of the 35th Canadian Chemical Engineering Conference, 1, p.2630, 1985. Kutowy, O., T.A. Tweddle, and J.D. Hazlett, "Method for the Molecular Filtration of Predominatly Aliphatic Hydrocarbon Liquids, " U.S. Patent 4,814,088, 1989.  142  Liu, J.K. and H.E. Gunning, Syncrude Analytical Methods Manual for Bitumen Upgrading, Alberta Oil Sands Technology and Research Authority (AOSTRA), Alberta, 1991. Markhasin, I.L., O.D. Svirskaya and L.N. Strads, Kolloid-Z, 31:299, 1969. Marshall, A.D., P.A. Munro and G. Tragdrdh, "The effect of protein fouling in microfiltration and ultrafiltration on Permeate Flux, Protein Retention and Selectivity: A Literature Review," Desalination, 91, p.65-108, 1993. Mehrotra, A.K. and W.Y. Svrcek, "Viscosity of Compressed Cold Lake Bitumen," The  Canadian Journal of Chemical Engineering, 65, p.672-675, 1987. Mehrotra, A.K. and W.Y. Svrcek, "Properties of Cold Lake Bitumen Saturated with Pure  Gases and Gas Mixtures," The Canadian Journal of Chemical Engineering, 66, p.656-665, 1988. Mehrotra, A.K., B.B. Nielsen and W.Y. Svrcek, "Effects of Temperature and Pressure on Asphaltene Particle Size Distributions in Crude Oils Diluted with n-Pentane, " Ind. Eng. Chem. Res., 33, p. 1324-1330, 1994. Porter, M.C., "Concentration Polarization with Membrane Ultrafritration, " Ind. Eng. Chem.  Prod. Res. Develop., Vol. 11, No. 3, 1972. Osterhuber, E.J. and N.J. Phillipeburg, "Upgrading Heavy Oils by Solvent Dissolution and  Ultrafiltration, " U. S. Patent 4,797,200,1989. Reerink, H., "Size and Shape of Asphaltene Particles in Relationship to High-Temperature  Viscosity, " Ind. Eng. Chem. Prod. Res. Develop., Vol. 12, No. 1, 1973. Reinsert, A.E. and J.I. Considine, After The Crisis: World Oil Market Projections 1991-2006. Canadian Energy Research Institute, Study No. 39, 1991. Renkin, E.M., "Filtration, Diffusion and Molecular Sieving Through Porous Cellulose  Membranes, " J. Gen. Physiol., 37, 1954. Sane, R.C., T.T. Tsotsis, I.A. Webster and V.S. Ravi-Kumar, "Studies of Asphaltene Diffusion and Their Implications for Resid Upgrading, " Chemical Engineering Science, Vol.  47, No. 9-11, p.2683-2688, 1992. "Effect of Catalyst Pore Structure on Hydrotreating of Heavy Oil, " Ind. Eng. Chem. Fundam., Vol 25, No. 3, 1986. Shimura, M . , Y. Shiroto and C. Takeuchi,  143  Sparks, B.D., J.D. Hazlett, O. Kutowy and T.A. Tweddle, "Upgrading of Solvent Extracted Athabasca Bitumen by Membrane Ultrafiltration," AIChE Symp., Vol.36, Series No. 8, p. 1279-1282, 1990. Speight, J.G., The Chemistry and Technology of Petroleum. 2nd ed., Marcel Dekker, Inc., New York, 1991. Speight, J.G., "Bitumens, Asphalts and Tar Sands," in: T.F. Yen and G. V. Chilingarian, eds.,  Developments of Petroluem Science. No. 7, Elsevier, New York, 1978. Speight, J.G. and S.E. Moschopedis, "On the Molecular Nature of Petroleum Asphaltenes,"  in: J.W. Bunger and N.C. Li (Eds.), Advances in Chemistry Series: Chemistry of Asphaltene. Series No. 195, American Chemical Society, Washington, D.C., 1981. Storm, D.A., R.J. Barresi and E.Y. Sheu, "Rheological Study ofRatawi Vacuum Residue in  the 298-673K Temperature Range," Energy and Fuels, 9, p. 168-176, 1995. Trambouze, P., J.P. Euzen, P. Bergez and M . Claveau, "Process for Deasphalting Hydrocarbon Oil, " U.S. Patent 4,816,410, 1989. Zeman, Leos J., "Adsorption Effects in Rejection of Macromolecules by Ultrafiltration  Membranes," Journal of Membrane Science, 15, p213-230, 1983.  APPENDIX A: 1. Solvent Deasphaltene Process  145  2. Fluid Coking Process  146  3. Partial Oxidation Process  147  145  1. Solvent Deasphaltene Process: Table A.l. Products Disposition Options and the Primary Advantages and Disadvantages of Solvent Deasphalting Process. PRODUCT  PRODUCT DISPOSITION  ADVANTAGE  DISADVANTAGE  Deasphalted Oil (DAO)  • FCC or Hydrocracking Feed • Asphalt Blending • Fuel • Coker Feed • Gasification  • High liquid yield • Separation metals from DAO • Simple  • Residue disposition required • DAO below pipeline quality, requires additional processing • High fuel cost • Not proven on bitumen  Asphaltenes  Figure A.l. A Schematic Diagram of A Solvent Deasphalting Process (Booz, 1980)  Feed  Asphalt and Resin  146  2. Fluid Coking Process:  Table A.2. Products Disposition Options and the Primary Advantages and Disadvantages of Coking Process. PRODUCT  PRODUCT DISPOSITION  ADVANTAGE  DISADVANTAGE  Light Ends  • Process Fuel or LPG Recovery • Processed to finished products • Fuel • Gasification  • Valuable fuels • Proven reliability  • Less valuable than liquid products • Coke high in impurities and difficult to use • Complex coke removal • Fuel gas desulfurization needed  Distillates Coke  Figure A.2. A Schematic Diagram of A Fluid Coking Process (Booz, 1980) Gn ScrubOr/ Fraciionator  Combuttion Gases Recycle  10S0* Raid Feed  Elutriator  Slum  Air Blown  Reactor  Burner  147  3. Partial Oxidation:  Table A.3. Products Disposition Options and the Primary Advantages and Disadvantages of Partial Oxidation PRODUCT  PRODUCT DISPOSITION  ADVANTAGE  DISADVANTAGE  Light Ends  N. A.  • Proven technology • Suitable for large scale gas product needs • Full range feedstock flexibility • No other upgrading needed • No residue by-product  • No liquid yield • Oxygen plant required • Excessive energy imbalance may force uneconomic export of gas product • H2S and CO2 removal needed  Distillates Coke  Figure A.3. A Schematic Diagram of Texaco Partial Oxidation Process (Booz, 1980)  Oil and Soot  CO2 to Vent HjS To Claus Unit  Oxygen  1  Steam  1 Waste  Syn Gas Generator  Heat Boiler  Soot Removal &  &  co  Gas  I Ash Removal  T  Ash  Residual Oil — Boiler Feed Water Make-Up"  Dry Gas  Hot Gas Water Treatment  2  Removal  Cooling  J  Fuel Gas  148  APPENDIX B:  A SEM Micrograph of the Cross-Sectional Area of a Commercial Membralox® Ceramic Membrane  150  APPENDIX C: 1. Description of Computer Hardware and Software Used  151  2. Description of Membrane Installation and Removal  151  3. Description of Pre-running Procedure  153  .  4. Description of Operating Procedure  153  5. Description of System Shut Down and Cleaning  155  6. Description of System Trouble Shooting  156  151  1. Computer Hardware and Software:  A computer was used to perform the data acquisition and control functions of the membrane testing unit. The software package used for this application was Labtech Notebook. It has two main parts, a build-time that allows the user to setup the system and a run-time that performs the operating functions, data-logging and data display.  In the current set up,  Labtech was used to control the four heating tapes and band heaters.  Data Acquisition Card: PC-LabCard, Model PCL-718. This card performs the data acquisition functions that are displayed by Notebook Amplifier/Multiplexer Board: PC LabCard Model PCLD-7898. This board allows for the multiplexing and signal conditioning of 16 differential inputs into the analog input channels of the PCL-718. Up to sixteen thermocouple measurements can be performed using this board.  2. Membrane Installation and Removal:  Membrane Installation 1. Lubricate the inside of each end of the membrane housing unit with silicone vacuum grease. 2. Place a new O-ring (size 114) onto the ferrule at one end of the membrane tube, very slightly lubricate the O-ring and slide the end without the O-ring into the housing first.  152  3. When the O-ring contacts the end of the module, slowly and gently push the end of the element with the O-ring inside the housing to a point that the ferrule on the other end is exposed so that the second O-ring can be placed on. Caution: Use slow, even and direct force only; a sudden shock or forcing of the tube at an angle may crack the ceramic element. 4. Place the second O-ring on the exposed ferrule, lightly lubricate the O-ring, and slowly push the element back into the housing. Then center the element within the housing. 5. Place one O-ring on each end of the element to avoid direct ceramic to metal contact in case the membrane shifts up or down during the experiment.  Membrane Removal 1. Use a wooden/plastic dowel and slowly push on one end of the ceramic element until its O-ring is exposed at the other end. Note: If having difficulty with one end, try pushing at the opposite end. Go back and forth from end to end until the element begins to move. 2. Remove the O-ring from the exposed ferrule, then push this ferrule back slowly and directly through the housing until the O-ring of the other ferrule is exposed. 3. Remove the second O-ring, and slide the membrane out of the housing. 4. Place the element into a cylinder and soak it in toluene replacing the toluene at least once in 24 hours to remove the heavy oilfromthe membrane.  153  3. Pre-running Procedure:  1. Install membrane housing unit in the system and ensure that allfittingsare tightened. 2. Wrap heating tapes around all piping tightly with insulation. Ensure that the retentate and permeate lines have flat thermocouples installed to control the heating tape. 3. Ensure fume hood is operating. Remove all combustible material from inside the fume hood. 4. Ensure all accumulation of solids deposited on the in-line Tee filter from the previous run have been removed. 5. Check the poppet seal in the pressure relief valve to ensure that it has not been damaged by previous operation of the unit. The high temperature oil eventually damages the seal and proper control of the pressure in the system cannot be achieved. 6. Close all sample valves and check that all three-way and on/off valves are in the correct position. 7. Check the feed tankfilledwith oil to about 2/3 full (to avoid oil bubbling up and flowing over the side of the tank while heating up).  4. Operating Procedure:  1. Open the build time Notebook and set the operating temperature for the heating bands and heating tapes.  154  2. Fill the feed tank with oil to about 2/3 full. Heat the oil of the feed tank but avoid overheating which may cause boiling of the entrained water (often present in the form of a water-oil emulsion) or low-boiling hydrocarbons contained in the feed oil. As a result, vapour in the tank will force the viscous oil upward and result in uncontrollable spillage through the top. Turn off the power and open the drain valve located at the bottom of the feed tank if this occurs. 3. Fully open the sample valve, VI9, to avoid pressure build up in the system while heating the air inside the empty tube using the heating tapes. 4. As the oil in the feed tank approaches the predetermined value (90 - 100°C), switch the 3way valve, VI, located directly underneath the feed tank to allow free flow of heated oil toward the designated feed pump. 5. Turn on the feed pump, PI, increase speed slowly and allowfluidto fill the whole system. The sample valve, VI9, can be used to bleed air out of the system during filling as well as to observe the oil level in the system. 6. When the system is full of liquid, manually open the pressure relief valve, VI3, to the desired cracking pressure and completely close the sample valve, VI9. 7. Set the feed rate and turn on the recycle pump; adjust the speed slowly to the predetermined rate. Adjust the pressure relief valve to the desired pressure by monitoring the membrane inlet pressure, PT1, inside the unit through the control panel as well as in the computer display. As the system reaches its final temperature, this valve will have to be adjusted as the adjusting spring inside the valve heats up.  155  5. Shutdown and Cleaning:  Extreme care should be exercised when shutting down and cleaning the system. The points to remember are to ensure that the heating tapes are all off upon shutdown and that the system has cooled sufficiently. 1. Turn off the pumps, all the heating sources as well as the main power switch and exit from Labtech. This step ensures that there will be no overheating of any components of the system and none of the equipment can be operated. 2. Immediately begin draining the oil from the system using the drain valve near the feed pump. Switch the 3-way valve, VI, to stopflowof oil toward the feed pump. 3. Drain the oil from the permeate line using the valve on the lower port of the membrane module. 4. Turn the recycle pump in reverse at a very low flow to drain any oil remaining in the line. When as much oil as possible has been drained from the system, the system must be allowed to cool sufficiently (<40°C) before any solvent can be introduced into the system for cleaning. 5. To clean the system, fill the recycle and feed lines with solvent and slowly circulate the solvent. The key is to leave the solvent in the system long enough to dissolve the oil which has remainedfromdraining the system. The pressure relief valve can be adjusted to increase the pressure in the system and force solvent into the permeate line to rinse this section. Drain the solvent/oil mixture and rinse twice more. 6. Drain solvent/oil mixture as much as possible by repeating steps 3 and 4.  156 7. Undo the insulation around the heating tapes and then remove the membrane module. 8. Carefully remove the membrane by following the membrane removal procedure described above. It is easiest to clean the membrane if the oil is not allowed to remain on the membrane after the experiment (i.e. remove the membrane from the module and soak in solvent as soon as possible after the experiment is over). 9. Clean the inside of the membrane module with toluene. 10. Remove and clean the on-line Teefilter,F l .  6. Troubleshooting:  By properly preparing the equipment before each experiment, potential problems that may arise can be minimized. Some of the operating conditions that were monitored to alert the operator of experimental problems are discussed below.  During Start-up of Experimental Run 1. No pressure in the system. The system should fill in less than 10 minutesfromthe start of the feed pump at 15% of D C input power. If there is no pressure or very low pressure observed compared to what should be observed in a normal run, stop the pump and check the on-line filter (Fl). Large particles and cake deposits on thefiltermight clog the line. This may cause serious damage to the feed pump due to pressure build up. 2. Abnormal temperature, pressure and/or flow rate readings. blown fuses,  unplugged heating-tape  connectors  This could be caused by  and/or electrical interference.  157 Connectors should be tightly plugged together. Check fuses and replace with new fuses as required.  Frequently, some electrical interference problems may occur due to poor  contact at the joint, turn off and on the main power switch.  During Experimental Run 1. Pressure loss. System leakages would result in pressure drop. If the feed pump has stopped operating or cannot generate enough pressure to keep the pressure constant in the recycle line, the temperature of the liquid in the retentate line will begin to drop as liquid will cease to flow through the relief valve. On the other hand, if the seal in the relief pressure is broken, the pressure will drop immediately. Stop pumping, dissemble relief valve from the system and replace seal kit. The other possibility is that the membrane itself or the O-rings securing the membrane have failed.  In this case, the permeate  flowmeter will show a dramatic increase in flowrate. Turn off the pumps immediately. 2. As the system is operated at high temperature, any piece of equipment containing a seal may require tightening of the screws/bolts around the seal. For some seals, preheating the component after installing a new seal will allow the seal to be properly seated and tightened.  APPENDIX D:  1. Derivation of Equation 4.1 2. Derivation of Equation 4.17 3. Sample Calculations and Model Data  159  1. Derivation of E q u a t i o n 4.1  Assuming straight cylindrical and uniform pores, and laminar flow through the pores, the initialflux,Jo, can be expressed by the Hagen-Poiseuille equation: Jn = T I T / APT . N  (D.l)  8pL,. A« where, APT  = the transmembrane pressure differential (Pa)  u  = thefluidviscosity (Kg/m.s)  L,  = the pore length (m)  Tpo  = the initial average pore radius (m)  N  = the number of pores  Ao  = the membrane area (m )  J  = the initial average permeateflux(volumetric rate/total membrane  0  2  area) (m /m /s) 3  2  The average permeate flux decline with time due to the adsorption/deposition of solutes (asphaltenes) in the membrane pore can be determined as follows,  J (t) = 7ir (f) AP . N  (D.2)  4  P  T  8p.L,. A<, where, r (t)  = the membrane pore radius at time t (m)  J (t)  = the average permeatefluxat time t (m /m /s)  p  3  2  By dividing Equation D.2 into Equation D. 1 and rearranging, the permeatefluxat time t can be expressed as follows, J (t) = J r (t) n  r  4 r 1  po  4  p.3)  160  2. Derivation of Equation 4.17  Assuming uniform pore distribution, the clean membrane pore area can be expressed as A p o ^ r ^ N . Thus, the clean membrane porosity, eo, is the ratio of the clean membrane pore area, Apo to the total membrane area, A, or e „ = TCrpn . N  (D.4)  2  A»  For a fouled membrane due to pore restriction, the porosity at time t can be expressed as: e(t) = 7tr (f) .N 2  P  (D.5)  A,  Dividing Equation D.5 into D.4, we have,  e(t)= eojr^trl 2  (D.6)  r  * po  The local permeateflux(volumetric rate per membrane pore area) at time t is equal to the average permeatefluxat time t divided by the membrane porosity at time t, or  J (t)= J(t)/e(t) L  (D.7)  Substituting Equations D.3 and D.6 into Equation D.7 and rearranging, the local permeate flux at time t can be expressed as follows, JL(0=  (D.8)  161 3. Sample Calculations * Using Figure 5.13: T = 383.15°K AP = 586 kPa U = 6.9m/s C = 19wt% dpo = 0.1pm 8 = 0.4 1.0057xlO in d« = 0.0052pm p. = 1100 kg/m D „ = 1.61xl0- m /s K* = l'.75xl0* m/s T  c  b  0  I2  1  3  10  2  u  1. Initial permeate flux, J„ (Eq. 5.5): J = AP/(R„p) = 8.36x10^ m/s = 722 L/m /day. 2  0  * ConvertfromL/m /day to kg/m /day 2  2  J = 722 x p/1000 = 687 kg/m /day. 2  0  where, p, = 69.8 mPa.s (Eq. 5.6); p = 952 kg/m (Eq. 5.7) 3  2. Initial asphaltene rejection coefficient (Eq. 4.15): Ro= 1- [2(1-X0 - (l-^) ] [1 -2.10471c +2.09A* - 0.95^] = 0.1183. 2  4  3  where, A* = oVdpo = 0.0052/0.1 = 0.052. 3. Mass Transfer Coefficient, k (Eq. 4.19): k = 1.62 [ U D _ ] dh Lt 2  £  where, d = 0.01m; Lt = 0.2m h  2  1 / 3  = 7.8x10^ m/s.  162  4. Initial asphaltene concentration at the membrane wall (Eq. 4.18): Cwo =  q expfJrVkl Ro + [1 - RolexptWk]  =203.21 kg/m . 3  where, Ju,= Ve . = 2.158xl0" m/s s  0  5. Initial permeate asphaltene concentration, Cpo (Eq. 4.10): Cpo = Cwo (1-R) = 179.2 kg/m  3  * Convert from kg/m to wt%: 3  Cpo = 179.2 /(p x 1000) = 0.1882 = 18.82 wt% 6. Pore volume loss in thefirst60 seconds (Eq. 4.5): 0i (60) = K* Sam C (t) At = 0.00684 p. Pv t  0  where, Sa(0) = TrdpoL. = 3.14 x lxlf/ X . = 3.14x10"* X . m ; C.(0) = Cp, = 179.2 kg/m ; Pv = 7idpo L, /4 = 3.14 x (lxlO" ) X , /4 = 7.85 xlO- .L, (m ). 6  2  6  2  2  3  13  3  0  7. Membrane pore restricted in thefirst60 seconds (Eq. 4.7): d (60) = dpofl - 6(60)]  I/2  p  =0.0997um.  8. Permeate Flux in thefirst60 seconds (Eq. 4.9): J(60) = J * dp (60) / dpo = 711.8 L/m /day = 677.4 kg/m /day. 4  4  2  2  0  9. Rejection coefficient, R, after 60 seconds (same as step two except with new d ): p  R(60) = 1 - [2(l->.) - (l-X) ] [1 - 2.104X + 2.09X - 0.95X ] = 0.1187. 2  4  3  where, X = cL/d (60) = 0.0052/0.0997 = 0.0522. p  10. Local permeatefluxafter 60 seconds (Eq. 4.17): h(t) = (Jo/6 ) * (d (t)/dp„) = 2.145 xlO" m/s. 2  0  p  5  5  163  11. Asphaltene concentration at the membrane wall after 60 seconds (Eq. 4.18): CU.60) =  C exp[J (60)/k1 R(60) + [1 - R(60)]exp[J (60)/k] h  =203.31 kg/m . 3  L  L  12. Permeate asphaltene concentration, Cp(t) after 60 seconds (Eq. 4.10): C (60) = Cw(60) (1-R(60)) = 179.1 kg/m  3  p  * Convertfromkg/m to wt%: 3  C (60) = 179.1 /(p x 1000) = 0.1881 = 18.81 wt% p  13. Pore volume loss in the additional second 60 seconds (Eq. 4.5): 0 (6O)= K* Sam C ft) At p. Pv 2  e  = 0.00682  0  where, Sa(60) = 7tdp(60)L = 3.14 x 9.97xl0" X . = 3.13X10" .L. m ; C.(60) = C (60) = 179.1kg/m ; Pv = %A^hJA = 7.85 xlO* .L.(m ). 7  6  2  a  p  3  13  3  c  14. Pore volume loss in thefirst120 seconds (Eq. 4.6): 0 (120) = 0 (60) + 0 (60) = 0.01366 2  15. Membrane pore restriction in 120 seconds (Eq. 4.7): d (120) = p  dpo[l  - 0(12O)]  1/2  = 0.0993pm.  16. Permeate Flux in thefirst60 seconds (Eq. 4.9): J(120) = J * dp (120) / 4  0  dpo  4  = 702.0 L/m /day = 668.2 kg/m /day.  17. Repeat steps 9 to 17 until the end of the experiment.  Complete calculated data can be seen in the following:  2  2  9.97E-08 9.93E-08 9.90E-08 9.86E-08 9.83E-08 9.79E-08 9.62E-08 9.45E-08 9.28E-08 9.11 E-08 8.94E-08 8.77E-08 8.60E-08 8.43E-08 8.26E-08 8.10E-08 7.93E-08 7.76E-08 7.60E-08 7.43E-08 7.26E-08 7.10E-08 6.93E-08 6.77E-08  1E-14 9.93E-15 9.86E-15 9.8E-15 9.73E-15 9.66E-15 9.33E-15 9E-15 8.68E-15 8.36E-15 8.06E-15 7.75E-15 7.46E-15 7.17E-15 6.89E-15 6.61 E-15 6.34E-15 6.08E-15 5.82E-15 5.57E-15 5.32E-15 5.09E-15 4.85E-15 4.63E-15  CM  —»  CO  < —» E  +  co <  E  2  17 686.81 677.45 668.18 659.01 649.94 640.96 597.49 556.31 517.35 480.53 445.77 412.98 382.09 353.03 325.72 300.09 276.07 253.59 232.57 212.96 194.68 177.67 161.88 147.22  <  721.63 711.79 702.05 692.41 682.88 673.45 627.77 584.51 543.58 504.89 468.36 433.91 401.46 370.93 342.23 315.30 290.06 266.44 244.36 223.75 204.55 186.68 170.08 154.69  Cp(t)  179.18 179.14 179.10 179.07 179.03 178.99 178.77 178.54 178.28 178.00 177.69 177.35 176.98 176.57 176.14 175.66 175.14 174.58 173.97 173.32 172.61 171.85 171.03 170.16  < E  0.006841 0.013658 0.02045 0.027216 0.033958 0.040675 0.073886 0.106472 0.138431 0.169761 0.200461 0.230529 0.259962 0.28876 0.31692 0.344441 0.371321 0.397558 0.423152 0.448101 0.472404 0.496061 0.51907 0.541432  kg/m2/day  co"  1 .OOE-07 0.006841 9.97E-08 0.006817 9.93E-08 0.006792 9.90E-08 0.006767 9.86E-08 0.006742 9.83E-08 0.006717 9.66E-08 0.006592 9.49E-08 0.006467 9.32E-08 0.006342 9.15E-08 0.006216 8.98E-08 0.006089 8.81 E-08 0.005963 8.64E-08 0.005836 8.47E-08 0.005709 8.30E-08 0.005581 8.13E-08 0.005453 7.96E-08 0.005325 7.80E-08 0.005196 7.63E-08 0.005067 7.46E-08 0.004938 7.30E-08 0.004809 0.00468 7.13E-08 0.00455 6.97E-08 6.80E-08 0.004421  d(t+1)  |  203.21 203.27 203.32 203.38 203.44 203.49 203.78 204.05 204.33 204.59 204.85 205.10 205.35 205.58 205.80 206.00 206.19 206.37 206.53 206.67 206.79 206.88 206.96 207.01  s (1e6m)  Volume fract. Loss  o  0.1183 0.1187 0.1191 0.1196 0.1200 0.1204 0.1227 0.1250 0.1275 0.1300 0.1326 0.1353 0.1381 0.1411 0.1441 0.1473 0.1506 0.1540 0.1576 0.1614 0.1653 0.1693 0.1736 0.1780  Cwft)  &  r-  18.82 18.82 18.81 18.81 18.81 18.78 18.76 18.73 18.70 18.67 18.63 18.59 18.551 18.51 18.46 18.401 18.34 18.281 18.21 18.14 18.061 17.97 17.881  CO  o o o o o o o o  co  CN  00  CM  o CO  o CO  o  CD  1200 1500 1800 2100 2400 2700 3000 3300 3600 3900 4200 4500 4800 5100 5400 5700  Time  Model Data : Fiqures 5.13(a) and I  co oo  "CT  •o  >.  CO  < E  +  co  < E oo o • LU  00 LO  0006  0066 0096  V) ^>  133.66 4.41 E-15 121.12 4.2E-15 109.56 3.99E-15 98.91 3.79E-15 89.12 3.6E-15 80.14 3.42E-15 71.92 3.24E-15 64.42 3.06E-15 57.57 2.9E-15 51.35 2.73E-15 45.70 2.58E-15 40.59 2.43E-15 35.97 2.29E-15 31.81 2.15E-15 28.07 2.02E-15 24.71 1.9E-15 21.71 1.78E-15 19.03 1.66E-15 16.65 1.56E-15 14.54 1.45E-15 12.67 1.36E-15 11.02 1.27E-15 9.56 1.18E-15 8.29 1.1 E-15 7.17 | 1.02E-15  CM  140.43 127.26 115.11 103.92 93.64 84.20 75.57 67.68 60.49 53.95 48.02 42.65 37.79 33.42 29.49 25.96 22.81 20.00 17.50 15.27 13.31 11.57 10.05 8.71 7.53  <  5.69E-08 5.53E-08 5.38E-08 5.23E-08 5.08E-08 4.93E-08 4.78E-08 4.64E-08 4.50E-08 4.36E-08 4.22E-08 4.08E-08 3.95E-08 3.81 E-08 3.69E-08 3.56E-08 3.43E-08 3.31 E-08 3.20E-08  6.61 E-08 6.45E-08 6.29E-08 6.13E-08 5.97E-08 5.81 E-08 5.66E-08 5.50E-08 5.35E-08 5.20E-08 5.05E-08 4.90E-08 4.75E-08 4.61 E-08 4.47E-08 4.33E-08 4.19E-08 4.05E-08 3.92E-08 3.79E-08 3.66E-08 3.53E-08 3.41 E-08 3.29E-08 3.17E-08  169.22 168.21 167.14 166.00 164.78 163.49 162.11 160.66 159.12 157.49 155.77 153.97 152.06 150.06 147.96 145.76 143.45 141.04 138.52 135.90 133.17 130.33 127.38 124.32 121.16 I  1  0.563148 0.584217 0.60464 0.62442 0.643557 0.662054 0.679914 0.69714 0.713736 0.729707 0.745058 0.759795 0.773924 0.787452 0.800388 0.812738 0.824514 0.835724 0.846378 0.856489 0.866068 0.875127 0.883679 0.891739 0.89932  kg/m2/day  <  10200 10500 10800 11100 11400 11700 12000 12300 12600 12900 13200  •a  0.004291 0.004162 0.004033 0.003904 0.003776 0.003648 0.003521 0.003395 0.003269 0.003144 0.003021 0.002899 0.002777 0.002658 0.00254 0.002424 0.00231 0.002197 0.002087 0.001979 0.001874 0.001771 0.001671 0.001573 0.001479  J(t) CO**  6.64E-08 6.48E-08 6.32E-08 6.16E-08 6.00E-08  d(t+1)  © E  207.03 207.03 207.01 206.95 206.87 206.75 206.61 206.44 206.24 206.01 205.75 205.46 205.14 204.79 204.42 204.03 203.61 203.16 202.70 202.22 201.72 201.21 200.68 200.14 199.59  Volume fract. Loss  ->  (1e6m)  CL  12.73|  17.78 17.67 17.56 17.44 17.31 17.18 17.03 16.88 16.72 16.55 16.37 16.18 15.98 15.77 15.55 15.31 15.07 14.82 14.55 14.28 13.991 13.69 13.38 13.061  s  9300  0.1827 0.1875 0.1926 0.1979 0.2034 0.2093 0.2154 0.2218 0.2285 0.2355 0.2429 0.2506 0.2587 0.2673 0.2762 0.2856 0.2955 0.3058 0.3166 0.3280 0.3398 0.3523 0.3653 0.3788 0.3930  Cw(t)  o  6000 6300 6600 6900 7200 7500 7800 8100 8400 8700  Time  Model Data: Continue (2).... O  TJ  +  3.06E-08 2.95E-08 2.84E-08 2.74E-08 2.63E-08 2.54E-08 2.44E-08 2.35E-08 2.27E-08 2.18E-08 2.10E-08 2.03E-08 1.96E-08 1.89E-08 1.82E-08 1.76E-08 1.71 E-08 1.65E-08 1.60E-08 1.55E-08 1.51 E-08 1.47E-08 1.43E-08  +  CO  < E CO OJ CO CO co CD CO •«* CN q  oo  LU cn cq  o i  1.35E-08 1.32E-08  CO  0.90 0.78 0.68 0.60 0.52 0.46 0.41 0.36 0.32 0.29 0.26 0.24 0.21  o> o oo CO co CN q  0.95 0.82 0.72 0.63 0.55 0.49 0.43 0.38 0.34 0.30 0.27 0.25 0.22  TJ  9.5E-16 8.82E-16 8.19E-16 7.6E-16 7.05E-16 6.54E-16 6.06E-16 5.62E-16 5.22E-16 4.84E-16 4.5E-16 4.18E-16 3.89E-16 3.62E-16 3.38E-16 3.15E-16 2.95E-16 2.76E-16 2.59E-16 2.44E-16 2.3E-16 2.17E-16 2.05E-16 1.95E-16 1.85E-16 1.76E-16  CN  6.19 5.34 4.60 3.96 3.41 2.93 2.52 2.17  <  6.51 5.61 4.84 4.16 3.58 3.08 2.65 2.28  117.89 114.53 111.07 107.53 103.90 100.20 96.44 92.63 88.79 84.93 81.07 77.22 73.40 69.63 65.92 62.30 58.78 55.37 52.09 48.94 45.93 43.08 40.37 37.82 35.42 33.16  s  0.906437 0.913106 0.919342 0.925162 0.930583 0.935622 0.940297 0.944624 0.948623 0.95231 0.955704 0.958823 0.961685 0.964306 0.966703 0.968894 0.970894 0.972718 0.974381 0.975896 0.977276 0.978534 0.979681 0.980726 0.981681 0.982552  kg/m2/day  O  3.08E-08 0.001387 2.97E-08 0.001299 2.86E-08 0.001213 2.76E-08 0.001132 2.65E-08 0.001053 2.56E-08 0.000978 2.46E-08 0.000907 2.37E-08 0.000839 2.28E-08 0.000774 2.20E-08 0.000713 2.12E-08 0.000656 2.04E-08 0.000603 1.97E-08 0.000552 1.90E-08 0.000506 1.84E-08 0.000462 1.78E-08 0.000422 1.72E-08 0.000385 1.66E-08 0.000351 1.61 E-08 0.00032 1.56E-08 0.000292 1.52E-08 0.000266 1.47E-08 0.000242 1.43E-08 0.000221 1.40E-08 0.000202 1.36E-08 0.000184 1.33E-08 0.000168  Volume fract. Loss  CL < E  199.04 198.48 197.92 197.36 196.80 196.24 195.69 195.15 194.62 194.09 193.59 193.09 192.61 192.14 191.70 191.27 190.86 190.46 190.09 189.73 189.39 189.07 188.77 188.48 188.21 187.96  TJ  (1e6m)  CO*  3  00081  18300 18600 18900 19200 19500 19800 20100 20400 20700 21000  Cw(t)  3.481  12.39 12.03 11.67 11.30 10.92 10.53 10.13 9.73 9.33 8.92 8.52 8.11 7.71 7.32 6.93 6.55 6.18 5.82 5.47 5.14 4.831 4.53 4.24 3.97 3.72  c  0.4077 0.4230 0.4388 0.4552 0.4721 0.4894 0.5072 0.5253 0.5438 0.5624 0.5812 0.6001 0.6189 0.6376 0.6561 0.6743 0.6920 0.7093 0.7260 0.7421 0.7575 0.7722 0.7861 0.7994 0.8118 0.8236  Rft)  3  13500 13800 14100 14400 14700 15000 15300 15600 15900 16200 16500 16800 17100 17400 17700  Time  Model Data: Continue (3).... O  —>  < E !3  +  •a  E H 0.983349 0.984078 0.984746 0.98536 0.985924 0.986443 0.986922 0.987364 0.987773 0.988153 0.988505 0.988833 0.989138 0.989423 0.989689 0.989939 0.990173 0.990393 0.9906 0.990795  1.29E-08 1.26E-08 1.24E-08 1.21 E-08 1.19E-08 1.16E-08 1.14E-08 1.12E-08 1.11 E-08 1.09E-08 1.07E-08 1.06E-08 1.04E-08 1.03E-08 1.02E-08 1.OOE-08 9.91 E-09 9.80E-09 9.70E-09 9.59E-09  >»  CO  0.20 0.19 0.17 0.16 0.15 0.13 0.13 0.12 0.11 0.10 0.10 0.09 0.09 0.08 0.08 0.07 0.07 0.07 0.06 0.06  CM  <  1.68E-16 1.61 E-16 1.54E-16 1.48E-16 1.42E-16 1.37E-16 1.32E-16 1.27E-16 1.23E-16 1.19E-16 1.16E-16 1.12E-16 1.09E-16 1.06E-16 1.04E-16 1.01 E-16 9.87E-17 9.65E-17 9.44E-17 9.24E-17 |  •a  0.19 0.18 0.16 0.15 0.14 0.13 0.12 0.11 0.10 0.10 0.09 0.09 0.08 0.08 0.07 0.07 0.07 0.06 0.06 0.06 I  A  (kg/m 3)  O to  p  I—  ro  CO  .< -E  or  o a  ai  o o  CO  CN CN  *  CO  21900 22200 22500 22800 23100 23400 23700 24000 24300 24600 24900 25200 25500 25800 26100 26400 26700 27000  0.000154 0.000141 0.000129 0.000119 0.000109 0.0001 9.27E-05 8.57E-05 7.94E-05 7.36E-05 6.84E-05 6.37E-05 5.94E-05 5.55E-05 5.19E-05 4.86E-05 4.56E-05 4.29E-05 4.04E-05 3.81 E-05  J(t) kg/m2/day  CL  1.30E-08 1.27E-08 1.24E-08 1.21 E-08 1.19E-08 1.17E-08 1.15E-08 1.13E-08 1.11 E-08 1.09E-08 1.08E-08 1.06E-08 1.05E-08 1.03E-08 1.02E-08 1.01 E-08 9.94E-09 9.82E-09 9.72E-09 9.61 E-09  d(t+1)  o  (1e6m)  Volume fract. Loss  5  187.72 187.50 187.29 187.09 186.90 186.73 186.56 186.41 186.26 186.12 185.99 185.87 185.76 185.65 185.54 185.45 185.35 185.26 185.18 185.10  1 Q.  0.8346 0.8448 0.8544 0.8634 0.8717 0.8794 0.8866 0.8932 0.8994 0.9051 0.9104 0.9153 0.9199 0.9241 0.9281 0.9317 0.9351 0.9382 0.9412 0.9439  cwm  Model Data: Continue (4)....  13  •M Oi O EO 3> N C CO CM o 3 DO O I CM < n CM CM CM rd CO CM CM CM  CO CO CO CO CM CO O CM in o> m O) in CO CM CM CM CM CM CM  ai  %  o >* 0> rCO  CO  m n CO co o co to in «t * CO  17.66]  ra>  co <u-  CO CO  in  14.88 14.09 13.35 12.67 12.03 11.44 10.89 10.38  to  CO o CM CM  p  < —>E  >  Model Data: Continue (5)... ds=0.0026jm dp=0.1um Determining the Best"rtted K* K* 5.00E-11 2.50E-11 2.25E-11  2.00E-11 1.75E-11  1.50E-11 1.25E-11  SSQ_J  SSQ_Cp SSQ  0.7102 0.5907 0.5746 0.5541 0.5314 0.5240 0.5951  0.4330 0.1529 0.1078 0.0704 0.0891 0.2427 0.6051  1.1432 0.7436 0.6824 0.6244 0.6204* 0.7666 1.2003  Model Data: Continue (6 K*=1.75E-11 m/s  Ji  JO)  TIME  Exp. kg/m2/day  (min) 0 4 11 20 26 32 47 66 90 93 105 109 110 139 140 155 156 200 201 260 262 280 299 300 329 330 397 400 419 420 450  668 383.2 301.1 251.3 219.8 222.8  187.5  57.9 49 52.6  57.2 55.8  46.2  Cal. kg/m2/day 686.81 649.94 589.07 517.35 473.42 432.42 341.90 249.27 161.88 152.95 121.12 111.79 109.56 58.89 57.57 40.59 39.63 12.67 12.32 2.17 2.04 1.20 0.70 0.68 0.33 0.32 0.10 0.10 0.08 0.08 0.06  Cp Exp. wt%  Cp(t) Cal. wt% 18.83 18.81 18.78 18.73 18.70 18.66 18.53 18.33 17.97 17.92 17.67 17.58 17.56 16.75 16.72 16.18 16.14 13.99 13.93 9.73 9.57 8.11 6.62 6.55 4.58 4.53 1.92 1.86 1.50 1.48 1.09  SSQj  SSQcp  0)  19 18.9  8.374E-05 0.0007312 0.2886272 0.5158349  18.9 18.2 18.8  4.122E-05 0.0001169  0.5194477 0.3085977 0.0141138  17.3  0.0003378 0.0006237 0.001267  0.1630232  16.4  0.0003771 0.0995905  10.6  0.560362  0.0988625  0.9237999 0.3323437  4.2 0.9884582 0.9963702 3.8  0.367564 0.9974615  Sum= sum/(n=1)  6.376418 0.5313682  0.8016176 0.0890686  170  APPENDIX E: Experiments Data 1. Run 1  171  2. Run 2  173  3. Run 3  .175  4. Run 4  177  5. Run 5  179  6. Run 6  182  7. Run 7  185  8. Run 8  188  9. Run 9  191  10. Run 10  193  11. Nickel and Vanadium Versus Asphaltene Contents  195  12. Heavy Oil Viscosity Versus Asphaltene Contents  196  13. Heavy Oil Density Versus Asphaltene Contents  196  1.  Run  1.  •  Heavy Oil Feed: Undiluted  •  Feed Asphaltene Content: 15.4 wt%  •  Cross-flow Velocity: 1.51m/s  •  Membrane Surface Area: 0.005 m  •  Membrane Nominal Pore Diameter: 0.1 um  2  Table E.l.a: Permeate Flux  Time  Permeate Flux  (min)  (Kg/m /day)  (0 - 220)  45.0  (220 - 340)  27.0  2  Table E.l.b: Asphaltene Content in the Permeate  Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0 - 180)  13.0  16  (180 - 340)  3.7  76  172  2.  Run 2.  •  Heavy Oil Feed: Undiluted  •  Feed Asphaltene Content: 17.1 wt%  •  Cross-flow Velocity: 2.26 m/s  ••  Membrane Surface Area: 0.005 m  •  Membrane Nominal Fore Diameter: 0.1 um  2  Table E.2.a: Permeate Flux Time  Permeate Flux  (min)  (Kg/m /day)  (0 - 330)  39.0  (330 - 395)  20.7  2  Table E.2.b: Asphaltene Content in the Permeate Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0 - 330)  8.2  52  (330 - 395)  4.5  74  Run 2. Operation Conditions 120  T  100  80  60  40 - Average Temp. = 100C 20 4-  —I  50  150  100  300  250  200  350  Time (min)  100  Transmembrane pressure— 59 psig  90 80 70 + 60 50 40 30 + x  20  Memln MemOut  10 0  —i 50  100  150  h— 200  Time (min)  250  300  350  3. Run 3. »  Heavy Oil Feed: Undiluted  o Feed Asphaltene Content: 16.3 wt% •  Cross-flow Velocity: 5.01 m/s  •  Membrane Surface Area: 0.005 m  •  Membrane Nominal Pore Diameter: O.lum  2  Table E.3.a: Permeate Flux Time  Permeate Flux  (min)  (Kg/m /day)  (0 - 210)  106.1  (210-320)  41.1  (320 - 418)  10.4  (418-466)  9.3  2  Table E.3 .b: Asphaltene Content in the Permeate Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0 - 210)  9.6  41  (210 - 466)  4.9  70  Run 3. Operation Conditions (Note: Computer errors due to insufficient REM occured 120 after 300 mininutes) T  100  o  8 0  3 £ o  60  a  E a> H  40  -Average Temp. = 98C  20  250  200  150  100  50  300  Time (min)  Transmembrane Pressure = 89psig 120  — Memln • MemOut  100  200  300  Time (min)  400  500  4.  Run 4.  •  Heavy Oil Feed: Undiluted  •  Feed Asphaltene Content: 17.3 wt%  •  Cross-flow Velocity: 6.90 m/s  » Membrane Surface Area: 0.005 m  2  © Membrane Nominal Pore Diameter: 0.1 um  Table E.4.a: Permeate Flux  Time  Permeate Flux  (min)  (Kg/m /day)  (0 - 149)  138.1  (149 - 269)  61.0  (269 - 329)  54.2  (329 - 372)  54.5  (372 - 436)  54.0  (436 - 494)  52.7  (494 - 571)  51.9  2  Table E.4.b: Asphaltene Content in the Permeate  Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0 - 149)  12.6  27  (149-269)  5.5  68  (269 - 372)  4.2  76  (372 - 494)  3.8  78  (494 - 571)  4.0  77  178  Run 4. Operation Conditions 120  100  o  8  0  3  jo OJ  60  a.  E a>  40 •Average Temp. = 110C 20  100  200  300  400  500  600  Time (min)  Transmembrane pressure = 87psig  at '35 Q.  P  S  co co  60  CI)  a. 40 o  20  Memln MemOut  0 100  +•  200  300 Time (min)  400  500  600  5. Run 5. •  Heavy Oil Feed: Undiluted  •  Feed Asphaltene Content: 18.3 wt%  •  Cross-flow Velocity: 8.78 m/s  •  Membrane Surface Area: 0.005 m  •  Membrane Nominal Pore Diameter: O.lum  2  Table E.5.a: Permeate Flux  Time  Permeate Flux  (min)  (Kg/mVday)  (0 - 60)  239.0  (60 - 180)  63.6  (180-240)  63.4  (240 - 302)  61.3  (302 - 360)  60.1  (360 - 427)  60.1  (427 - 486)  60.0  (486 - 527)  61.8  (527 - 587)  60.0  180  Table E.5.b: Asphaltene Content in the Permeate Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0 - 60)  17.7  3  (60 - 180)  5.3  71  (180 - 302)  3.6  80  (302 - 427)  3.4  81  (427 - 527)  3.7  79  (527 - 587)  3.4  81  j  181  Run 5. Operation Conditions  2 2.  60 +  •Average Temp. = 118C  H  40 + 20  100  200  300  400  500  600  Time(min)  0  Transmembrane pressure = 78 psig  A Memln MemOut 100  200  300 Time (min)  400  500  600  I 182  6. Run 6.  •  Heavy Oil Feed: Undiluted  •  Feed Asphaltene Content. 19.0 wt%  •  Cross-flow Velocity: 6.90 m/s  •  Membrane Surface Area: 0.005 m  •  Membrane Nominal Pore Diameter: 0.1pm  2  Table E.6.a: Permeate Flux Time  Permeate Flux  (min)  (Kg/m /day)  (0-7)  668.0  (7 - 15)  383.2  (15 - 26)  301.1  (26 - 37)  251.3  (37 - 56)  219.8  (56 - 76)  222.8  (76 - 143)  187.5  (143 - 170)  57.9  (170 - 232)  49.0  (232 - 292)  52.6  (292 - 366)  57.2  (366 - 428)  55.8  (428 - 472)  46.2  2  Table E.6.b: Asphaltene Content in the Permeate  Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0 - 15)  18.9  0.5  (15 - 37)  18.9  0.5  (37 - 56)  18.2  4  (56 - 76)  18.8  1  (76 - 143)  17.3  9  (143 - 170)  16.4  14  (170 - 232)  10.6  44  (232 - 366)  4.2  78  (366 - 472)  3.8  80  Run 6. Operation Conditions 120  T  100 +  O  80 +  3 JB  60  o. E £  40  a>  -Average Temp. = 105C 20  100  200  500  400  300  Time (min)  Transmembrane pressure = 85 psig 120 100 cn  4  80  (0 Q.  + 20 +  Memln MemOut  -+-  100  200  300  Time (min)  400  500  7. Run 7.  •  Heavy Oil Feed: Undiluted  •  Feed Asphaltene Content: 19.2 wt%  •  Cross-flow Velocity: 8.78 m/s  •  Membrane Surface Area: 0.005 m  •  Membrane Nominal Pore Diameter: 0.1pm  2  Table E.7.a: Permeate Flux  Time  Permeate Flux  (min)  (Kg/m /day)  (0-7)  668.0  (7 - 13)  359.5  (12-19)  315.8  (19 - 36)  301.2  (36 - 54)  313.9  (54 - 63)  88.1  (63 - 96)  60.6  (96 - 155)  51.3  (155 - 186)  48.6  (186-242)  52.6  (242 - 296)  58.8  (296 - 337)  60.0  (337 - 389)  54.7  (389 - 437)  55.3  (437 - 483)  58.4  2  Table E.7.b: Asphaltene Content in the Permeate Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0 - 19)  19.0  0  (19 - 36)  19.1  0  (36 - 54)  18.6  4  (54 - 96)  18.6  1  (96 - 186)  17.7  9  (186 - 296)  12.8  14  (296 - 389)  4.5  44  (389 - 483)  3.6  78  187  Run 7. Operation Conditions  100  200  300  400  500  Time (min)  Transmembrane pressure = 78 psig 120  T  x Memln MemOut 100  200  300  Time (min)  400  500  8.  Run 8.  •  Heavy Oil Feed: Undiluted  •  Feed Asphaltene Content: 18.3 wt%  •  Cross-flow Velocity: 6.90 m/s  •  Membrane Surface Area: 0.005 m  •  Membrane Nominal Pore Diameter: 0.05pm  Table E.8.a: Permeate Flux Time  Permeate Flux  (min)  (Kg/m /day)  (0 - 37)  132.3  (37 - 47)  85.0  (47 - 60)  78.0  (60 - 74)  72.0  (74-131)  57.9  (131 -222)  30.3  (222 - 267)  42.9  (267 - 293)  39.9  (293 - 334)  48.5  (334 - 365)  48.3  (365 - 396)  44.6  (396 - 423)  46.9  (423 - 477)  48.5  2  2  Table E.8.b: Asphaltene Content in the Permeate Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0 - 74)  13.9  24  (74-131)  7.5  59  (131 -222)  5.5  70  (222 - 334)  5.1  72  (334 - 423)  4.3  77  (423 - 477)  4,1  78  Run 8. Operation Conditions  120  T  100 4O  80  3  « 60 Q.  E  £  40 •Average Temp. = 108C 20 0  100  200 300 Time (min)  500  400  Transmembrane pressure = 85 psig  120  T  100 4, at '55  a. CD i_  3  co (0  a>  o Memln MemOut 100  200 300 Time (min)  400  500  9.  Run  9.  » Heavy Oil Feed: Undiluted © Feed Asphaltene Content: 16.4 wt% © Cross-flow Velocity: 6.90 m/s © Membrane Surface Area: 0.005 m  2  o Membrane Nominal Pore Diameter: 0.02um  Table E.9.a: Permeate Flux  Time  Permeate Flux  (min)  (Kg/m /day)  (0 - 150)  62.2  (150-270)  39.1  (270 - 330)  27.4  (330 - 390)  36  (390 - 450)  41.8  (450- 510)  41.8  (510 - 570)  35  2  Table E.9.b: Asphaltene Content in the Permeate  Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0-150)  3.4  79  (150- 330)  1.9  88  (330 - 570)  1.6  90  Run 9. Operation Conditions 120 100 -O  80 \  t  ? 3  ]5  60  E £  40  o> a.  20  •Average Temp. = 107C  4-  0  100  200  300  400  500  600  Time (min)  Transmembrane pressure = 87 psig 120  T  100  200  300 Time (min)  400  500  600  10.  Run  10.  •  Heavy Oil Feed: Undiluted  •  Feed Asphaltene Content: 19.5 wt%  «  Cross-flow Velocity. 8.78 m/s  •  Membrane Surface Area: 0.005 m  •  Membrane Nominal Pore Diameter: 0. lum (RegeneratedfromRun 5)  2  Table E.lO.a: Permeate Flux  Time  Permeate Flux  (min)  (Kg/m /day)  (0 - 174)  104  (174 - 191)  66.3  (191 - 239)  63.9  (239 - 300)  63.1  (300 - 417)  62.0  (417 - 482)  61.1  2  Table E.lO.b: Asphaltene Content in the Permeate  Time  Asphaltene  Asphaltene  (min)  Content (wt%)  Reduction (%)  (0 - 191)  6.8  65  (191 - 300)  5.6  71  (300 - 366)  4.4  77  (366 - 482)  4.4  77  194  R u n 10. Operation Conditions  140 -r  100  200  300  400  500  Time (min)  Transmembrane pressure = 76 psig 120  100  200  300  Time (min)  400  500  11. Nickel and Vanadium Versus Asphaltene Contents  Table E . l l : Data for Figure 5.2  Asphaltene  Nickel  Vanadium  Content (wt%)  (ppm)  (ppm)  17.3  76  190  16.3  70  170  12.6  61  153  9.6  52  135  4.6  32  77  1.6  27  60  1  1  2  2  2  2  1. Measured asphaltene content of heavy oil feed. 2. Measured asphaltene content of the permeate.  12. Heavy Oil Viscosity Versus Asphaltene Contents  Table E.12: Data for Figures 5.3(a) and (b) j  Viscosity (mPa.s) Asphaltene  T=40°C  T=110°C  16.3  5825  67.5  12.6  1581  35  9.6  900  27.5  5.3  710  22.5  3.6  700  22.5  1.6  428  20  1  Content (wt%) 1  2  2  2  2  2  1. 2. 3. 4.  3  4  I  Measured asphaltene content of heavy oil feed. Measured asphaltene content of the permeate. Average value of the repetitive data of the same sample, i.e. 1525 and 1637.5 mPa.s Average value of the repetitive data of the same sample, i.e. 32.5 and 37.5 mPa.s  13. Heavy Oil Density Versus Asphaltene Contents  Table E. 13: Data for Figure 5.4 Asphaltene  Density (kg/m )  Content (wt%)  (T=15.6°C)  16.3  999.4  12.6  987.4  9.6  984  3.6  974  1.6  967.9  1  2  2  2  2  3  1. Measured asphaltene content of heavy oil feed. 2. Measured asphaltene content of the permeate.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0058522/manifest

Comment

Related Items