Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Internally circulating fluidized bed membrane reactor for high-purity hydrogen production Boyd, David Anthony 2007

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


831-ubc_2007-266861.pdf [ 40.12MB ]
JSON: 831-1.0059011.json
JSON-LD: 831-1.0059011-ld.json
RDF/XML (Pretty): 831-1.0059011-rdf.xml
RDF/JSON: 831-1.0059011-rdf.json
Turtle: 831-1.0059011-turtle.txt
N-Triples: 831-1.0059011-rdf-ntriples.txt
Original Record: 831-1.0059011-source.json
Full Text

Full Text

Internally Circulating Fluidized Bed Membrane Reactor for High-Purity Hydrogen Production by D A V I D A N T H O N Y B O Y D B . A . S c . (Department of Chemical Engineering) University o f British Columbia, 1990 A T H E S I S S U B M I T T E D I N P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F D O C T O R OF P H I L O S O P H Y in T H E F A C U L T Y OF G R A D U A T E S T U D I E S Chemical and Biological Engineering T H E U N I V E R S I T Y OF B R I T I S H C O L U M B I A February 2007 © David Anthony Boyd, 2007 Abstract A novel reactor configuration, the internally circulating fluidized bed membrane reactor (ICFBMR), was studied in an experimental program for the steam reforming (SMR) of natural gas to produce hydrogen. This work builds on previous fluidized bed membrane reactor (FBMR) research, in which H2-selective membranes were located within a fluidized bed of catalyst to produce a H 2 stream-directly from the reactor, thereby shifting the chemical equilibrium of the reforming reaction forward. The ICFBMR advances this concept by modifying the reactor geometry in order to induce circulation of catalyst solids up a central core draft assembly, which house vertical planar H 2 membranes, and down an outer annular region. The catalyst solids circulation has a number of benefits, especially when the reactor is applied to autothermal reforming (ATR), where the endothermic reforming heat is supplied by direct addition of air to the reactor. In this case, the circulating solids transfer heat from the upper oxidation zone to the core reforming zone, with very little circulation of the nitrogen entering with the oxidation air. The hydrodynamics of the I C F B M R geometry were studied using a Plexiglas cold model. Dimensionless variables were used in an attempt to match key scaling parameters between the cold model, which used air and fluidized catalytic cracking (FCC) solids, and the hot reformer. Solids circulation was studied as a function of the main and annular gas feed rates for three different membrane panel geometries. It was found that solid membrane panels, which prevented communication between the core flow slots, led to maldistribution of solids and gas. Helium tracer studies confirmed that only a small portion (-10%) of the N 2 in the oxidation air fed to the upper ' reactor transferred to the reactor core with the returning solids. Solids circulation was found to increase linearly with the main feed rate up to a core superficial gas velocity of -0.3 m/s, and tended to level off after a superficial gas velocity of -0.5 m/s. The experimental data were used to find-predictive equations for solids circulation that could be used for the hot reformer design. Double-sided planar H 2 membranes (each 83 x 280 x 6 mm) were prepared using 50-um thick palladium alloy foil using techniques of Membrane Reactor Technologies Ltd. Six membranes were installed in a pilot reactor (diameter 0.135 m, height 2.3 m) and a number of pilot reforming experiments were performed. The reformer was successfully operated up to 650°C and 1,500 kPa with a feed of natural gas and steam, under both S M R (external heat) and A T R (direct air addition). Helium tracer studies were performed on the hot reformer, and internal solids circulation was measured to be 0.21 kg/s at a typical operating condition, closely matching the value predicted from cold model experimentation. Pure H 2 (>99.999%, excluding N 2 ) was produced for the first -180 ii hours of testing, after which the H 2 purity from two of the six membranes dropped to -99.7% for the remaining -150 hours of hot operation. The highest hydrogen production from the pilot reactor was -1.06 Nm3/h. The highest measured ratio of permeate H 2 to feed natural gas was 1.17 Nm3/Nm3, well below the value required for economic operation (-2.5), highlighting how the reactor performance was limited by the installed membrane area. ATR operation showed that permeate H 2 production is only marginally affected by the rate of air addition. Two types of catalyst powders, a SMR (NiO) catalyst and a novel ATR catalyst, were used in the pilot reformer. Low catalyst activity affected a number of the experimental runs. The ICFBMR reactor was simulated using a commercial process simulator (HYSYS) to study the influence of a number of variables on a reactor producing 30 Nm3/h of H2. The simulation ignored reaction kinetics, a reasonable assumption for this reactor configuration as reactor performance is overwhelmingly controlled by membrane performance and reactor geometry, with reactor gases near equilibrium. Simulations indicate that the predicted solids circulation rate is sufficient to maintain the core temperature drop to below 30°C, and that there would be limited reduction in membrane area if the circulation rate were to be increased. The reactor model was incorporated within a simulation for the complete system, leading to a predicted overall energy efficiency of 69%, based on utility consumption and the higher heating value of reactants. An economic evaluation of the ICFBMR system was performed and compared with published data from a conventional small-scale SMR system. Results indicate that the ICFBMR can achieve higher process efficiencies, but that membrane cost and longevity are critical to making the process economically viable. i i i Table of Contents Abstract . ii Table of Contents -iv List of Tables vii List of Figures ••• bx Nomenclature • xiv Acknowledgements xvii Chapter 1. Introduction.... 1 1.1 Hydrogen -. • 1 1.1.1 Hydrogen Economy 2 1.1.2 Small-Scale Thermal Processes for H 2 Production 4 1.2 Steam Methane Reforming '.. 6 1.2.1 SMR Catalysis 8 1.2.2 Limitations of Conventional SMR 9 1.2.3 Small-Scale SMR 10 1.2.4 Membrane Reforming .-. 11 1.2.5 Fluidized Bed Membrane Reactor 12 1.3 Internally Circulating Fluidized Bed Membrane Reactor 15 1.4 Research Objectives 18 1.5 Thesis Outline 19 Chapter 2 . Hydrogen Selective Membranes 2 1 2.1 Background • 21 2.1.1 Oxygen Permeable Membranes 22 2.1.2 Types of Hydrogen Permeable Membranes 24 2.1.3 Ceramic Hydrogen Permeable Membranes 25 2.2 Metallic Hydrogen Membranes 27 2.2.1 Dense Palladium-Based Membranes 27 2.2.2 Dense Palladium-Based Membranes: Mechanical Issues 31 2.2.3 Composite Membranes /. 33 2.2.4 Palladium Membranes in SMR 34 2.3 ICFBMR Membranes 34 2.3.1 Planar Membrane Design.. 36 2.3.2 Membrane Fabrication 38 2.3.3 Membrane Costs 42 2.4 Membrane Characterization 43 2.4.1 Effect of Hydrogen Partial Pressure 44 2.4.2 Effect of Temperature 47 2.4.3 Variation between Membranes 49 2.5 Conclusions 50 iv Chapter 3. Cold Model Testing . 52 3.1 Introduction 52 3.2 Fluidized Beds 53 3.3 Cold Model Hydrodynamics 54 3.3.1 Scaling Parameters 54 3.3.2 Catalyst Powder Characteristics 56 3.4 Experimental Apparatus 60 3.4.1 Reactor Geometry 60 3.4.2 Instrumentation 67 3.5 Experimental Results 69 3.5.1 Fluidization Observations 70 3.5.2 Solids Circulation 71 3.5.3 Pressure Data 76 3.5.4 Helium Tracer Experiments 80 3.6 Discussion '. 86 3.6.1 Solids Circulation 86 3.6.2 Other Geometry Aspects 89 3.6.3 Design Implications • 91 Chapter 4. ICFBMR Pilot Plant 93 4.1 Introduction • 93 4.2 I C F B M R Reactor 94 4.2.1 Process Design 94 4.2.2 Reactor Layout ........96 4.2.3 Membrane Assembly Layout 100 4.2.4 Reactor Instrumentation ; 103 4.3 Pilot Plant Description •• 104 4.4 Pilot Plant Operation 106 4.5 Catalyst 106 4.6 Pilot Plant Experiments without Membranes 108 4.7 Pilot Plant Experiments with Membranes I l l 4.7.1 Membrane Performance : 111 4.7.2 SMR Operation (External Heating Only) 115 4.7.3 A T R Operation (Air Addition) 119 4.7.4 Solids Circulation Tests 121 4.8 Catalyst Testing 124 4.8.1 SMR Catalyst Testing. 125 4.8.2 A T R Catalyst Testing .' 126 4.9 Discussion '. 128 4.10 Conclusions • 132 Chapter 5. Post-Run Membrane Testing.. • 133 5.1 Membrane Removal and Observations 133 5.2 Membrane Flux Testing '. • • 135 5.3 Permeate Purity..: 136 5.4 Foil Analysis with S E M / E D X 137 5.5 Discussion 142 v Chapter 6. Process Simulation and Economics , 144 6.1 Introduction 144 6.2 Reactor Simulation 145 6.2.1 Simulation Base Case 149 6.2.2 Effect of Reactor Variables 153 6.3 Design of a 30 Nm 3/h H 2 Unit 157 6.3.1 Design Basis 158 6.3.2 Process Simulation 159 6.3.3 Reactor Layout 162 6.3.4 Economics 166 6.4 Discussion 169 Chapter 7. Conclusions and Recommendations 171 7.1 Conclusions 171 7.2 Alternative Reactor Applications 173 7.3 Recommendations 174 References 176 Appendix 1: Introductory Details 185 Appendix 2: Membrane Details ..187 Appendix 3 : Cold Modeling A3.1 Cold Model Installation Instructions 192 A3.2 Cold Model Testing Details 195 Appendix 4: I C F B M R Pilot Reactor A4.1 ICFBMR Reactor Details 205 A4.2 ICFBMR Assembly Procedure 233 A4.3 ICFBMR Operating 242 A4.4 Catalyst Microreactor 249 A4.5 Commissioning 251 A4.6 ICFBMR Pilot Plant Data 253 A4.7 Microreactor Data 261 A4.8 T G A Analysis of NiO Catalyst 263 A4.9 X R D Analysis of N iO Catalyst 265 Appendix 5: Post-Run Membrane Testing Details 266 Appendix 6: Process Simulation and Economics A6.1 Simulation Details 273 A6.2 30 Nm 3/h Cash Flow Analysis 281 A6.3 Alternate 30 Nm 3 /h ICFBMR Cost Analysis ; 287 vi List o f Tables 1.1 Hydrogen unit equivalence of 100 S L M : 4 1.2 Selected thermal processes to produce hydrogen 5 1.3 Common S M R reactions 7 2.1 Parts list for ICFBMR panel.. 37 2.2 Summary of membrane panel areas 41 2.3 Approximate ICFBMR membrane panel costs 42 2.4 Calculated effective membrane areas for ICFBMR panels 50 3.1 Cold model and pilot reactor properties and characteristics 59 3.2 Cold model scaling parameters at a core superficial velocity of 0.1 m/s 60 3.3 Comparison of cold model and ICFBMR reactor geometries 65 3.4 Summary of cold model helium tracer results 84 4.1 Summary of ICFBMR experimental runs 109 4.2 Permeate purity from membranes operating in ICFBMR pilot reactor 113 4.3 Calculated effective permeation area of the membranes installed in the pilot reactor 114 4.4 Local measured and corresponding equilibrium temperatures 124 4.5 Equilibrium methane conversions for Figure 4.19 127 5.1 Calculated effective membrane permeation area, before and after pilot runs 136 5.2 Permeate purity post-pilot runs 137 6.1 ICFBMR base case simulation 150 6.2 Stream summary for 30 Nm 3/h reactor base case simulation 151 6.3 Stream table for 30 Nm 3/h flowsheet 161 6.4 PFD stream summary for 30 Nm 3/h design 162 6.5 Figures of merit for 30 Nm 3/h H 2 ICFBMR and conventional SMR production units 170 A l . l Steps employing catalysts in SMR 185 A1.2 SMR reactions with oxygen input 185 A1.3 SMR reaction rates ; 186 A l .4 SMR reaction rate parameters '. 186 A2.1 Permeation rig components 189 A2.2 Gas chromatograph details 189 A2.3 Hydrogen permeation data for Panel A - effect of hydrogen partial pressure 190 A2.4 Effect of temperature on permeation for Panel A 191 A2.5 Effect of temperature on permeation for Panels 1-8 191 A3.2.1 Cold model feed air rotameters 197 A3.2.2 Cold model pressure transducer data 198 A4.1.1 ICFBMR thermocouples , 223 A4.1.2 ICFBMR pressure transducers 223 A4.1.3 ICFBMR pressure taps and sample points 223 A4.1.4 Equipment list for ICFBMR pilot plant 225 v i i A4.1.5 Natural gas analysis •.• 232 A4.3.1 Operator task checklist 248 A4 .6 .1 Gas chromatograph analyses from ICFBMR helium tracer testing 253 A4.6.2 Membrane flux testing during run #5 253 A4.6.3 Membrane flux testing during run #6 254 A4.6.4 Summary of selected ICFBMR pilot tests 255 A4.6.5 Calculation of solids circulation in pilot reactor from helium tracer tests 258 A4.6.6 Data from A T R experiments with 0 2 : C H 4 molar ratio varied from 0 to 0.6 258 A4.6.7 Temperature profile of the reactor core during high temperature A T R tests 259 A4.7.1 Microreactor catalyst test data 261 A4.7.2 Data from virgin A T R microreactor tests 262 A4.8.1 T G A SMR catalyst samples. 263 A4.9.1 X R D samples 265 A5.1 Flux data for ICFBMR panels #6 and #8 after service in pilot plant 272 A6.1.1 H Y S Y S simulation details 273 A6.1.2 Stream table for 30 NmVh base case reactor simulation 276 A6.1.3 30 Nm 3/h process inputs 278 A6.1.4 Pinch analysis summary 279 A6.1.5 System losses for 30 Nm 3/h design 279 A6.2.1 Equipment list for 30 Nm 3/h ICFBMR design 283 A6.2.2 Assumptions for Net Present Value analysis 284 A6.2.3 Net Present Value calculations for 30 Nm 3/h ICFBMR design 285 A6.2.4 Summary of N P V analysis for 30 Nm 3/h ICFBMR design 286 A6.3.1 Cost basis 287 A6.3.2 Process basis 287 A6.3.3 Conventional SMR vs. ICFBMR cost comparison 288 v i i i List of Figures 1.1 World hydrogen market summary 1 1.2 Cost of delivered H 2 of various technologies 3 1.3 Carbon emissions from current H 2 technologies 3 1.4 Equilibrium methane conversion in SMR as a function of temperature and pressure 7 1.5 Conventional SMR process schematic 8 1.6 Equilibrium methane conversion in SMR as a function of temperature and in-situ hydrogen removal 12 1.7 Simplified process schematics for conventional SMR, membrane SMR and membrane A T R 14 1.8 ICFBMR schematic : 17 1.9 ICFBMR plan view 18 2.1 Types of hydrogen permeable membranes 24 2.2 Hydrogen permeation through dense metallic and porous inorganic membranes 25 2.3 Four-layer alumina membrane 26 2.4 Relative permeability of hydrogen in various palladium alloys at 350°C and 2.2 MPa.... 30 2.5 Thermal expansion coefficients of palladium and various membrane support materials.. 32 2.6 Fabrication techniques of supported palladium membranes 33 2.7 Membrane panel cross section 36 2.8 ICFBMR panel assembly drawing 37 2.9 Membrane bonding apparatus schematic 39 2.10 Three membrane panels in preparation for bonding 40 2.11 Finished membrane panel (Panel A) 41 2.12 Membrane permeation rig schematic 43 2.13 Hydrogen flux on Panel A as a function of difference between square root of hydrogen partial pressures 46 2.14 Permeation of Panel A at three different upstream pressures with permeate at 101 kPa.. 47 2.15 Arrhenius plots of hydrogen permeation on Pd 7 5 Ag25 Panel A (475-564°C) at three different upstream pressures with permeate at atmospheric pressure 48 2.16 Flux data for six membrane panels as a function of temperature 49 3.1 Gas-solid flow regimes 54 3.2 Cumulative size distributions for the three catalyst powders studied.... 57 3.3 Cold model pressure drop vs. superficial air velocity for the F C C 57 3.4 Cold model system schematic 62 3.5 Plan view of cold model core box and panels (solid panels shown) 63 3.6 Front views of cold model panels 64 3.7 Details of bottom section of cold model column 66 3.8 Helium tracer system for cold model column 69 3.9 Average downward solids velocity in the annular regions with solid panels as a function of main column flow and secondary air flow 73 3.10 Average downward solids velocity in the annular regions with horizontally slotted panels as a function of main column flow and secondary air flow 74 3.11 Average downward solids velocity in the annular regions with vertically slotted panels as a function of main column flow and secondary air flow 74 3.12 Average downward solids velocity in the annular regions with a secondary air flow of 20 S L M for all three panel geometries 75 ix 3.13 Average solids internal circulation rate with a secondary air flow of 20 S L M for the three panel geometries 76 3.14 Absolute pressure measurements from the middle of the core channels with solid panels (480 S L M main air, 20 S L M secondary air) 77 3.15 Absolute pressure measurements from the middle of the core channels with horizontally slotted panels (480 S L M main air, 20 S L M secondary air) 78 3.16 Pressure drop across channel 4 in cold model core with solid panels 79 3.17 Pressure drop across channel 4 in cold model core with vertically slotted panels 79 3.18 Measured helium concentration at various points in the cold model with 0% and 3.28% helium in the main feed for the vertically slotted panels 81 3.19 Measured helium concentration at different sample points in the upper core box after 10 min. continuous injection of helium into the bottom of channel 4 (vertically slotted panels) 85 3.20 Measured cold model solids flux in core draft box (Gp,core) for horizontally slotted and vertically slotted panels vs. product of superficial gas velocity in core , (Ucore) and particle density (pp), with an annular air flow of 20 S L M 87 3.21 Sketches showing solids entrainment in the wake of a rising bubble 89 3.22 Some alternative ICFBMR layouts 90 3.23 Alternate annular distributors 91 3.24 Effect of solids circulation on core reactor temperature drop and normalized membrane area 92 4.1 Process flow diagram (external heating) 97 4.2 Process flow diagram (direct air addition) 98 4.3 ICFBMR pilot reactor 99 4.4 Location of membranes in ICFBMR reactor 101 4.5 Plan view of ICFBMR reactor internals 102 4.6 End view photograph of final membrane assembly 103 4.7 Pilot plant schematic 104 4.8 Purity of combined hydrogen permeate (excluding N 2 ) and core reactor temperature over the duration of the ICFBMR pilot reactor testing 113 4.9 Permeate hydrogen recovery relative to gas feed rate as function of reactor temperature and N 2 sweep .116 4.10 Permeate hydrogen recovery relative to natural gas feed rate as function of reactor temperature and sweep gas 117 4.11 Permeate hydrogen recovery relative to natural gas feed rate as function of steam-to-carbon molar ratio reactor temperature and sweep gas 117 4.12 Hydrogen recovery as a function of natural gas feed rate for all data in Table A4.6.4..... 118 4.13 H 2 partial pressure in a well-mixed reactor for A T R experimental conditions predicted by equilibrium model 119 4.14 Effect of air addition on H 2 flux 120 4.15 Permeate hydrogen flow (with and without sweep gas) with varying air addition rate.... 121 4.16 I C F B M R helium tracer test schematic 122 4.17 Measured helium concentration and methane conversion from helium tracer tests 123 4.18 CH4 conversion in microreactor with SMR catalyst 125 4.19 CH4 conversion in microreactor with virgin A T R catalyst 127 4.20 Microreactor testing of diluted (10%) A T R catalyst after use in ICFBMR pilot plant 128 4.21 Alternate ICFBMR reactor layout 131 5.1 ICFBMR internals being removed from horizontal pilot plant reactor 134 5.2 ICFBMR membrane assembly after service in pilot plant reactor 135 x 5.3 Post-pilot run flux test data for membrane panels #6 and #8 at 560°C 136 5.4 Peeling palladium alloy foil from membrane panel #1 138 5.5 S E M of fresh P d 7 5 A g 2 5 foil at x300 magnification 139 5.6 S E M of scaled membrane foil (front) at x300 magnification 139 5.7 S E M of polished membrane foil (front) at x300 magnification 140 5.8 S E M of back of membrane foil at x300 magnification 140 5.9 S E M of back of membrane foil at x 1000 magnification 141 5.10 Summary of E D X analyses of used membrane foil 142 6.1 ICFBMR simulation schematic 146 6.2 Permeation stage temperature and reactor hydrogen partial pressure (base case) 152 6.3 Calculated permeation stage area (base case, 50 yaa. Pd 7 5 Ag 25 membrane) 153 6.4 Effect of reactor pressure and steam-to-carbon ratio on normalized base case membrane area 154 6.5 Effect of hydrogen recovery on normalized base case membrane area 154 6.6 Effect of membrane operating temperature on normalized base case membrane area 155 6.7 Effect of feed preheat temperature on normalized membrane area 156 6.8 Effect of normalized solids circulation rate on normalized base case membrane area 156 6.9 Effect of permeate H 2 partial pressure on normalized base case membrane area 157 6.10 Process flow schematic for 30 Nm 3/h production rate 160 6.11 Plan view of option 1 layout - 30 Nm 3/h reactor (Ucore = 0.22 m/s) 163 6.12 Plan view of option 2 layout - 30 Nm 3/h reactor {ucore = 0.12 m/s).. 164 6.13 Elevation drawing for Option 2 - 30 Nm 3/h reactor (Ucore = 0.12 m/s) 165 6.14 Estimated bare equipment cost components for 30 Nm 3 /h unit 167 6.15 Breakdown of variable operating costs for 30 Nm 3/h ICFBMR (base case) 168 7.1 Sorption enhanced ICFBMR concept : 174 A l . 1 Equilibrium methane conversion in SMR as a function of steam-to-carbon ratio and temperature 185 A2.1 Palladium pricing 1992-2007. 187 A2.2 Crease defects in ICFBMR membrane panels after bonding 188 A3.1.1 Cold model assembly schematic 194 A3.2.1 Cold model primary and secondary air distributors 195 A3.2.2 Cold model viewed from bottom flange 195 A3.2.3 Cold model main distributor plate 196 A3.2.4 Elevation view of cold model oxidant distributor.. 196 A3.2.5 Pressure drop over cold model main distributors 197 A3.2.6 Pressure drop over cold model secondary and oxidant distributors 197 A3.2.7 Plan view of cold model pressure tap connections with vertically slotted panels. 198 A3.2.8 Elevation view of cold model pressure tap connections (looking east) 199 A3.2.9 Cold model pressure drop vs. superficial air velocity for SMR (NiO) catalyst 200 A3.2.10 Cold model pressure drop vs. superficial air velocity for A T R catalyst 200 A3.2.11 Calibration of FI-06 used for helium injection in cold model testing at 207 kPag as measured against a mass flow meter operating with helium 201 A3.2.12 Typical calibration curve of thermal conductivity detector (TCD) with helium used in cold model system 201 A3.2.13 Ratio of maximum to minimum quadrant solids velocity for cold model data 202 A3.2.14 Pressure drop over channel 4 in cold model core with horizontally slotted panels 202 xi A3.2.15 Gage pressure at mid-height of cold model channel 4, 20 S L M air to annulus 203 A3.2.16 Pressure of annulus vs. pressure in channel 4 in with vertically slotted panels 204 A4.1.1 ICFBMR main distributor drilling pattern 205 A4.1.2 ICFBMR distributor pressure drop 205 A4.1.3 Photograph of ICFBMR main and annular distributor assembly 206 A4.1.4 ICFBMR external pressure vessel drawing (1) 207 A4.1.5 ICFBMR external pressure vessel drawing (2) 208 A4.1.6 ICFBMR support steel drawing..; 209 A4.1.7 I C F B M R internal vessel drawing (1) 210 A4.1.8 ICFBMR internal vessel drawing (2) 211 A4.1.9 ICFBMR main distributor mechanical drawing 212 A4.1.10 ICFBMR internal preheater 213 A4.1.11 ICFBMR draft core box 214 A4.1.12 Membrane assembly general arrangement .215 A4.1.13 Dummy membrane panel. • • 216 A4.1.14 Membrane backing strips and full-length dummy panel 217 A4.1.15 Membrane connector piece 218 A4.1.16 P&JD for pilot plant (gas cylinders) 219 A4.1.17 P&ID for pilot plant (feeds and condensers) 220 A4.1.18 P&ED for pilot plant (external preheater) 221 A4.1.19 P&ID for new ICFBMR reactor 222 A4.1.20 ICFBMR instrument locations '. 224 A4.1.21 ICFBMR pilot plant control screen #1 226 A4.1.22 I C F B M R pilot plant control screen #2 227 A4.1.23 ICFBMR pilot plant control screen #3 228 A4.1.24 ICFBMR pilot plant control screen #4 ..: 229 A4.1.25 ICFBMR pilot plant control screen #5 230 A4.1.26 I C F B M R dummy internals for commissioning 231 A4.1.27 ICFBMR off-gas filter pressure drop 232 A4.2.1 ICFBMR block positions 239 A4.2.2 ICFBMR preheater 240 A4.2.3 Inserting ICFBMR internals into horizontal reactor 240 A4.2.4 ICFBMR internal reactor shell and heaters 241 A4.2.5 Raising the ICFBMR reactor 241 A4.4.1 Microreactor flow schematic : 249 A4.4.2 Microreactor geometry 250 A4.5.1 Original pilot ICFBMR secondary distributor (dimensions in mm) 251 A4.5.2 Original pilot ICFBMR air distributor 251 A4.5.3 Calibration of helium rotameter (CHE 4251C) at 1,380 kPag.... 252 A4.6.1 Measured and predicted methane conversion as a function of reactor temperature 259 A4.6.2 Methane conversion as function of steam-to-carbon ratio, reactor temperature and sweep gas... 259 A4.6.3 Hydrogen flux as a function of natural gas feed rate for all data in Table A4.6.4 260 A4.8.1 Temperature profile of T G A tests : 263 A4.8.2 T G A weight profiles of NiO samples, normalized to initial sample weight 264 A4.8.3 T G A differential weight profiles of sample #1 (fresh NiO catalyst) 264 A4.8.4 T G A differential weight profiles of sample #3 (after final ICFBMR run) .264 A4.9.1 X R D analysis of fresh NiO catalyst (sample #1, top spectra) and used catalyst 265 xii A5.1 Post-pilot run membrane panel refrigerant leak test observations 266 A5.2 S E M of fresh Pd 7 5 Ag25 foil at xlOOO magnification 267 A5.3 S E M of scaled membrane foil (front) at xlOO magnification 267 A5.4 S E M of scaled membrane foil (front) at x3000 magnification 268 A5.5 S E M of polished membrane foil (front) at xlOOO magnification 268 A5.6 S E M of polished membrane foil (front) at x3000 magnification 269 A5.7 S E M of back of membrane foil at x3000 magnification 269 A5.8 E D X scan #1 of scaled front of membrane 270 A5.9 E D X scan #2 of polished front of membrane 270 A5.10 E D X scan #3 of smooth zone of back of membrane 271 A5.11 E D X scan #4 of scaled zone of back of membrane 271 A6.1.1 HYSIS reactor simulation flowsheet (Sheet 1: Left hand side) 274 A6.1.2 HYSIS reactor algorithm schematic 278 A6.1.3 Pinch analysis diagram for 30 Nm 3 /h design 279 A6.1.4 HYSIS 30 Nm 3/h process simulation flowsheet 280 A6.2.1 Simplified P&ID for 30 Nm 3/h 281 xiii Nomenclature Letters A am, Aeff A gap A Ar A r A B C,,C2 Cm CHI C0 D DH DAH dn, d * Fr g ^p, core JH Area [m2] Total cross-sectional area of annular regions [m2] Open cross-sectional area of membrane core box [m2] Effective permeable area of membrane [m2] Cross-sectional area of gap at annular turn into core draft box [m2] Area of membrane [m2] Archimedes number (Ar = d = ——-—r-5 —) [-] P Archimedes number boundary between Geldart's group A and B particles [-] Constants in equation 3.5 [-] Concentration of atomic hydrogen on high concentration side of membrane [mol/m3] Concentration of atomic hydrogen on low concentration side of membrane [mol/m3] Discharge coefficient in equation 3.11 [-] Reactor internal diameter [m] Hydraulic diameter [m] Atomic hydrogen diffusivity [m2/s] Particle diameter [m] Particle diameter for a certain size fraction [m] -, 1/ Dimensionless particle diameter (J* = dp P/(PP-Pf)g P )[-] Activation energy for hydrogen permeation [J/mol] Particle Froude number (Fr Ml •)[-Bubble wake fraction [-] Acceleration due to gravity (9.81) [m/s2] Mass flux of solids [kg/s m2] Mass flux of solids up the ICFBMR core based on area open to flow [kg/s m2] Hydrogen flux through membrane [mol/s m2] Permeation flux constant [mol/s m Pa0 5] xiv Ks Sievert's constant [mol/m3 Pa0 5] L Generic bed dimension [m] Wp.core Net solids circulation rate up reactor core (kg/s) MWX Molecular weight of species x [g/mol] n Exponent for hydrogen partial pressure in flux equation [-] PHh Partial pressure of hydrogen on high-pressure (reactor) side [Pa] PHeh Partial pressure of helium on high-pressure (reactor) side [Pa] PHI Partial pressure of hydrogen on low-pressure (permeate) side [Pa] • PHei Partial pressure of helium on low-pressure (permeate) side [Pa] Qb Bubble volumetric flow [m3/s] Qcore Gas volumetric flow in reactor core [m3/s] Qg.ann Gas volumetric flow dragged down in annulus [m3/s] QH Hydrogen permeation rate [mol/s] r Square of ratio of hydraulic diameter of the gap and hydraulic diameter of annulus in equation 3.11 [-] R Gas constant (8.314) [J/mol K] pfUm;d Rem/ Particle Reynolds number at minimum fluidization (Rem / = 1 m j / ) [-] M s Reactor space velocity [h"1] T Temperature [K] U Superficial gas velocity [m/s] Uann Superficial gas velocity in the annulus [m/s] Uc Superficial gas velocity at which pressure fluctuations in a fluidized bed reach a maximum [m/s] Ucore Superficial gas velocity in reactor core based on area open to flow [m/s] Umf Superficial gas velocity at minimum fluidization [m/s] U, Terminal particle settling velocity [m/s] U* Dimensionless terminal particle velocity (U* = U, Vp Particle velocity in annulus [m/s] V Catalyst volume [m3] Xi Mass fraction for a certain size fraction [-] M(Pp-Pf)g 3 XV Greek Letters AP„ Pressure difference between annulus and core, equation 3.11 [Pa Bed voidage [-] Bed voidage at minimum fluidization [-] n Energy efficiency [-] p Gas viscosity [Pa s] Pann Bed density in annulus region [kg/m3] Pf Gas density [kg/m3] PP Particle density [kg/m3] X Thickness of palladium membrane layer [m] Volumetric flow of reactants [m3/h] t Particle sphericity [-] Abbreviations ATR Autothermal reforming EDX Energy dispersive x-ray FBMR Fluidized bed membrane reactor FCC Fluidized catalytic cracking ICFBMR Internally circulating fluidized bed membrane reactor MRT Membrane Reactor Technologies Ltd. Pd Palladium PID Piping and instrumentation drawing ROG Non-permeate reactor off-gas SEM Scanning electron microscope SLM Standard liters per minute SMR Steam methane reforming XVI Acknowledgements The genesis of this project arose from collaboration between Membrane Reactor Technologies Ltd. (MRT) and Noram Engineering and Constructors Ltd. This project would not have been possible without the generous assistance and expertise of the engineers and scientists of MRT for membrane preparation and pilot plant operations, especially from Dr. Anwu Li, Ali Gulamhusein and Hongbin Zhao. I am very grateful to my co-supervisors, Dr. John Grace, Dr. Jim Lim and Dr. Alaa-Eldin Adris, for helping me to further their previous research in this area. The staff and my fellow graduate students of my department made my stay at UBC very enjoyable and helped me survive the pilot plant wjrk. I am thankful for the direct financial and material support received from NSERC, BC Science Council, UBC UGF, MRT and Noram Engineering. Finally, my greatest source of support has been my parents and my family, Jen, Gavin, Morgan and Jocelyn, who have been extremely patient during my extended stay at UBC. Without their love and inspiration, completion of this project would have been impossible! xvi i Chapter 1. Introduction This chapter presents the background and motivation for the present research; improvement of the conventional steam methane reforming (SMR) process with a novel membrane reactor for small-scale production of high-purity hydrogen. Previous high-temperature membrane reactor research is discussed. The concept for a new reactor, an internally circulating fluidized bed membrane reactor (ICFBMR), is also presented. 1.1 Hydrogen Hydrogen is one of the basic building blocks of the chemical and petroleum industry. It is produced in large quantities for a variety of processes, with ammonia production and fuel upgrading being the largest consumers (Figure 1.1 A). The world's annual hydrogen production is currently estimated at approximately 50 million tonnes (Stoll and von Linde, 2000). Canada's current annual production is estimated at 3.8 million tonnes, and is growing rapidly to meet demand in heavy oil upgrading (NRCan, 2005). gas A. Global H 2 demand (SRI, 2004) B. Feedstocks for global H 2 supply (NAE, 2004) Figure 1.1: World hydrogen market summary (2003) 1 Hydrogen can be produced by a wide range of processes, including electrolysis of water, many thermo-chemical processes and some anaerobic biological systems. However, the majority of the world's hydrogen demand is supplied by steam reforming of hydrocarbons, principally of natural gas (Figure 1 .IB), and is likely to remain so for the foreseeable future. 1.1.1 Hydrogen Economy As oil reserves dwindle and the effects of global warming become evident, there has been considerable interest in employing hydrogen as an energy carrier. When used in a fuel cell, hydrogen produces electricity, with water being the only emission. Other fuel cell feedstocks, such as methanol, have also been extensively studied, but hydrogen currently remains the most practical fuel source. The promise of hydrogen powered fuel cell vehicles has yet to be commercially realized, and some are pessimistic that it ever will (e.g. Roram, 2004). However, many believe that switching to a "hydrogen economy" is needed to satisfy the world's future energy requirements in an environmentally sustainable manner (e.g. see National Academy of Engineering, 2004, Sperling and Ogden, 2004). There has been considerable debate on where the hydrogen needed to power fuel cell devices will come from (Ewan et al., 2005). Many researchers agree that "green" hydrogen produced from a renewable energy source such as hydro or wind-powered water electrolysis and biomass gasification is desirable in the long term, while a lesser number advocate coal or nuclear-derived hydrogen. However, hydrogen production from hydrocarbons, such as natural gas, is likely to remain the most practical and economic solution in the near term (i.e. Thomas et al., 1998, NAE, 2004). Figure 1.2 presents delivered hydrogen cost estimates for a variety of technologies, at the present and projected into the future, for large and small-scale facilities. The reforming technology studied in this work falls under "small scale natural gas - future" in Figure 1.2, and highlights the potential to provide distributed hydrogen reformed from natural gas at an economically attractive price, both in the short and longer terms. "Gasoline" in Figures 1.2 and 1.3 is the gasoline equivalent for a hybrid vehicle compared to a H2-powered fuel cell vehicle. 2 12 -Ol 10 jc: 55 3 8 «» Hi CO 6 -c 0> Ol o 4 -•a >s X 2 0 Large scale Midsize Distributed - small scale Q current Hfuture I o O 1 a — (0 o 3 O cr > -° OL Technology o CO ro O Figure 1.2: Cost of delivered H 2 of various technologies (NAE 2004, gasoline values adjusted for hybrid vehicle efficiency, "PV" = photovoltaics) Figure 1.3 presents estimated carbon emissions for a number of the current hydrogen technologies presented in Figure 1.2. Grid-based technologies show relatively high emissions due to the current electrical generation supply in the US. If the electrical supply were less reliant on carbon fuels, the C 0 2 load of the grid-based technologies would be lower (Ewan et al., 2005). x o 15 3 a n 2. 2 o oi * 1 Large scale Midsize Distributed - small scale H indirect release through electicity O direct release i CO o O c x: g 1 g — to ro <n o n E g bo CO CD >. to S £ LU 3 1 o f 2 > -° a. CD C O W) ro Technology Figure 1.3: Carbon emissions from current H 2 technologies (NAE, 2004, gasoline values adjusted for hybrid vehicle efficiency) One of the technical challenges in commercializing the proton exchange membrane (PEM) fuel cell has been hydrogen purity. Very low levels of impurities, specifically carbon monoxide (<1 ppm), are required to prevent catalyst poisoning. High purity hydrogen is also needed in other applications, such as semiconductor manufacturing. This study investigates a new reactor to produce high-purity 3 hydrogen with the SMR process. It is envisioned that this process could be a viable option for small-scale hydrogen production, such as hydrogen fuelling stations, or as an on-site alternative for liquid hydrogen consumers. In the literature and industry, hydrogen production figures are quoted in a wide variety of units. In this work, hydrogen flows are given in NrnVh or standard liters perminute (both referenced to 0°C, 101.3 kPa). Table 1.1 below presents some of the more common units for the nominal hydrogen design rate for the ICFBMR pilot reactor (100 SLM). Large-scale SMR plants typically produce hydrogen in the range of 25,000 to 100,000 NmVday. Small-scale reformers for distributed hydrogen-fuelling stations are envisioned to be in the range of 100 to 1,000 NmVday, sufficient to supply a fleet of 900 to 9,000 fuel cell cars (Ogden, 2001). 1 Nm3 of hydrogen is equivalent in energy to approximately 0.3 liters of gasoline. Table 1.1: Hydrogen unit equivalence of 100 SLM 1 Molar Basis Mass Basis 100 SLM (0°C basis) 0.0089 kg/min 6.0- Nmj/h (0°C basis) 0.536 kg/h 144 NtrrVd 12.9 kg/day 1.1.2 Small-Scale Thermal Processes for H 2 Production Steam methane reforming, or other thermal processes based on hydrocarbons, is an appropriate technology for developing a distributed hydrogen infrastructure. Reformers can be readily installed at service stations using conventional utilities such as natural gas and water, and are cheaper to operate than electrolyzers. Although SMR is the benchmark technology, several other thermal processes are being considered for small-scale distributed hydrogen production (Armor, 2005). See Ogden (2001) for a review of commercial-ready technologies and suppliers and Pena et al. (1996) for a review of new catalytic routes to H 2 . Table 1.2 summarizes the main chemical reactions of the processes listed. • Steam methane reforming (SMR): See reaction #1 in Table 1.2. SMR is the most commercialized of the listed technologies. Further details on the conventional SMR process are given in the next section. • Autothermal reforming (ATR): In autothermal reforming, the SMR process is modified by adding air or oxygen to oxidize a portion of the methane (reaction #2 in Table 1.2), thus supplying the endothermic heat for the SMR reaction and allowing the reactor to run adiabatically. This 1 In this document "NmVh" and " S L M " conditions are referenced to 0°C, 101 kPa. Unless otherwise specified, pressure units in this report are absolute. 4 eliminates many of the heat transfer constraints of the SMR process, potentially making the process more compact and reducing capital costs. Table 1.2: Selected thermal processes to produce hydrogen Reaction > A H 2 9 8 (kJ/mol) Notes 1 CH4 + 2H 20 <-> C 0 2 + 4H2 165 Methane reforming with water gas shift 2 CH4 + '/202 —> CO + 2H2 -36 Partial oxidation of methane 3 C H 4 —> C + 2H 2 67 Methane cracking 4 C H 3 O H + H 2 0 <-» C 0 2 + 3H2 49 Methanol reforming with water gas shift 5 NH 3 —>• V2N2 + V/2H2 -46 Ammonia cracking • Partial oxidation (POx): See reaction #2 in Table 1.2. Methane, or other hydrocarbons, can react with oxygen (with and without steam) to form hydrogen at high temperatures. The process can be performed non-catalytically, but a catalyst is commonly used in commercial units. The process can be operated adiabatically, thus eliminating many of the heat transfer steps present in the SMR process. The POx process is also compact and has fast response times. However thermal efficiency is generally lower than for SMR (Ogden, 2001). • Thermocatalytic cracking of methane: See reaction #3 in Table 1.2. Methane can be catalytically decomposed to hydrogen and carbon at high temperatures. Catalyst fouling with carbon and low conversion are the main barriers to commercialization. • Methanol steam reforming: See reaction #4 in Table 1.2. Methanol can be readily reacted with steam at around 300°C to form hydrogen over copper or zinc-based catalysts. The major advantage of methanol is that it is a liquid fuel, allowing existing fuel distribution infrastructure to be readily utilized. The chemical processing conditions for methanol reforming are also less severe than for SMR. A significant disadvantage of methanol for distributed hydrogen production is that it is produced from reforming of natural gas, and the extra processing steps make it more expensive and less thermally efficient that SMR. Methanol reforming provides an approximate net production of only 1.7 moles of H 2 per mole of CH4 (Stott and von Linde, 2000). Methanol has also been studied for on-board reforming to power fuel cell vehicles. • Gasification of coal: Coal or biomass can be gasified with oxygen and steam to produce synthesis gas rich in H 2 . This process can potentially produce large amounts of H 2 very cheaply, but with large carbon emissions unless coupled with C0 2 sequestration. 5 • Ammonia cracking: See reaction #5 in Table 1.2. Ammonia can be catalytically cracked to nitrogen and hydrogen. However, once again methane is conventionally used to make ammonia, so the use of ammonia is likely to have lower thermal efficiency than SMR. 1.2 Steam Methane Reforming In addition to hydrogen production, the SMR process is important for production of synthesis gas for production of methanol, Fisher-Tropsch synthesis and other economically important petrochemical operations (Rostrup-Nielsen, 2002). The steam reforming process has been employed for many years to produce hydrogen, originally using naphtha as a feedstock, but now more commonly with natural gas. In the SMR process, natural gas (methane) is reacted with steam over a catalyst, generally nickel-based, to form reformate, a gaseous mixture of CFL, H 2 , CO, C 0 2 and water. Reviews of the steam reforming process have been published by Rostrup-Nielsen (1984) and Ridler et al. (1996). The SMR process is highly endothermic. The first three reversible reactions presented in Table 1.3 are the most common reactions in the SMR process. Note that only two of the first three reactions are independent. Formation of elemental carbon (reactions 4 and 5) is possible, but is generally avoided by operating with a sufficiently high steam-to-carbon molar ratio. Reactions 6 to 9 are not applicable in conventional SMR reactors, as there is no oxygen. However, air was added to the reformer in this work to supply heat in-situ (autothermal reforming). As the intrinsic reaction kinetics are rapid in the presence of reforming catalysts, the conventional SMR process generally operates close to chemical equilibrium. As it is an endothermic process, high temperatures favour conversion. Typical SMR reactor operating temperatures are in the range of 750 to 900°C. From the net reforming reaction (reaction 3 in Table 1.3), it can be seen that extra moles of gas are produced, and thus by Le Chatelier's principle, increasing pressure tends to favour the reverse reaction. In practice, conventional reformers are operated at pressure (typically 2 MPa, but ranging from 1 to 4 MPa) in order to produce a compressed product and to reduce the size of the processing equipment. When operated for hydrogen production, a steam-to-carbon molar feed ratio between 2.5 and 3 is commonly used (Armor, 1999). Figure 1.4, produced with a commercial process simulator (HYSYS, Gibbs free energy minimization), presents the SMR equilibrium methane conversion as a function of temperature and pressure. This highlights the need for elevated operating temperatures in conventional primary reformers to achieve high methane conversion. 6 Increasing the steam-to-carbon ratio increases the equilibrium methane conversion modestly, as shown in Figure A 1.1, Appendix 1. However, increasing steam usage tends to decrease overall energy efficiency. Table 1.3: Common SMR reactions (all species are gaseous) Reaction ' A H 2 9 8 (kJ/mol) Notes 1 CH, + H 2 0 <-> CO + 3H2 206 Methane reforming 2 CO + H 2 0 ~ C 0 2 + H 2 -41 Water gas shift 3 CH 4 + 2H 20 <-> C 0 2 + 4H2 165 Net SMR reaction 4 2CO ^ C + C 0 2 -172 Disproportionation 5 CH4 <-> C + 2H 2 75 Decomposition 6 CH4 + V2O2 — CO + 2H2 -36 Partial oxidation 7 CH, + 1V2O2 -»• CO + 2H20 -519 Partial CH4 combustion 8 C H 4 + 20 2 -*• C 0 2 + 2H20 -802 Total C H 4 combustion 9 H 2 + '/202 - » H 2 0 -242 Hydrogen combustion 0 I , r , — ' 500 600 700 800 900 Temperature (°C) Figure 1.4: Equilibrium methane conversion in SMR as a function of temperature and pressure (steam-to-carbon molar ratio = 3; no inert gases or oxygen present) Conventional SMR catalysts are sensitive to sulfur, and thus a gas desulfurization step is performed prior to feeding the reformer. The endothermic reaction heat is supplied to banks of reactor tubes, typically ~12 m long, located within a fired furnace. The gas produced in the reformer contains significant amounts of CO and therefore shift reactors are installed after the primary reformer to convert CO to C 0 2 and H 2 (water gas shift, reaction 2 in Table 1.3). Hydrogen is then separated from 7 the gas mixture in a pressure swing adsorption (PSA) system or an amine scrubber. A methanation or preferential oxidation reactor may also be installed to convert trace CO to C O 2 (Cromarty and Hooper, 1996). Figure 1.5 presents a SMR process schematic for a typical large-scale facility. Feed pretreatment Reformer High temperature Low temperature Hydrogen 1—r— . \ shift reactor shift reactor purification ] r U 6 l / A i r y j j j I Steam Figure 1.5: Conventional SMR process schematic (adapted from Stitt et al., 2000) Depending on the heat recovery scheme and accounting for gas burnt in the furnace and steam export, conventional SMR systems produce approximately 2/2 moles of H 2 per mole of methane consumed (Stoll and von Linde, 2000). Steam is typically exported from the SMR system, as it is difficult to utilize all of the waste heat due to temperature pinches in the process. The energy efficiency of a modern large-scale SMR plant can reach 85%-90% when the export steam is credited, but values in the range of 75% are more common (Ogden, 2001). Thermal efficiency is significantly lower (-60%) if only the energy value of the product hydrogen is compared to the feed natural gas (Spath and Mann, 2001). 1.2.1 SMR Catalysis Armor (1999) lists nine steps in the SMR process that use catalysts (see Table A 1.1 in Appendix 1). In the reformer proper, the most common type of catalyst is nickel (12-20% as NiO) supported on alumina or ceria. A variety of promoters, including potassium and calcium, are often used to reduce carbon formation. The catalyst is commonly formed into rings or spoked wheels in order to minimize pressure drop and enhance mass and heat transfer within the fixed bed. SMR catalysts are an established, cost-effective product, but have some drawbacks including Ni sintering, carbon formation and heat and mass transfer constraints (Rostrup-Nielsen, 1984). The NiO catalyst must be 8 reduced prior to use. This is typically carried out industrially in a H /^steam or CFL/steam mixtures at elevated temperature. Tables A 1.2 to A 1.4 in Appendix 1 summarize a number of the SMR kinetic equations proposed by Xu and Froment (1990). For autothermal reforming, Ni-based catalysts have been shown to lose activity through sintering and other mechanisms (Qi et al., 2005). An autothermal SMR catalyst must be capable of supporting the SMR reaction (reduction) and be able to withstand oxidation conditions. New catalysts are being developed, mostly based on noble metals such as Pt, Pd, Rh and Ru, supported on alumina or ceria. Both a conventional SMR catalyst (NiO) and a new ATR (proprietary noble metal) catalyst were used in the ICFBMR pilot work. However, as the main objective of this study was investigation of the reactor configuration, and in view of the proprietary nature of the ATR catalyst, details on the catalysts in this thesis are limited. 1.2.2 Limitations of Conventional S M R Although it is a mature and common technology, there are several areas where the SMR process could be improved, including (Roy, 1998): • Catalyst performance: Improved catalysts have been studied to reduce the potential for carbon formation and increase resistance to sulphur poisoning. • Fixed bed reactor: The conventional SMR reformer consists of large number of long, thin reactor tubes arranged in parallel in a fired furnace. The tubes are filled with high-voidage catalyst pellets in order to limit the reactor pressure drop to a reasonable value. o Catalyst effectiveness: Because of mass transfer constraints within the pellets, catalyst effectiveness is very low, with up to 95% of the catalyst reported not utilized (Soliman et al., 1988). Chen et al. (2005) demonstrated the advantages of fluidized bed catalysts over fixed bed catalysts. o Reactor heat transfer: Heat transfer from the furnace to the fixed bed reformer bed is relatively low. In order to keep the temperature profile relatively constant within the reformer, small diameter (50-125 mm) reactor tubes are required, resulting in a large number of parallel tubes. • Unit operations: The SMR process involves a number of sequential reactors and other operating units. This results in a large plant footprint and higher capital costs, especially for smaller scale plants. 9 • Thermodynamic limitation of the SMR process: In order to achieve acceptable methane conversion, the reformer must operate at high temperatures. However, this poses some technical challenges, including: o Materials of construction: The reformer tubes run very hot, requiring expensive, high-nickel steel, resulting in finite service life. o Heat recovery: High operating temperatures reduce the amount of heat that can be internally recovered from the process, and thus steam must be exported to achieve high thermodynamic efficiency. Although this may be practical at an integrated refinery or chemical plant, it is not usually feasible for a stand-alone or small-scale plant. o Carbon formation: Increasing the reformer temperature increases the potential for formation of carbon on the catalyst. Relatively poor heat transfer within the fixed reactor bed increases the potential for hot spots, and thus carbon formation. • Hydrogen purity: The hydrogen purity from a typical PSA system is 99.95%, though higher purities are possible at the expense of hydrogen recovery (Biegler et al., 2004). PEM fuel cell applications require higher purity, especially with respect to CO impurities. The novel reactor investigated in this study attempts to address the above limitations, with the exception of catalyst performance. 1.2.3 Small-Scale SMR Scaling down the conventional SMR process to production rates less than about 1,000 Nm3/day is possible and is commercially offered. However, the high operating temperatures and high capital costs of this approach are often thought not to be the best solution for distributed hydrogen fuelling stations (Ogden, 1996, Thomas et al., 1997). Many industrial and academic groups have developed new SMR-based systems for small-scale production, often incorporating some of the following design approaches: • Less severe operation conditions: Operating the reformer at lower temperature (<700°C) permits lower-cost materials and can improve the overall system heat recovery. However, the reactor pressure must be reduced significantly to achieve acceptable conversion. • Alternative reactor geometries: Rather than the traditional fired tubular reactor, other reactor geometries have been proposed, including fixed bed annular reactors and plate-type configurations. • Autothermal reforming: In order to avoid the problems and costs of reactor heat transfer, autothermal reforming has been studied for small-scale SMR. An oxidant, usually oxygen or air, 10 is added to the reactor to allow adiabatic operation. Both direct oxidant addition and 02-permeable membranes have been studied. • CO2 removal: An adsorbent such as calcined limestone or dolomite can be incorporated to trap CO2 within the reactor (Yi et al., 2005). The spent absorbent can be regenerated, potentially producing a high-purity G0 2 stream that would be appropriate for one of several sequestration options, thus removing the carbon penalty of the SMR process. In-situ C 0 2 removal can also shift the reaction equilibrium forward. Fluidized bed reactors may be used to move absorbent particle between a reformer and regenerator (Johnsen et al., 2006). • H2 membranes: Hydrogen-permeable membranes have been extensively studied in the SMR process. One approach is to simply replace the conventional H 2 purification stage (PSA) with H 2 perm-selective membranes. This reduces the footprint of the plant and produces a high-purity hydrogen stream (Edlund, 1999). • H2 membrane reactor. A more interesting approach (from a chemical engineering perspective) to the previous option is to place the H 2 perm-selective membranes directly in the reaction zone, creating a "membrane reactor". In addition to producing high-purity hydrogen, removal of hydrogen from the reaction zone shifts the chemical equilibrium forward (Adris et al., 1994). Benefits of this approach are detailed in the next section. 1.2.4 Membrane Reforming Selectively adding a reactant or removing a product to a reacting system with a species-selective membrane can significantly enhance the reactor characteristics for some processes. Thermodynamically limited reactions, such as SMR, are often candidates for membrane reactors, as addition or withdrawal of a species can favourably shift the chemical equilibrium. Removal of hydrogen from the SMR reactor significantly enhances the equilibrium conversion of methane, as illustrated in Figure 1.6, prepared from output from a commercial process simulator (HYSYS, Gibbs free energy minimization). The lower curve is the SMR equilibrium with no hydrogen removal. The upper curves represent equilibrium conversion for several hydrogen recoveries, represented as the molar ratio of hydrogen removed to methane in the reactor feed. As progressively more hydrogen is removed from the reacting system, conversion of methane is increased, permitting high conversions at significantly lower operating temperatures than in conventional SMR. Ceramic-based and dense metallic membranes are the most common types of hydrogen permeable membranes employed in SMR research. The membranes used in this study are thin metallic membranes based on palladium. Dense palladium membranes have the ability to produce ultra-pure 11 H 2 , as only atomic hydrogen can diffuse through the metal lattice. Details on H2-permeable membranes are provided in Chapter 2. SMR membrane reactors constitute an active field of research (see recent review by Uemiya, 2004). Most work has focussed on fixed bed reactor configurations. Reviews of past palladium-based membrane reactor research are given by Shu et al. (1995), Gryaznov (2000), Paglieri and Way (2002) and Dixon (2003). The most prominent commercial membrane reformer demonstration to date was undertaken by Tokyo Gas, who recently operated demonstration fixed bed reformer units in Japan with a hydrogen production capacity of the order of 40 Nm3/h (Mitsubishi Heavy Industries Ltd.; Yasuda et al., 2004). Their design appears to be based on fixed catalyst bed monolith sandwiched between planar, palladium-alloy membranes. 0.0 I : 1 1 ! 500 600 700 800 900 Temperature ( °C ) Figure 1.6: Equilibrium methane conversion in SMR as a function of temperature and in-situ hydrogen removal (1,000 kPa, steam-to-carbon molar ratio = 3; no inerts or oxygen present) H2-permeable membranes eliminate the need for shift reactors and hydrogen purification. Figure 1.7 contrasts simplified process flow schematics for (a) conventional SMR, (b) a SMR membrane reactor and (c) an autothermal membrane reactor. Compression is required for the membrane reactor configurations to bring the product hydrogen pressure up to conventional SMR levels. Further compression is required for all three configurations to raise the hydrogen pressure to current fuel cell vehicle storage standards (> 3,000 kPa). 1.2.5 Fluidized Bed Membrane Reactor (FBMR) Fluidized bed reactors have some characteristics that are favourable for implementing a membrane reactor in SMR service (Adris et al., 1996, Grace et al., 2005, literature review by Deshmukh et al., 12 2007). A single reaction vessel can be used, rather than the multiple reactor tubes in the conventional SMR process. Other advantages of fluidized bed reactors over fixed bed configurations are: • Higher external heat transfer coefficient: The SMR process is highly endothennic and significant amounts of heat must be transferred to the reactor, even for small-scale units. Fluidized beds have higher surface-to-bed heat transfer coefficients, typically in the range of 250 to 400 W/m2K (Kunii and Levelspiel, 1991), approximately an order of magnitude better than for the same particles in fixed beds. The average heat flux in a modern tubular reformer is less than 100 W/m2K (Usami et al., 2003). The higher heat flux in the fluidized bed can reduce the heat transfer area and temperature driving forces as compared to fixed bed reactors. • Reducing temperature gradients: Due to the high rate of particle mixing and high particle surface area, temperature gradients in fluidized beds tend to be very low. This reduces hotspots in the reactor, decreasing the potential for coke formation. For autothermal reforming, the heat released by direct oxidant addition can be rapidly dissipated in the fluidized bed, reducing local temperature gradients. Another feature of fluidized beds is their ability to effectively move heat in the reactor through movement of catalyst particles. • Catalyst effectiveness: In order to fluidize the catalyst bed at reasonable gas velocities, catalyst particles are small (typically 75 to 250 um diameter). Such small catalyst size virtually eliminate the internal diffusion limitations of fixed bed catalysts. Pressure drop over the fluidized catalyst bed is also typically very low. • High bed to membrane mass transfer: Analogous to high heat transfer, fluidized beds exhibit high bed-to-surface mass transfer, preventing bulk mass transfer from being a significant resistance in fluidized bed membrane reactors. Disadvantages of applying fluidized beds to small-scale SMR include: • Fluidized beds tend to be best suited to large scale applications. • Bubbling fluidized bed reactors tend to deviate from plug flow, potentially leading to inefficient contacting of reactants. • Entrainment of catalyst fines from the reactor must be addressed. • Wastage of vessel internals, such as hydrogen permeable membranes, through particle abrasion can be problematic. • Catalyst attrition can occur due to particle-particle and particle-wall collisions. 13 Natural gas>-Water ^>-Sulfur removal Heat exch. / steam gen. Burner PSA purge gas Heat "1 Reformer Shift reactors A. Conventional SMR H 2 separation (PSA) - Flue gas^> - Steam exporfr H 2 producT)> Natural gas>-Water ^> Oxidant Natural gas>-Water > (start-up only?) Sulfur removal Sulfur removal ± Heat exch. / steam gen. Burner Heat Reactor off-gas Reformer Membrane permeate H 2 compression B. Membrane SMR Heat X Burner ± Reactor off-gas Heat exch. / steam gen. Reformer Membrane permeate H 2 compression C. Autothermal membrane SMR Flue gas ^ > H 2 product^ Flue gas ^ H 2 product^ Figure 1.7: Simplified process schematics for conventional SMR, membrane SMR and membrane A T R Recognizing the benefits of combining a SMR fluidized bed with membranes, Adris (1994) undertook a pilot plant study, using dense palladium tubes immersed in a bubbling fluidized bed reactor. High purity hydrogen was produced and a small equilibrium shift was demonstrated. Subsequent modelling showed the potential benefits of the process (Adris and Grace, 1997, Abba et al., 2003, Abashar et al., 2003). The process was also patented (Adris et al., 1994, Adris et al., 2002). Roy (1998) utilized the same pilot reactor as Adris to study thinner, higher-flux membrane tubes. Direct injection of oxygen into the fluidized bed was also tested and found to be practical and safe. Direct oxidant addition (autothermal reforming) permits the SMR reactor to run adiabatically, eliminating external heat transfer. An economic evaluation of the process was also performed (Roy et al., 1998). Following the work of Adris and Roy, a company called Membrane Reactor Technologies Ltd. was started to commercialize the FBMR technology. MRT is developing an autothermal fluidized bed membrane reactor in the capacity range of 15-50 Nm3/h (Klassen, 2005), the configuration of which differs from the internally circulating geometry studied in this work. Others, e.g. Jarosch et al. (1999) and Elnashaie and co-workers (Prasad and Elnashaie, 2002, Chen and Elnashaie, 2005), have proposed circulating fluidized bed membrane reformer designs. In these configurations, catalyst solids circulate between a reforming reactor containing hydrogen permeable membranes and a second vessel where heat is added. Patil et al. (2005) proposed a bubbling bed reactor with separate membranes to add oxygen and remove hydrogen. Patil (2005) published experimental data from a fluidized bed reactor with tubular Pd membranes, though operation coupled with oxygen membranes was not performed. 1.3 Internally Circulating Fluidized Bed Membrane Reactor The fluidized bed system has some distinct advantages for membrane SMR compared to fixed bed configurations. The work of Adris (1994) and Roy (1998) demonstrated the feasibility of incorporating membranes into a fluidized bed SMR reactor, and indicates that oxygen addition may be practical. In considering ways of developing a larger-capacity, potentially commercial process, some challenges in the existing FBMR configuration were identified: • Separation of oxidant addition and membranes: Although heat is dissipated quickly in fluidized beds, high temperatures are present locally when oxidation gas is injected as jets. As palladium-based membranes havea practical upper operating temperature of ~650°C, it is desirable to keep the membranes well away from the oxidation zone. 15 • Internal heat transfer: Membranes within the reactor impede the movement of catalyst and gases, leading to larger temperature gradients within the fluidized bed. This becomes more acute with larger reactor diameters and increased packing of membranes within the bed. • Nitrogen in oxidant air: Oxygen can be used as an oxidant for autothermal reforming, but an oxygen separation unit would then be required, adding complexity and cost for small-scale hydrogen installations. If air is the oxidation gas, the nitrogen entering with the air can significantly dilute the reactor gases. If the system is well mixed, this reduces the hydrogen partial pressure in the reactor and thus decreases the membrane flux of hydrogen. For autothermal operation with air addition, nitrogen could account for almost 50% of the gases leaving the reactor. For a well-mixed reactor, this nitrogen represents a significant reduction in the H 2 driving force over the membrane, and more membrane area would be required to achieve the desired hydrogen flux. • Tubular membranes: The sealing, manifolding and mechanical support of many small diameter palladium tubes are challenging. This study extends the research of Adris and Roy by proposing a new reactor configuration, the internally circulating fluidized bed membrane reactor (Boyd et al., 2005). A patent on this reactor concept has been granted (Grace et al., 2006). When adding oxidant to the fluidized bed membrane reactor, there are two distinct reaction zones within the bed, a reforming/hydrogen permeation zone and an oxidation zone. It is desirable to keep these zones separate to protect the membranes, but heat must then be transferred from the oxidation zone to the reforming/permeation zone. However, it is undesirable for nitrogen from the oxidation air to reach the reforming/permeation zone, as the reduced hydrogen partial pressure would diminish the membrane flux. The challenge of separating two distinct reaction zones within a fluidized bed system is not unique and is often addressed by moving catalyst solids between two vessels, for example in fluid catalytic cracking (Kunii and Levenspiel, 1991) and fluid coking. Internally circulating beds for non-membrane reactors have also been studied (e.g. Jung et al., 2004). The reactor concept presented here is an internally circulating fluidized bed reactor. To achieve this, the reactor internals of the fluidized bed membrane reactor were significantly modified (Figure 1.8): • Planar membranes are employed instead of tubular membranes. • The planar membranes are oriented vertically within an open-ended draft box in the core of the reactor. 16 • Oxidant air is added to the fluidized bed above the draft box, well away from the membrane panels. Effluent Gases Fluidized catalyst bed Core draft box Planar membrane J - -[*^  modules Permeate H 2 I Oxidation air External heat (if used) t Reactants (natural gas & steam) Figure 1.8: ICFBMR schematic The steam and methane reactants are fed to the core draft box via the reactor windbox. The reforming reaction proceeds in the core box and hydrogen is produced. The membranes withdraw hydrogen from the reactor as it is produced, shifting the reforming reaction forward. The upward flow of gas in the core box induces internal circulation of catalyst up the core box and down the annular regions, as indicated in Figure 1.8. Air is introduced to the fluidized bed above the core draft box and reacts with the residual reformate, releasing heat. The partially oxidized reformate disengages from the catalyst particles in the freeboard and leaves the reactor through a fdter or cyclone. Circulating hot catalyst particles from the oxidation zone carry the heat from the oxidation zone to the core reforming zone via the outer annular zone, which acts like a downcomer. If desired, external heat can also be transferred to the circulating solids through the reactor wall, as shown in Figure 1.8. If the performance is as anticipated, the catalyst returning in the annular regions should drag very little gas 17 downward, ensuring that the nitrogen entering the reactor with the oxidation air does not significantly dilute the gases in the core reforming/permeate zone. Figure 1.9 presents a plan view of the reactor. The square draft box allows a simple layout for the planar membranes, but other configurations, such as a circular draft tube, could also be implemented. Reactor shell Double-sided membrane panels Core draft box Downward catalyst flow in annular region Upward catalyst / gas flow in core slots H2 permeation through membranes Figure 1.9: ICFBMR plan view Planar membranes made from thin palladium alloy foil were used in this study. These metallic membranes theoretically permeate only atomic hydrogen, producing a high-purity hydrogen product. The planar membranes were fabricated by techniques developed by Membrane Reactor Technologies Ltd. Background on hydrogen perm-selective membranes is given in Chapter 2. 1.4 Research objectives The overall objective of this study is to evaluate the potential of the I C F B M R concept for production of high-purity hydrogen at a scale appropriate for the developing distributed hydrogen market. Specific objectives include: • Investigate the hydrodynamics of the ICFBMR concept using a cold model. • Fabricate novel planar membranes from thin palladium alloy foil. 18 • Pilot the ICFBMR process at a semi-industrial scale. Investigate reactor and membrane performance under both SMR and autothermal reforming conditions. • Simulate the ICFBMR process in sufficient detail to permit the design of an industrial-scale system. • Evaluate the economics and commercial potential of the ICFBMR process. 1.5 Thesis outline Chapter 2 introduces the various types of hydrogen permeable membranes, and provides more details on dense, palladium-based membranes. The design and fabrication of seven, double-sided membrane panels using proprietary techniques of Membrane Reactor Technologies Ltd. are then described. The hydrogen flux characteristics of the membrane panels were determined prior to their installation in the ICFBMR pilot reactor. Efforts to simulate the hydrodynamics of ICFBMR reactor in a cold Plexiglas model are described in Chapter 3. Fluidization characteristics and scaling parameters for the particles tested in the cold model and ICFBMR pilot reactor are discussed. The cold model Plexiglas reactor was assembled, filled with FCC particles and fluidized by air. Internal solids circulation rates were determined as a function of air flow and membrane panel geometry. Design equations are developed to correlate circulation rate as a function of fluidization velocity. The process and mechanical design basis for the new ICFBMR pilot reactor is presented in first part of Chapter 4. Reactor fabrication details and assembly instructions are described, as are the experimental methods and operating procedures for the modified pilot plant. The latter section of Chapter 4 describes the results from seven ICFBMR pilot plant runs. Both SMR and ATR experiments were performed. Despite operational challenges, high-purity hydrogen was produced from the ICFBMR and catalyst circulation confirmed. Observations from the dismantling of the ICFBMR reactor after operation are given in Chapter 5. The six membrane panels were inspected for damage and tested for leaks. Flux test data from two membrane panels are summarized, showing that permeation rates were slightly higher than before the pilot reactor testing. The foil from one membrane was removed and analyzed. SEM images show significant metallic scaling on the backside of the membrane foil, thought to be the main cause of the reduction in hydrogen permeation below predicted values. 19 Chapter 6 presents process simulation and economic analyses of the ICFBMR process. The reactor was simulated with commercial process software (HYSYS). The effects of major process variables on the ICFBMR design were investigated. A design for a 30 Nm3/h ICFBMR system was developed and costs are estimated. Economic analysis indicates that the H 2 permeable membranes are a significant capital and operating cost expense. Chapter 7 includes the main conclusions of our study, and recommendations for future work. 20 Chapter 2. Hydrogen Selective Membranes This chapter provides background on various types of hydrogen permeable membranes, including palladium-based membranes, which have the unique feature of being able to produce ultrapure hydrogen. The design, fabrication and testing of the novel planar palladium membranes used in the ICFBMR pilot reactor are then described. Seven double-sided membrane panels were successfully prepared, each using 50 pm thick palladium alloy (25% silver) foil The palladium alloy foil was bonded to a stainless steel substrate using a proprietary bonding technique (patent pending) of Membrane Reactor Technologies Ltd. (MRT). The hydrogen flux characteristics of each membrane panel were measured before installation in the ICFBMR pilot reactor. 2 . 1 Background Membrane research is a rich and growing field of study, especially in the area of process intensification, where membranes are employed to improve performance of reacting systems by addition or withdrawal of selected chemical species. Applications of membrane reactors vary widely, and include biological, electrochemical and high temperature catalytic processes. The form of membrane systems differs significantly, depending on the permeate characteristics and the reactor operating conditions. The two types of membranes that have been most commonly applied S M R process are hydrogen permeable membranes and, to a lesser extent, oxygen permeable membranes. In the SMR process, selective removal of hydrogen favourably affects the thermodynamic equilibrium and produces pure hydrogen directly. Oxygen permeable membranes have been studied as a means of supplying an oxidant to support autothermal reforming. In reviewing the variety of hydrogen permeable membranes that can be applied to the SMR process, the following desirable characteristics and constraints should be considered: 21 • Hydrogen selectivity: Ultrahigh purity or PEM-fuel-cel l quality hydrogen is preferred. • Permeability: High hydrogen permeability is desirable to lower membrane costs and to reduce reactor volume. • Temperature: The conventional S M R process operates at 700 to 900°C due to the equilibrium constraints of the chemical equilibrium. B y using hydrogen permeable membranes, S M R operating temperatures can be reduced. Nevertheless, membranes need to be able to withstand operating temperatures in the range of 500 to 650°C. • Pressure: The conventional S M R process is pressurized in order to reduce equipment costs and produce a high-pressure hydrogen product. Due to an increase in the number of moles, high pressure shifts the chemical equilibrium towards reactants. B y removing hydrogen in-situ, the negative effect o f pressure on the chemical equilibrium can be reduced. High pressure is often preferred for membrane reactor operation, as it increases the hydrogen membrane flux for a given reactor gas composition. Though the I C F B M R pilot plant is limited to 1,500 kPa, membranes for the I C F B M R process would ideally be able to operate with higher trans-membrane pressures, perhaps up to 3,000 kPa. • Chemical stability: The membranes must be chemically stable to the main gas species present in the S M R process, including carbon monoxide and water, and to the fluidized S M R catalyst. Susceptibility to trace gas components that might poison membrane surfaces, such as sulphur, is an important consideration for long-term service. The membranes used in this work were fabricated from palladium alloy fo i l , and have the potential to meet most, i f not a l l , of the above criteria. 2.1.1 Oxygen Permeable Membranes This work investigates hydrogen permeable membranes for both conventional S M R and autothermal reforming. However, the published literature on high temperature oxygen permeable membranes is also reviewed briefly, as the I C F B M R concept is especially suited to autothermal reforming and could be modified to incorporate oxygen permeable membranes, rather than direct air injection as studied in this work. Oxygen permeable membranes operating at high temperature have been extensively studied, most commonly for partial oxidation reactions, with several systems involving a fluidized bed reactor (i.e. Mleczko et al., 1996, Deshmukh, 2004). Both porous ceramic membranes and ion conducting ceramic membranes have been proposed for methane oxidation and syngas production. A review by L i n (2001) provides an overview of research in this area. 22 Fluidized bed membrane reactor concepts with both hydrogen and oxygen permeable membranes have been recently proposed (Pradeep and Elnashaie, 2002, Patil et al., 2005). The potential advantages of oxygen permeable membranes over than direct oxygen or air addition include: • If air is used, nitrogen in the oxidation air dilutes the reactor gases, reducing the hydrogen partial pressure and permeation driving force over the hydrogen permeable membranes; • The reactor off-gas (i.e. non-permeate stream) would contain significant amounts of nitrogen, making carbon dioxide sequestration difficult; • If direct oxygen addition is used, the need for upstream air separation is eliminated; • If an ion-conducting oxygen membrane is used, the need for compression of oxygen or air is eliminated. The ICFBMR reactor concept addresses the first point, as oxidation air is added to the top of the reactor, thus greatly reducing the amount of nitrogen circulating to the hydrogen permeable membranes. As membrane-assisted reforming is likely to be first commercialized on a small scale, the carbon credits of sequestering the reformer off-gas are likely to be very modest. The most tangible benefit of oxygen permeable membranes may be the last point, as a simple air fan can potentially replace an expensive multi-stage air compressor, but this saving must be balanced against the cost of the oxygen membranes. Perhaps the most extensively studied oxygen ion-conducting membranes are perovskite-type, zirconia and bismuth oxide based ceramics (Lin, 2001). In perovskite-type ceramics, oxygen permeates as an ion, and the driving force is the difference in oxygen partial pressure over the membrane. Therefore oxygen from an atmospheric air supply can permeate through the membrane into a high-pressure reactor, if the oxygen is rapidly consumed in the reactor. There has been considerable patent activity in this field since the early 1990s, much of it led by commercial interests (i.e. Amoco: Balachandran, et at, 1997; Praxair: Dyer et al., 2000; Eltron: Schwartz et al., 2000). The major drawback to integrating perovskite-type oxygen membranes into a SMR membrane reactor is that these membranes typically operate at around 900°C, significantly hotter than the practical operating limit for palladium membranes (< ~650°C). Combining two distinct temperature zones, one for palladium (H2) membranes and the other for perovskite (02) membranes, into one reactor, as proposed by Patil et al. (2005), is challenging. Another drawback of oxygen membranes is the extra reactor volume required to house the membranes. However, as development in oxygen permeable 23 membranes progresses, incorporation o f oxygen permeable membranes in the I C F B M R design may become practical. 2.1.2 Types of Hydrogen Permeable Membranes Many types of membranes have been used for gas-phase hydrogen separation. These are broadly summarized in Figure 2.1. The hydrogen permeable membranes used in this work are dense (i.e. non-porous) metal foils based on palladium, the most commonly studied metallic membrane material. Hydrogen permeable membranes I Organic (polymeric) 1 Inorganic ~ T ~ Proton Gaseous H 2 "Ceramic" conducting separation Glass (microporous) Metallic (dense) Composite • microporous silica • microporous alumina • polycrystalline zeolites • perovskite (proton-conducting) Figure 2.1: Types o f hydrogen permeable membranes (Islam 1997, L i n 2001 and L i n et al., 2002) Polymeric hydrogen membranes, such as hollow fibre membranes have been commercialized, and they have been applied to membrane reactors (Vankelecom et al., 2000), but cannot be used at the elevated temperatures required for the S M R process. In contrast, inorganic membranes have the potential to be used at elevated temperatures for hydrogen separation, but until recently, have remained mostly in the realm o f research and development. O f the inorganic membranes, microporous ceramic and metallic based membranes, or a combination o f the two (composite membranes), have been most widely applied to membrane reactors. Figure 2.2 schematically shows the permeation mechanism of dense metallic membranes, as used in this work, and microporous membranes. Dense metal membranes permeate hydrogen as protons through the metal lattice structure, and are therefore theoretically 100% selective for hydrogen. This ability to produce ultrapure hydrogen is the major feature o f dense metallic membranes. 24 Dense Metal Membrane Porous Membrane Purified H2 Membrane] y V Figure 2.2: Hydrogen permeation through dense metallic and porous inorganic membranes (adapted from Uemiya, 2004) 2.1.3 Ceramic Hydrogen Permeable Membranes Inorganic ceramic membrane research has been reviewed by Coronas et al. (1999) and Lin et al. (2002). Inorganic ceramic membranes for hydrogen separation can be broadly grouped into three classes (Lin, 2001): • Dense ceramic proton-conducting membranes. • Polycrystalline zeolite membranes. • Microporous amorphous membranes. Dense, proton-conducting inorganic membranes, often ceramics mixed with metals (cermet), have been proposed for hydrogen separation. These membranes have the benefits of near 100% hydrogen selectivity as the membranes have no porosity (Morreale et al., 2005). However, microporous amorphous membranes have been the most common type of ceramic membrane applied to high-temperature reactors. These membranes feature very small pores, typically smaller than 2 nm, and generally separate hydrogen from heavier gases based on molecular size. Many formulations have been tested, with the most commonly on alumina, zirconia, silica and other metal oxides. Some of the benefits of microporous ceramics compared to dense metallic membranes include relatively high hydrogen permeance, much lower cost base materials and, in some cases, good mechanical, chemical and thermal stability at high operating temperatures. The major drawback of microporous ceramics is hydrogen selectivity, as a porous membrane always permeates some non-hydrogen species. Sealing of brittle ceramic membranes, typically in tubular form, in a pressurized, high temperature reactor can be a difficult, though surmountable, engineering task. Chemical stability of some ceramics in the presence of water, as is abundant in the SMR process, has also been a 25 challenge (Imai et al., 1997). As a result, hydrogen permeable ceramic membranes have been more commonly applied to membrane reactor processes where there is little or no water, such as in dehydrogenation (i.e. Ziaka et al., 1993), dry reforming (i.e. Onstot et al., 2001, Galuszka et al., 1997) and thermal decomposition of hydrogen sulphide (i.e. Ohashi et al.). Examples of ceramic microporous membranes applied to the SMR process include works reported by Minet et al. (1992), Nijmeijer (1999) and Tsuru et al. (2004). Microporous amorphous ceramic membranes are commonly prepared as very thin films (typically between 20 nm to 50 (xm), often silica-based, deposited on a porous support. The support is usually inorganic, commonly alumina or zirconia, and provides the mechanical strength for the membrane. Sol-gel techniques remain the most successful method for depositing the silica membrane layer, typically on a high-quality sol-gel support. Figure 2.3 shows a common strategy of preparing a ceramic membrane with layers of progressively smaller pore sizes capped with a thin final layer of very small pore diameter material. 0"* 100 CO E 3 99 o > u. 98 O Q. > 97 E 96 o 95 10 100 - I ~> I -- t Bulk M e m b r a n e S u p p o r t - layer First -support layer _ Second s u p p o r t l layer i 1,000 Pore diameter (A) 10,000 100,000 Figure 2.3: Four-layer alumina membrane (from Lin et al., 2002, after Hsieh et al., 1988) Gas transport through ceramic membranes may involve a mesoporous support (20-1000 nm thick), where flow may be controlled by viscous flow, molecular diffusion or Knudsen diffusion. If Knudsen diffusion is controlling, the ideal separation factor between two gases is the ratio of the square roots of the molecular weight of the gas species (Kast and Hohenthanner, 2000): Separation factor = JMW2 (2.1) From equation 2.1, the ideal separation factor for hydrogen and methane under Knudsen diffusion is therefore 2.8. Though fluxes through these types of ceramic membranes may be reasonably high, hydrogen separation factors are certainly too low to produce fuel-cell-quality hydrogen from the SMR process. 26 1 For very small micropores, perhaps less than ~5 nm diameter, transport is controlled by configurational effects, such as molecular-sieving or adsorption processes (see Benes et al., 1999 and Kast and Hohenthanner, 2000 for reviews of microporous transport), and higher gas separation factors can be achieved. Achieving defect-free layers of this pore diameter is challenging. Despite this, some groups have reported separation factors for H 2 :N 2 as high as 3,000 (Nam and Gavalas, 1989, Gavalas etal., 1989). 2.2 Metallic Hydrogen Membranes Metallic membranes have been extensively studied for membrane reactor application due to their very high gas selectivity. In theory, palladium has infinite selectivity for hydrogen, and its application in high temperature membrane reactors is extensive (Shu et al., 1991). This selectivity to hydrogen is important in the context of hydrogen production for PEM fuel cells, where carbon monoxide limits are very low, on the order of 0.2 ppm. Other metals have been proposed for hydrogen separation, such as niobium and tantalum (Buxbaum and Kinney, 1996), but surface oxides on these refractory metals have generally precluded their use for hydrogen membranes. Palladium and its alloys are currently the most practical metallic membrane materials for hydrogen separation (Uemiya, 2004). However, the purity of hydrogen produced with palladium comes with a cost, as it is an expensive material with limited mechanical strength at the temperatures present in the SMR process. Maintaining high-purity hydrogen, but at a thinner layer of metal to reduce cost and increase flux, is one of the major goals in palladium membrane research. 2 .2 .1 Dense Palladium-Based Membranes Palladium has been known to absorb and permeate hydrogen since first observed in the mid-19th century. Application of palladium to membrane reactors accelerated in the late 1960's, initially centred in the Soviet Union, and later in Japan, with hydrogenation and dehydrogenation reactions being the most commonly studied processes. Extensive reviews of past and present palladium membrane research have been given by Paglieri and Way (2002), Gryaznov (2000) and Shu et al. (1991). Hydrogen is transported atomically through dense metallic membranes through a solution-diffusion mechanism. Hydrogen permeation is a multi-step process, generally acknowledged to include: • Gas transport of hydrogen molecules to the upstream surface of the metallic membrane; • Reversible chemisorption of hydrogen molecules on the metal surface; • Reversible dissolution of atomic hydrogen on the membrane surface into the bulk metal; 27 • Diffusion of atomic hydrogen through the metal lattice of the membrane; • Reassociation of atomic hydrogen on the surface of the downstream metal surface; • Desorption of adsorbed molecular hydrogen from metal surface; • Gas transport away from the downstream surface of the membrane; In theory, only hydrogen can be transported through palladium through the above diffusion mechanism, and thus 100% selectivity is conceivable. The presence of other gas species in the permeate gas from a palladium membrane generally indicates a fissure or pinhole in the palladium layer or membrane seals. Producing defect-free palladium layers becomes increasingly difficult with progressively thinner membranes. For the thickness of the palladium alloy membrane used in this work (50 (xm), the rate-controlling step for hydrogen permeation is expected to be bulk diffusion of atomic hydrogen through the membrane. Below a palladium thickness of perhaps 10 urn, other steps of the permeation process appreciably influence the overall hydrogen flux (Ward and Dao, 1999). If bulk diffusion of atomic hydrogen is the rate-controlling step, the steady-state flux (JH) can be approximated from Fick's Law using the concentrations of atomic hydrogen in the palladium (CHh, CHI) on either side of a membrane with thickness r and diffusivity of atomic hydrogen DAH (Shu et al. 1991). J H = — ( C H h - C H 1 ) (2-2) x If atomic diffusion of hydrogen is rate-controlling, the concentration of atomic hydrogen at the surface of the membrane can be assumed to be in near-equilibrium with molecular hydrogen in the gas phase. The atomic hydrogen concentration at the surfaces can then be related to the partial pressure of hydrogen in the gas with Sievert's equation (Shu et al., 1991): CHh = KsPHh ,CHl = KsPHl (2.3) Ks is known as Sievert's constant and follows an Arrhenius relationship with temperature (7). The overall flux can then be described by: J H = ^ e x p ( ^ ) ( P ° h 5 - P ° , 5 ) (2.4) T RT Ep is the apparent activation energy for permeation and kH is the diffusion constant. The permeation flow (QH) can then be calculated by multiplying the specific membrane flux (JH) by the membrane area (Amem). Reported values for Ep and kH for palladium vary. The results of Holleck (1970) (22,000 28 J/mol) are commonly used (Ward and Dao, 1999). From equation 2.4, the effect of some important variables on the hydrogen flux can be seen: • Thickness (r): Hydrogen flux is inversely proportional to membrane layer thickness, providing an incentive to create thinner membranes. However, this must be balanced against the risk of pinholes and cracks. • Temperature (J): Hydrogen flux increases with increasing temperature. Thus it is desirable to operate palladium membranes at the highest practical temperature. • Hydrogen partial pressure: Hydrogen flux increases with increasing reactor hydrogen partial pressure (Pm) and decreasing permeate hydrogen partial pressure (PHI)- Increasing the total reactor pressure can therefore sometimes be effective. In order to reduce the partial hydrogen pressure in the permeate, vacuum or an inert sweep gas can be used. In practice, many researchers have found that the best-fit partial pressure exponent in equation 2.4 is somewhat larger than the theoretical value of Vi, and therefore equation 2.4 is often written with a generic partial pressure exponent («): JH=^eXp{Z^.)(P^-P^) (2.5) t KI Deviation of n from Vi may be due to a number of factors, including rate-influencing surface processes, external mass transfer resistances, blockages of membrane area with structural supports and membrane leaks. A value for n of 1 would be expected if the rate-determining step were due to surface effects. Intermediate values between Vi and 1 are often reported (Ward and Dao, 1991). Below the critical temperature of 293°C, the palladium-hydrogen system can form two different hydride phases (a and P). Palladium readily absorbs significant amounts of hydrogen at elevated temperatures. Operating at temperatures and pressures where the two phases are present can lead to embrittlement and severe distortion (Lewis, 1967, Shu et al., 1991), and for this reason, care must be taken to purge pure palladium membranes with inert gas during heat-up and cool-down. To avoid this phase transition issue, various palladium alloys, most commonly palladium-silver, often replace pure palladium. The critical temperature of the o>p phase transition for Pd7 7Ag23 is near room temperature. Many palladium alloys also demonstrate higher strength and hardness than pure palladium, and do not exhibit some of the severe thermal cycling characteristics of pure palladium, which can result in warping and cracking (Paglieri and Way, 2002). 29 Some palladium alloys show higher hydrogen permeation than pure palladium. Figure 2.4 presents relative permeability data for a number of palladium alloys at 350°C, a much lower temperature than is required in the ICFBMR reactor. Not shown in Figure 2.4 are permeability data for alloys of palladium with rhodium, ruthenium and platinum, which can also exhibit high hydrogen permeabilities (Gryaznov, 2000). 0 10 20 30 40 50 60 Alloy component (wt%) Figure 2.4: Relative permeability of hydrogen in various palladium alloys at 350°C and 2.2 MPa (after Shu et al., 1991 from data of Knapton, 1977) Although Pd6oCu40 is more difficult to apply to the SMR process due to temperature limitations, its resistance to sulphur poisoning is reported to be significantly higher than palladium-silver alloys, and it has been proposed for commercial hydrogen gas purifiers (i.e. Juda et al., 1999). As shown in Figure 2.4 above, palladium-silver exhibits higher hydrogen permeation than pure palladium up to a silver content of about 30% by weight, and reaches a maximum around 23 wt% due to variations of hydrogen solubility (which increases with increasing silver content) and diffusivity (which decreases with increasing silver content). For this reason, and the important practical mechanical issues discussed above, palladium-silver (usually 23 or 25 wt%) is the most common palladium alloy currently used in membrane reactor research. Pd75Ag25 foil is used is this work. It should also be noted that silver is a much cheaper metal than palladium, and thus silver alloys should have reduced material costs compared to pure palladium of the same dimensions. 30 2.2.2 Dense Palladium-Based Membranes: Mechanical Issues Support and sealing of thin palladium foil membranes present difficult engineering challenges, especially when the size of the reactor increases and many membrane foils or tubes need to be manifolded into a vessel, as in the ICFBMR pilot reactor. The form of the membrane (tubular or planar) and the form of the membrane support, greatly influence the reactor design. Some practical mechanical engineering issues that need to be addressed to utilise membranes in a larger scale reactor include: • Membrane sealing: Membrane modules or tubes must be sealed, whether through welds, bonds or gaskets. As ultrapure hydrogen is the major feature of dense metallic membranes, a small seal leak could compromise product quality. Membrane tubes may be welded or braised to a header. Planar membranes may be bonded, braised, welded or gasketed. • Support: The support of a network of membranes, especially small tubes, packed into a reactor can be challenging. Thin palladium-based foil is usually formed by cold rolling thicker sheet. This process can produce defect-free foils down to perhaps 25 urn, though cold-rolled foils of 50 um thickness are more common. Cold-rolled Pd 7 5Ag25 foils of 15 um thickness are being now offered commercially (Klassen, 2005). The rolling process can affect grain size and hydrogen permeation characteristics, and annealing is often necessary (i.e. Tosti et al., 2000). Tubular membranes are typically formed by welding or bonding cold-rolled foils. Although stronger than pure palladium, Pd75Ag25 still has low yield stress at elevated temperatures. Given this low strength, use of unsupported tubes or foils becomes impractical for thin palladium membranes if they are to be installed in a pressurized reactor. Due to mechanical strength, many published palladium membrane reactor research papers have been conducted at relatively low pressure. However, if thin (<75 urn) palladium-based membranes are to be used in practical, pressurized reactors, they must be mechanically supported. For thin-walled, dense membrane tubes, internal springs (Roy, 1998, 75 um, 3 mm OD) and porous ceramics (Gallucci et al., 2004, 50 urn PdAg alloy, 10 mm OD) have been used to withstand external pressure. The internal supports in the above examples were not physically bonded to the palladium membrane. Porous ceramics and expanded metal (Tosti, 2003) have also been used to support planar foils. Whatever the mechanical support, its inclusion into the membrane system can reduce hydrogen permeation through a number of mechanisms, including: • Physically blockage of the permeation area; 31 • Addition of a mass transfer resistance on the permeate side of the membrane; • Chemical modification of the palladium by diffusion of undesirable chemical species, for example nickel, into the palladium from the support. Another consideration for the dense membrane support is thermal expansion. As palladium membranes operate at high temperatures in the SMR process, significant thermal growth is expected. In addition, some swelling of the membrane is expected due to hydrogen absorption. Matching the thermal growth characteristics of the support to that of the palladium can be difficult, but may be important if the palladium foil is bonded to the membrane support, as is the case with the ICFBMR membranes, hence metallic supports may be preferred. The thermal expansion coefficient of palladium and various support materials is presented in Figure 2.5. As shown, the thermal expansion coefficients of stainless steel and alumina are significantly higher than for palladium. If the palladium foil is not bonded to the support, but rather is loosely overlaid, thermal expansion may be less critical, and ceramics may be considered. Inconel 625, a high nickel alloy, has thermal expansion coefficients very similar to palladium, and its use was considered for the membrane substrate for the ICFBMR panels. However, it is expensive and has poor welding characteristics compared to stainless steel. The 50 pm thick, palladium alloy foils used in this work were supported by porous sheets of sintered 316 stainless steel, coated with a thin layer of alumina to prevent inter-diffusion of the support metal with the palladium membrane. 300 400 Temperature ( °C) 700 Figure 2.5: Thermal expansion coefficients of palladium and various membrane support materials (Sources: palladium: Touloukian et al., 1970; alumina: Pal et al., 1999; SS 316: 2004 ASME boiler and pressure vessel code, part D; Inconel: Nicofer® 6020 hMo datasheet) 32 2.2.3 Composite Membranes As discussed above, microporous ceramic membranes cannot produce a pure hydrogen product, unlike dense metallic membranes. Dense metallic membranes typically suffer from lower fluxes than ceramics, as sheets cannot be reliably formed with thicknesses much below 25 um. A compromise between microporous ceramic and dense metallic membranes is the composite membrane, in which a very thin layer of metal, typically palladium or palladium-silver alloy, is deposited on a porous support. Both microporous ceramics and metallic supports have also been used. The thin palladium layer, typically 0.5 to 15 urn thick, offers the potential of higher hydrogen fluxes and lower material costs than thicker, dense metallic membranes, while providing higher hydrogen purities than ceramic membranes. Figure 2.6 summarizes some of the techniques used to prepare supported palladium membranes. The three methods most commonly used to prepare metal fdms on porous supports are electroless plating, chemical vapour deposition (CVD) and physical sputtering (Lin, 2001). Cold rolling Electroless plating Electroless plating + electroplating Reactive ion plating + electrodeposition CVD Magnetron sputtering (RF, DC) Flush evaporation Spray hydrolysis Solvated metal atom deposition 0.01 0.1 1 10 Metal thickness (um) 100 Figure 2.6: Fabrication techniques of supported palladium membranes (from Uemiya, 2004) Pagliari and Way (2002) reviewed previous research on composite palladium membranes, and Rothenberger et al. (2004) give a good summary of past data from composite membrane research. Although there has been significant progress in composite membranes research, major challenges remain to incorporating this type of membrane into larger reactors such as the ICFBMR: Selectivity: Composite membranes with very thin palladium layers tend to have small pinholes or other defects, allowing passage of non-hydrogen gas species. 33 • Thermal expansion: The metal layer in composite membranes is usually tightly bonded to the underlying support, which may not have the same thermal expansion properties as the palladium layer, leading to defects from thermal cycling. • Robustness: Very thin layers of palladium may be abraded or eroded in the reactor. Despite these technical challenges, if palladium-based membrane reactors and purifiers are to be commercially viable, development of robust, selective, thin-film (<10 um) composite membranes may well be essential. 2.2.4 Palladium Membranes in S M R Uemiya (2004) briefly summarized progress and issues with respect to dense and composite metal membranes for steam reforming. In addition, there are technical issues for palladium alloy membranes that are typically not addressed in short-term membrane reactor research, including: • Surface poisoning: Impurities in the feed, such as sulphur, halogen compounds or mercury, can reduce hydrogen permeability of palladium membranes, sometimes irreversibly. Gas pretreatment for these impurities is therefore necessary. • Robustness: Physical abrasion of membranes and repeated thermal cycling may damage membranes. • Alloy segregation: It has been reported that palladium-alloy membranes in hydrogen service can undergo compositional rearrangement, with palladium segregating at the membrane surface on the high hydrogen concentration side, and silver on the low hydrogen concentration side (Shu et al., 1993). Despite several analyses (i.e. Aasberg-Petersen et al., 1998, Onstot et al., 2001) that outline the challenging economic issues faced in applying palladium membranes to the SMR process, several commercial enterprises are transforming membrane reactors into a practical reality.' The current work does not attempt to answer the lingering questions of long-term application of palladium in the SMR process, but rather to show the potential of coupling the ICFBMR concept with palladium alloy membranes to produce high-purity hydrogen. 2.3 ICFBMR Membranes The high operating temperatures (500-650°C) and pressures of the ICFBMR pilot reactor provide significant mechanical challenges for the membranes. As the permeate side of the membrane operates at near atmospheric pressure, the membrane must withstand a very high trans-membrane pressure 34 gradient. The ICFBMR pilot reactor operates up to 1,500 kPa, but a commercially designed system may operate up to -3,000 kPa. The pressure differential over the membranes is only in one direction, from reactor to permeate. Previous research on the original FMBR pilot plant employed palladium-based tubes. Adris (1994) utilized pure palladium tubes (4.7 mm OD, 200-um thick). Roy (1998) and Islam (1997) also used commercial tubular membranes, believed to be fabricated from a noble-metal based alloy (3.2 mm OD, 76-um thick). Roy inserted a spring inside the membrane tubes to provide sufficient mechanical strength to withstand a trans-membrane pressure difference up to 1,330 kPa at 600°C. Although used in previous FMBR research, tubular membranes suffer from a number of disadvantages, including sealing, manifolding, strength and support within the fluidized bed. A new style of planar membrane, developed by MRT, replaced the earlier tubular membranes, and are described in the next section. Based on the experience of MRT, palladium alloyed with 25 wt% silver (Pd75Ag25) was selected as the membrane alloy material for the ICFBMR pilot work. This material is available commercially from several suppliers. Selection of the thickness of the palladium foil is a balancing act between the competing requirements of high permeation and acceptable mechanical strength. For high hydrogen flux and cost, thinner is clearly better. For example, a 25 um thick foil permeates essentially twice as much as a 50 urn thick foil, with the palladium costs halved for the same membrane area, or decreased by a factor of 4 for the same permeation rate. Since palladium costs are a significant component of a membrane reactor, minimizing palladium usage is a key goal in membrane reactor research and development. Despite the incentive to use thinner foils, they are prone to mechanical failure and may develop pinholes, permitting non-hydrogen gases to reach the permeate side of the membrane. 25 urn and 50 um thick Pd7 5Ag25 foils were considered for the ICFBMR pilot reactor. Experience from MRT suggested that commercial 25 um foils sometimes contain defects, which could occasionally lead to pinholes in the membrane. Rather than risk the hydrogen purity of the permeate stream, it was decided to make all of the ICFBMR membranes for this project from 50 um thick Pd75Ag25 foil. There was concern that abrasion of the membrane surfaces with fluidized catalyst particles could lead to membrane failures. A protective cover consisting of a thin ceramic felt and a metallic screen was considered, but ultimately rejected, so that the ICFBMR membranes were used with the palladium surfaces unprotected from the fluidized catalyst. It should be noted that the planar membrane surfaces 35 were installed vertically in the ICFBMR, parallel to the flow of fluidizing gas. Horizontally placed tubes (or bends in vertical U-tubes) are much more prone to erosion than vertical surfaces (Zakkay, 1986). In addition, gas velocities were relatively low by fluidization standards, and palladium is a relatively hard metal. 2.3.1 Planar Membrane Design The membrane panel design was based on proprietary and patent-pending (Li, 2005b) procedures developed by MRT. The novel features of the ICFBMR planar membranes are: • Pd75Ag25 foil is bonded directly to a stainless steel support to form a leak-tight seal around the edge of the membrane surface. • A thin sheet of porous sintered stainless steel is located under the membranes to support the thin Pd foil against the high trans-membrane pressure gradient, allowing hydrogen to permeate. • The Pd foil is isolated from the stainless steel sintered metal support by a thin layer of alumina to prevent metal inter-diffusion between the stainless steel and the Pd foil, which would reduce the hydrogen permeability of the palladium. • A double-sided membrane is created by bonding Pd foil to both sides of the panel support. Permeate hydrogen on the back side of both foils is collected in a serpentine groove cut into the panel. Tube connections are installed for the inlet sweep gas and permeate outlet at opposite ends of the serpentine groove to generate plug flow of the sweep gas on the backside of the membrane. Figure 2.7 presents a cross-section (not to scale) of the membrane and highlights the main components of the panel assembly. Figure 2.8 below shows a cut-away assembly of the ICFBMR membrane panels. Items 1 to 4 are welded to form the completed stainless steel panel support, to which item 5, the Pd 7 5Ag25 foil, is bonded. bonding region / — Pd foil (2) SS substrate permeate flow channel Y//////y/////////////////////////y/w alumina layer (2) SS sintered metal support (2) Figure 2.7: Membrane panel cross-section (not to scale) 36 The membrane panels in the ICFBMR pilot reactor are installed vertically inside an open-ended core box. As the inner dimension of the core box was 86 mm, the width of the membrane panel was fixed at VA" (83 mm). This width also suited the palladium foil as it was supplied in sheets of 11" x 6" (279 mm x 152 mm), allowing the sheet to be cut in half lengthways with no palladium wastage. The length of the panel was set at 12" (305 mm) to suit the foil length. Figure 2.8: ICFBMR panel assembly (see Table 2.1 for identification of numbered components) Table 2.1: Parts list for ICFBMR membrane panels (numbers refer to Figure 2.8) Item Qty Description 1 1 Membrane panel (304 SS) 2 2 Sintered membrane substrate (316SS) 3 1 Sweep gas inlet (0.125" OD SS tube) 4 1 Permeate gas outlet (0.1875" OD SS tube) 5 2 Pd 7 5Ag 2 5 foil (11" x 3") 37 2.3.2 Membrane Fabrication Substrate Fabrication Preparation of the stainless steel membrane substrate involves a number of fabrication steps. Although each step is relatively straightforward, preparation of a defect-free substrate was the most challenging task in preparing the membrane assemblies, partly due to the number of different workshops involved in preparing each panel. The panel base is (6.4 mm) thick stainless steel sheet, type 304. Each sheet was sized to the overall panel dimensions of 12" x 3%" (305 mm long x 83 mm) by machining or water-jet cutting. Next a recess for the sintered metal sheet was machined on each face. Two different methods were used to cut the lA" (6.4 mm) thick serpentine slot. In several panels the serpentine was machined, while other panels were water-jet cut. Both methods proved successful, though vibration of the unsupported tines proved to be troublesome for the machining process. The ports for the 0.125" (3.2 mm) sweep gas inlet tube and the 0.1875" (4.8 mm) permeate outlet tube were also machined. The sintered metal sheets were cut to size, placed in machined recesses in the base panel, and then welded. Obtaining a weld that was crack-free was challenging due to the thinness and the poor thermal heat transfer characteristics of the sintered metal. The sweep gas inlet and permeate outlet tubes were welded on opposite ends of the membrane panel. After the panels were welded, the sintered metal seal weld was ground smooth. Several panel substrates were rejected at this stage due to gouges in the sintered metal caused in grinding or cracks in the sintered metal weld. Seven substrate panels were judged to be acceptable and were bonded with palladium foil. Palladium Bonding The palladium alloy foil was bonded to the stainless substrate using a proprietary and patent-pending technique of Membrane Reactor Technologies Ltd (Li, 2005). The details that follow outline the general steps used in bonding the membrane: • The polished membrane support panel was cleaned with in an ultrasonic bath to remove dirt and polishing wax. A thin alumina layer was deposited on the surface of the porous sintered metal using a sol-gel technique. 38 Palladium alloy foil (25% Ag, 50 urn thick) was purchased from a commercial supplier (Alfa Aesar, stock #42682, Lot# G30N14). The foil was supplied in sheets of 11" x 6" (280 mm x 152 mm), cut in half longitudinally to form sheets of 11" x 3" (280 mm x 76 mm). The foils were then laid onto both sides of the support panel. The bonding border around the palladium foil was approximately 9 mm. The bonding border on the foil was first covered with alumina paper gasket, then a graphite gasket. The membrane panel was placed between two steel end plates and bolted tight. The entire assembly was inserted into the bonding reactor, a stainless steel vessel with external heaters. Figure 2.9 shows a schematic of the bonding vessel apparatus. Details of the bonding process are proprietary to MRT. Figure 2.9: Membrane bonding apparatus schematic Figure 2.9 illustrates the bonding assembly of a single membrane panel. In practice, several membrane panels can be bonded concurrently by stacking the panels between the steel endplates. Figure 2.10 shows a stack of three membrane panels being assembled prior to being placed in the bonding vessel. 39 Figure 2.10: Three membrane panels in preparation for bonding As stainless steel ferrule fittings are known to leak at the temperatures of the reformer pilot plant, extension tubes were welded to both the sweep and permeate tube connections. This ensured that all permeate connections within the reactor were welded. Both sides of the membrane were checked for leaks using a refrigerant leak detector (Bacharach Instruments, model H-10PM) and a refrigerant (Dupont Suva® 134a, 1,1,1,3 tetrafluoroethane). This procedure proved to be a quick and reliable method for detecting cracks or pinholes in the palladium membranes. The refrigerant supply was connected to the 0.125" (3.2 mm) inlet tube and the 0.1875" (4.8 mm) outlet tube dipped in a small beaker of water. A small flow of refrigerant was started to the backside of the membranes, confirmed by bubbles in the outlet beaker. The leak detector wand was then slowly run over the membrane surfaces and tube welds. No leaks were detected in any of the membrane surfaces of the seven panels. Several leaks detected in the stainless tube welds were subsequently repaired and confirmed to be leak-free. A photograph of one of the finished membrane panels is presented in Figure 2.11. Surface defects on the palladium surfaces were noted on several of the bonded panels (see Figure A2.2 in Appendix 2). A few of the panels had small linear indentations due to small cracks in the underlying stainless steel substrate. Others had creases in the palladium foil, most commonly along the long edge of the membrane. These creases were likely due to insufficient bolting force on the long edge of the panels, permitting the palladium foil deform in during bonding. As can be seen in Figure 2.11 above, the end plates used to bond the membranes were oversized, likely resulting in warping of the endplate and uneven bolt torque. 40 Figure 2.11: Finished membrane panel (Panel A) The total foil area installed on one panel is 426 cm2. The area open to hydrogen permeation is significantly less than the total foil area due to the sealing zone around the edge of the foil, and is estimated to be 258 cm2, 61% of the palladium surface. The reduction in permeation area would be less for larger panels and it may be possible to decrease the palladium bonding area in future designs. However, the reduction in palladium effectiveness in the ICFBMR panels is significant. This has been addressed by MRT in their recent membrane designs. Table 2.2: Summary of membrane panel areas Component area Dimensions (mm) Area per panel (cm2) Fraction of foil area Pd75Ag25 foil area 279 x 76 426 100% Sintered metal substrate 260 x 57 296 69.9% Sintered metal substrate unaffected by seal weld (H2 permeation area) 254x 51 258 60.6% Six of the fabricated seven membrane panels were installed in the ICFBMR pilot reactor. During initial testing in a permeation rig (see below), all panels were proven to be leak free up to a differential pressure of 1,200 kPa at a temperature of 550°C. Higher pressure testing was not attempted. MRT has found that similarly constructed 50 pm Pd75Ag25 panels can withstand a trans-membrane pressure difference of 2,500 kPa (Li, 2003). The ultimate rupture pressure of the ICFBMR membranes is not known. 41 2.3.3 Membrane Costs Palladium is a very expensive material and can be a significant cost component of a membrane reactor. Palladium peaked at over $l,000US/oz ($32US/g) in 2001, but has averaged $260US/oz over the last five years (2002 to 2007, source: Johnson Matthey, Figure A2.1 in Appendix 2). Minimizing palladium costs by using thinner membranes is a key goal for many researchers. The cost of each ll"x 6" (280 mm x 152 mm) Pd75Ag25 foil sheet used to make each ICFBMR membrane panel was roughly $600CAD. If the foil cost is converted to a weight basis, the cost of the palladium in the foil is approximately $800US/oz, roughly four times the base metal cost. Lower foil costs may be possible for larger quantity orders. Costs for manufacturing the ICFBMR panels are estimated in Table 2.3. The value of the individual components may appear high, as the panels were built in a one-off manner, but it is instructional to see that the palladium foil is by far the largest single cost component. The palladium foil in the membrane can, in theory, be recovered and recycled after use. If we assume that 80% of the palladium in the ICFBMR membrane can be recycled, and that the value of the recycled palladium is $100US/oz, the value of recycling the palladium is only about $70CAD/panel. Table 2.3: Approximate ICFBMR membrane panel costs Description Unit cost ($CAD) 50'umPd75Afesfoil(H"x6") " $600 Sintered stainless metal sheet (0.05" thick) $50 Supply, cut and machine V" stainless sheet support (12"x3V4") $150 Cut, weld and grind sintered metal sheet to base support $50 Labour to polish and clean metal substrate prior to bonding $100 Labour to bond Pd75Ap,25 foil to base support $100 Total $1,050 Active membrane area cost $40,700/m2 Limited data on membrane costs has been published, but the palladium component is expected to dominate the cost of dense foil or tubular membranes. Composite membranes, made by depositing palladium on a mechanical support, have been prepared with significantly thinner palladium layers, typically less than 15 urn. At very low palladium loadings, the palladium material cost will likely be dwarfed by the membrane fabrication cost. Tong et al. (2005) estimated that the cost of fabricating a 0.5-urn thick Pd75Ag25 composite wafer (0.15 x 0.15 m) to be $340 USD (ca. $18,000 CAD/m2). Clean-room costs were estimated to be the most significant cost component of this membrane. Palladium and silver materials represented less than 3% of the total fabricated cost. This membrane had roughly 50 times the permeation rate of the 42 ICFBMR membranes, but did not produce ultrapure hydrogen and broke under a trans-membrane pressure difference of only 400 kPa, thus making it inappropriate for the ICFBMR process. However this illustrates the cost reduction potential of composite palladium alloy membranes. 2 .4 Membrane Characterization Seven double-sided membrane panels were manufactured, six of which were installed in the ICFBMR pilot reactor. After fabrication, the membranes were then individually tested to establish their hydrogen flux characteristics at elevated temperatures in the permeation rig shown in Figure 2.12. Table A2.1, Appendix 2 details the main components of the permeation rig, which included: A heated pressure vessel, which could house up to three membrane panels at one time. The temperature in the vessel was monitored by two thermocouples. Two mass flow controllers (FIC-01), which set the feed flow of hydrogen or inert gas, either argon or nitrogen, to the main vessel. A mass flow controller (FIC-02) to set the flow of sweep gas to the membrane. Vessel gas to vent / GC> Permeate Sweep gas Argon or N2>—[FIC-02. V-10: S S vessel (0.25 m diam x 0.64 m high) Figure 2.12: Membrane permeation rig schematic 43 • The membrane permeate stream could be operated at either atmospheric pressure or under vacuum. If at atmospheric pressure, the permeate flow was measured with a bubble meter and stopwatch. If under vacuum, permeate flow was measured with either a mass flow meter (FI-03) or a rotameter (FI-04) before being sent to a vacuum pump. Gas samples were piped directly to the sample valve of a Shimadzu model GC-8AIT gas chromatograph equipped with a thermal conductivity detector. See "GC#1" in Table A2.2 in Appendix 2 for analyzer details. The GC calibration was checked daily during permeation testing using a certified standard gas mixture (1.00% 0 2, 3.09% N 2 , 4.01% CO, 11.0% C0 2 , balance H2). The start-up procedure for the permeation rig was as follows: • Nitrogen was introduced to the membrane sweep inlet to purge the backside of the membranes. • The permeation vessel was purged with an inert gas, either argon or nitrogen, and heated to 250°C at a maximum rate of 2°C/min. • At 250°C, hydrogen was introduced and the vessel temperature was then increased under a hydrogen atmosphere to the desired operating conditions at a maximum rate of 3°C/min. • Before permeation experiments started, the nitrogen sweep to the membrane was stopped. The reactor gas composition was measured by gas chromatograph and confirmed to be 100% hydrogen. A continuous flow of hydrogen was maintained to the reactor. • Operating conditions were then stabilized and flux testing started. With an equimolar mixture of hydrogen and nitrogen in the permeation vessel at 545°C, no nitrogen was found in the hydrogen permeate for all seven membrane panels. It is believed that the nitrogen detection limit of the gas chromatograph is approximately 0.001% (Li, 2005). 2.4.1 Effect of Hydrogen Partial Pressure Experiments were conducted on the first membrane fabricated (panel A) to find the partial pressure exponent that best described the hydrogen flux for the ICFBMR panels. From equation 2.5, the hydrogen flux driving force is proportional to the difference in the hydrogen partial pressures to an exponent, n: As discussed previously, the theoretical value for the partial pressure exponent (n) is lA, assuming that hydrogen diffusion through the metal is the rate-controlling step. This may not be the case for very thin composite membranes, where external mass transfer and diffusion through porous pore supports 44 may contribute a significant resistance. Exponents close to 1 can be expected for very thin membranes (Tong et al., 2005). Many researchers have found that an exponent of Vi accurately describes hydrogen permeation (e.g. Holleck, 1970). Despite this, deviation from Sievert's law (n = Vi) has been reported, ascribed to such factors as surface effects, surface poisoning and grain boundaries (Ward and Dao, 1999). Some previously reported values for the partial pressure exponent in thick film palladium include 0.62 (Morreale et al., 2003), 0.68 (Hurlbert and Konecny, 1961) and 0.72 (Roy, 1998). In order to find the partial pressure dependence of the ICFBMR membranes, experiments were carried out in the permeation rig at 545°C with high pure hydrogen (Praxair UHP 5.0) in the permeation vessel. The hydrogen partial pressure driving force over the membrane was varied by changing the vessel total pressure, which is also equal to PHh, or by adjusting the hydrogen partial pressure on the permeate side, PHI. The permeate hydrogen partial pressure was varied using a dilution sweep gas, or by lowering the permeate pressure with a vacuum pump: • No sweep gas: Initially, the permeate stream was pure hydrogen (no sweep gas) at atmospheric pressure. The reactor pressure (PHh) was then varied from 133 and 204 kPa. • Sweep gas: The vessel pressure was held at 170 kPa. The permeate stream was maintained at atmospheric pressure, but nitrogen sweep gas was fed to the backside of the membrane to dilute the permeate, thus reducing the hydrogen permeate partial pressure (PHI)- The hydrogen concentration in the permeate was varied from 64 to 92%. • Vacuum, no sweep gas: The vessel pressure was held at 170 kPa. No sweep gas was used and therefore the permeate was pure hydrogen. The absolute pressure of the permeate was reduced using a vacuum pump to between 33 to 67 kPa. Table A2.3 in Appendix 2 summarizes the flux test data. The flux on the panel varied between 0.78 and 3.18 SLM during the tests. The flux data are plotted as a function of the square root of the hydrogen partial pressures in Figure 2.13. 45 0.003 0.002 o E, x CM I 0.001 0.000 y = 1.06E-05X R 2 = 0.985 • No sweep A Vacuum O Inert sweep 50 100 150 200 250 Partial H2 pressure, highA0.5 - Partial H2 pressure, lowA0.5 (PaA0.5) Figure 2.13: Hydrogen flux on Panel A as a function of difference between square root of the hydrogen partial pressures (545°C, pure hydrogen in permeation vessel) Linear regression of the above flux data gives good agreement, and it is tempting to conclude that the data confirm a partial pressure exponent of Vi. However, the data also fit a range of pressure exponents, with values of 0.45 to 0.65 all giving good regression coefficients. The range of differential pressures tested was too low to permit an adequate analysis, and it cannot be definitely concluded from the above data that n = Vi is valid. However, the range of hydrogen partial pressures in the ICFBMR pilot reactor is close to the range in Figure 2.13, and it was therefore concluded that an exponent of Vi can be practicably applied. If a partial pressure exponent of Vi is assumed, the slope of above regression line is equal to: slope = k H A , - e x p ( - ^ ) R T (2.7) The membrane thickness, r, is 50E-6 m and the permeation temperature in the above experiments was 818K. If we assume values for kH and Ep of 3.43E-7 mol/m s Pa 0 5 and 9,180 J/mol respectively (Li, 2003), a value for the permeation area,v4mem, can be calculated by rearranging equation 2.7. Using the regression slope in Figure 2.13 of 1.06E-5 mol/s Pa 0 5, the membrane area, Amem, is calculated at 0.0189 m2. This value is 73% of the sintered metal area previously calculated from the panel geometry (0.0258 m2). 46 2.4.2 E f f e c t o f T e m p e r a t u r e Panel A was flux-tested in the permeation rig to see the effect of temperature. For these tests, the permeation vessel was under pressure (238-376 kPa) and contained pure hydrogen. The permeate stream was at atmospheric pressure (101 kPa). The membrane flux was measured for a range of temperatures (475-564°C). Figure 2.14 summarizes the results of this testing (see Table A2.4, Appendix 2 for data). As expected, the hydrogen flux increases with increasing temperature and increasing reactor pressure. 0.003 3» 0 0 0 2 o E. x 3 X 0.001 0.000 • X • X o X • X 0376 k P a vessel pressure XC307 k P a vessel pressure 0 238 k P a vessel pressure .720 740 760 780 . 800 Temperature (K) 820 840 860 Figure 2.14: Permeation of Panel A at three different upstream pressures with permeate at 101 kPa Flux data from Figure 2.14 are transformed into an Arrhenius plot in order to estimate the activation energy (Ep) and permeation constant (kH) for the membrane. From equation 2.7 with an exponent («) of Vi, the slope and intercept of regression lines in the Arrhenius plot are: slope : R A intercept = l n [ k H - J = - ( P ^ - P ^ ) ] T (2.8) (2.9) 47 .7.0 J • • • • 1.15E-03 1.20E-03 1.25E-03 1.30E-03 1.35E-03 1.40E-03 Inverse temperature (1/K) Figure 2.15: Arrhenius plots of hydrogen permeation on Pd 7 5Ag 2 5 Panel A (475-564°C) at three different upstream pressures with the permeate at atmospheric pressure Values for Ep from the three slopes are calculated to be 13,600, 12,900, 14,800 J/mol, giving an average of 13,800 J/mol. From the three regression lines in Figure 2.15, and assuming a panel permeation area of 0.0189 m2, an average value for kH of 2.2E-7 mol/m s Pa0 5 was obtained. It should be noted that the span of the temperature range of the permeation data (475-564°C) was quite limited, and did not cover some of the higher temperatures tested in the ICFBMR pilot plant (up to 650°C). The quality of the data in Figure 2.15 above, especially for the lower curve, does appear to be slightly inconsistent. Many researchers have measured the hydrogen diffusion activation energy of pure palladium, with values of 22 kJ/mol (Holleck, 1970) and 13.8 kJ/mol (Morreale et al., 2003) commonly utilized for application to the temperature range of interest in the ICFBMR pilot reactor. There is considerable spread in the activation energies reported for palladium and palladium alloys, likely due to hydrogen diffusion being a multi-step process, in which small changes in surface chemistry and testing conditions can significantly affect experimental results (Ward and Dao, 1999). Values for the hydrogen diffusion activation energy in Pd7oAg3o of 2.2, 6, 6.3, 9.4 and 23 kJ/mol have been reported (Amandusson et al., 2001). Holleck (1970) predicts Ep = 23,500 J/mol for Pd 7 5Ag 2 5. From literature reviews and testing, MRT has found the values kH = 3.43E-7 mol/m s Pa 0 5 and Ep = 48 9,180 J/mol to reasonably describe hydrogen flux through their membranes fabricated with Pd75Ag25 foil (Li, 2003). Given the variability in the published data, it is perhaps not surprising that the experimental values obtained for kH and Ep differ from the values recommended by MRT. It should be noted that the estimate for kH depends on the active permeation area (Amem), which is not known accurately for the ICFBMR panels. The recommended MRT values for kH and Ep were therefore employed in subsequent permeation modelling. 2.4.3 Variation between Membranes All seven ICFBMR panels were tested to establish their flux characteristics. Although fabricated in an identical manner, it became apparent that the seven membranes displayed significant differences in hydrogen permeation, despite being made in a near-identical manner. This is highlighted in Figure 2.16 below, which shows permeation data for six panels over the temperature range of 460 to 570°C. For these tests, the permeation vessel contained pure hydrogen and was held at 300 kPa. The permeate streams were at atmospheric pressure, with no sweep gas. 5.0 4.0 E 3.0 0.0 Theoretical flux (area = 0.0258 m2) • Mem#1 X M e m # 3 + M e m # 4 • M e m # 5 A M e m # 6 O M e m # 8 450 475 550 575 500 525 Temperature (°C) Figure 2.16: Flux data for six membrane panels as a function of temperature (vessel pressure at 300 kPa, pure hydrogen, permeate at 101 kPa, see Table A2.5 for data) The measured H 2 permeation on the highest flux panel was more than twice that for the lowest flux panel. There were no obvious reasons for the wide variation in permeation behaviour. The two most likely factors are likely to be: • An insufficient thickness of alumina separating the palladium from the metal substrate may have led to metal-to-metal interdiffusion and bonding, reducing the area available for permeation. 49 • Surface poisoning from the bonding procedure. Subsequent work by MRT has demonstrated that post-bonding oxidation with air can remove some reversible surface poisons (Li, 2005). The permeation data in Figure 2.16 show similar dependence of flux on temperature for all six panels. It was concluded that the best way to characterize the varying permeation of the ICFBMR panels was to calculate an effective membrane area (Aejj) for each membrane based on permeation equation 2.4: (2.10) With values for r, kH and Ep of 50E-6 m, 3.43E-7 mol/m s Pa 0 5 and 9,180 J/mol respectively, the effective membrane area was calculated for each of the seven ICFBMR membrane panels. The resulting values are summarized in Table 2.4 below. Also shown is the "batch number" in which each membrane panel was bonded in the permeation bonding vessel. The six numerically labelled panels were installed in the ICFBMR pilot reactor. Table 2.4: Calculated effective membrane areas for ICFBMR panels Membrane j Effective membrane area (m ) Fraction of expected Batch A 0.0189 73% 1 1 0.0121 47% 2 3 0.0101 39% 3 4 0.0075 29% 3 5 0.0142 55% 3 6 0.0059 23% 2 8 0.0101 39% 2 During subsequent ICFBMR experiments, the permeation characteristics of the six membranes were periodically tested. The relative permeation of the six panels stayed roughly the same (as shown in Chapter 6), but the absolute permeation values (or the calculated effective membrane area) did vary somewhat. 2.5 Conclusions Although ceramics have been studied for high-temperature hydrogen membranes, palladium and its alloys have been more commonly used. Dense Pd-based foils have the potential to produce very high hydrogen purity from membrane reactors. A e f f ~ e x P ( - ^ ) ( P l 5 - P H ? ) 50 50 um thick Pd7 5Ag25 foil was used to prepare seven planar membrane panels using MRT's proprietary procedures. The membranes proved to be pinhole-free and produced a pure hydrogen permeate in permeate rig tests. Sievert's law could accurately describe the measured membrane flux, though the value of the partial pressure exponent QA) was not statistically proven over the experimental pressure range. The measured hydrogen fluxes on the membranes were somewhat lower than expected, and varied among the panels. Surface defects, such as wrinkles and bulges, were present on a number of the membrane faces. Six of the membrane panels were subsequently installed in the ICFBMR pilot plant, as detailed in Chapter 4. 51 Chapter 3. Cold Model Testing As the hydrodynamics of the ICFBMR system were difficult to test in the pilot plant, a Plexiglas cold model was build to simulate and test reactor operations. This chapter discusses some of the hydrodynamic considerations for the ICFBMR design, including particle characteristics and the scaling parameters between the cold model and the hot reformer. Solids circulation rates in the ICFBMR were determined for a range of gas flow rates and for three different membrane panel configurations. Testing clearly indicated that slotted membrane panels are preferred to solid (non-communicating) panels. Data from helium tracer experiments supported the circulation rates calculated from physical observations. Calculations indicate that adequate solids circulation rates can be achieved in the ICFBMR system to support reforming operations. 3.1 Introduction Some limited hydrodynamic and gas distribution testing was attempted on the ICFBMR pilot plant reactor. However given the complexity of the pilot reactor, it was difficult to undertake a systematic hydrodynamic study on the effects of reactor geometry and flow conditions. A Plexiglas cold model mock-up was therefore assembled and tested. Direct observations of internal solids circulation within the column were possible with the Plexiglas system. Solids circulation is a key feature of the ICFBMR concept, as this transports heat from the oxidation (exothermic) reactions in the top dense zone and splash zone to the reforming (endothermic) reactions in the core permeation zone. The cold model tests were carried out in a 0.285 m diameter Plexiglas column operating at ambient conditions. A Plexiglas core draft box was installed in the column containing seven vertical panels to represent hydrogen permeable membranes. The column was filled with fresh fluidized catalytic cracking (FCC) catalyst and fluidized with compressed air. Solids circulation and gas distribution were studied with three different panel geometries for a range of feed gas flow rates. 52 The objectives for the cold model work included: • Qualitative evaluation o f fluidization and solids flow patterns; • Quantification o f solids circulation as a function o f gas flow and panel geometry; • Determination o f whether membrane geometry (solid vs. or slotted panels) leads to flow instability; • Evaluation o f gas distribution using helium gas tracing. 3.2 Fluidized Beds Fluidized beds are used extensively in the chemical and process industries for reaction, drying, particle contacting and numerous other applications. The I C F B M R is an example o f gas-solid fluidization, but there are also important industrial applications for liquid-solid and gas-liquid-solid systems. Some o f the reasons for using a fluidized bed for gas-solid catalytic reactions over a fixed bed include improved heat transfer, good solids mixing, lower pressure drop and the ability to move particles within the reactor or between vessels. Gas contact time tends to be low in fluidized beds, and therefore they are mainly applied to fast reactions. A major drawback o f using a catalytic fluidized bed reactor in many cases is the non-plug flow gas movement through the reactor. The fluidization regimes for a gas-solid system (gas upflow) are shown in Figure 3.1. As gas flows upward through a bed o f particles, gas initially trickles through the bed with no movement of the particles (fixed bed). A s the flow o f gas is increased, the pressure drop over fixed bed increases, until the pressure drop over the fixed bed matches the weight o f the particles. A s flow is increased past this point, bubbles form and rise through the bed (bubbling regime) and the bed behaves like a fluid. The bubbles contain few particles. The gas flow in the bed is divided into the bubble phase and the dense particulate phase, where the close-packed particles are supported by an upward flow of interstitial gas. The amount o f gas in the dense phase is approximately equal to the flow of gas needed for minimum fluidization, with the balance of the gas flow partitioning to bubble phase. Particles are entrained in the wakes of the rising bubbles, causing solids mixing. Bubbles tend to coalesce and grow as they rise through the bed. If the rising bubble diameter is close to that of the column, slugging may occur. A s the gas velocity is further increased, the bubbling regime gradually undergoes a transition to the turbulent regime, in which the gas voids tend not to look bubble-like, but rather are unstable, transient and streaky. If the gas velocity is further increased, the bed enters the fast fluidization regime, where a dilute suspension surrounds solids clusters. 53 BUBBLING AGGREGATIVE FLUIOIZATION INCREASING U. c . Figure 3.1 : Gas-solid flow regimes (Grace, 1986) The previous FMBR work by Adris (1994) and Roy (1998) operated their fluidized bed in the bubbling regime with superficial gas velocities of the order of 0.03 m/s. The ICFBMR concept requires higher operating gas velocities in the reactor core, in the range of 0.1-0.5 m/s, in order to generate a sufficient internal circulation of solids within the reactor. 3.3 Cold Model Hydrodynamics 3.3.1 Scaling Parameters The pilot ICFBMR reformer operates at high temperatures and pressures, conditions that cannot be duplicated in a Plexiglas system. The ICFBMR cold model attempts to simulate operation of the ICFBMR pilot plant, but fluidization characteristics of a system operating at high temperature and pressure can change considerably from ambient conditions and with different powders (Yates, 1996). Scale-up of gas-solid fluidized beds is an area of considerable study (e.g. Kelkar et al., 2002, Knowlton et al., 2005). Despite this, there remains a high degree of uncertainty in fluidized bed scale-up, especially when applied to reacting systems (Matsen et al., 1997). Numerous researchers have proposed dimensionless hydrodynamic scaling parameters for cold modelling of gas-solid fluidized beds, including Glicksman (1984, 1993) and Horio et al. (1986). The gas superficial velocity in the core of the ICFBMR pilot plant ranges from about 0.05 to 0.3 m/s. At these velocities, the particle Reynolds numbers are low (<4). In this hydrodynamic region, viscous forces are more significant than inertial forces. Starting with the Ergun equation, Glicksman et al. (1993) recommended equivalence of the following seven dimensionless parameters for scale-up of gas-solid fluidized beds at low Reynolds numbers: 54 U2 PP U Lx Gp ,(/> , dimensionless particle size distribution (3.1) gL' P/'Um/ L2' ppU Here L is a generic bed dimension. If L in the first term is replaced with dp, the particle Froude number is obtained. Roy and Davidson (1989) found that it was not necessary to match the particle-to-fluid density ratio, the second term above, when scaling gas-solid fluidized beds with low Reynolds numbers. However, this term is retained to extend the scaling to higher Reynolds numbers. The fourth term represents the ratio of any two dimensions and hence geometry scaling. It (Li/L2) could be applied as the ratio of bed height to particle diameter (H/dp) or bed diameter to the particle diameter (D/dp) (He et al., 1997). As the cold model is geometrically similar to the pilot reactor, this scaling term is subsequently ignored. The fifth term includes the solids flux in the column (Gp), and was originally envisioned for a circulating fluidized bed. In a conventional bubbling bed, the solids flux is very nearly zero, and thus this scaling term can be neglected. In the ICFBMR, however, there is a flux of solids up the core of the reactor, and down the annulus, and therefore this term must be retained. The sixth term in equation 3.1 is the particle sphericity. The sphericity of the three types of catalyst particles used in the cold model and ICFBMR pilot reactor is not exactly known, but observations through a microscope indicate a range between 0.6 to 0.8. The sphericity factor is ignored in further scaling calculations. The particle size distributions of the three powders studied in this work are included in the next section, and were judged sufficiently similar, that on a non-dimensional basis this term could also be eliminated. The three areas of vigorous fluidization in the ICFBMR concept are (i) above the distributor plate, (ii) inside the reactor core box, and (iii) above the core box. The feed gas distributor is drilled to match the cross-section of the core box, so that the superficial velocity above the distributor is similar to that in the core box. The fluidized area above the core box, which includes the oxidant distributor, does not significantly influence the key system hydrodynamic parameters of interest, such as solids circulation and voidage in the core. Therefore the superficial gas velocity in the core (Ucore) is applied where a velocity is needed in the scaling parameters. With the above simplifications, the seven Glicksman scaling terms are reduced to the following four terms for the ICFBMR cold model: 55 Ucore Pp Ucore ^'p.core There are other considerations when using a cold model to simulate the hydrodynamics of a hot, reacting system such as the steam reforming process: • Temperature entry effects: If cold gas is introduced into a hot reactor, thermal gradients may develop that cannot be simulated in an isothermal cold model. In the ICFBMR pilot reactor, the feeds are preheated to within 50-100°C of the core reactor temperature (typically 550-650°C) and become quickly equilibrated with the circulating solids, so this is not likely to be a significant concern for this system. • Change in molar flow rate: The volumetric gas flow in a fluidized bed reactor can change due to changes in the number of moles due to reaction or absorption onto the bed solids. In the conventional SMR process, three moles of feed gas react to form five moles, which results in an increase in flow of 50% if a feed steam-to-carbon ratio of 3 is used. This represents a significant increase in superficial gas velocity as the reaction proceeds, difficult to simulate in a cold model. In the ICFBMR system however, hydrogen is removed from the reactor core through the membrane surfaces. If 2/4 moles of permeate hydrogen per mole of feed methane were recovered, there is actually a small decrease (-10%) in the number of moles in the reactor core, and therefore a constant superficial velocity in the cold model is a reasonable assumption. 3.3.2 Catalyst Powder Characteristics In order evaluate the relevance of the cold model testing to the operating conditions in the ICFBMR pilot plant, hydrodynamic characteristics and scaling parameters are evaluated for the two catalyst powders used in the ICFBMR pilot plant (crushed nickel oxide SMR catalyst and autothermal (ATR) catalyst) as well as FCC catalyst used in the cold model work. The measured cumulative size distributions for the three powders are shown in the Figure 3.2. The average particle diameters (dp) of the SMR, ATR and FCC catalyst powders were found to be 107, 87 and 62 um respectively, calculated from the average particle size (dpi) and mass fraction (*,-) on each sieve according to: The SMR (NiO on alumina) material has a wider particle distribution than the other two catalyst powders. This is not unexpected as this material was obtained by crushing and screening 56 conventional SMR catalyst pellets, whereas the ATR and FCC powders were designed and produced to be fluidized bed catalysts. 300 Particle diameter (pm) Figure 3.2: Cumulative size distributions for the three catalyst powders studied The fluidization pressure drop vs. superficial velocity curve for the FCC catalyst in air is shown in Figure 3.3. The corresponding curves for the SMR and ATR catalyst powders in atmospheric air are shown in Appendix 3.2 (Figures A3.2.9 and A3.2.10). The minimum superficial fluidization velocities (UmJ) of the three catalyst powders in atmospheric air were found to be 6.0E-3, 3.0E-3 and 2.6E-3 m/s for the SMR, ATR arid FCC powders respectively. ° o o o 0 o o ° o ° ° ° o o o o 0 0.0025 0.005 0.0075 0.01 Superficial velocity (m/s) Figure 3.3: Cold model pressure drop vs. superficial air velocity for the FCC ~ 1.5 n 0_ X. a. o k. •a o> 3 (A <fl CD 1.0 0.5 0.0 57 Table 3.1 summarizes a number of fluid and particle characteristics for the cold model (FCC) and the pilot reactor operating with two different catalysts (SMR and ATR catalyst). The reformer operating conditions were assumed to be 600°C and 1200 kPa, typical operating parameters for the ICFBMR pilot plant. The Archimedes number for the Geldart A-B powder boundary is calculated from an equation propose by Grace (1986). A r ^ = 1..03E6I P P - Pf -1.275 (3.4) All of the catalysts had an Archimedes number below ArAB boundary value, so all may be considered Geldart Type A powders. The particle Reynolds number at minimum fluidization (Remf) in air was estimated from another equation proposed by Grace (1982). Re mf •• yjc2 +C2Ar - C , , where C, = 27.2, C2 = 0.0408 (3.5) For FCC, the calculated Um/ in air using the following equation matched the measured value. However, the predicted Umf for the SMR and ATR catalysts in air was roughly twice the measured value. The Reynolds numbers at minimum fluidization are low for all of the powders, so the minimum superficial fluidization velocity can also be estimated by simplifying the Ergun equation to a relationship suggested by Kunii and Levenspiel (1991): •mf 150// l-£ (3.6) mf Given the error in predicting the minimum superficial fluidization velocity of the SMR and ATR catalysts in air from equation 3.5, Umf at reformer conditions were predicted by scaling the Umf measured in atmospheric air by the density and viscosity ratio using equation 3.6: l^™/ \cFBMR L ^ V l '(Pp-pf) M ICFBMR (Pp-Pf). (3.7) 58 Table 3.1: Cold model and pilot reactor properties and characteristics Variable Units ICFBMR Pilot Plant Cold Model Particle data Bed material - Ni oxide on alumina ATR on alumina FCC catalyst Particle data Particle density (pp) kg/m3 2,150 2,090 1,600 Particle data Loose packed inter-particle voidage - 0.5 0.5 0.45 Particle data Average particle diameter (dp) m 107E-6 87E-6 62E-6 Fluid data Temperature °C 600 20 Fluid data Pressure, absolute kPa 1,200 101 Fluid data Fluidization gas - Reformate mixture Air Fluid data Gas molecular weight kg/kmol 15 29 Fluid data Gas density (pj), assuming ideal gas kg/m3 2.48 1.20 Fluid data Gas viscosity (p) Pas 2.90E-5 1.74E-5 < Archimedes number - 76 40 15 Archimedes number for A-B powder boundary, Eqn. 3.4 - 185 192 107 Geldart powder classification - A A A Predicted R e m / r i n air, Eqn 3.5 - 0.077 0.040 0.011 Umf in atmospheric air (predicted) m/s 0.010 0.067 0.0026 Umf in atmospheric air (measured) m/s 0.0060 0.0030 0.0026 Umf at reforming conditions, Eqn. 3.7 m/s 0.0036 0.0018 -Remf at reforming conditions - 0.033 . 0.013 -As the gas velocity in a bubbling bed is increased, the onset of the turbulent fluidization flow regime may be reached (see Figure 3.1). The superficial gas velocity at the initial transition to turbulent fluidization (Uc) is defined as the point at which the standard deviation of pressure fluctuations reaches a peak. Ellis (2003) found that the value for Uc for FCC particles in the same 0.285 m ID column used in the cold model testing, but with no internals, ranged from 0.35 to 0.6 m/s, depending on the bed height. Core superficial gas velocities in this range were studied in the cold model testing, though the system geometry is differed markedly from that in the cited work. The channels in the cold model reactor core have a much lower hydrodynamic diameter that the main column. Decreasing reactor diameter is reported to increase Uc, but column internals may cause a decrease in Uc (Bi et al., 2000). Table 3.2 indicates the values of the retained dimensionless Glicksman scaling parameters for the reforming catalysts in the ICFBMR pilot plant and the FCC powder in the cold model, assuming a superficial core velocity (Ucore) of 0.1 m/s. The first hydrodynamic scaling term is the Froude number. At a core superficial velocity of 0.1 m/s, the Froude number for the cold model is slightly larger than for the reforming catalysts. A core velocity of only 20% lower in the cold model would produce a similar Froude number to the values for the reforming catalysts. 59 Table 3.2: Cold model scaling parameters at a core superficial velocity of 0.1 m/s Scaling Variable ICFBMR Pilot Plant Cold Model Term Bed material Ni oxide on ATR on FCC catalyst alumina alumina - Fluid Reformate at 600°C, 1,200 kPa Ambient air 1 Particle Froude number in core 9.5 12 16 2 Density ratio (ppl pf) 866 842 1,330 3 Velocity ratio (Ucore 1 Umj) 28 56 33 4 Gp,core / Pp Ucore TBD TBD TBD The values for the density ratio, the second scaling term, for three powders are indicated in the Table 3.2 above. The cold model ratio (FCC / air) is greater than for the two reforming catalysts at reformer operating conditions, but lower than would be the case if the cold model column was filled with alumina. In any event, the low particle Reynolds numbers in these systems likely means that this scaling term can be ignored (Roy and Davidson, 1989). The minimum fluidization velocity (Umj) of FCC in air (2.6E-3 m/s) is close to the predicted UmfOf the two reforming catalysts at the assumed ICFBMR operating conditions (3.6E-6, 1.8E-6 m/s). Therefore the third scaling term (UcorJUmj) can be kept approximately constant if the same core velocities are used in the cold model as used in the ICFBMR pilot reactor. The fourth term scaling term could not be calculated with confidence at the outset as it includes the solids flux in the reactor core. Finding the solids flux at different operating flow conditions was a key goal of the cold model work. From the above scaling parameter analysis, it is concluded that it is appropriate to use similar core velocities in the cold model as those in the ICFBMR pilot plant. 3.4 Experimental Apparatus 3.4.1 Reactor Geometry Figure 3.4 contains a schematic of the Plexiglas cold model system adapted from a column previously constructed for work by Abba (2001) and Ellis (2003). The main section of the column had an inner diameter of 0.285 m. There was a small disengagement section at the top of the column (0.4 m ID). The core box was hung in the column by a rod and was centred with setscrews. Panels were installed into the core box from the bottom, secured by grooves and bolts. There was a 100 mm gap between the main distributor and the bottom of the core box to permit circulation of solids from the annuli to the core. The column was filled with approximately 95 litres of FCC solids to give a settled bed depth of 25 mm above the top of the core box. In order to reduce electrostatic charges, approximately 0.5 60 wt% of Larostat® 519, an antistatic agent, was added to the FCC solids. In addition, copper tape was applied to the exterior of the column and then grounded, together with any installed metal fittings and flange bolts. Assembly instructions and photographs of the cold model system are provided in Appendix 3.1. Compressed building air was used for fluidization. Air flow to the column was manually controlled with throttle valves and measured by rotameters. The column operated at atmospheric pressure and ambient temperatures. Air exited the column via two cyclones and a filter baghouse before being vented to atmosphere. As in the ICFBMR pilot reactor, gas was introduced into the cold model column through three separately controlled distributors: • Main feed distributor (0-750 SLM) • Secondary feed distributor to four annulus sections, see Figure 3.4 (0-30 SLM in total) • Oxidant distributor to splash zone (0-280 SLM) There was a concern that solid panels (membranes) may lead to hydrodynamic instability within the core box. If the panels are solid (i.e. there is no gap allowing communication among the channels), gas and particles cannot move between the channels and the system is equivalent to the tube side of a heat exchanger. If only gas were to flow up the core box, the pressure drop over the channels would lead to a stable distribution of gas. As each of the channels in the ICFBMR reactor is identical, even distribution of gas through the core would be expected if no solids were present. However, for a fluidized gas-solid system, solids can distribute non-uniformly among the channels, leading to many possible gas-solid flow patterns that could satisfy the system pressure drop (Bolthrunis et al., 2004). Variable solids loading between the channels would then lead to a redistribution of gas between the channels, making it unlikely that each channel would contain an identical flow of gas and solids. In the extreme, solids could block one of more channels, leading to very uneven gas and solids flow. Uneven flow of gas in the membrane core could lead to varying hydrogen partial pressures for adjacent channels and still satisfy the pressure balance requirement. This could result in hydrogen flux variations between the membrane panels, causing underutilization of the installed membrane surface and reduced hydrogen recoveries. 61 Plexiglas cold model column - 285 mm ID main section - 400 mm ID expanded 3260 mm Building AJj>-740 mm 04, Oxidant Feed Annulus for return of particles Secondary distributor (4) Secondary Feed Main distributor Cyclones Baghouse Column Vent Main Feed Cyclone bottom valves closed Splash zone / freeboard Oxidant distributor Unexpanded bed height (25 mm above core box) Plexiglas core box (178 mm square) with 7 membrane panels Channels (8) for passage of gas and particles e-Multiple pressure transmitters Pressure Data Acquisition Figure 3.4: Cold model system schematic 62 Figures 3.5 shows a plan view of the cold model column. The core box was centred in the column. Seven panels were slotted into the core box, creating eight identical channels, in which the bulk of the column gas flowed upward, carrying entrained solids. Solids return to the bottom of the column in the four outer annular quadrants. In order to investigate fluidization stability within the core, three different panels were separately tested in the cold model: solid (i.e. no gaps, slots or holes), vertically slotted and horizontally slotted panels (Figure 3.6). The panels were 100 mm shorter than the outer core box. This created a 50 mm long entry and exit zone within the core box, similar to the ICFBMR pilot plant membrane assembly. The area open to inter-channel flow for the vertically and horizontally slotted panels is the same, 9% of the area of the solid plate. The horizontally slotted panel is similar to the membrane assembly in the ICFBMR pilot reactor. The tabs on the vertically slotted panels are to hold the panel within the grooves of the core box. The vertically slotted panels were installed in alternating direction (slot on the right, slot on the left, etc.), creating a serpentine-line connection within the core box (Figure A3.2.7, Appendix 3.2). After insertion into the core box, the width of the communicating slot in the vertically slotted panels was 21 mm. Figure 3.5: Plan view of cold model core box and panels (solid panels shown) 63 All dimensions in mm All panels are 6.4 mm thick k—164—H 1422 typ' for all 3 panels 164-Vertical slots for communication between flow channels 1 C -139-4 51 typ' A. Solid (impervious) panel R 10--164—H Horizontal slot for communication between flow channels A 38 typ' 254 typ' (SIS typ'"" B. Vertically slotted panel C. Horizontally slotted panel Figure 3.6: Front views of cold model panels The diameter of the cold model column (0.285 m) was approximately twice that of the ICFBMR pilot reactor (0.135 m). A larger diameter column was selected for the cold model work in the hopes of reducing the influence of wall effects and the numerous pressure impulse lines in the column annulus. The cold model channel width and height were identical to those in the pilot reformer, but the channels and channel lengths were doubled, with the result that the area in the core box available for flow in the cold model was approximately four times greater in the cold model. The fractions of the total column cross-section open to flow in the annulus and core were roughly the same in the cold model and pilot reactor. The "annulus turn" area indicated in Table 3.3 is the curtain flow area under the perimeter of the core box, where solids from the annulus flow into the core region. Note that the annulus turn area is larger than the annulus area by a factor of approximately 2 for the cold model and a factor of 4 for the 64 ICFBMR pilot plant. The hydraulic diameters in Table 3.3 are calculated from flow area and the wetted perimeter: 4 A DH= (3.8) perimeter Table 3.3: Comparison of cold model and ICFBMR reactor geometries Cold Model Pilot Reformer Column inner diameter (m), D 0.286 0.135 Core box height (m) 1.52 1.52 Channel width (m) 0.014 0.014 Channel length (m) 0.158 0.083 Channel height (m) 1.42 1.42 Number of channels 8 4 Free flow area in core box (m2), Acore 0.0188 0.0048 ——. •— j Free flow area in annulus (m ), Aan„ 0.0354 0.0074 — • 1 — T Flow area of annular gap (m ), Agap 0.064 0.032 Fractional free core area of column 28% 31% Fractional annulus area of column 53% 51% Hydraulic diameter of channel (m) 0.026 0.025 Hydraulic diameter of annulus (m) 0.087 0.039 Hydraulic diameter of turn (m) 0.20 0.20 . — — ~ Solids volume in column (m ) 0.095 0.020 Uniform flow of gas and solid within the core of the ICFBMR is very much desired in order to maximize the hydrogen membrane flux and minimize temperature variations within the vessel. It may be possible to design a gas distributor that would inject identical flows of gas into all core box channels. However, this may still not lead to an even distribution of gas and solids among the eight channels and implementing a complex design in a high temperature reactor would be challenging. A simple gas distributor is very much preferred. The distributor plate installed in the cold model is shown in Figure A3.2.3, Appendix 3.2. It is a simple orifice plate distributor equipped with a mesh screen to prevent back sifting. The orifice holes are 5.8 mm in diameter on a 39 mm triangular pitch. Only the distributor area directly under the core box contains these holes, as this directs the main gas flow into the core box, rather than into the annular region. A small amount of secondary air aids in fluidizing particles descending in the outer region, increasing the solids circulation. Aeration tubes were therefore installed near the centre of each of the four 65 annular quadrants. A photograph of the complete secondary distributor assembly is provided in Figures A3.2.1 and A3.2.2, Appendix 3.2. The four quadrants of the column each had a 12 mm diameter vertical tubular distributor, with twelve 0.8 mm diameter holes at three levels. The holes were drilled at a steep angle to direct the air downward at an angle of 20° to the vertical. A fine metal screen was then wired to the outside of the tube to prevent back sifting into the orifice holes. Figure 3.7 shows details of the bottom section of the cold model column. The oxidant distributor was a 12 mm OD horizontal tube drilled with two opposing rows of 13 holes (each 1 mm in diameter) oriented horizontally (see Figure A3.2.4 in Appendix 3.2). The distributor rube was fed from both ends to improve gas distribution through the orifices. As in the ICFBMR pilot reactor, the oxidant distributor was perpendicular to the core panel channels, and the gas distribution interval was limited to the width of the core box. Pressure data drop for the three cold model distributors (Figures A3.2.5 and A3.2.6) appear in Appendix 3.2. Upper holes (two rows of 4 x 0.8 mm) Annular nozzle (4) Lower holes (4 x 0.8 mm) Secondary feed Core box 12> 115 45 1 ' Mesh screen | 100 -285 column ID-178 14 Panels (7) — Flow channels (8) V Entry region Main distributor ~ Windbox Annular downflow region Main feed • All dimensions in mm Figure 3.7: Details of bottom section of cold model column 66 3.4.2 Instrumentation Pressure and Flow The flows of air to the various parts of the column were measured with four rotameters (Table A3.2.1, Appendix 3.2). Flow readings were manually recorded during experiments. The calibrations of these feed rotameters were confirmed with a dry gas meter. Absolute and differential pressure transducers (Omega PX140) were connected to various ports along the wall of the cold model via V" (6.4 mm) impulse lines made from plastic or stainless steel tubing. A fine stainless steel screen (38 pm) was glued to the tip of each impulse line to prevent solids back-sifting. Impulse lines were connected to the core box to measure the pressure within the eight channels at three elevations: • Bottom of core box (230 mm above main distributor)- flow channels 4, 5. • Mid-height of core box (862 mm above main distributor) - flow channels 1 to 8. • Top of core box (1494 mm above main distributor) - flow channels 4, 5, 6. Due to the cold model geometry, the core box impulse lines were very long (7-10 m). All of the core box impulse lines were grouped together in the north and south annular quadrants. Pressure transducer details and their locations within the cold model column are presented in Appendix 3.2 (Table A3.2.2, Figures A3.2.7 and A3.2.8). The pressure transducers were connected to a computer via an A/D converter (DAS08-EXP32). The pressure signals were monitored and logged on the computer by means of one of the two programs listed below. The two data acquisition programs could not be operated concurrently. • An existing Visual Basic® program logged pressure at a relatively low frequency (<1 Hz). These data were used to provide trends of steady state pressure signals over longer periods of time, usually five or more minutes. The program converted the pressure transmitter signals (0-5 VDC) to pressure (kPa) with user-inputted calibration values. • An existing in-house program created in LabTECH® by Dr. C.J. Lim logged data from the pressure transducers at a higher frequency (48 Hz) for short periods of time, typically one minute. These data were used to analyze dynamic pressure fluctuations within the column. The program logged the pressure transducer voltage, which was then converted to pressure during post-run analysis. 67 He l ium T r a c i n g A number of steady-state helium tracer tests were carried out to investigate how gas was circulated and distributed within the fluidized bed. Typically this involved continuous injection of helium into one part of the bed, usually through the oxidant distributor, and determining its concentration at points in the vessel through a thermal conductivity detector (TCD). The system is shown schematically in Figure 3.8. The helium flow to the column was measured with a rotameter (FI-05) at a supply pressure of 307 kPa. A calibration curve for the helium rotameter appears in Figure A3.2.11, Appendix 3.2. A vacuum pump was installed to draw sample gas and atmospheric air through the TCD. As the TCD signal was very sensitive to changes in flow, it was critical to maintain the same sample and air flow when taking data. This was achieved by throttling the sample gas inlet valve to maintain the TCD pressure, obtained by a water manometer, at the desired value. The sample and reference gas flows were measured with two rotameters (FI-06 and FI-07). The TCD signal voltage was logged in the system computer at a frequency of 0.2 Hz. The recorded TCD voltage was converted to the helium concentration in the post-run data treatment. Figure A3.2.12 (Appendix 3.2) shows a typical TCD calibration curve obtained during the testing. The calibration generally produced a very linear response, so long as the flow rates of sample and reference air were maintained at the same conditions for all tests. However, the TCD signal response was not always reproducible from the previous day's work, and the TCD was therefore re-calibrated each morning. The procedure for the TCD calibration was as follows: • The TCD system was warmed up for at least 90 minutes prior to starting calibration. • A constant flow of 250 LPM of air was maintained to the windbox to fluidize the column. No air was added to the secondary or oxidant feed distributor at this stage. • TCD samples were taken from the column freeboard. The TCD signal was zeroed by adjusting the amplifier and then monitored for 30 minutes to ensure a stable baseline. • Helium was introduced by increments into the main column feed upstream of the windbox to produce a gas mixture of up to 15% helium in the column. The signal was allowed to stabilize for 10 minutes prior to recording the TCD signal. A linear equation for the TCD response was developed and used for subsequent helium tracer experiments. The zero of the TCD system was tested periodically throughout the day to check for instrument drift. 68 Oxidant distributor Oxidant feed Water manometer I < § i > ^ | - ( ^ — TCD detector - I X } -Vacuum Pump Reference Air Sample throttle valve (multiple sample points) 0-5 V DC • TCD Signal Amplifier/ Converter TCD Data Acquisition | Main feed Figure 3.8: Helium tracer system for cold model column 3.5 Experimental Results A series of experiments was carried out on the cold model column using the three different panel geometries described above. The solid panels were tested first, followed by the horizontally slotted, and finally the vertically slotted panels. The column solids were dumped after testing one panel geometry, and then refilled once the core panels had been changed. All other system components remained unchanged for the cold model tests. The main variables in the cold model experiments were: • Main air flow to windbox • Secondary air flow to annular quadrants 69 • Oxidant air flow to splash zone • Membrane panel geometry (solid, horizontally slotted and vertically slotted) 3.5.1 Fluidization Observations Initially, the cold model column was filled with FCC solids to a depth of 50 mm below the top of the core box, which contained the solid panels. It was observed that solids circulation did not start until the total expanded bed height was above the top of the core box. In order to maintain solids circulation at low feed rates, additional particles were added to the column to bring the settled bed depth to 25 mm above the core box. This settled bed depth was maintained for all subsequent tests. When the settled bed depth was initially below the top of the core box, it was observed that once the flow of fluidization gas was stopped, several of the core channels were completely full and others empty. This implied a poor distribution of gas and solids among the channels of the solid-panel core. This maldistribution was subsequently confirmed by pressure data. In general, the column operated very smoothly. As soon as the main flow of gas was initiated, gas would flow up the reactor core and solids would start to circulate up the core and down the annular regions. No net upward flow of solids in any of the annular regions was observable from the outside in the range of column flows investigated. No significant bypassing of the main air flow to the annular regions was noted during the testing. This implies that the simple orifice plate distributor of the cold model was at least adequate for the system. Circulation of solids improved significantly once a small flow of air was introduced to the bottom of the annular regions through the secondary distributor. No bubbles were seen in the annular regions until the secondary air flow reached ~10 S L M (0.6 NmVh). Assuming an even distribution of gas throughout the annular region, 10 S L M corresponds to a superficial gas velocity in the annular region of 0.005 m/s, roughly double the measured minimum superficial gas velocity (Um/) for the FCC particles. Periodic bubbles were noted when the secondary air flow was increased to 20 S L M (1.2 NmVh), and more significant gas flow up the annulus was noted at 30 S L M (1.8 NmVh). Upward bubble movement in the annulus was hindered by the downflow of solids, leading to occasional stagnation and bubble flattening. No downward bubble flow in the annulus was observed. There was very significant bubbling at a secondary air flow of 40 S L M (2.4 NmVh). The overall solids movement at this flow was still downward in all quadrants, but there were regions where solids were occasionally dragged upwards by air bubbles. The bubbles in the annulus region were visually observed to be approximately 2 to 5 cm in diameter. 70 Numerous V" (6.4 mm) pressure impulse tubes were located in two of the annular quadrants (north and south). In these quadrants, the solids flow was markedly lower than in the two empty annular quadrants (east and west), sometimes even close to stagnant. It was also observed that impulse lines in the north and south quadrants occasionally caused the gas bubbles to coalesce, allowing the annular gas to bypass the particles and flow quickly upward. The bed typically expanded by about 150-180 mm above the initial settled bed depth, representing an expansion of roughly 10% over the original settled bed depth. Bed expansion was slightly greater at high feed rates, but was difficult to quantify due to the very vigorous fluidization in the splash zone and adhesion of particles to the column wall. Once the main feed flow increased to about 100 SLM, corresponding to a core gas superficial velocity of 0.08 m/s, periodic sprays from the reactor core channels into the splash zone became evident. At this feed rate, the height of the sprays above the expanded bed surface ranged from about 100 to 150 mm. As the main feed flow was increased, the frequency and height of the core sprays also increased, with the spray height reaching up to 600 mm above the bed surface at high feed rates (> 600 SLM, core gas superficial velocity > 0.48 m/s). Solids carryover from the column only became significant when the total column flow exceeded about 700 SLM. At this feed rate, the superficial velocity in the main column was 0.18 m/s, and 0.09 m/s in the expanded column freeboard (0.4 m ID). The collected entrained solids appeared to be mostly fines, however the particle size distribution was not analysed. The terminal velocity ((/,) of the average diameter of FCC particles in the cold model was calculated to be 0.20 m/s, based on the dimensionless terminal velocity (£/,*) and dimensionless particle diameter (dp*) and an equation given by Haider and Levenspiel (1989) for a particle sphericity of 1. - l 18 0.591 2+ / J * \ 0 . 5 (dpy (dp) (3.9) 3.5.2 Solids Circulation A predictable, high solids circulation rate is key to the ICFBMR concept. Circulating solids transport heat from hot zones, the reactor wall or oxidation zone in the upper reactor, to the reactor core. The temperature decreases up the reactor core, as the main SMR reactions are endothermic. A high solids circulation rate is therefore essential to ensure that the temperature gradient in the reactor core remains low and that the temperature is sufficient for the reaction. Too low a temperature in the core would reduce the hydrogen generation and recovery of the overall system: 71 • Cooler reaction temperatures shift the SMR reaction equilibrium to the left (see Figure A 1.1, Appendix 1), reducing the hydrogen partial pressure in the reactor. • Cooler temperatures reduce the rate of the reactions. • Cooler temperatures reduce the permeability of palladium alloys. The solids circulation rate in the cold model column was estimated: • Through physical observations at the column wall, based on multiple measurements of the downward solids velocity in each of the four annular quadrants at their widest section (north, south, east and west); • Then averaging the annular solids velocity from the four quadrants; • Finally multiplying the average annular downward velocity by the total annular cross-sectional area to obtain the solids flux. The above assumes that the solids velocity throughout each annular compartment is uniform, i.e. the same throughout the region as measured at the column wall. In reality, the solids flow was observed to be slightly lower in the regions where the corners of the core box approached the column wall. The solids flow is also likely slightly higher in the interior of the annular region, than at the column wall. The overall error is likely to be small in view of these offsetting factors, probably <15%. Solids circulation was not affected by flow of oxidant air added to the splash zone in the upper reactor. Solids circulation was measured for the following three variables: • Panel type (solid, horizontally slotted and vertically slotted) • Main air flow to windbox (100-700 SLM) • Secondary air flow to annular quadrants (10-40 SLM) The average downward solids velocity in the annulus was quite reproducible for a given set of operating conditions. However, there was considerably variation in the solids velocity among the four annular quadrants. Generally the east and west quadrants had significantly higher solids flow than the north and south quadrants, presumably due to the presence of the pressure impulse lines in the north and south quadrants. Figure A3.2.13 (Appendix 3.2) presents the ratio of the maximum to minimum quadrant velocity for all data points taken during the cold model testing, plotted against the main air flow to the column. The average maximum-to-minimum quadrant velocity of all the cold model data points was slightly greater than 3. 72 Summary data of the solids circulation testing are presented in Figures 3.9 to 3.11 for solid, horizontally slotted and vertically slotted panels. The lines on the three graphs are best-fit quadratics and are shown to aid data viewing. The vertical error bars on these figures indicate the 95% confidence interval (t-test) of the average annular solids velocity. It is possible that the error bars underestimate the data variance, but the figures do show clear trends that are consistent for all three panel types: • Solids circulation tended to increase linearly with the main feed flow up to about 300 SLM, which corresponds to a superficial gas velocity in the core (Ucore) of 0.25 m/s. • Solids circulation tended to level off once the main feed reached -500 SLM (Ucore = 0.42 m/s). • Solids circulation increased significantly, even doubling in some cases, when the annular air flow was increased from 10 to 20 SLM. • Solids circulation only increased slightly, or remained the same, when the secondary air was increased from 20 to 30 SLM. Increasing the secondary air above 30 SLM did not increase the solids circulation. 0.05 A 'o o ffi > </> •u Solid panels < o o 100 200 300 400 500 600 700 800 Main air flow (SLM) Figure 3.9: Average downward solids velocity in the annular regions with solid panels as a function of main column flow and secondary air flow (Best fit quadratic line shown) 73 0.05 E 0.04 o o CU u> 0.03 !5 o co | 0.02 to CD D) 2 0.01 > < 0.00 A 30 S L M annular air • 20 S L M annular air O 10 S L M annular air Horizontally slotted panels 100 200 600 700 800 300 400 500 Main airf low (SLM) Figure 3.10: Average downward solids velocity in the annular regions with horizontally slotted panels as a function of main column flow and secondary air flow (95% confidence interval) 0.05 1 0.04 u o cu > </) 0.03 ;o o I 0.02 c c CO cu CO cu > < 0.01 0.00 • 30 S L M annular air A 20 S L M annular air 0 1 0 S L M annular air Vertically slotted panels 100 200 300 400 500 Main air flow (SLM) 600 700 800 Figure 3.11: Average downward solids velocity in the annular regions with vertically slotted panels as a function of main column flow and secondary air flow (95% confidence interval) Solids velocity data at an annular air flow of 20 SLM for the three different panel geometries are plotted in Figure 3.12. The circulation for the vertically slotted panels was higher than for the other two panels within the flow range of 200 to 600 SLM. A best-fit quadratic curve for each panel type is plotted. 74 0.050 • solid panels O horizontally slotted panels 0.000 -I , 1 1 1 r- , , 1 0 100 200 300 400 500 600 700 800 Main air flow (SLM) Figure 3.12: Average downward solids velocity in the annular regions with a secondary air flow of 20 S L M for all three panel geometries (95% confidence interval) The data in Figure 3.12 are now transformed to an internal circulation rate in Figure 3.13. The main air flow is converted to a superficial gas velocity in the core, based on the total cross-sectional area open to flow in the core (sum o f the 8 separate channels) of 0.0188 m 2 . Annular solids velocity is converted to a mass and volumetric flow, assuming a F C C bulk density o f 880 kg/m 3 and an annular cross-sectional area of 0.0354 m 2 . A s can be seen, the total internals solids circulation rate for the vertically slotted panels reached a maximum of about 1.2 kg/s at a core velocity o f about 0.5 m/s. The circulation rates for the solid and horizontal slotted panels were slightly lower, and appear to have levelled beyond a core velocity of 0.5 m/s. 75 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Core superficial gas velocity (m/s) Figure 3.13: Average annular solids flow rate with a secondary air flow of 20 SLM for the three panel geometries The curves in Figure 3.13 are nearly linear at relatively low superficial core gas velocities (<0.3 m/s). The core gas velocities studied in the ICBFMR pilot plant fall within this range. The slope of the linear section of the best-fit curves suggests that the volumetric flow of circulating solids is roughly 15 to 20% of the volumetric flow of the gas feed flow rate to the reactor core. Excess air flow to the annular regions leads to more gas ascending in the outer quadrants, thus bypassing the membranes in the reactor core. It is therefore desirable to minimize this flow, and the above data suggest that 20 SLM annular air is an optimal value for the cold model configuration. 20 SLM evenly distributed over the annular area corresponds to a superficial gas velocity in the annulus of 0.010 m/s, approximately four times the minimum fluidization velocity (Umj) for the cold model system. 3.5.3 Pressure Data A number of absolute and differential pressure transmitters monitored the cold model system. High (48 Hz) and low (< 1 Hz) frequency pressure data were separately recorded for the three different panel geometries at a variety of air flows. The low frequency data were used to investigate overall bed properties, such as bed voidage and flow stability. The high frequency data were transformed into 76 the frequency domain to investigate hydrodynamic regime features such as slugging and onset of turbulent fluidization. Core Channel- to-Channel Flow Stability (Low Frequency Data) Observations while operating the cold model with solid panels suggested that there were non-uniform solids loadings among the core channels. Long-term pressure data from within the core channels confirmed that the solid panel configuration led to unstable flow patterns among the 8 channels. Some channels had much higher solids hold-up, or perhaps were plugged, compared to neighbouring channels. This is demonstrated in Figure 3.14, where the absolute pressures in each of the eight core channels were monitored over an extended period of time. The large number of data points makes viewing of individual channel pressures difficult, but every channel is seen to have had episodes of large pressure swings, representing significant variations in solids loading. This channel pressure instability for the solid (impervious) panels was noted over all flow ranges studied. This instability is quite different from Figure 3.15, which shows core channel pressure data for the horizontally slotted panels at the same flow conditions. Note the substantial difference in ordinate scale of the two graphs. 3 o 5 in channel 1 channel 3 - - channel 5 channel 7 channel 2 -channel 4 - channel 6 -channel 8 Solid panels, 480 SLM main, 20 SLM secondary air 10 15 Time (mins) 20 25 30 Figure 3 14: Absolute pressure measurements from the middle of the core channels with solid panels (480 SLM main air, 20 SLM secondary air) 77 6.5 o i a a. o < 5.5 / ' w , . ' l V •fV I ' V '•; t > v • 1 -A A . w V V Horizontally slotted panels, 480 SLM main, 20 SLM secondary air 10 15 Time (mins) 20 25 30 Figure 3.15: Absolute pressure measurements from the middle of the core channels with horizontally slotted panels (480 SLM main air, 20 SLM secondary air) When operating with the horizontally slotted panels, all channels were found to have nearly the same pressure, with only small pressure fluctuations. The lower pressure curve, channel 5, is likely lower than the others due to a small error in signal zeroing. Core pressure data for the vertically slotted panels are not shown, but were very similar to those for the horizontally slotted panels. The slots clearly improved flow distribution among the core channels compared to the non-communicating solid panels. Core Voidage (Low-Frequency Data) It is important to estimate the voidage in the reactor core, as this aids in quantifying the catalyst loading and volume available for reaction in the main reforming section of the ICFBMR. The voidage in the core was estimated from the pressure drop within channel 4 over a height of 1.27 m for a number of flow conditions. Figures 3.16 and 3.17 show the channel pressure drop as a function of the main and secondary air flows for the solid panels and for the vertically slotted panels. 78 9 a R Q. o Q . O t-T3 « k— 3 • in 7 in ' a> a. "aj c c .c 6 O • 03 Solid panels y = -0.0027x + 8.18 R 2 = 0.900 m i A 30 SLM annular air • 20 SLM annular air O 10 SLM annular air - Linear (20 SLM annular air) 100 200 300 400 500 Main Feed (SLM) 600 700 800 Figure 3.16: Pressure drop across channel 4 in cold model core with solids panels (95% confidence interval) « 8 a. o •a o k. 3 cn 7 B) ' 0> L . CL ~a c c n S 6 1 " H . y = -0.0025x + 9.06 ^ R 2 = 0.992 tii «• •• A A 30 SLM annular air • 20 SLM annular air o 10 SLM annular air Linear (20 SLM annular air) Vertically slotted panels 100 200 300 400 500 Main Feed (SLM) 600 700 800 Figure 3.17: Pressure drop across channel 4 in cold model core with vertically slotted panels (95% confidence interval) The channel pressure drop data from the solid panel (Figure 3.16) are widely scattered and have larger error bars compared to the vertically slotted panels due to periodic channel blockage and de-79 blockage as described above. The data for the vertically slotted panels are much more consistent and indicate that the voidage increased nearly linearly with the main gas feed flow. The channel pressure drop decreased by 20% as the main air flow was increased to 700 SLM, with the secondary air maintained at 20 SLM or 30 SLM, corresponding to an increase in voidage. The voidage increased slightly more (23%) over the same flow range when the secondary air was reduced to 10 SLM. Data for the horizontally slotted panels are presented in Figure A3.2.14 (Appendix 3.2) and are virtually identical to the data for the vertically slotted panel presented in Figure 3.17 above. Higher Frequency Pressure Data Higher frequency pressure data were recorded from the cold model to investigate the fluidization behaviour at various flow rates. Fast Fourier Transformation (FFT) was used to convert selected pressure data from the time domain to the frequency domain. The frequency range of interest in fluidized beds is usually 0 to 10 Hz. As data collection below 0.5 Hz is susceptible to signal noise, the frequency range of interest for the cold model work was taken to be 0.5 to 10 Hz. See Ellis (2003) for a discussion of FFT data treatment of pressure signals. Few significant conclusions could be drawn from the pressure-frequency data. Sample data are shown in Figure A3.2.15 in Appendix 3.2. This figure presents absolute pressure spectra for core channel 4 with the solid and vertically slotted panels for a variety of flow conditions. No significant peaks are noted for the vertically slotted panels. There is an increased frequency response for the solid panels at approximately 1 Hz, perhaps indicating some degree of slugging in the channels. Because of the ICFBMR geometry, the pressure impulse lines for the cold model pressure gauges were very long (7-10 m), and this may have resulted in some dampening of the measured pressure fluctuations. 3.5.4 Helium Tracer Experiments Helium tracer experiments were undertaken to investigate the movement of gas within the fluidized bed for all three panels geometries types (solid, horizontally slotted and vertically slotted) under different flow conditions. Two sets of tracer experiments were performed: • Helium was added to the oxidant air line and samples were extracted throughout the bed to investigate recirculation of gases from the top to the bottom of the internally circulating fluidized bed. • Helium was added to the bottom of a core channel and samples were then taken at various points higher in the core to investigate gas mixing within the core. 80 The helium T C D response was very sensitive to flow, and thus care had to be taken to ensure that all sampling was performed in a reproducible way. Even with this, there was considerable scatter in the T C D results. Several tests were performed to check the reproducibility o f the T C D system. Figure 3.18 shows the measured helium concentration at various locations within the cold model column when the bulk helium concentration in the bed was zero and 3.28%. There are deviations between the measured and actual values beyond the standard deviation o f the measured T C D signal. It was concluded that the T C D data can be relied upon to detect trends, but that absolute values need to be treated with a degree of caution. It should also be noted that the gas sample points in the core were at the edge o f the channel (see Figure A3.2.7 in Appendix 3.2), and therefore in some cases may not be representative o f the interior o f the channel. Figure 3.18: Measured helium concentration at various points in the cold model with 0% and 3.28% helium in the main feed for the vertically slotted panels Helium Addition to Oxidant Distributor In this series o f tests, helium tracer was added to the cold model column through the oxidant distributor, located just above the top of the core box. The helium concentration in the oxidant air supply ranged from 9 to 25%. Gas was then sampled to see how much helium was being circulated from the top of the column to lower points in the bed. 81 • Gas from the annulus was sampled near the bottom of the four quadrants (225 mm above the bottom of the core box). • Gas from the eight core channels was sampled near the bottom of the channels 4 and 5, mid-height in channels 1 to 8 and near the top of the channels 4. to 6. During the tracer tests, the main air flow was varied for all three panel geometries, whereas the secondary air to the annulus was maintained at 20 SLM for all tests. A summary of all data is presented in Table 3.4. Some notable trends include: • The concentration of helium in the annulus was typically close to the helium concentration at the column exit, much higher than in the core. This is expected, as the air that is dragged down by the annular solids is from the well-mixed zone in the upper column. In several of the tests with solids panels and low air flows, the annular helium concentrations were close to the helium concentration in the oxidant feed. Clearly gas mixing in the upper column was relatively poor in these cases, and the gas dragged down by the annular solids was mostly from the oxidant supply. • Assuming that all of the main air feed reports to the core and knowing the average helium concentrations in the annulus and core, the amount of gas dragged down by the annular solids circulation can be estimated. The estimates are shown in Table 3.4. The average gas velocity in the annulus should be less than the average particle velocity in the annulus by the value of Umfl emf. If bubbles in the annulus are ignored, the volumetric flow of gas (Qg,ann) dragged down by the annular flow of particles is given by: Qs,ann = UannAam = (Vp-^L)Aamemf (3.10) £mf As Umf for the FCC (0.0026 m/s) is significantly lower than the range of measured particle velocities (0.01 to 0.04 m/s), the effect of this gas slip is ignored. If a voidage of 50% in the annular solids is assumed, the volumetric flow of gas dragged down in the annulus should be just slightly less than the measured solids circulation, expressed in the same units. The helium tracer data in fact supports this. In all but one run, the flow of annular gas circulation estimated from the helium tracer data is within ±30% of the solids circulation obtained by measuring the downward velocity of the annular solids at the wall. • One key benefit claimed for the ICFBMR concept is that addition of air does not significantly reduce the membrane flux, as the bulk of the nitrogen in the air quickly leaves into the freeboard 82 and beyond without affecting the hydrogen partial pressure in the reactor core where the membranes are located. This feature is verified in the helium tracer data. In all but one run, the amount of helium circulated to the core from the annulus was only approximately 10% of the total helium added to the column. As the data from run #2 illustrate, however, it is clear that the top of the column needs to be well mixed with uniform flow through the core to benefit fully from this reactor feature. The helium concentration measured at the top of the core channels tended to be higher than at the middle. From this finding, it seems likely that some solids from above the core box were drawn down into the upper regions of the core flow channels. Measured helium concentrations in the core with solid panels tended to vary more than for the slotted panels, especially for channels 1 and 8 at either edge of the core box. This is as expected, as gases within the core channels can interchange with the slotted panels, but not with the solid (impervious) panels. In addition, channels may plug when operating with the solid panels, leading to uneven gas distribution, whereas the flow is much more uniform when adjacent panels can "communicate" with each other. 83 Table 3.4: Summary of cold model helium tracer results (all % listed in table are vol%) Core panel geometry Solid Danels Horizontally slotted panels Vertically slotted panels Run number 1 2 3 4 7 6 8 9 Main airflow (SLM) 94 240 480 713 240 480 240 480 Seconday air flow (SLM) 20 20 20 20 20 20 20 20 Oxidant air flow (SLM) 70 70 140 210 90 155 70 140 Helium flow (SLM) 22.7 22.7 28.3 21.4 22.7 28.3 22.7 24.2 Gas velocity in core (m/s) 0.09 0.23 0.46 0.68 0.23 0.46 0.23 0.46 [Helium] in oxidant supply 24.5% 24.5% 16.8% 9.3% 20.1% 15.4% 24.5% 14.7% [Helium] in column exit 11.0% 6.4% 4.2% 2.2% 6.1% 4 .1% 6.4% 3.6% Helium measurements [He] % St. dev. [He] % St. dev. (He] % St. dev. [He] % St. dev. [He] % St. dev. [He] % St. dev. [He]% St. dev. [He] % St. dev. Sample location (mm above main distributor) East quadrant - bottom (325) 27.91 2.20 20.64 1.12 3.02 1.54 5.83 0.32 4.68 0.43 3.11 0.43 3.71 0.31 3.15 0.39 Sample location (mm above main distributor) West quadrant - bottom (325) 33.38 0.16 17.63 2.21 4.11 0.88 2.76 1.03 3.27 0.39 4.83 0.26 6.26 0.33 4.80 0.18 Sample location (mm above main distributor) North quadrant - bottom (325) 29.62 0.74 23.40 0.29 5.70 0.71 5.06 0.41 4.00 0.21 Sample location (mm above main distributor) South quadrant - bottom (325) 33.59 0.08 19.08 2.29 4.61 4.11 5.11 0.40 3.03 1.44 0.96 0.31 4.68 0.36 Sample location (mm above main distributor) Channel 4 - bottom (230) 15.10 1.45 7.68 2.55 0.56 0.15 Sample location (mm above main distributor) Channel 5 - bottom (230) 11.66 0.80 8.64 1.71 Sample location (mm above main distributor) Channel 1 - mid height (862) 8.57 1.74 0.46 0.26 0.98 0.13 0.42 0.28 0.36 0.43 0.19 0.13 0.46 0.06 Sample location (mm above main distributor) Channel 2 - mid height (862) 2.33 1.38 -0.58 0.16 0.02 0.23 1.46 0.11 0.18 0.07 0.10 0.09 0.13 0.14 Sample location (mm above main distributor) Channel 3 - mid height (862) 5.48 0.33 0.62 0.19 0.21 0.04 0.41 0.05 0.60 0.14 0.60 0.09 0.71 0.08 Sample location (mm above main distributor) Channel 4 - mid height (862) 2.90 0.40 0.39 0.22 0.16 0.04 0.56 0.15 0.37 0.09 0.35 0.09 0.36 0.12 Sample location (mm above main distributor) Channel 5 - mid height (862) 5.10 0.76 0.54 0.19 0.58 0.12 0.32 0.13 -0.22 0.02 0.59 0.11 ' 0.73 0.13 Sample location (mm above main distributor) Channel 6 - mid height (862) 4.11 0.89 0.51 0.23 0.21 0.10 0.39 0.14 0.30 0.06 0.57 0.11 0.54 0.16 Sample location (mm above main distributor) Channel 7 - mid height (862) 1.36 0.19 0.11 0.16 0.00 0.13 0.54 0.12 0.26 0.05 0.67 0.15 0.31 0.08 Sample location (mm above main distributor) Channel 8 - mid height (862) 4.21 0.09 1.86 0.20 0.37 0.20 0.58 0.16 0.24 0.10 2.56 0.59 1.47 0.22 Sample location (mm above main distributor) Channel 4 - top (1494) 3.43 0.12 4.87 0.72 1.45 0.43 0.49 0.16 Sample location (mm above main distributor) Channel 5-top (1494) 3.38 0.16 3.42 2.24 0.92 0.27 0.83 0.36 Sample location (mm above main distributor) Channel 6 - top (1494) 4.34 0.98 1.26 0.29 0.38 0.08 Average of all quadrants - bottom 31.1 20.2 2.47 3.9 0.81 4.8 1.43 3.7 0.89 4.3 1.90 5.0 1.80 4.2 0.76 Average of all channels at mid-height 4.3 2.23 0.49 0.68 0.32 0.33 0.58 0.36 0.26 0.23 0.70 0.78 0.46 0.41 Estimate of annular gas downflow (SLM) 64 69 50 46 31 39 60 Measured solids circulation (LPM) 14 39 64 71 36 55 51 82 Fraction of helium recycled 57% 10% 11% 7% 5% 9% 10% Helium Addition to Core - Vertically Slotted Panels An investigation into gas mixing with the reactor core equipped with the vertically slotted panels was also attempted. For this set of experiments, helium was added to the bottom of channel number 4 for a period of 10 minutes, and the helium concentration was monitored at locations higher up the core box. The cold model operating conditions for the tests were: • Main air flow 240 SLM; • Secondary air 20 SLM; • No oxidant air flow; • 6.0 SLM helium flow added to flow channel 4, 130 mm above the bottom of the core box. The 10-minute helium addition test was repeated four times for four different core sampling points: at top of channels 4, 5 and 6 and at mid-height of channel 8. See Figure A3.2.7 in Appendix 3.2 for the plan view of the core box and the locations of the sample points. Results are presented in Figure 3.19. Figure 3.19: Measured helium concentration (vol%) at different sample points in upper core box after 10 minute continuous injection of helium into bottom of channel 4 (vertically slotted panels) As all sampling exhibited a system response delay of several minutes due to the long sampling lines and low sampling flow rates, no conclusions on the dynamic response of the system could be drawn. The average helium concentrations at the top of channels 4 and 5 were 2.2% and 2.1% respectively, close to the concentration in the column exit. The average helium concentrations at the top of channel 6 and the middle of channel 8 were significantly lower, 0.8% and 0.7% respectively. As noted above, the positions of the vertically slotted panels were alternated so as to create a back-and-forth serpentine-line connection between the channels. The sample point for channels 4 and 5 were thus effectively on opposite sides of the same channel, so the similar helium concentrations are Time after He injection (mins) Time after He injection (mins) 85 not unexpected. The sample points for channels 6 and 8 were several "U-turns" away from the channel 4-injection point, clearly reducing gas dispersion. If the helium in channels 6 and 8 were due solely to the circulation of gas from the annular regions, a concentration of roughly 0.4% would be expected, so some dispersion of helium from channel 4 likely occurred through the intervening slots. 3.6 Discussion The cold model testing clearly showed that solid (i.e. impervious, non-communicating) membrane panels within the core box led to poor gas-solid distribution among the core channels. Slotted, communicating membrane panels greatly improve the uniformity within the core, and are thus preferred in the ICFBMR design. The pressure impulse lines in the north and south annular quadrants added flow resistance, and thus these quadrants exhibited significantly lower solids circulation than the other two quadrants. Such tubing and obstacles should be minimized, especially if external heating of the reactor is used, as this will lead to non-uniform heat transfer to the annulus solids. 3.6.1 Solids Circulation Increasing the main feed gas flow to the core draft box increased solids circulation up to a core gas superficial velocity of approximately 0.5 m/s, at which point circulation tended to level off. Increasing annular air from 10 SLM to 20 SLM led to a significant increase in the solids circulation rate. A further increase in the annular air to 30 SLM resulted in only a marginal increase in solids circulation, or in some cases, no significant change. The cold model testing clearly shows the importance of independent control of the annular and main gas flows, especially if the reactor is to operate at variable feed rates. An annular flow of 10 SLM corresponds to a superficial velocity in the annulus (Uann) of 0.0051 m/s, 1.9 times Umf of the FCC particles. However, it is very likely that some of the air introduced to the annulus did not flow up the annulus, but rather was dragged into the core draft box, as noted in earlier ICFB studies (e.g. Song et al., 1997, Namkung, 2000). An estimate of the amount of annular gas bypassing to the core is difficult to infer from the cold model data. Annular gas bypassing to the core would be an important design parameter if the annular gas was to be chemically used in the annulus, such as air for oxidation. In the ICFBMR design, however, the annulus gas can be compositionally identical to the main feed (i.e. steam and methane) and transfer of the annulus gas to the core improves utilization of the reactor feed. Bypassing of feed gas up the annulus would result in lower utilization of natural gas for reforming and H 2 permeation, but this quantity is judged to be small 86 (<5% of total natural gas feed) and the annular gas would still be available for oxidation in the upper reactor. Of the four retained hydrodynamic scaling parameters discussed earlier in this chapter, only the last term, Gpcore / pp Ucore, contains the solids flux (Gpcore). This scaling parameter suggests that a plot of circulation flux vs. pp Ucore in the cold model should provide a reasonable estimate of that in the hot reforming reactor. Figure 3.15 presented above showed the solids circulation rate (mpcore, kg/s) as a function of the core superficial velocity (Ucore). Given an open area available to flow in the core (Acore = 0.0188 m2), the FCC particle density (1,600 kg/m3) and assuming an annular voidage of 0.5, these data are transformed into solids fluxes (see Figure 3.20 below). 0 200 400 600 800 1,000 1,200 Product of core superficial velocity and particle density (kg/m2 s) Figure 3.20: Measured cold model solids flux in core draft box (Gpcore) for horizontally slotted and vertically slotted panels vs. product of superficial gas velocity in core (Ucore) and particle density (pp), with an annular air flow of 20 SLM If the geometry of the cold model is maintained, the regression curves in Figure 3.20 are useful design tools to estimate the solids circulation rate in the hot ICFBMR. Note that the calculated solids flux is based on the cross-sectional area open to flow in the core (i.e. inner area of core draft box less area blocked by panels). The catalyst particle density (pp) in the ICFBMR pilot reactor was approximately 30% higher than that of the FCC particles in the cold model. This implies that in order to achieve a certain solids flux, the required core velocity in the hot ICFBMR would be less than for the cold model. 87 The question of what factors control the solids circulation rate remains. Several studies have been conducted on fluidized beds or spouted beds fitted with draft tubes to generate an internal circulation of particles similar to the ICFBMR concept. Most of these studies have used a configuration where the annulus turn into the core draft tube restricts the flow, whether through orifices or by the gap under the draft tube (i.e. Milne et al., 1992, Song et al., 1997, Shih et al., 2003). In this case, the annular turn area (orifice or gap) is smaller than the annular area, and the pressure drop over the gap between the annulus and the draft tube (AP0) determines the solids flow rate, Gp (Song et al., 1997, adapted from Judd and Dixon, 1978): C0 is the particle discharge coefficient, pan„ is the density of the annulus upstream of the gap or orifice and DH is the hydraulic diameter. Equation 3.11 also indicates why, after a certain point, some researches have found that the rate of solids circulation with additional annular gas levels off, as bubbles in the annulus reduce the annular density (pann), decreasing the density difference (AP0) between the core and annulus (Song et al, 1997). However in our cold model, the annular gap area (Agap) is almost twice the annular area (Aann) (r = 5.3); hence the pressure drop over the annular gap is very low and thus equation 3.11 cannot describe the solids circulation. Figure A3.2.16 in Appendix 3.2 presents the measured annular pressure vs. the measured pressure in the bottom of Channel 4 corrected to the same elevation, and shows, as expected, that the annular pressure closely follows the core pressure. As shown in Figures 3.11 to 3.13 presented earlier, solids circulation increases significantly when the annular gas flow increased from 10 to 20 SLM, but did not change significantly when it was increased beyond 20 SLM. Clearly the flow of particles down the annulus and the annular turn into the core, restricted solids flow at 10 SLM. Perhaps an increase in bubbles near the bottom of the annular increased the mobility of solids as they turn into the vessel core. The extra annular air may have also helped to prevent local defluidization in the annulus. Figures 3.11 to 3.13 also clearly show a consistent trend of core velocity on solids circulation with all three core panels tested. For a given annular air gas flow, solids circulation increased nearly linearly with core gas superficial velocity up to approximately 0.3 m/s, after which solids circulation tended to level out. In a conventional bubbling fluidized bed, solids are transported vertically within the bed in the wake of the rising bubbles, as demonstrated in Figure 3.21. (3.11) 88 Colorless solid -Colored solid ~ 7 7 Figure 3.21: Sketches showing solids entrainment in the wake of a rising bubble (from Kunii and Levenspiel, 1991, after photographs from Rowe and Partridge, 1965). If solids circulation up the vessel core is primarily due to this mechanism, the solids flux in the vessel core (GpXOre) could then be described as a function of the wake fraction of the bubbles (fw) and the volumetric bubble flow up the vessel core (Q0). p,core = (Ucore - Umf )pp (!-*)« UcorePp (1 - e) (3.12) At core superficial velocities below 0.3 m/s, the fluidization flow regime is likely the bubbling regime, with the wake fraction relatively constant. For most of the testing in the cold model, Ucore » Umf, so the circulation rate is therefore approximately proportional to Ucom. It is further postulated that as the core superficial gas velocity is increased above 0.3 m/s, a gradual transition to turbulent fluidization occurs, and spherical-cap bubbles are increasingly replaced by streaking voids (see Figure 3.1), whose carrying capacity levels off to a constant value. As previously shown in Figure 3.13, the measured solids circulation for the vertically slotted panels was higher than for the solid panels and the horizontally slotted panels. It is not known what factors reduced the solids circulation in the horizontally slotted panels, but perhaps it is due to disruption of the particle-containing wakes in the rising bubbles at the communicating gaps between the channels. The solid panels demonstrated lower circulation rates than the vertically slotted panels. This is likely due to periodic blockages of the core flow channels. Why the solid panels show a very slightly higher circulation rate than the horizontally slotted panels is not known. 3.6.2 Other Geometry Aspects Some geometric aspects that could be adjusted in the ICFBMR design, but were not tested, are as follows: Height of core box: In the ICFBMR design, the height of the core draft box is primarily set by two parameters: the amount of membrane surface required for permeation and the design superficial gas 8 9 velocity in the core (Ucore). The height of the core draft box in the cold model, 1.52 m, is the same as in the ICFBMR pilot reactor. Song et al. (1997) found that doubling the height of the draft tube in the ICFB system from 0.3 to 0.6 m had a negligible effect on solids circulation. If the circulation rate up the core depends on bubble wakes as suggested above, it follows that solids circulation would be relatively insensitive to the height of the core box. Ratio of annular and core draft box area: If a square core draft box is used in the ICFBMR, the dimensions of the annulus are set, with the only adjustable parameter being the distance from the corners of the core box to the column wall. Thus, if square core box is used in the ICFBMR design, the ratio of the annular area to the core box area will be approximately equal to that of the cold model. However, other core box geometries, such as a rectangular or circular draft tubes, may be preferred, as this would allow the annular area to be reduced, thus shrinking the overall volume of the reactor (see Figure 3.22). The cold model testing indicated that for a given annular gas flow, the solids circulation rate is a function of the core velocity. However, it is not known how much the annulus area can be shrunk before it starts to reduce solids circulation. Arrangement of planar membranes in a non-rectangular draft tube is an additional challenge of the alternate designs, and testing would be required to establish the solids circulation characteristics of the different geometries. A. Square core B. Circular core C. Chord layout (cold model) Membranes — \ (annulus) Figure 3.22: Some alternative ICFBMR layouts Gap between distributor and draft tube: The gap below the core draft box was 100 mm in the cold model. As the gap area under the core box is larger that the annular area, there is unlikely to be any significant improvement on solids circulation by increasing this distance. A very large gap could lead to increased bypassing of core gas to the annulus and possible start-up issues. The "turn" of solids from the annulus to the core may well restrict solids circulation, but it is believed that this resistance could be better addressed with an alternate annular air distributor design. 90 Annular distributor design: The cold model had a relatively complex secondary air distributor. The cold model design was chosen as it approximated the design of the ICFBMR pilot reactor, where four small sintered plugs distributed the annulus gas. In many ICFB studies, a conical distributor has been used (i.e. Milne et al., 1992, Chu et al., 2002, Marschall et al., 1999). Song et al. (1997) tested three distributors similar to those displayed in Figure 3.23, and reported that the solids circulation for the flat plate and conical distributor were similar, and higher than for the conical sparger. Intuitively, the conical distributor appears to be a better design, and indeed Song et al. concluded that the conical distributor equipped with tuyeres was the preferred arrangement. However, this was because less bypassing of annular gas to the core was measured with this distributor (Song et al, 1997). In the case of the ICFBMR, bypassing of annular gas to the core is not critical. If the conical distributor could be modified with angled jets to accelerate the solids from the annulus to the core, an increase in circulation might be possible (Xie, 2002), but may not merit the extra fabrication complication. More study is needed A. Flat plate B. Conical plate C. Ring sparger Draft tube t t t t t t / I I \ Annular ring t t t o Annular windbox Main windbox Figure 3.23: Alternate annular distributors, adapted from Song et al. (1997) 3.6.3 Design Implications Higher design gas velocities in the core reactor reduce the overall reactor volume. The cold modelling suggests that a design core velocity below approximately 0.3 m/s is appropriate, as this is the highest gas velocity at which the solids circulation flux varies nearly linearly with velocity. Above 0.3 m/s, the solids flux levels off, which would correspond to higher temperature gradients in the reactor core. Of course the core reactor volume also need to be sufficient for the reaction kinetics. Figure 3.24, generated from HYSYS simulations (see Chapter 6 for description), indicates the expected temperature gradient from the bottom to the top of the ICFBMR core as a function of solids circulation rate for one design condition producing 50 Nm3/h of permeate hydrogen. If the core 91 velocity in this case is specified to be 0.25 m/s, the flow area in the hot reactor will be nearly identical to that in the core model (0.026 m2). As solids circulation increases, the temperature gradient in the reactor core is reduced. The predicted normalized membrane area is also plotted on the figure. At lower solids circulation rates, more membrane area would be required due to three phenomena: • At lower temperatures, the permeance of the palladium foil is reduced. • At lower temperatures, the SMR equilibrium is shifted towards reactants, reducing the hydrogen partial pressure. • Though not accounted in the data presented in Figure 3.24, reaction kinetics are slower at lower temperatures. 50 I — • • 1 1.0 0 I , . . —I 0.7 0 2 4 6 8 Catalyst solids circulation (kg/s) Figure 3.24: Effect of solids circulation on core reactor temperature drop and normalized membrane area (conditions: 50 NmVh H 2 , steam-to-carbon molar ratio = 3, methane feed to reactor = 25 NmVh, reactor pressure 1500 kPa, feeds preheated to 650°C) From the core velocity (Ucore = 0.25 m/s) and the reformer particle density (pp= 2,100 kg/m3), the predicted solids circulation rate is calculated to be 1.4 kg/s for vertically slotted panels based on the regression equation in Figure 3.20 presented above. As shown in Figure 3.24, the predicted temperature gradient in the reactor core is approximately 10°C, a very reasonable value given the large heat input required to support the reforming process. Increasing the circulation rate beyond 1.4 kg/s results in only a marginal reduction in membrane area, indicating that appropriate solids circulation rates can be achieved in the ICFBMR for the SMR process. 92 Chapter 4. ICFBMR Pilot Plant This chapter provides the basis on which the ICFBMR pilot reactor was designed, constructed and tested. Process flowsheets for the pilot plant are provided, both for external reactor heating (conventional SMR) and with direct air addition (autothermal reforming). A semi-industrial size reactor (135 mm ID) was constructed and integrated into an existing pilot plant, which was upgraded with an improved control system. The nominal design basis of the updated pilot plant was 100 SLM (6 Nm3/h) of permeate H2, although the six planar membranes installed into the new pilot reactor were insufficient to produce this much hydrogen. The ICFBMR pilot reactor proved to be very stable to operate. However the large scale of the equipment made reactor modifications, such as internals and catalyst change-outs, challenging. In addition, low catalyst activity affected the reactor performance for several experiments, resulting in the pilot plant test data being somewhat incomplete and below initial expectations. The pilot testing did confirm the key features of ICFBMR concept and showed the potential for the planar membranes to withstand the fluidized bed reforming environment. H2 permeate purity was initially very high (>99.999%, excluding N2) for up to -180 hours of testing, after which the purity dropped to 99.8%. The hydrogen recovery relative to the feed natural gas was relatively low (maximum of 1.17 Nm3 H^/Nm3 of natural gas), as was the maximum H2 production (17.6 SLM or 1.06 Nm3/h), limited by the installed membrane area. 4.1 Introduction As described in the previous chapter, cold modelling proved that the ICFBMR concept was viable from a hydrodynamic perspective. The ICFBMR concept was tested in a large-scale pilot reactor. Six of the new membranes described in Chapter 2 were integrated into the reformer system to investigate their performance under actual operating conditions. Much of the original pilot plant utilized in the work of Adris (1994) and Roy (1998) was re-used in this work. An externally heated bubbling bed reactor (97 mm ID, 0.66 m tall) had been employed in 93 their tests, with small-diameter tubular membranes for hydrogen withdrawl. This reactor was too small to house the new internals (core draft box, planar membranes) o f the I C F B M R design, so a larger reactor was designed and built. 4.2 ICFMBR Reactor Combining the process design requirements with mechanical engineering considerations and space constraints made the design and installation of the new I C F B M R reactor a significant undertaking. One of the main design challenges was the operating temperature o f the reactor shell, as radiant reactor heaters were to be strapped to the exterior o f the reactor, potentially raising the wall temperature to values not applicable for stainless steels under the A S M E pressure vessel code. A s a result, it was decided to house the I C F B M R reactor (V-02 in Figure 4.7, presented later in this chapter) within a large carbon steel shell (V-01), which acted as the pressure vessel. A n automatic system was designed to pad the carbon steel shell with nitrogen to a pressure approximately 100 kPa higher than the operating pressure o f the inner I C F B M R reactor. The inner I C F B M R reactor only saw a small differential pressure, and thus could be mechanically designed to pressures much lower than the external shell design pressure (3,450 kPag). Padding the external shell also ensured that any leaks in the inner I C F B M R reactor would result in ingress o f inert nitrogen, rather than egress of flammable gases, such as methane and hydrogen, into the confined laboratory work space. 4.2.1 Process Design Design of the new reactor was an iterative procedure, influenced by the capacity o f the existing pilot plant equipment, the characteristics of the new planar membranes and information gained during cold modelling experimentation. The I C F B M R reactor was designed and supplied by Noram Engineering and fabricated by Axton Manufacturing. The final design criteria for the new pilot plant were: • Feed rate: A s the maximum natural gas methane feed rate o f the existing natural gas compressor and mass flow controller was 50 S L M (3 NmVh), this became the design gas feed rate. With this gas feed flow and assuming a steam-to-carbon molar ratio o f 3, a design steam rate o f 150 S L M (7.2 kg/h) was established. • Hydrogen production: A n economic reformer design should have product hydrogen recoveries approaching 214 moles o f permeate H 2 per mole o f CFL, fed to the reactor. It was decided to design the pilot system for a permeate hydrogen-to-methane feed ratio o f 2, equivalent to 100 S L M (6 NmVh) of H 2 . 94 • Heat input: The reactor was equipped with a new preheater, external reactor heaters to supply the process heat required for conventional SMR operation, and an oxidant delivery system which could supply air directly to the reactor for autothermal operation. • Mechanical: Although existing components of the pilot plant constrained the operating pressure to 1,500 kPa, the reactor design was to be sufficient to permit the reformer to operate at approximately 3,000 kPa at 650°C. The above established the design envelope for the new system. Two process flow diagrams (PFDs) were prepared, one showing operation with external heat input, and the other with direct air addition (Figures 4.1 and 4.2 respectively). The autothermal PFD has an air addition rate equivalent to 0.4 moles of 0 2 per mole of CH4 feed, slightly higher than that calculated to operate the reformer adiabatically in the absence of heat losses. The PFDs indicate that steam was to be the membrane sweep gas, but in practice nitrogen was used. The mass and energy balances presented in the PFD stream tables were based on HYSYS simulations. From the PFDs, the required heat input for the internal preheater and reactor heaters was established at 3.0 and 6.7 kW respectively (external heating, conventional SMR mode). In order to account for heat losses and facilitate start-up, these heaters were made significantly larger than the calculated load. The required membrane area to produce 100 SLM of hydrogen was calculated based on the conditions outlined in the externally heated PFD (Figure 4.1) and the membrane flux equation 4.1: = 0.35 m2 (4.1) for z — 50E-6 m, Q„ = 0.0744 mol/s, kH = 3.43E-7 mol/m s Pa 0 5, and Ep = 9,180 J/mol, T= 923K, PHh = 297.7E3 Pa, PHl = 50E3 Pa In the above calculation, the H 2 partial pressure was assumed to be equal to the reactor exit H 2 partial pressure (298 kPa). The permeate H 2 partial pressure was set at 50 kPa, assuming a 50% dilution of the permeate stream with sweep gas. These H 2 partial pressure assumptions are slightly conservative, as the H 2 partial pressure should be slightly higher in the lower sections of the reactor, and the permeate H 2 partial pressure lower at the sweep end of the membrane panel. A permeation temperature of 650°C was assumed. Given a nominal permeation area of 0.0258 m2 for each membrane, -14 double-sided membrane panels would be needed to provide sufficient area. There A e f f = - ^ - e x p ( — ^ - P H D k u R T ,0.5 > 9 5 were locations for 12 double-sided and 8 single-sided panels in the ICFBMR pilot reactor, but only 6 double-sided membranes were installed due to budget and time constraints. 4.2.2 Reactor Layout The ICFBMR reactor is tall and narrow, resulting from two main criteria: • Diameter - set by fluidization velocity: As described in the previous chapter, cold model testing indicated that the solids circulation was a near-linear function of the superficial gas velocity in the core up to a value of approximately 0.3 m/s, at which point solids circulation levelled off. The design core velocity for the ICFBMR pilot reactor was set at 0.3 m/s at the maximum design feed rate (50 SLM CH4, 150 SLM H 20) and an operating pressure of 1,300 kPa. The design free core area in the reactor was thus established at 0.0025 m2. The dimensions of the core box was set at ~85 mm square, and the reactor diameter was then fixed at 135 mm to fit the core box comfortably within the reactor shell, leaving sufficient annular space at the corners of the core box to permit solids to pass through the gaps. • Height - set by membrane area: The surface area of each ICFBMR membrane panel was 0.0258 m2, based on the panel geometry. Twelve panels of this area would give an installed reactor membrane area of 0.31 m2, close to the value of 0.35 m2 calculated above in equation 4.1. However, as described in Chapter 2, the effective membrane area of each panel was in significantly less than the geometrical permeation area. Figure 4.3 presents the dimensioned layout of the ICFBMR reactor. The main components included: • A - Reactor shell: The reactor vessel was fabricated from stainless steel. Grade 310 was used for the main 135 mm diameter section, as this component was exposed to the external heaters. All other vessel components were grade 304 stainless steel. • B - Core draft box: A rectangular, open-ended duct housed the membranes. • C - Membrane panels: As indicated earlier, only six panels were installed in the reactor; the balance of the membrane places were filled with stainless steel dummy plates. Each installed membrane panel had separate reactor connections for sweep gas and hydrogen permeate. Sweep gas entered the top of the membrane panel, counter-current to the up-flowing reactor gases, whereas permeated hydrogen left the bottom of the panel. • D - Main feed distributor: Preheated reactor feed was introduced into the reactor through a simple orifice plate distributor in which 49 - 1 mm diameter holes were drilled in a pattern to match the area of the core draft box (see Figure A4.1.1 in Appendix 4.1). 9 6 FBMR SKID ICFBMR SKID ELECTRIC HEATER f l Stream t 2 i 4 5 6 7 8 9 10 11 12 14 15 16 17 18 20 21 22 23 24 30 31 32 33 MOM Ho- 2.04 2.04 7.0a 7.65 5.12 9.I2 9.21 8.21 0.91 6.31 6.06 6.00 8.57 8.57 4.55 3.55 6.66 4.55 4.86 5\4i - 0.50 4.61 266 26S 361 361 Mox Flow kq/h Volumetric Flow mi/h i.3d 6.22 0.007 6.U 0.56 1.39 1.25 2.76 0.14 0.31 6.60 6.6o 2.63 1.47 0.46 0.664 6.66 6.665 5.96 27.62 4.94 6.665 0.20 6.26 6.36 6.36 Molar Flow kmol/h 0.13 0.13 0.39 0.39 0.52 0.52 0.47 0.47 0.05 0.05 0.00 0.00 0.45 0.45 0.23 0.22 D.00 0.27 0.27 0.54 0.28 0.26 11.4 11.4 20.1 20.1 cH4 mol fract 1 0.244 6,244 6.244 6.244 0.244 0.244 0.06b 0.665 0.126 CO 6.043 0.043 O.OBJ C02 6.176 6.175 6.341 H20 1 1 0.756 6.756 6.736 0.756 6.756 6.756 0.486 0.4B6 0.006 1 1 1 0.500 6.056 1 1 1 1 N2 0.790 0.730 1 02 6,210 0.210 H2 0.229 0.229 0.443 0.500 . 0.950 Vapour Fraction 1 0 0.124 6.376 1 1 1 1 1 i i 1 1 1 6 1 0 i - 1 1 0 6 6 6 6 Temperature *C 10 15 15 193 147 180 180 650 180 650 15 650 650 250 40 40 15 15 133 650 40 40 15 30 15 30 Pressure kPa (abs) 150 1350 1350 1350 1350 1350 1350 1300 1350 1300 1350 1300 1300 1300 1300 1300 2000 150 150 150 150 150 150 150 150 150 MW kq/kmol 16.0 16.0 1B.0 18.0 17.3 17.5 17.5 17.5 17.5 17.5 28.9 28.9 19.2 19.2 20.4 18.0 28.0 18,0 18.0 10.0 2.8 18.0 18,0 18.0 18.0 18.0 Density kq/m3 1.03 9.3 999 51.4 18.4 6.57 6.57 2.98 6.57 3.0 16.39 4.87 3.26 5.83 10.29 992 23.7 999 0.82 0.20 0.16 992 999 995 999 995 Viscosity CP 0.011 0.011 1.14 - - 0.013 0.0)3 0.028 0.013 0.028 0.019 0.045 0.030 0.018 0.015 0.65 0.017 1.1 0.013 0.024 0.0087 6.65 1.1 1.1 0.80 THermol Cona. W/m K 0.032 0.033 0.60 - - 0.039 0.039 0.097 0.039 0.097 0.026 0.063 0.116 0.D67 6.661 0.63 0.025 6.60 0.027 0.18 0.1? 0.63 0.60 0.62 0.60 0.62 Heat Copacity kJ/kqX 2.22 2.31 4.05 4.29 3.62 2.20 2.20 2.72 2.20 2.72 1.04 1.13 2.18 1.93 1.66 4.04 1.04 4.05 1.95 3.51 10.2 4.04 4.05 4.04 4.05 4.04 Enthalpy kW -2.66 -2.66 -3i.ia -29.22 -31.8S -28.23 -25.41 -22.73 -2.B2 - 2 . 5 3 0.00 0.00 -21.41 -23.36 -9.66 -16.97 0.00 -21.41 -17.81 -14.98 -0.93 -20.15 -905.6 -902.2 -1591.2 -1585.1 NOTE I: OXYGEN / AIR DISTRIBUTOR REQUIRED NOTE 2: MULTIPOINT THERMOCOUPLES REQUIRED NOTE 3: PRESSURE TAPS REQUIRED NOTE 4: MAIN. CORE riUlDIZATION CAS NOTE 5: ANNULUS GAS NOTE 6: HEATER DUTIES DO NOT INCLUDE HEAT LOSSES INTERNAL CIRCULATING FLUIDIZED BED MEMBRANE REFORMER PROCESS FLOW DIAGRAM 50 SLM CH4 FEED, 100 SLM H2. NO OXIDANT ADDITION 50% SWEEP STEAM DILUTION B-Z0B45-13B5 Figure 4.1: Process flow diagram (external heating) FBMR SKID ICFBMR SKID NOTE * /?> /tf> NOTE 5 Stream 1 i 1 2 i 4 5 6 7 8 9 10 11 12 14 IS 16 17 18 20 21 22 2i 24 30 31 32 33 2.04 2.04 7.6fl 7.06 9.12 9.12 B.21 B.21 6.9 1 6.91 S.fiS S.65 17.23 17.23 12.69 5.14 5.66 4.86 4.Bb b.41 4,61 286 361 361 § 7 S Volumetric Flo* — l.9fl 0.218 6.007 6.41 673 1.3S 1.25 176 0.14 6.31 6.S3 1.76 4.34 2.43 0.89 6.665 0.00 0.005 5.96 27.62 4.94 0.005 0.29 0.29 0.36 0.36 Motor How kmol/h 0.13 0.13 0.39 0.39 O.S2 0.52 0.47 0.47 0.05 0.05 0.30 0.30 0.73 0.73 0.45 0.29 O.OO 0.27 0,27 0.54 0.28 0.26 15.9 15.9 20.1 20.1 CM*1 mol Iroct 1 1 0.244 6.244 6.244 6.244 6.244 6.244 6.06S 6.665 6.065 £6 0.022 6.622 6,635 Cb2 6.147 6.147 0.246 H20 1 1 0.756 6.756 6.756 6.756 0.756 6.756 6.3S3 6.393 6.667 1 1 1 0.500 0.050 1 1 1 1 N2 6.796 0,790 6.323 6.323 0.52S 1 02 6.216 6.216 H2 0.112 0.112 0.1 B2 0.500 0.950 Vopour rraction 1 0 6.387 6.531 1 1 1 1 1 t 1 0 1 0 1 1 1 0 0 U 0 0 Temperature "C 16 i i 15 193 166 186 160 650 180 650 15 650 650 250 40 40 15 15 133 650 40 40 15 30 15 30 Pressure kPo (obs) i5o 1350 1350 1350 1350 1350 1350 1300 1350 1300 1350 1300 1300 1300 1300 1300 2000 150 150 150 150 150 150 150 150 150 MW ko/krrtol 16.0 16.0 18.0 1B.0 17.5 17.5 17.5 17.5 17.5 17.5 28.9 28.9 23.3 23.5 26.9 18.0 28.0 18.0 18.0 10,0 2,8 18.0 18.0 18.0 18.0 16.0 Density 1.03 9.3 999 17.1 12.5 6.57 6.57 2.98 6.57 3.0 16.39 4.87 3.97 7.09 13.60 992 23.7 999 0.B2 0.20 0.16 992 999 995 999 995 CP 0.011 0.011 1.14 _ - 0.013 0.013 0.028 0.013 0.028 0.019 0.045 0.033 0.021 0.018 0.65 0.017 1.1 0.013 0.024 0.0087 0.65 1.1 0.80 1.1 0.80 Thermal Cond. W/m K 6.032 0.033 0.60 - 0.039 0.039 . 0.097 0.639 6.69? 0.026 0.063 6.687 0.051 0.037 6.63 0.025 0.60 0.027 0,18 0.17 0.6J 0.60 0.62 0.60 0.62 Heat Capacity kJ/kg'C i.ii 2.31 4.05 3.65 3.31 2.26 2.20 2.72 2.20 2.72 1.64 i . i i 1.64 1.49 1.20 4.64 1.04 4.05 1.95 3.51 10.2 4.04 4.05 4,04 4.05 4.04 Enthalpy -2.66 -2.66 -28.18 -30.85 -28.23 -25.41 -22.78 -2.82 -2.53 -0.03 1.60 -27.14 -30.13 -12.4a -22.48 0.00 -21.41 -17.81 -14.98 -0.93 -20.15 -1259.1 -1254.3 -1591.2 -1585.1 NOTf 1: OXYGEN / AIR DISTRIBUTOR REQUIRED NOTE 7: MULTIPOINT THERMOCOUPLES REQUIRED NOTE 3: PRESSURE TAPS REQUIRED NOTE 4: MAIN CORE ELUIDIZATION CAS NOTE 5: ANNULUS GAS NOTE 6: HEATEP. DUTIES DO NOT INCLUDE HEAT LOSSES NORAM INTERNAL CIRCULATING FLUIDIZED BED MEMBRANE REFORMER PROCESS FLOW SO SLM CH4 FEED. 100 SLM H2. AIR ADOmON SOX SWEEP STEAM DILUTION B-Z0B45-1396 Figure 4.2: Process flow diagram (direct air addition) N2 sweep^> (multiple) H2 permeate)> (multiple) Reactor Exit G: exit gas filter A: vessel shell settled catalyst bed height E: oxidant distributor C: membrane panels with sweep inlet and permeate outlet B: open ended core draft box H: external reactor heaters E: annular distributor D: main distributor Figure 4.3: ICFBMR pilot reactor (dimensions in mm, not to scale 99 • E - Secondary feed distributor: Preheated annular gas was distributed to the annular quadrants via four Vi" (6 mm) OD stainless steel sintered metal plugs (20 um nominal pore rating, Mort Corporation model 6500-1/4-V/8-20). A photograph of the assembled main and secondary distributors is presented in Figure A4.1.3 in Appendix 4.1. • F - Oxidant feed distributor: Preheated air was added to the fluidized zone above the core box through a VS" (13 mm) OD sintered Hastelloy® C276 tube (50 mm long, 2 um nominal pore rating, Mott Corporation). The oxidant distributor was installed horizontally in the reactor, perpendicular to the membrane panels. Pressure drop data for the three feed distributors are presented in Figure A4.1.2 in Appendix 4.1. • G - Exit fdters: Non-permeate reactor gases left the reactor via two tubular metal fdters made from sintered stainless steel (300 mm long, 38 mm OD, 5 urn nominal pore rating, Mott Corporation). These fdters ensured that only traces of catalyst solids left the reactor. The filter pressure drop for a flow of ambient temperature nitrogen is presented in Figure A4.1.27, Appendix 4.1. • H - Heaters: External heaters were clamped onto the reactor. Initially ceramic fibre heaters were used. However, these failed during the initial pilot plant run, likely due to oil binders in the insulation short-circuiting the heating wires. The original ceramic heaters were then replaced by tubular heaters. The total internal volume of the reactor shell is 0.045 m3. The volume of catalyst required to fill the reactor to the top of the core draft box is approximately 0.021 m3. Detailed mechanical drawings for the external pressure vessel (Figures A4.1.4 to A4.1.6), internal reactor (Figures A4.1.7 to A4.1.9), internal preheater (Figure A4.1.10) and core draft box (Figure A4.1.11) are provided in Appendix 4.1. ICFBMR reactor assembly instructions and installation photographs appear in Appendix 4.2. A formal hazard and operability study (HAZOP) was conducted on the reactor design and pilot plant modifications. The reactor and pilot plant were inspected by the MRT and CHBE safety committees prior to operations. 4.2.3 Membrane Assembly Layout Rather than risk damaging the palladium membranes in the initial commissioning runs, dummy internals were initially installed in the reactor. These internals were made from carbon steel and were dimensionally similar to the membrane assembly presented in Figures 4.5 and 4.6. However, the dummy membrane panels were thicker (9.5 mm) than the actual membrane panels (6.3 mm) and had 100 no inter-channel communication slots. A drawing for the dummy internals is presented in Figure A4.1.26, Appendix 4.1. After commissioning, the carbon steel internals were removed and replaced with the membrane assembly. Of the twelve membrane positions in the ICFBMR pilot reactor, membrane panels occupied six spaces for a total of 0.15 m2 membrane area, based on panel geometry. The balance of the membrane spaces were filled with '/i" (6 mm) thick stainless steel dummy plates. It would have also been possible to install six additional single-sided membranes on two of the reactor walls, bringing the total possible installed membrane area to 0.41 m2, in excess of the 0.35 m2 initially estimated to produce the design pure H 2 flow of 100 SLM. Core box :4 £ • 3 A. Possible membrane positions B. Installed ICFBMR panels Figure 4.4: Location of membranes in ICFBMR reactor (numbered items are installed membranes as described in Chapter 2, "X" and "Y" are solid V*" thick stainless steel plates. Possible single-sided membranes on two core box outside walls are not shown.) Cold modelling (Chapter 3) experiments clearly showed that communication slots between the membrane channels are preferred to solid (non-communicating) membrane plates. The membrane panels in the ICFBMR were assembled with horizontal communication slots between panels, as this maximized the width of the membrane panels. There were three vertical panel rows in the pilot reactor, with the six membranes installed in the first and third row. The middle (second) panel row was filled by a plate of stainless steel with no communication slots. Therefore, channels 1 and 2 101 communicated, and channels 3 and 4 communicated, but the middle plate prevented gas and solid exchange between channels 2 and 3. In retrospect, it would have been preferable to have also had communication gaps in the middle panel. Figure 4.4 presents a schematic of the membrane installation showing the location of the six membrane panels described in Chapter 2. The three membranes and one dummy panel were connected into a row (Figure A4.1.12) by three "H" shaped pieces (Figure A4.1.15 in Appendix 4). This formed six communication gaps in the assembly row, each with dimensions of 57 x 22 mm. The communication gaps accounted for 6.5% of the total row area, slightly less than the 9% fraction in the cold model tests. Figure 4.5 presents a plan view of the reactor internals. The flow channels within the core box were identical (16 mm wide). The minimum gap for solids flow between the core box corner and the vessel wall was approximately 8 mm, more than 60 times greater than the maximum catalyst particle size used in the reactor (125 pm), large enough to prevent catalyst bridging (Grace, 1982). The three rows of vertical membranes were held in place by a series of threaded rods. In order to prevent stainless steel parts from contacting the palladium membrane surfaces, a thin graphite gasket was inserted between the membrane and the stainless steel backing strip required to secure the panels. Detailed drawings of the membrane assembly components are provided in Appendix 4.1 (Figures A4.1.12 to A4.1.15). An end view photograph of the final membrane assembly prior to installation in the reactor is presented in Figure 4.6. Figure 4.5: Plan view of ICFBMR reactor internals 102 Figure 4.6: End view photograph of final membrane assembly 4.2.4 Reactor Instrumentation The reactor instrumentation is summarized in the ICFBMR reactor piping and instrument drawing (Figure A4.1.19 in Appendix 4.1). The two types of instruments used to monitor the ICFBMR reactor were: • Thermocouples: Long, type K, multipoint thermocouples were inserted vertically downward into the reactor from the upper reactor flange in order to measure temperature profiles in the annular region, reactor core and reactor gas space. See Table A4.1.1 in Appendix 4.1 for full thermocouple descriptions. • Pressure transducers: The reactor pressure (gauge) was measured in the vent space of the internal vessel. Differential pressure taps were installed at several points in the reactor. See Table A4.1.2 in Appendix 4.1 for full transducer descriptions. Five sampling points were also placed in the reactor. Three differential pressure transducers were used (Omega PX-750, 0-2.54m WC). The sampling line tips were covered with a sintered stainless steel plug (nominal 20 pm, Mott 6500-1/4-1/8-20) to prevent catalyst entry (see Table A4.1.3 in Appendix 4.1 for locations). 103 Instrument locations within the ICFBMR reactor are summarized in Figure A4.1.20 (Appendix A4.1). Pressures and temperatures of the external pressure vessel were monitored to ensure that the reactor operated within its mechanical design limits. Instrument measurements were logged in the control computer twice per minute. 4.3 Pilot Plant Description Figure 4.7 presents a process schematic of the pilot plant. The main feeds to the system are compressed natural gas, deionized water and, if running ATR, compressed air. Table A4.2.4 in Appendix 4.1 provides a description of the main equipment components. The main process blocks of the pilot plant included: • Natural gas feed: Natural gas was first compressed to approximately 2,000 kPa, then desulfurized at ambient temperature in an adsorbent bed filled with CuO (8 wt%) impregnated into granulated activated carbon (Calgon® Sulfasorb-8, 4x10 mesh). A typical analysis of the natural gas supply for the Vancouver area is shown in Table A4.1.5, Appendix 4.1. As the methane content is high (>95%), the natural gas was assumed to be 100% methane for all process calculations. | Sweep N2^-from cylinders j Oxidation Air/-from cylinders Natural gas^ ( V frnm mains ^ Gas compressor (C-1) City water Desulfurizer (DSU) Deionizer Water "pump" (V302A/B) (DIU) Condenser (E-2) Hydrogen^ to vent / GC Figure 4.7: Pilot plant schematic Off-gas >^ to vent / GC j Condensate^ to drain 1 0 4 Water feed: City water was deionized in a mixed-bed ion exchange system (US Filter, activated carbon bed followed by two mixed resin beds). The water was then pressurized to approximately 2,000 kPa in V302A/B and metered to a preheat interchanger (E-l). Reactor feed: Natural gas and steam were mixed, and then preheated in the external electrical preheater (E-3) and then in an internal electrical preheater (H-l). Reformer: The natural gas and steam reacted on the fluidized catalyst to produce a reformate mixture (H2, CO, C0 2 , CH4, H20). The endothermic reaction heat was supplied by external heaters or by direct air addition to partially oxidize the reactor gases. Membranes: H 2 permeable membranes removed produced hydrogen directly from the fluidized bed reformer. An inert sweep gas (N2 ] Praxair industrial grade, >99.995%) could be used to enhance hydrogen permeation. The combined permeate H 2 flow from the active membranes was measured with a dry gas meter (American Meter, model AM-250) before being vented to atmosphere. The permeate H 2 was analyzed in a gas chromatograph ("GC #2" in Table A2.2 in Appendix 2) by injecting from filled sample bags. • Off-gas treatment: The hot non-permeate reactor gases leaving the reformer were cooled to ambient temperature and condensate was separated before the off-gas was vented. The off-gas could also be sent directly to a gas chromatograph for analysis ("GC #1" in Table A2.2 in Appendix 2). Piping and instrumentation drawings (PIDs) for the pilot plant are presented in Figures A4.1.16 to A4.1.18 in Appendix 4.1. The pilot plant was controlled by a programmable logic controller (PLC, GE Fanuc, series 90-30), which ran both the control loops and the safety shutdown logic. The pilot plant operator had access to the PLC program through a control computer, which ran a custom-built program created using commercial HMI (human machine interface) software (Cimplicity). The five operating screens from the HMI program used to operate the pilot plant are presented in Figures A4.1.21 through A4.1.25 in Appendix 4.1. The main control variables in the pilot plant were: • Natural gas flow (mass flow controller). • Water flow (measured by a turbine flow meter, flow manually adjusted). • Oxidant air flow (mass flow controller). • Preheat temperature (adjusted by heater input). • Reactor temperature (adjusted by heater input or by air addition). • Reactor pressure (automatic control valve). 105 4.4 Pilot Plant Operation Detailed operating instructions for the pilot plant are located in Appendix 4.3. The general procedures to start-up the system are listed below. The shutdown procedure was a reverse of the start-up procedure. • Start the control computer and PLC. • Open H 2 and N 2 gas cylinder supplies. • Pressurize external shells of preheater and ICFBMR reactor with N 2 . • Fluidize the ICFBMR with N 2 at low pressure (< 300 kPa). . • Turn on the process heaters and slowly raise the reactor temperature at a rate of <2°C/minute until the bed reaches 250°C. • Introduce H 2 , then raise slowly raise the reactor temperature to 500°C at a rate of <3°C/minute in a N 2 /H 2 atmosphere. • If required, hold the reactor at conditions suitable for catalyst reduction. • Set a small flow of water, converted to steam in the feed preheaters, followed by natural gas. • Sample the non-permeate reactor off-gas (ROG) until CO present, indicating CFL reformation. • Shut off H 2 , then stabilize operations at the desired temperature, pressure and feed rate. • Measure permeate flow using a stopwatch and dry gas meter. • Measure permeate purity (GC #2). 4.5 Catalyst The total catalyst charge in the ICFBMR pilot reactor was 0.021 m3, which brought the settled bed depth just over the top of core box. The free volume in the core box (the volume not occupied by membranes and dummy panels) was 0.0085 m3. It is in this volume that the bulk of the reforming reaction occurs. The SMR reaction volume (V) increases to 0.0090 m3 if we include the reactor volume immediately above the bottom distributor, but below the core box. Under the flowsheet conditions presented in Figure 4.2, the methane/steam feed has a volumetric flow of 2.8 m3/h. If we assume that a bubble void volume of -20% of the core box volume not available for reaction, an estimate of the SMR space velocity (s) at the ICFBMR flowsheet operating conditions (1,300 kPa, 650°C) can be made: b»>f \ £ ~ £ m f J 2 - S m l h 1 =390/ . " ' (4.2) 0.0090w 3 0.8 106 The relatively low space velocity in the ICFBMR pilot reactor is a result of the need to accommodate the hydrogen membranes in a practical layout for fluidization. Commercial NiO catalysts can be operated at space velocities above 200,000 h"', but conventional fixed-bed SMR reformers operate with significantly lower space velocities due to heat transfer and mass transfer constraints. Given the relatively large volume available for reaction in the ICFBMR pilot reactor, it was decided that it would be acceptable to dilute the active catalyst with alumina to conserve catalyst stock. It should be noted that ICFBMR operates ~300°C colder than the conventional SMR process, and therefore reaction kinetics will be slower. Two types of catalyst were employed separately in the ICFBMR pilot plant: SMR catalyst: A conventional SMR catalyst, Sud Chemie G-91 5/8" x 1/4" EW, was used initially. The SMR catalyst (18% NiO, 1.6% K 2 0, supported on alumina) was designed for a conventional fixed bed reformer, and was in the form of spoked cylindrical rings. The catalyst rings were crushed in a grinder, then lightly fluidized for several days in atmospheric air. The resulting material was then, sieved to obtain a powder in the particle range of 50 to 125 urn. The crushed SMR catalyst was then diluted with an equal volume of alumina powder before being introduced into the pilot reactor. Autothermal reforming (ATR) catalyst: A novel ATR catalyst was prepared by a commercial catalyst supplier on MRT-suppIied alumina substrate. The catalyst supplier indicated that this ATR catalyst formulation had been successfully tested for ATR in the form of a fixed-bed monolith, but had not been previously used in a fluidized bed. By agreement with the catalyst supplier, no chemical analysis could be undertaken on this ATR material. It is likely that the formulation was based on a noble metal, such as rhodium, palladium or platinum. See Ayabe et al. (2003) for a discussion of various noble metal ATR formulations. In order to produce a fluidizable ATR catalyst, a high-quality y-alumina powder (Alcoa HiQ 7S19CC) was supplied to the catalyst manufacturer for impregnation. Its average particle diameter was 87 urn (see Figure 3.2 in Chapter 2 for particle size distribution). Before impregnation, the alumina powder was preconditioned at UBC by lightly fluidizing with air for several days in a Plexiglas column. The alumina was then screened (+38 urn, -250 um). This preconditioning was an attempt to remove any fines that might be initially produced by fluidization, but in fact very few fines were found in the final screening. For the ICFBMR experiments, the ATR catalyst was diluted 9:1 by the same alumina as had been utilized as the catalyst substrate. 107 Particle size distributions and hydrodynamic characteristics for the two catalysts are summarized in Chapter 3. Both catalysts can be considered Geldart type A powders. The activity of the catalyst was tested in a microreactor prior to loading the pilot reactor and after pilot test runs. Appendix A4.4 describes the microreactor system set-up. The microreactor tests did not attempt to obtain intrinsic kinetic data as their purpose was to check catalyst and compare activity of fresh and used catalyst at space velocities significantly higher in of the ICFBMR pilot reactor. 4.6 Pilot Plant Experiments without Membranes A number of experimental reforming runs were conducted in the ICFBMR pilot plant, initially with dummy steel internals installed, then with palladium membranes. The overall objectives of the ICFBMR pilot testing were: • Qualitative evaluation of the ICFBMR reactor concept, including reactor stability, reactor isothermality, catalyst circulation and segregation of nitrogen in the oxidation air from the reactor core. • Quantification of the reactor performance, including hydrogen recovery and demonstration of an equilibrium shift. • Evaluation of the planar palladium membranes in a fluidized bed reformer. The process variables explored in the pilot plant work included: • Reactor pressure and temperature. • Natural gas feed rate. • Steam-to-methane molar feed ratio. • Operation with and without sweep nitrogen. • External heating (SMR) and autothermal operation (ATR) with varying air feed rate. Table 4.1 summarizes the pilot plant experimental program. No membranes were installed in the first run, as its purpose was to commission and prove the pilot system. A carbon steel dummy internal was inserted in the ICFBMR to simulate the membranes, similar to the cold model testing. After the first run, the carbon steel dummy assembly was replaced by membrane assembly, which remained installed for the duration of the experimental program. Not all of the data from the pilot runs are described in this chapter; only results that are believed to give insight into the ICFBMR performance are presented. 108 Table 4.1: Summary of ICFBMR experimental runs Run Dates Catalyst Comments Run time (h) 1 July - Sept, 2003 Diluted NiO Commissioning runs, dummy internals installed (no membranes) 2 Aug 9-12, 2004 As above Reactor heaters replaced Membranes installed 90 3 Sep 20-22, 2004 As above Heluim tracer study SMR & A T R testing 65 4 Oct 26-28, 2004 As above Low catalyst activity experienced 65 5 Dec 13-14, 2004 Used A T R (11%) diluted with alumina Low catalyst activity experienced 24 6 Aug 18-21,2005 Virgin A T R (10%) diluted with previous A T R material SMR & A T R testing Up to 650°C operation Low catalyst activity experienced after -40 hours 90 334 A series of experimental runs was conducted on the ICFBMR pilot plant with dummy internals installed in the reactor. The specific objectives of these runs were: • To prove operation of the pilot plant, including the PLC control system and ICFBMR heaters at reforming temperatures and pressures. • To investigate hydrodynamic characteristics, such as bed and distributor pressure drops. • To perform reforming runs on the reactor to confirm catalyst activity and gas chromatography procedures. The core draft box of the steel dummy matched the dimensions'of draft box used to house the membrane internals in subsequent ICFBMR experiments. However, the dummy internals had three solid (i.e. non-communicating) plates to simulate the membrane surfaces, while the membrane assembly had horizontal communications slots (Figure 4.4). As discussed in Chapter 3, results from cold model experimentation indicated that the non-communicating panel design is prone to maldistribution of solids and gas among core flow channels, leading to flow instability. The reactor was filled with 21 liters of the diluted NiO catalyst powder mixture (50 wt% SMR catalyst, balance alumina). The physical properties and fluidization characteristics of the NiO catalyst mixture are summarized in Table 3.1 in Chapter 3. This batch of catalyst had been previously used in reforming experiments conducted by MRT. As indicated in Table 4.1, this catalyst mixture was used for ICFBMR pilot plant runs #1 through #4. The settled catalyst bed depth was confirmed to be 25 109 mm above the top of the core draft box. The catalyst was reduced by fluidizing for four hours in a 30% H 2 , 70% N 2 mixture at 500°C and 300 kPa. Notable results of the commissioning runs include: • Although the chemical duty of the reforming reaction was modest, the temperature profile within the reactor core was very uniform (less than ~10°C variation). • Overall heat losses from the reactor were found to be -2 kW. • The freeboard of the reactor was about 30-50°C colder than the main bed, resulting in reaction reversal in the freeboard. The gas analysis of the outlet gases therefore tended to show significantly lower methane conversions than expected based on the main reactor operating temperature. This phenomenon was present in all of the experimental runs to some degree. • The pressure drop over the two sintered metal outlet filters was stable and acceptable. As a result, the back-pulsed nitrogen cleaning system was not required. Several components in the ICFBMR system did not perform as well as expected and had to be replaced after the commissioning runs: ICFBMR heaters: After approximately 36 hours of service, the resistance of the four ceramic reactor heaters (Watlow VS108A30S) dropped by -25%. Their resistance continued to decrease to about -50% of their original value. At the end of the commissioning runs, the internal ICFBMR was removed from the pressure shell, and the ceramic heaters were replaced with four new 0.5 kW tubular heaters (Watlow Incoloy RDN134J10S). It is believed that organic binders present in the insulation blanket permeated the ceramic heaters, leading to short-circuiting of the heater wires. The ceramic heaters clamped onto the internal preheater were of design similar to the failed units, but surprisingly suffered no performance deterioration during the ICFBMR pilot program. Plugged distributors: The secondary feed distributor initially installed in the ICFBMR incorporated a ring header with an angled distribution point in each annular quadrant (Figure A4.5.1, Appendix 4.5). The 0.6-mm diameter holes in the distributor partially plugged with catalyst solids, and the assembly was then replaced by the sintered metal distributor presented in Figure A4.1.3 in Appendix 4.1. Similarly, the original orifice air distributor (Figure A4.5.2, Appendix 4.5) also partially plugged with catalyst solids during commissioning runs, and was replaced by the sintered metal distributor described in Section 4.2.2. Although orifice distributors are generally preferred for fluidized beds, the many start-ups and varied gas flows presented opportunities for backsifting, which was prevented by the sintered metal distributors. 110 Thermocouples: Several of the multipoint thermocouples failed in service, likely a result of the bending required by the rather complication installation procedure. Disassembly and re-assembly of the pilot plant reactor was challenging and very time-consuming. As a result, the failed multipoint thermocouples were not replaced, and temperature profile information from the pilot plant tests was incomplete. 4.7 Pilot Plant Experiments with Membranes Although the installation of membranes did not complicate reactor control, the extra flow and gas analyses added extra operational tasks. Additional upgrades to the pilot control system, such as a permeate H 2 mass flow meter and an automatic gas chromatograph sampling system, would have facilitated generation of reliable experimental data. A mass balance closure around the pilot plant was attempted for several atomic species (carbon and nitrogen) during a number of the experimental runs, but was not possible for hydrogen and oxygen, as the condensed water in the reactor off-gas was not measured. The error in the atomic carbon balance (carbon in - carbon out) was generally within ±15%. Although the reactor feed gas flow was known accurately, the reactor off-gas flow was measured manually with a dry gas meter and stopwatch, and there was a potential for leaks from the pressurized system. 4.7.1 Membrane Performance The performance of the H 2 membranes is the key to the ICFBMR system. The three most important membrane performance parameters are: • Purity - Ultrahigh purity is desirable in order to meet fuel cell specifications. • Flux - Sets the H 2 production rate of the system. • Effectiveness - Defined here as the fraction of measured H 2 flux relative to the predicted flux, thus a measure of the excess membrane required above theoretical predictions. Membrane Performance The total cumulative time that the membranes were exposed to hot, fluidized operating conditions in the pilot reactor was approximately 14 days. The membranes cycled from ambient conditions to hot temperatures (>500°C) six times. As presented in Figure 4.8, the combined hydrogen permeate was initially found to be "pure" H 2 (>99.999%, excluding N 2 present from sampling or sweeping) until the cumulative run time was approximately 180 hours, from which point the averaged measured purity of the combined permeate was 99.8% (dry basis). Given the presence of methane, it is likely that there were also traces of steam in the permeate H 2 at concentrations of -0 .2%. The times of the six start-111 ups / shutdowns of the pilot reactor are also shown on Figure 4.8, as are the reactor core operating temperature (right hand axis). Membrane impurities were detected midway through the third test run, just after the reactor core was operated at 575°C. However, a link between operating temperature and H 2 purity cannot be concluded. The maximum core operating temperature was 650°C, near the end of the last run. Table 4.2 summarize permeate testing results for each of the six membranes panels installed in the reactor. No leaks were detected in four membranes during the pilot testing, and the two leaking membranes produced relatively high purity hydrogen throughout the entire pilot test. Table 4.2: Permeate purity (dry basis) from membranes operating in ICFBMR pilot reactor (excluding N 2 , see Figure 4.4 for location of membranes in the reactor) Membrane Location Run time to 1st detected leak (hours) Lowest detected H 2 purity 1 3rd from top (right) No impurities detected >99.999% 3 2nd from top (right) 253 99.72% 4 Top (right) No impurities detected >99.999% 5 3rd from top (left) 173 99.67% 6 2nd from top (left) No impurities detected >99.999% 8 Top (left) No impurities detected >99.999% Membrane Flux Testing In Table 2.4 (Chapter 2), the effective area of each of the membranes was calculated from permeate test data. From this, the total permeation area of the six membranes installed in the pilot reactor was calculated to be 0.060 m2. Individual flux testing of the membranes installed in the pilot reactor was periodically performed at elevated temperature (>500°C) under a pure H 2 atmosphere with no N 2 sweep to the membranes, to ensure that the partial pressure of hydrogen in the system was equal to the absolute pressure. H 2 was fed to the pilot reactor in excess of the membrane flux and above the minimum fluidization velocity. The effective permeation area of each membrane was calculated based on the measured flux and Sievert's law permeation equation (equation 4.1). Table 4.3 summarizes the calculated effective area of each of the six membranes during two such tests, as well as the pre-pilot data from Chapter 2. As previously noted, the expected permeation area of each membrane based on the panel geometry was calculated to be 0.0258 m2 (six panel total =0.155 m2). 112 100.0 99.8 b 99.6 • a CM 99.4 I 99.2 © • Combined H2 permeate (left axis) © Startup / Shutdown A Core temperature (right axis) A A £\ A A A A A .-> • AAA • • /ft 2A A 650 600 ~ o 3 +•> re 550 • E o L . o 500 ° 99.0 50 100 150 200 Run time (hours) 250 300 450 350 Figure 4.8: Purity (dry basis ) of combined hydrogen permeate (excluding N 2 ) and core reactor temperature over the duration of the ICFBMR pilot reactor testing The data from the reactor flux testing generally gave calculated membrane areas slightly higher than the pre-reactor permeation testing. The flux data are somewhat inconsistent, especially for panel #8, but for the most part the relative areas of the membranes are approximately the same within each data set. It is likely that some of the measured variation is real (i.e. adsorbed surface impurities during the pre-pilot permeation tests), but experimental error such as temperature variation and gas leaks in permeation testing equipment probably also contributed to the inconsistency. Table 4.3: Calculated effective permeation area of the membranes installed in the pilot reactor (see Tables A4.6.2-3, Appendix 4.6 for data, 0.0258 m2 is expected permeation area of 1 panel based on membrane geometry) Membrane Average of calculated effective permeation area (m2) Pre-pilot testing (Table 2.4) After run #5 During run #6 1 0.0121 0.0154 0.0142 3 0.0101 0.0124 0.0118 4 0.0075 0.0102 0.0099 5 0.0142 0.0184 0.0165 6 0.0059 0.0075 0.0064 8 0.0101 0.0113 0.0080 Total 0.0599 0.0752 0.0668 Fraction of 1 0.0258 m2 38.7% 48.6% 43.2% Flux Under SMR Conditions The effective membrane area for a number of ICFBMR operating runs is presented in Table A4.6.4, Sheet3 (Appendix 4.6). The effective membrane area was calculated based on equation 4.1 and the reactor operating variables for each operating point. The variables needed to calculate the effective membrane area were: • Temperature - The average reactor core temperature was used. • H2 partial pressure in the permeate: o Assumed to be 101 kPa for all non-sweep operating points. o For the N2-sweep cases, the H 2 partial pressure was assumed to be 101 kPa multiplied by the measured H 2 mole fraction of the combined permeate stream, which was typically in the range of 35 to 50%. • H 2 partial pressure in the reactor: o For SMR operation (no air addition), the H 2 partial pressure was estimated by calculating the equilibrium reactor gas composition from the reactor feeds, temperature and pressure 114 at two points within the reactor (inlet and exit, after H 2 permeation). The permeate H 2 partial pressure used for permeation was the average of the two values, o For ATR operation (air addition to top of reactor), the reactor equilibrium was calculated as in the SMR case, except that 0 2 was excluded (as it is consumed in the upper reactor, away from the membranes) and only 10% of the N 2 in the oxidation air was assumed to reach the core permeation zone, based on results from the cold model work. The balance of the N 2 in the oxidation air was assumed to not circulate to the reactor core, but rather to leave the reactor directly. The average permeation area calculated from the data listed in Table A4.6.4, sheet 3 (Appendix 4.6) for the case of no sweep gas is 0.075 m2, similar to the areas calculated in Table 4.3 based on the pilot flux tests. This implies that the membrane "efficiency" was close to 100% in the pilot reactor for the no-sweep operating points. The average calculated permeation area calculated for cases with sweep nitrogen was lower (0.059 m2). This is likely due to the assumed H 2 permeate partial pressure being less than under actual operating conditions. As the H 2 flux on each panel varied somewhat, the H 2 permeate partial pressure on any individual membrane would not necessarily be equal to the measured combined H 2 permeate concentration. In addition, the gas mass transfer resistance on the permeate side of the membrane means that the H 2 partial pressure will be higher than the permeate gas leaving the membrane panel. Clearly the N 2 membrane sweep was less than fully effective. 4.7.2 SMR Operation (External Heating Only) The highest membrane flux was expected for SMR operation at the highest reactor operating temperature and pressure, with sweep gas fed to the membranes. However, this turned out not to be the case, as the catalyst activity at the highest core operating temperature (650°C) was low. The maximum recorded H 2 production rate was 17.6 SLM (1.06 Nm3/h) at 589°C and 1,209 kPa using nitrogen sweep to the membranes. Table A4.6.4, Appendix 4.6, summarizes the measured hydrogen flux different operating points. As discussed above, the measured exit gas composition of the reactor is not a true measure of the gas composition in the reactor core as the temperature drop in the freeboard reverses the SMR reaction. The only practical way of estimating the theoretical membrane flux is to assume that the non-permeate reactor gases are at equilibrium at the reactor temperature and pressure. However, in a significant number of the data points in Table A4.6.4, the measured methane conversion in the reactor 115 off-gas exceeded the equilibrium methane conversion with no H 2 removal. This is evidence of an equilibrium shift caused by in-situ H 2 removal. Experimental data showing the effect of the main process operating variables is now presented. In some cases the pilot plant results are somewhat sparse, as it proved difficult to operate the pilot single-handedly. However, in general the reactor performance followed trends suggested by equilibrium reactor modelling. Effect o f Temperature Increasing temperature increases the H 2 flux by shifting the SMR equilibrium towards products and by increasing the permeance of the palladium foil. Figure 4.9 presents selected H 2 recovery data as a function of reactor temperature from run #3, with and without sweep N 2 to the membranes. The enhancement of H 2 recovery by sweep N 2 is evident. In addition to experimental error, some of the data scatter is due to the fact that other operating variables (such as reactor pressure, steam-to-carbon ratio and permeate H 2 partial pressure) could not be controlled to be identical for each operating point. Nevertheless, both sets of experimental data reasonably follow the predicted trend lines. 1.2 0.4 -I 1 , , , , 1 550 560 570 580 590 600 610 Temperature (C) Figure 4.9: Permeate hydrogen recovery relative to gas feed rate as function of reactor temperature and N 2 sweep (run #3, 15 SLM natural gas, reactor pressure ranges from 1,090 to 1,275 kPa, S:C = 2.1 to 4.0, predicted lines are 15 SLM CR,, 2.4 SLM N 2 , S:C = 3.0, 1,200 kPa, permeate pressure = 101 and 50 kPa for sweep / no-sweep respectively, area = 0.060 m2) Figure 4.10 presents another set of H 2 recovery data as a function of temperature at slightly higher temperatures. Again, hydrogen recovery increased with temperature, though it should be cautioned that the catalyst activity for these operating conditions was low. The enhancement of the H 2 recovery with sweep N 2 is again shown. Figure A4.6.1 (Appendix 4.6) presents the measured and predicted methane conversion for these operating conditions. 116 1.0 > o u <D I o C N X 0.8 0.6 0.4 0.2 0.0 —I— model for A = 0.06 m2, 50 kPa permeate —X— model for A = 0.06 m2, no sweep A no sweep - measured • sweep - measured • • Q • g A n A A 560 580 640 660 600 620 Temperature (°C) Figure 4.10: Permeate hydrogen recovery relative to natural gas feed rate as function of reactor temperature and sweep gas (1,060 kPa, steam-to-carbon = 3.0, run #6) Effect of Steam-to-Carbon Ratio The steam-to-carbon feed ratio was varied in some experimental runs. As shown in Figure 4.11, the predicted and measured effect of this variable on H 2 production is modest. It should be noted that the catalyst activity was again low during these test runs. Although the effect on the H 2 flux is modest, lower steam-to-carbon ratios tend to improve the thermal efficiency of a SMR system. Figure A4.6.2 (Appendix 4.6) presents the corresponding measured and predicted methane conversions. 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 > o o k. X O cvi I model for A = 0.06 m2, 50 kPa permeate model for A = 0.06 m2, no sweep • sweep A no sweep 2.0 2.5 3.0 3.5 4.0 Steam-to-carbon molar ratio Figure 4.11: Permeate hydrogen recovery relative to natural gas feed rate as function of steam-to-carbon molar ratio and sweep gas (1,060 kPa, 570°C, run #6) 117 Effect of Feed Rate At very high hydrogen recoveries (> 2V2 moles H 2 per mole of natural gas feed) the H 2 partial pressure in a plug flow membrane reactor drops significantly at the reactor exit, decreasing the flux through the membranes located near the reactor exit (see modelling output in Chapter 6) . However, at low hydrogen recoveries (less than ~IV2 moles H 2 per mole of natural gas feed), the H 2 partial pressure does not decrease significantly with in-situ H 2 removal. This is the situation in the pilot plant, as the hydrogen flux was found to only decrease marginally at lower feed rates. Figure 4.12 presents all of the Table A4.6.4 hydrogen recovery data as a function of the natural gas feed rate. There is variation in the data, as there was no attempt to match the operating conditions (e.g. temperature, pressure, sweep gas), but the trend is clearly to higher H 2 recovery at lower natural gas feed rates. This figure also highlights that the pilot reactor performance was flux limited. The installed membrane area governs the H 2 production rate; varying operating conditions only modestly affects the membrane flux. Figure A4.6.3 (Appendix 4.6) presents the same data, but expressed as the absolute permeate H 2 flux, showing that this was almost independent of the natural gas feed rate. 1.4 T . 1.2 H 0.0 "I 1 1 1 , , r — ~ \ 10 15 20 25 . 30 35 40 45 Natural gas feed rate (SLM) Figure 4.12: Hydrogen recovery as a function of natural gas feed rate for all data in Table A4.6.4 The highest measured hydrogen recovery was 1.17 Nm3/Nm3 of feed natural gas (data set #20, Table A4.6.3, Appendix 4.6). The feed flow of natural gas was 15 SLM for this operating point. Higher recoveries were anticipated for the high temperature operating points in run #6, but poor catalyst activity hindered reactor performance. 1 1 8 4 . 7 . 3 A T R Operation (Air Addition) A series of ATR experiments were performed to determine the effect of the rate of air addition on hydrogen production. The reactor was first stabilized under SMR conditions at 1,100 kPa and a core temperature at ~560°C. The natural gas feed was set at 20 SLM and the steam-to-carbon ratio at 3.0 for all test points. Air was then introduced in increments up to a maximum feed oxygen-to-methane molar ratio (02:CH}) of 0.6. A constant flow of sweep nitrogen was maintained to the membrane panels. No methane or carbon monoxide impurities were detected in the permeate testing. A summary of the operating conditions and experimental results are presented in Table A4.6.6, Appendix 4.6. Figure 4.13 presents the H 2 partial pressure in the reactor predicted by the Gibbs reactor equilibrium model if the system were well mixed for operating conditions #1-8 in Table A4.6.6, and account for the measured permeate H 2 removal. As the air addition rate increased, the H 2 partial pressure in the reactor decreased, reducing the membrane permeation driving force. 250 j <D 3 225 n (A tn 0) l _ 200 -Q. re 0. 175 -ro o o. -o o 150 -cto E re o 125 -i _ CN X 100 --0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 2 : C H 4 Molar Ratio Figure 4.13: H 2 partial pressure in a well-mixed reactor for ATR experimental conditions predicted by equilibrium model (560°C, 1100 kPa, NG feed 20.1 SLM, S:C = 3.0, sweep N 2 to membranes) Figure 4.14 presents the measured hydrogen flow as a function of the air addition rate. The measured flow of permeate H 2 was not significantly affected by air addition. This insensitivity of H 2 flux to air addition is a key advantage of the ICFBMR configuration. Also plotted on Figure 4.14 is the predicted H 2 flux for a well-mixed system, based on the H 2 partial pressures from Figure 4.13, normalized to the measured H 2 flux with no air addition. If the entire system were well mixed, the lower H 2 concentration in the reactor would reduce the membrane flux by almost 40%. However, the H 2 permeate flow did not change significantly in the ICFBMR pilot testing, confirming that only a small portion of the nitrogen in the oxidation air circulated to the reactor core. 119 w ra E CD Q. 5 o H — CM X 12 10 8 6 4 2 0 A H2 flux (measured) A Predicted H2 flux (normalized) 0.0 0.1 0.2 0.5 0.6 0.7 0.3 0.4 02.CH4 Molar Ratio Figure 4.14: Effect of air addition on Ff2 flux (measured and predicted for well-mixed reactor, 560°C, 1100 kPa, NG feed 20.1 SLM, steam-to-carbon = 3.0, sweep N 2 to membranes) As the oxidant air flow increased, the reactor heaters were reduced from an initial power of 4.2 kW (SMR, 02:CH4 = 0) to 0 kW at the highest air addition rate (0 2:CH 4 = 0.6). Simulations presented in Chapter 7 show that an 02:CH4 molar ratio in the range of 0.35 to 0.4 should be sufficient for the reactor to run adiabatically. However, the pilot reactor has significant heat losses, so that adiabatic 02:CH4 ratio would be even higher than 0.6 if commercially viable H 2 recovery ratios could be achieved in the pilot plant. In any event, the testing did show that the reactor duty, although modest, could be fully sustained by air addition. Under SMR conditions, the temperature profile within the reactor core was flat, with a temperature variation of approximately 10°C. This increased to ~20°C at the highest air addition rate (see Table A4.6.7, Appendix 4.6 for temperature profiles at different air addition rates). Figure 4.15 presents additional test data that again indicate that the hydrogen flux is relatively insensitive to the air addition rate. It should be noted that it likely that the catalyst activity was somewhat low during these tests. Nevertheless, the H 2 flux was not reduced significantly with air addition. Also shown is the significant increase in hydrogen flux when sweep N 2 is used. 120 14 12 — 1 0 s W 8 X £ 6 CN 1 4 • sweep • no sweep 0 0.1 0.4 0.5 0.2 0.3 02.CH4 Molar ratio Figure 4.15: Permeate hydrogen flow (with and without sweep gas) with varying air addition rate (645°C, 1060 kPa, N G feed = 20 S L M , steam-to-carbon = 3, A T R catalyst) 4.7.4 Solids Ci rcu la t ion Tests Similar to the cold model work presented in Chapter 3, a series of helium tracer experiments was conducted to investigate internal movement of gas and solids within the reactor. Helium was metered to the reactor through the main oxidant distributor with a rotameter (Figure A4.5.3, Appendix A4.5). Gas samples were taken from various points in the reactor, then analyzed for their helium concentration in a gas chromatograph. Figure 4.16 presents the reactor schematic and sample point locations for the tracer tests. Reactor gas was directly sampled from the annulus, core and freeboard (Points D, E, H , I in Figure 4.16) and the reactor outlet. During these tests, the reformer was operated under SMR conditions (no air addition) at ~595°C (core), 1,100 kPa with a natural gas feed rate of 42 S L M and a steam-to-carbon molar ratio of 3.0. The permeate hydrogen flow (with sweep nitrogen) was measured to be between 16 and 17 S L M (as 100% H 2 ) during the helium trials. Figure 4.17 presents the measured helium concentration and methane conversion at various points within the reactor (see Table A4.6.1 for data). The measured helium concentrations provide strong evidence of positive solids circulation. Several points are notable: • The helium concentration in the annulus at mid-height was approximately half the reactor exit concentration. At the top of annulus, the gas composition would be expected to be similar to the reactor exit. The fact that the annular helium concentration was lower may result from uneven mixing of the helium in the upper reactor, or modest stripping due to the secondary feed gas. 121 Oxidant feed 1,380 kPag He Sample "D" — in slot 2 Sample "I" — \ in annulus Sample "E" —-in slot 2 Reactor off-gas (to G C / vent) Secondary feed (4) Oxidant distributor Solid divider Solid dummy plate (1 of 2) — Sample "H" in slot 3 " Membrane (1 of 6) — Core draft box -tXh To GC Permeate H2 Main feed Figure 4.16: ICFBMR helium tracer test schematic The helium concentration in all of the core samples was 5 to 7% of the helium content at the reactor exit. The fact that the helium concentration was similar in the top and bottom of the core strongly suggests that the great majority of helium enters the core from the solids circulated from the bottom of the annulus, rather than due to axial dispersion from the top of the reactor. The helium tracer testing in the cold model (Section 3.5.4) did find slightly higher helium concentrations in the top of the core box, likely due to the cold model sample points being located 122 very close to the top of the core box (90% of core height), whereas the pilot reactor sample points were located somewhat lower (75% of core height). Non-permeate off- Annulus (I) Slot 2 bottom (E) Slot 2 top (D) Slot 3 top (H) gas Figure 4.17: Measured helium concentration (black bars) and methane conversion (open bars) from ICFBMR He tracer tests (see Table A4.6.1, Appendix 4.6 for data) The helium concentrations provide strong evidence of the intended up-the-core, down-the-annulus solids circulation pattern. If the reactor were running under ATR conditions, the helium concentration profile suggests that only 5 to 7% of the nitrogen in the oxidation air would be circulated to the reactor core. This corresponds to findings in the cold model helium tracer tests, where approximately 10% of the helium fed through the oxidant distributor was found to circulate to the reactor core (see Section 3.5.4). The core helium concentration was an average of 12% of the measured mid-height annular concentration. From this, an estimate of the solids circulation rate can be made using the same equations as in the cold modelling chapter (Section Assuming that the entire reactor feed flows up the reactor core (42 SLM natural gas, 126 SLM steam), the following estimates of various reactor hydrodynamic parameters were obtained (see Table A4.6.5, Appendix 4.6 for calculation details): Core superficial velocity (Ucore) = 0.18 m/s Solids circulation rate (mp core)= 0.22 kg/s Solids flux up core (Gpcore) = 45 kg/m2s Downward velocity of solids in annulus = 0.027 m/s 123 • Predictive scaling correlations were developed for the solids flux (positive when flux upwards) for both horizontally and vertically slotted panels in the cold model testing presented in Chapter 3 (Figure 3.20): Vertically slotted panels Gpcore = -1.03E-4(t/core ppf + 0.180(£/ c o r e pp) -11.3 (5.1) Horizontally slotted panels Gpcore = -4.14E-5(<7core ppf + 0.083(<7corepp) +0.65 (5.2) From the calculated value of Ucart, (0.18 m/s) and the catalyst particle density (2,150 kg/m3), the predicted solids fluxes (Gpcore) are 42 and 28 kg/m2s for vertically and horizontally slotted panels respectively, close to the estimate obtained from the experimental data (45 kg/m2s). The pilot plant membrane assembly most closely matches the horizontally slotted panels in the cold model. Although this is but one data point, the result suggests that the cold modelling results can be used for hydrodynamic evaluations of the hot, pressurized system. • The measured methane conversion from the in-bed samples was significantly higher than measured in the reactor off-gas. As discussed above, the reactor freeboard operates cooler than the main reactor due to heat losses, causing a reverse shift in the equilibrium. Gas analyses are presented in Table A4.6.1 (Appendix 4.6). The highest measured conversion was in the annulus, adjacent to the heated reactor wall, as this was the hottest part of the reactor. Table 4.4 presents the calculated equilibrium temperature needed to match the measured methane conversion at each point, given the species composition measured by the gas chromatograph. Except for the non-permeate reactor off-gas sample, the predicted equilibrium temperatures are reasonably close to what was measured in the reactor. Table 4.4: Local measured and corresponding equilibrium temperatures Sample location Temperature (°C) Measured Equilibrium predicted from gas analysis Reactor off-gas (freeboard 579 597 Point I (annulus) 628 621 Point E (bottom slot 2) 593 589 Point D (top slot 2) 595 597 Point H (top slot 3) 590 583 4.8 Catalyst Testing The catalyst used in the reformer was also tested in a microreactor to its confirm activity. The test conditions were not in the kinetically controlled region; rather the test conditions were chosen to 124 permit comparison to previous catalyst testing performed by MRT. The tests allowed the relative activity of the catalyst materials before and after use in the pilot reformer to be ascertained. 4.8.1 S M R Catalyst Test ing As indicated earlier, the NiO on alumina catalyst is designed for SMR conditions (no oxidant addition), and it was not unexpected that the catalyst deactivated after prolonged exposure to ATR conditions. Figure 4.18 summarizes catalyst microreactor testing under SMR conditions of fresh catalyst before ICFBMR service, and the SMR catalyst removed from the ICFBMR reactor after the first and third runs (see Table A4.7.1, Appendix 4.7 for data). The fresh and first-run catalysts show near-equilibrium conversion. The reason for the initial conversion being slightly above equilibrium and the slow decline in conversion of the fresh catalyst is not known, but is likely attributable to small variations in the microreactor temperature. The catalyst performance after the last SMR catalyst run (run #4 in Table 4.1) clearly shows declining activity, confirming pilot plant observations. Some residual activity exists, but clearly the catalyst is no longer useable. 100 -, 1 0 4- , , r ; 1 1 0 25 50 75 100 125 150 Time on stream (h) Figure 4.18: CFLt conversion in microreactor with NiO catalyst (0.3 g of 50% NiO catalyst and 50% alumina, 550°C and 201 kPa, CFL, feed = 30 Nml/min, H 2 feed = 30 Nml/min, steam-to-carbon ratio = 3.5, space velocity at inlet conditions = 99,400 h"1) SMR Catalyst Characterization with T G A and XRD Analysis When operating the ICFBMR in ATR mode, the catalyst is cycled through oxidizing and reducing zones. Although the fluidized bed dissipates heat much more quickly than fixed beds, the oxidation 125 zone still exposes the catalyst to elevated temperatures near the air distributor for brief periods of time. The causes for deactivation of NiO in ATR service have been attributed to nickel oxidation, carbon formation, sintering and spinel formation (Tomishige et al., 2004). In an attempt to determine the mode of deactivation for the SMR catalyst, thermo-gravimetric analysis (TGA) and X-ray diffraction (XRD) analyses of the spent catalyst were undertaken. TGA analyses of the SMR catalyst are presented in Appendix 4.8. Catalyst samples (~18 mg) were heated in an air/nitrogen mixture from ambient temperature to 800°C, with the sample mass continually measured. This technique detects sample loss or gain due to oxidation or volatilization. The fresh catalyst lost almost 4% of its weight, perhaps due to moisture and residual volatiles. The spent catalyst samples showed no significant weight change, indicating that there were no large deposits of carbon. However, it should be noted that the spent catalyst had a grey-black appearance, and small traces of carbon may have been present, below the detection limit of the TGA system used (< -0.25%). XRD analyses of several SMR catalyst samples are presented in Appendix 4.9. There were no significant differences in the spectra from fresh and spent catalyst, suggesting that spinel formation was not significant. 4.8.2 A T R Catalyst Testing Two ICFBMR runs were carried out with the novel ATR catalyst from a large commercial supplier. The first run used a mix of previously-used ATR diluted with alumina. The second ATR run used a mix of virgin ATR blended with bed material from a previous run. Low catalytic activity was experienced in both runs, resulting in disappointing reactor performance. The activity of the virgin ATR catalyst (diluted 50% with alumina) was tested in a microreactor under ATR and SMR conditions (Figure 4.19, data in Table A4.7.2 in Appendix 4.7). The space velocity of the catalyst was 12,000 h"1 under SMR conditions. The methane and steam flow to the reactor were kept constant during the various operating modes. Air was used in the ATR mode. Alternating SMR and ATR operating conditions were also tested by cycling the air on for one minute, then off for one minute over a 45-minute period (see "02 cycling" in Figure 4.19). The catalyst activity was relatively stable over the 100-hour test run. There is some variation in the data, especially the initial "ATR" run, likely attributable to deviations of the microreactor temperature from 550°C. As the space velocities in these tests were relatively low, the measured 126 methane conversions were close to equilibrium values (see Table 4.5), except for the last "ATR" run, where the measured methane conversion was slightly lower. 80 -, - I I - B - 2 . S M R With H2 O ! A 3.SMR Without H2 j X 4.02 Cycling I A 5.ATR 20 -I , : , : ; , 1 0 30 60 90 120 Run time (hours) Figure 4.19: CFL, conversion in microreactor with virgin ATR catalyst (mix of 0.3 g ATR catalyst and 0.3 g alumina at 550°C and 200 kPa, space velocity = 12,000 h"1 under "SMR" conditions, steam-to-carbon = 3.5 and CH 4 feed = 60 Nml/min for all tests, "ATR" used air addition with 02:CFL, molar feed ratio = 0.36, all tests except "SMR without H 2 " used H 2 :CH 4 molar feed ratio =1.0) Table 4.5: Equilibrium methane conversions for Figure 4.19 (equilibrium conversions are calculated at average measured bed temperature) Test Description Equilibrium CH 4 conversion Measured C T4 conversion Average Std. Deviation ATR 57.6% 58.6% 3.8 SMR with H 2 41.1% 42.2% 0.96 SMR without H 2 53.1% 53.9% 0.11 0 2 cycling (same as data #2) 41.1% 39.8% 0.75 ATR (similar conditions as data #1) 55.1% 48.7% 0.29 As discussed above, the catalytic properties of the ATR bed materials in the ICFBMR pilot plant were poor. This was confirmed in post-run microreactor testing of the bed materials. Both the used and virgin ATR catalyst mixtures (1st and 2nd samples respectively in Figure 4.20) showed extremely low reforming activity. The methane conversion measured during ATR operation is due to reaction of methane with oxygen. The cause of the loss in catalytic activity is not known, and as the material was subject to a non-analysis agreement, further testing of the used ATR catalyst was not performed. The 127 catalyst supplier has indicated that this catalyst formulation has been proven to be stable in the form of an impregnated monolith under both SMR and ATR conditions. 50 40 '£ 30 > C o o T t X o 20 H 10 • 1st sample - ATR - A - 1st sample - SMR - • - 2 n d sample - ATR -B-2nd sample - SMR 10 20 Run time (hours) 30 Figure 4.20: Microreactor testing of diluted ATR catalyst after use in ICFBMR pilot plant ("1st sample" is used ATR mixture, "2nd sample" is virgin ATR mixture, 1.65 g of catalyst mixture at 550°C and 200 kPa, space velocity = 12,000 h'1 under "SMR" conditions, steamxarbon = 3.5, CFLt feed = 30 Nml/min, H 2 :CH 4 molar feed ratio = 1.0 for all tests, "ATR" used air addition with 02:CH4 molar feed ratio = 0.5) The composition of the ATR is not known, but it is possible that it contained rhodium (Rh), as this is the most common precious metal reforming catalyst described in the literature. Instability of Rh-catalysts in membrane (Kurungot and Yamaguchi, 2004) and non-membrane reformers (Watson and Daly) have been described. Sintering of the dispersed Rh particles due to thermal instability of the catalyst substrate (alumina, ceria and others) is main cause attributed to the Rh deactivation. Perhaps the y-alumina provided to the ATR catalyst supplier for impregnation was not stable at the elevated temperatures present in the oxidation zone, but we are unable to draw definitive conclusions without analysis of the material. In addition to this instability, Beurden et al. also identify the high cost of Rh-based catalysts as a barrier to commercialization of membrane reformers. Perhaps a more economic alternative to Rh can be developed, such as bimetallic catalysts based on Ni modified with small amounts of Pd, Pt or Rh (Mukainakano, 2007). 4.9 Discussion The membrane area installed in the reactor limited reactor performance, and only a fraction of the nominal H 2 design capacity was achieved. However, the H 2 production could be reasonably predicted using a simple equilibrium model coupled with the membrane permeation equation, so it is likely that 128 the design production rate could be achieved if sufficient membrane area were to be installed. The effect of the main process variables (temperature, pressure, steam-to-carbon ratio, N 2 sweep) generally followed model predictions. The permeate H 2 production was relatively insensitive to the natural gas feed rate, as the flux was set by the limited membrane area in the reactor. As a direct result, the highest hydrogen recovery was achieved at the lowest natural gas feed rates. This is characteristic of membrane reactors, and highlights how the reactor performance is overwhelmingly governed by permeation. Only through increasing the permeation capacity, either by increasing area or decreasing foil thickness, can H 2 production be significantly increased. The H 2 flux increased significantly when a N 2 membrane sweep was used as it decreased the H 2 partial pressure on the backside of the membrane. Use of an inert gas such as nitrogen for membrane sweep is not practical for a commercial unit, as an impure hydrogen product would be produced. Use of sweep steam, as was originally intended for the pilot tests, would be possible as it can be condensed from the permeate stream. Residual moisture in the H 2 product is acceptable for some applications such as fuel cells. However, there is a significant energy penalty with use of sweep steam, and the overall thermal efficiency of the system would suffer. In addition, the water from which the sweep steam is generated would have to be of extremely high quality, as trace impurities would end up in the product hydrogen. Another method of decreasing the H 2 partial pressure in the permeate stream is to lower the absolute pressure with a vacuum pump. This has been used successfully in pilot tests at both MRT and Tokyo Gas (Klassen, 2005). It theoretically introduces no impurities into the permeate hydrogen. There are obvious safety issues with vacuum operation, as there is an inherent risk of air ingress into the permeate line, posing an explosion risk. This was of special concern here, as the permeate stream would have to pass through a vacuum pump, a potential source of ignition. Air ingress would also affect product quality. The permeability of the membranes did not decrease during the pilot testing. Another important finding is that the measured permeation fluxes during the pilot reforming tests were reasonably close to the predicted value, indicating that "effectiveness" of the membrane was close to 100% and suggesting that the mass transfer resistance within the fluidized bed was low. Value of less than. 100% may be due to: • Mass transfer resistance of hydrogen on the reactor side of the membrane. 129 • If using an inert sweep gas, mass transfer resistance on the permeate side. • Pressure drop in the permeate system, so that the absolute pressure on the backside of the membrane is higher than the measured pressure at the outlet. • Fouling, sintering or other degradation of the palladium foil. • Low catalyst activity. The calculated membrane effectiveness was close to 100% in the pilot tests, higher than compared to the work of Adris (1994). Perhaps the flow channels in the membrane assemble contributes to near plug-flow conditions, reducing axial dispersion and enhancing bed-to-membrane mass transfer. It is difficult to project whether the membrane effectiveness would decrease if the membrane flux and FL. recovery were increased to economically practical values. Further experimentation is needed. Theoretically, dense Pd membranes produce ultra-high purity H2. Given the tight feed specifications of fuel cell manufacturers, this is a very attractive feature of membrane reactors. Although the H 2 quality decreased to -99.8% midway through the pilot tests, high-purity product was initially produced, and there were no catastrophic membrane failures, even when the temperature was increased to 650°C. This shows the potential of the uncovered planar membrane to withstand the fluidized bed environment, but more work is needed to show longevity over months of operation and to establish a practical upper operating temperature. Low catalyst activity resulted in poor results for some experimental runs. The deactivation of the SMR (NiO on alumina) catalyst was not unexpected, as this catalyst is not designed for autothermal reforming. However, the poor performance of the ATR catalyst was a surprise, and limited the extent of autothermal reactor testing. No conclusive reasons for catalyst deactivation were found. More discussions with the ATR catalyst supplier and further testing are needed to better understand the catalyst performance. One advantage of fluidized beds is that catalyst can be replaced on-line, but this was not practical with our pilot reactor configuration. Finally, note that the while the mechanical design of pilot reactor allowed economical carbon steel to be used for the pressure shell, the complicated geometry was impractical for research purposes. A simpler, smaller reactor using higher-quality materials may have been more expensive, but would have greatly helped experimentation. Other areas of the reactor design that could be improved include: • Reactor connections - There were a very large number of connections to the reactor, especially on the top flange, primarily for membranes and instrumentation. A change in the membrane 130 design, for example increasing the membrane panel height or eliminating membrane sweep, would reduce the number of connections and simplify reactor assembly. Membrane access - The core assembly was housed in a rectangular core box, which hung from the top reactor flange. Assembly of the core internals was very time consuming. A mechanical design that would allow improved access to the membranes would be desirable. An alternative reactor geometry is proposed in Figure 4.21. It uses the chord layout proposed in Figure 3.22 (Chapter 3), in which the catalyst solids return to the bottom of the reactor via two chord segments, rather than via an annulus. Membranes could then be access through flanges on the side of the cylindrical reactor. Another benefit is that the membranes would then occupy a greater fraction of the reactor volume, reducing the overall size of the reformer vessel. Oxidation air Reactor off-gas Top of core divider Membranes NG / steam feed Catalyst returns (2) Membrane flanges Bottom of core divider Figure 4.21: Alternate ICFBMR reactor layout (elevation and plan view) 131 4.10 Conclusions Although operation of the pilot plant proved to be challenging due the scale of the equipment, and lack of automation, sufficient results were obtained to validate the main features of the ICFBMR configuration. Conclusions from the pilot testing include: • The uncovered membrane panels withstood ~14 days of hot operation and six temperature cycles with no catastrophic failures. The permeate hydrogen was initially very pure (>99.999%, excluding N2), but dropped to 99.8% (dry basis) after approximately 180 hours of testing. Four of six membrane panels did not exhibit any measurable leakage during the pilot tests. Membrane longevity in hot fluidized bed environments requires further study. • The permeability of the membranes was not significantly affected by pilot testing. • The membrane "effectiveness" was close to 100% in the reforming experiments when no nitrogen sweep was used. The effectiveness of the N 2 sweep was slightly lower than predicted by modelling. • Solids were shown to circulate within the reactor at a rate close to that predicted by cold model scaling correlations. • Air addition had only a modest impact on membrane flux, confirming cold model testing that showed that only a portion of the N 2 in the oxidation air circulated to the reactor core. • A positive shift in the equilibrium due to in-situ H 2 removal was demonstrated. 132 Chapter 5. Post-Run Membrane Testing The ICFBMR pilot reactor was disassembled after the experimental runs described in the previous chapter. The six membranes were removed from the internal reactor assembly, inspected for damage and tested for leaks. Two of the membrane panels were placed in a permeation vessel and flux tested. The results showed that the membrane flux was slightly higher than measured before those membranes were installed in the pilot reactor. Permeate purity testing on these membranes indicated small leaks, though measured hydrogen separation factors were still very high. The membrane foil of one membrane was removed from its stainless steel substrate. SEM-EDX analyses indicated significant scaling on the backside of the membrane foil, likely the cause of the lower-than-expected permeation flux of the ICFBMR panels. 5.1 Membrane Removal and Observations After the pilot tests, the core draft box containing the membrane assembly was removed from the pilot reactor together with the top reactor flange. Figure 5.1 presents a photograph of the ICFBMR internals being removed from the reactor. Note the grey catalyst scale on all surfaces. A large lump of catalyst was found adhering to the reactor outlet filters, but this did not totally block the filter surfaces, and was easy to remove. The outlet filter pressure drop was acceptable during all pilot plant runs (<100 kPa). No other significant lumps of catalyst were observed on any other internal surfaces. A l l installed parts were still firmly attached, and no degradation of any components could be seen. Threaded nuts could be unbolted normally. The pass-through connections on the top flange (i.e. thermocouples, pressure taps, sweep gas lines, hydrogen permeate) were removed or cut, and the core box was disconnected from the top flange. Next the membrane assembly was removed from the core box by first disconnecting the rods securing the assembly to the box, then by pushing the assembly out of the top of the core box. The assembly was firmly stuck in the core box, and was removed by knocking the bottom of the assembly with a heavy hammer. 133 Figure 5.1: ICFBMR internals being removed from horizontal pilot plant reactor The internal membrane assembly held up well, with all clips, wires, gaskets and nuts still firmly in place. All metal surfaces were covered with a uniform grey scale, with the exception of half of one side of membrane panel #8, which was relatively clear. Figure 5.2 shows a photograph of the internal assembly prior to it being dismantled. The six membrane panels were removed from the membrane assembly, and then individually leak tested using the refrigerant detection procedure described in Chapter 2. Significant leaks were detected on membranes #1, 3, 4 and 5, but not always on both faces of the panel. A possible leak was found on panel #8. No leaks were found on the membrane surfaces of panel #6. In addition, a small leak was found on the permeate tube weld on panels #1 and 8. Figure A5.1 in Appendix 5 summarizes the membrane observations and refrigerant leak detection results. 134 Figure 5.2: ICFBMR membrane assembly after service in pilot plant reactor 5.2 Membrane Flux Testing Panels #6 and 8, thought to be leak-free from the refrigerant check, were flux-tested in the membrane permeation rig shown in Figure 2.12. Panel #8 was installed at the top of the core draft box, while panel #6 was located immediately below (see Figure 5.2 above). The permeation vessel contained pure hydrogen and was heated to 560°C. The vessel pressure was varied between 170 and 585 kPa, and the flux on the two panels measured. The permeate pressure was 101 kPa, and no sweep gas was used. Figure 5.3 presents the resulting flux data as a function of the difference of the square root hydrogen partial pressures (see Table A5.1 in Appendix 5 for data). The data in Figure 5.3 are very linear, and the trend line passes near the origin, as would be expected if the hydrogen permeation followed Sievert's equation. The membrane panel permeation area was previously calculated to be 0.0258 m2 from geometry. Based on membrane permeation equation 2.10, the effective permeation area of the two panels, were found to be 0.0081 and 0.0167 m2 for membrane #6 and 8 respectively. These values are significantly 135 higher than previously determined in flux testing prior to the membranes being installed in the pilot reactor and pilot reactor flux testing (see Table 5.1). The data for panel #8 is especially inconsistent. 0.004 (A 15 0.003 0.002 5 o 0J -4-» CO I 0.001 0 . R 2 = 0.9987 ... x Panel #8 - Panel #6 _ — " " ' R 2 = 0.9995 0 100 200 300 400 500 Partial H2 pressure , h igh A 0 .5 - Partial H2 pressure , l o w A 0 . 5 (Pa A 0.5) Figure 5.3: Post-pilot run flux test data for membrane panels #6 and #8 at 560°C Table 5.1: Calculated effective membrane permeation area, before, during and after pilot runs Panel Pre-pilot testing Average of pilot flux Post-pilot testing testing in run 5 and 6 Effective Fraction of Effective Fraction of Effective Fraction of area (m2) expected area (m2) expected area (m2) expected #6 0.0059 23% 0.0069 27% 0.0081 31% #8 0.0101 39% 0.0096 37% 0.0167 65% The increases in the calculated effective permeation area are not due to membrane leaks, which were found to be small (see next section). The low initial membrane permeation rate may have been due to organic impurities initially adsorbed onto the membrane surface, later removed during the ICFBMR pilot tests. It is not likely that wear of the palladium foil by particle fluidization thinned the membrane, as a heavy catalyst scale was present on the membrane surface. Subsequent to the pilot testing, it was found that exposure of the membrane to air at high temperatures (>500°C) can oxidize organic contaminants, restoring the membrane flux (Li, 2005). Another possible explanation for the higher flux in the post-pilot tests is that surface catalyst scale may have been dislodged during removal of the membrane assembly from the reactor. 5.3 Permeate Purity The purity of the permeate gas from panel #6 and 8 was also tested. An equimolar mixture of hydrogen and helium was fed to the permeation rig (Figure 2.12 in Chapter 2). The rig was then held 136 at the same temperature (560°C) as in the flux test described in the previous section. The purity of the permeate hydrogen was measured with a gas chromatograph, which detected small amounts of helium in the permeate stream from both panels, indicating pinhole leaks. Table 5.2 summarizes the permeate purity data, in which the hydrogen-to-helium separation factors are calculated from: H 2 : He separation factor = {^"/[p _ p jj^PHeh ~ ^J/j j (5-1) Equation 5.1 is simply a ratio of the flux to partial pressure driving force for the two species. The partial pressure difference, rather than the difference in the square root partial pressures, is used as this permits comparison with microporous membranes. It should be noted that helium has a very small atomic radius, and thus leaks of heavier compounds such as methane and carbon monoxide should be lower than for helium. The concentrations of helium in the permeate stream are still low, and thus did not significantly affect the analysis of hydrogen flux data presented in the previous section. Table 5.2: Permeate purity in post-pilot runs (560°C, 50:50 mixture of H 2 and He in permeation vessel, 101 kPa permeate pressure) Panel Vessel pressure (kPa) [He] in permeate H2:He separation factor #6 439 0.031% 6071 #6 439 0.029% 6360 #6 439 0.028% 6525 #8 494 0.25% 671 #8 494 0.26% 660 Gas chromatograph analyses from membrane panels #6 and #8 during the last pilot test did not indicate any non-hydrogen species. It is not known whether the leaks found during the above post-run testing were present during the pilot work. They could have been a result of the heavy handling required to remove the membrane assembly from the ICFBMR core box. > 5.4 Foil analysis with SEM / EDX The palladium alloy foil from membrane panel #1 (side A) was removed from the stainless steel substrate by peeling off one corner using a knife. The foil was bonded firmly to the sealing edges, but could be removed with a firm tug. The permeation area, which was covered by alumina, also appeared to be bonded to the foil, though more loosely than the seal edges. Figure 5.4 shows a corner of the membrane foil peeled away, exposing the alumina substrate. The scale on the front of the membrane could be readily brushed off with tissue paper. 137 Figure 5.4: Peeling o f the palladium alloy foil from membrane panel #1 Three samples were cut from the palladium foil and analysed by a scanning electron microscope ( S E M ) and E D X (energy dispersive x-ray) using a Hitachi model S-2300 analyzer. S E M photographs of the used palladium alloy foil were taken at x300 (Figures 5.5 to 5.8), xlOOO (Figures 5.9 and Figures A5.2 , A5.3 and A5.5 in Appendix 5) and x3000 magnification (Figures A5.4, A5.6 and A5.7 in Appendix 5). S E M images o f fresh, unused Pd 7 5 Ag 2 5 foil are also presented: • Fresh palladium alloy foil - Note the smooth surface appearance in Figure 5.5 compared to the subsequent images o f the used foi l . The scratch-like marks may be due to cold rolling imperfections. • Scaled front (retentate side) - S E M images show heavy scale deposits across the entire membrane surface (Figure 5.6). • Polished front, exposing metallic foil (retentate side) - The surface morphology is much more regular than the previous scaled image, but smaller scale deposits are still present on the surface (Figure 5.7). • Back o f membrane (permeate side) - Heavy, irregular deposits are interspersed with smoother, metallic surfaces, covering approximately two-thirds o f the surface area (Figure 5.8). 138 Figure 5.5: S E M of fresh Pd75Ag25 foi l at x300 magnification X 3 Q 0 0000 2 0 1 0 0 P m Figure 5.7: SEM of polished membrane foil (front) at x300 magnification • • ; Is' if* ; 11 .* SSRrf v» -SSC^Wt - -. - > • - , - • : • : • x30© 0000 20NV 100^m Figure 5.8: SEM of back of membrane foil at x300 magnification 140 The SEM image in Figure 5.9 is a xlOOO magnification of the back of the membrane foil. Two distinct surface zones are readily apparent. The rectangular boxes indicate areas where separate EDX scans were taken. Figure 5.9: SEM of back of membrane foil at xlOOO magnification Box 1 and 2 are areas of EDX scan #3 and #4 respectively Data from four EDX scans of the used membrane are presented in Figure 5.10 below. Metal concentrations varied considerably in the four zones analyzed. Scan #1 - scaled front of membrane (Figure A5.8 in Appendix 5): It is apparent that most of the scale on the front of the membrane is due to nickel, and to a lesser extent alumina, present in the SMR catalyst. Very little palladium was detected. Scan #2 - polished front of membrane (Figure A5.9 in Appendix 5): The surface of the foil is detected in this scan, as evidenced by the large concentrations of palladium and silver. The measured palladium to silver ratio is approximately 2. Significant amounts of nickel are present, likely resulting from incomplete removal of the surface scale. Scan #3 - smooth zone of back of membrane (box 1 in Figure 5.9, EDX scan in Figure A5.10 in Appendix 5): This area is clearly the metallic foil area of the membrane. Some iron was also detected, likely from the underlying stainless steel support. Note that almost equal amounts of 141 silver and palladium were detected, a much higher proportion of Ag than detected in scan #2 (polished front) or present in the original foil (25%). This is consistent with other research (Shu et al., 1993, Amandusson et al., 2001) that indicates silver migration to the permeate membrane surface. This has been explained as either a thermally induced process to minimize the free energy of the surface, or induced by surface chemisorption. • Scan #4 - scaled zone of back of membrane (box 2 in Figure 5.9, EDX scan in Figure A5.11 in Appendix 5): This area contained significant amounts of iron, nickel and chromium, with the stainless steel sintered support being the obvious source. Palladium concentrations were also high. Silver was detected, but at lower concentrations relative to palladium than in the original foil. 1.Front (scaled) 2.Front (polished) 3.Back (smooth zone) 4.Back (scaled zone) Figure 5.10: Summary of EDX analyses of used membrane foil removed from membrane #1 5.5 Discussion All six of the membranes removed from the ICFBMR pilot plant appear to have developed leaks, though it is not known whether all leaks can be attributed to service in the reformer (some may have been caused by the rough handling required to. dissemble the reactor). From pilot plant results and post-run permeation tests, it appears that there were no catastrophic membrane failures. Permeate concentrations of non-hydrogen species were low. Instead, very small pinholes are likely to have developed, though no pinholes were detected in the SEM scans of the used membranes. Flux testing on two membranes indicated that permeation was greater after service in the pilot plant, perhaps attributable to removable impurities, such as carbon, being initially present. It appears that 142 the heavy catalyst scale formed on the membrane surface did not significantly hinder permeation. The hydrogen flux of the used membranes could be well described with a hydrogen partial pressure exponent («) of Vi, indicating that diffusion of atomic hydrogen likely remained the main resistance to hydrogen permeation. SEM images of the backside of the membrane indicates the presence of a heavy metallic scale, likely due to direct contact of the palladium foil with the stainless steel support. The alumina inter-layer deposited on the sintered metal support was intended to prevent this. The metallic scale covered a significant portion of the foil, roughly two-thirds in the SEM images. It is likely that this is the main contributor to a reduction in the membrane permeation relative to that predicted by the Sievert's permeation equation. The membrane foil may be only releasing hydrogen at the unsealed surface, thus greatly reducing the effective membrane area. It is concluded that calculation of an effective membrane area for each membrane, as in Chapter 2, is an appropriate means of characterizing the loss in permeation. The mode of pinhole development in the membranes is not known. Perhaps the adhesion of the metallic scale to the backside of the membrane reduced the foil mobility and led to local stresses in the palladium as it expanded and swelled with hydrogen and thermal expansion. What is clear is that the very thin alumina layer deposited below the foil was inadequate to prevent contact between the palladium and the stainless steel support. Since the ICFBMR membranes were fabricated, MRT has increased the thickness and modified the composition of the alumina layer, and this has resulted in membrane permeation areas remaining close to that expected from the geometry (Li, 2005). 143 Chapter 6. Process Simulation and Economics A commercial process simulator was used to model the ICFBMR reactor and perform a sensitivity analysis on the major reactor variables. The reactor model accounted for the internal circulation of solids and gas, but did not include reaction kinetics or hydrodynamics. A base case model was created for a unit producing 30 Nm3/h of high-purity H2, which was then developed into a process flow diagram. Performance parameters and economics of the 30 Nm3/h unit are discussed. 6 . 1 Introduction The I C F B M R pilot plant results demonstrated the potential to produce high-purity H 2 , but hydrogen recoveries were low due to the installed membrane area and, in some runs, poor catalyst performance. In order to extrapolate pilot plant experience to a commercial design for small-scale hydrogen production, process simulations were undertaken. The main objective was to create a design tool for future reactor and process designs. The simulation was also used to reconcile data from the I C F B M R pilot plant. Following the experimental F B M R work o f Adris (1994) and Roy (1998), several reactor models have been developed to simulate the F B M R process using two-phase fluidized bed reactor models: • Adris and co-workers (1997a, 1997b) presented findings from a fluidized bed reactor model, identifying effects o f permeation rate and pressure on methane conversion and hydrogen production. • Model l ing by Roy et al. (1998) studied various means of supplying heat to the F B M R process and concluded that the new process had the potential to reduce capital and operating costs compared to conventional S M R fixed bed reactors. • Abba et al. (2003) adapted a generalized fluidized bed reactor model to autothermal reforming in a F B M R . Results indicated a significant reduction (-30%) in permeate H 2 when air rather than 0 2 144 is used as an oxidant in the well-mixed reactor. It was predicted that superficial gas velocities in excess of 0.3 m/s would be of little benefit. • Dogan et al. (2003) summarized autothermal FBMR modelling results with direct oxygen addition. Varying the steam-to-carbon ratio showed only a mild effect on hydrogen permeate yield. The benefits of high temperature operation to produce higher H 2 membrane fluxes were also highlighted, although the operating temperatures investigated (650 to 850°C) are excessive for the palladium-silver membranes of interest in my work. Rakib and Alhumaizi (2005) also modelled oxygen addition to an FBMR. • Elnashaie and co-workers have published a number of modelling papers on a circulating fluidized bed membrane reactor for reforming of natural gas and higher hydrocarbons (e.g. Prasad and Elnashaie, 2002, Chen and Elnashaie, 2005). In all of the above models, the kinetics of Xu and Froment (1990) were adapted and combined with the hydrodynamics of a fluidized bed reactor. All assumed vertical membrane tubes (or U-tubes) and considered permeate sweep gas to increase the hydrogen flux. In some cases, simulations have suggested that the use of long vertical tubes in the FBMR would lead to reverse permeation of hydrogen in the relatively cool freeboard of the reactor (e.g. Abashar et al., 2003). This can be avoided with planar membrane modules that extract hydrogen over a more limited reactor height interval. The above studies also indicated that the kinetics of the SMR process are sufficiently rapid, at least for temperatures larger than about 550-600°C, that chemical equilibrium is very nearly achieved for FBMR heights and catalysts which are likely to be of practical interest. In addition, the above models indicate that for Pd membranes with current thicknesses (25 to 75 pm), the membrane flux is the controlling parameter of the reactor performance. In view of this, it was decided to ignore the SMR reaction kinetics and simulate the ICFBMR process using equilibrium reactors in a commercial software environment (HYSYS, version 3.4) in order to create a flexible process design tool. A commercial simulator permits rapid development of the overall process flow diagram and facilitates sensitivity analyses and equipment sizing. The only hydrodynamic feature considered in the simulation is internal circulation of solids, gas and heat. 6.2 Reactor Simulation The HYSYS model included interna! circulation of catalyst and gas down the annulus and up the reactor core. SMR equilibrium values from the HYSYS simulation were found to closely match those predicted by an independent kinetic model. Details on the HYSYS reactor model are presented in 145 Table A6.1.1, Appendix 6.1. The ICFBMR was simulated with a series of Gibbs reactors and hydrogen separators, as shown schematically in Figure 6.1. The HYSYS flowsheet and algorithm are presented in Figures A6.1.1 and A6T.2, Appendix 6.1. The reactants modelled were C H 4 , C 2 H6, CO, C0 2 , H 2 and H 20. Carbon formation is not thermodynamically favoured at these conditions, and was thus ignored. I C F B M R Preheated air)—d 1 Core draft box (10 repeated membrane/ reactor pairs) Stream key Gas stream Solid stream • Heat stream > Preheated C H 4 / H 2 Q E. Gas split D. Gibbs reactor (oxidant) ,9.10> <10.1 C10 Gibbs reactor (adiabatic) <6.10> <10.9 B10. Membrane k? i? rb.n Gibbs reactor"1 L (adiabatic) I j- B.n Membrane |-<^> C1. Gibbs reactor (adiabatic) A. Gibbs reactor (adiabatic) 14V-I Rxtr off-gas Heat loss Annular circulation (recycle) Permeate H^) Figure 6.1: ICFBMR simulation schematic Preheated steam and methane (1) are mixed with hot circulating solids (2) from the oxidation zone (D) in a Gibbs reactor (A), representing the area immediately above the main feed 146 distributor. The circulating solids also drag down a portion of the gases (3) from the top of the reactor (E). The temperature drops as methane and steam react adiabatically in the first Gibbs reactor (A). Higher catalyst circulation reduces the temperature drop. The temperature at the exit of the first Gibbs reactor (4) is considered to be the maximum operating temperature of the membranes. The gas stream (4) from the Gibbs reactor (A) then flows to the first membrane (Bi) in the draft tube, where pure, low-pressure hydrogen is removed (8.1). The hydrogen removal shifts the SMR equilibrium, and the non-permeate gases (6.1) and solids (5) further react in an adiabatic Gibbs reactor (Ci). A series of ten membrane / Gibbs reactor pairs then follows, each removing an equal fraction of the permeate hydrogen. By using a series of reaction / separation steps, the partial pressure driving force in each membrane set is known, allowing the membrane permeation area to be calculated. Full chemical equilibrium is assumed for the product stream after each H 2 removal stage. Gas (9.10) and solids (10.10) from the last core Gibbs reactor (C.10) leave the core draft box. These streams react with preheated air (11) in another Gibbs reactor (D). In order to account for ambient reactor heat losses, a heat stream (12) is rejected by the oxidant reactor (D). The flow of air (11) is adjusted until the reactor energy balance is satisfied. Note that the simulation reacts methane with the oxygen, as this minimized the Gibbs free energy. In practise, a portion of the hydrogen is also likely to be oxidized. In any event, the question of which chemical species is oxidized is does not practically matter, as the reactor off-gas is combusted in a commercial system to recover heat for raising steam and preheating feeds, as depicted in Figure 6.10. Hot catalyst (2) is circulated back to the bottom of the reactor via the annular region of the reactor. The solids mass flux (Gpcore, kg/m2s) is assumed to be a function of the superficial gas velocity in the core (Ucore, m/s) and the catalyst particle density (pp = 2,100 kg/m3), as described by the correlation developed in Chapter 3 (Figure 3.20) for vertically-slotted panels. As discussed in Chapter 5, this equation matched the deduced circulation rate in the pilot reactor within 10%: G p j a n = - 0 . 0 0 0 1 0 3 ( ^ L / c o r e ) 2 + 0 . 1 8 0 ( P p L / c o r e ) - 1 1 . 3 (6.1) In order to set the solids circulation flow (mpcore, kg/s), the flow area in the core draft box (Acore) is required. This is calculated from the user-inputted superficial core velocity (Ucore) and the volumetric flow of gas entering the core draft box (stream 4): A r~* Qcore s~* ff\0\ p,core core p,core j j p,core v^'^v 147 The circulating solids act only as a heat carrier in the simulation. Calcium was selected to represent the circulating catalyst solids, as AI2O3 was not available in the HYSYS property database. Calcium is chemically inert in the simulation. As the specific heat of calcium in HYSYS is slightly lower than that of alumina, the circulation rate expression (equation 6.2) was multiplied by a specific heat adjustment factor (1.15) to match the heat flow of the circulating alumina catalyst. • The gas leaving the oxidant Gibbs reactor (13) splits into two compositionally identical streams; a reactor off-gas stream (14) and a stream recycled back to the bottom of the reactor with the circulating solids (3). Cold model testing (Chapter 3) suggested that the volumetric downward flow of gas was approximately equal to the volumetric flow of solids down the annulus. The volumetric flow of the circulating gas is therefore set by dividing the solids circulation rate (equation 6.2) by the catalyst bulk density (2,100 kg/m3). The simulation application is somewhat limited as reaction kinetics and hydrodynamics are not accounted for. In addition, the simulation does not address the following features of the ICFBMR system: • A very small portion of the methane may flow up the annulus, bypassing the reactor core. • Some heat transfer must occur from the hot annulus zone to the cooler reactor core. • Heat conduction vertically along the membrane panels and supports is neglected. • The simulation effectively assumes plug flow (ten well-mixed stages) of gas and solids up the reactor core. More than ten stages did not significantly alter the simulation results and added to computation time. In cold model experiments, a small fraction of the gas and solids were found to circulate from the oxidation zone to the upper sections of core. • The pressure drop in the reactor is ignored, i.e. the pressure in each of the compartments was assumed to be the same. The effective membrane area (Aej) for each of the permeation zones can then be calculated from the simulation output using the hydrogen flux equation for Pd75Ag25: -1 T Q H exp( ) ( P H , ? - P H , 5 ) (6.3) k H R T where T is the membrane thickness (base case 50E-6 m). QH is the H 2 permeation per stage (base case 30 Nm /h/10 stages = 0.037 mol/s). • kH = 3.43E-7 mol/m s Pa 0 5 (see Chapter 2) 148 • Ep = 9,180 J/mol (see Chapter 2) • T is the permeation temperature (K), assumed to be equal to the temperature of the reactor gas entering each membrane separation stage. • PHI, is the hydrogen partial pressure in the reactor (Pa). For each permeation zone, Put, is assumed to be the average of that zone's inlet and outlet equilibrium partial pressures. • PHi is the hydrogen partial pressure in the permeate stream (Pa). The base case is atmospheric pressure (101 kPa), without any sweep gas. 6.2.1 Simulat ion Base Case The main reactor simulation variables and their base values are listed in Table 6.1. A summary of the main process streams for the base case is presented in Table 6.2. Refer to Figure 6.1 for nomenclature. Full stream output data for the base case simulation are presented in Table A6.1.2, Appendix 6.1. As indicated in Table 6.1, the target value for the permeate hydrogen to methane molar ratio was 2/2. Higher hydrogen recovery is achievable if more membrane area is installed. However, when the overall system is considered, a hydrogen recovery between 2.5 to 2.7 is likely to be an optimal, as the reactor off-gas containing methane, CO and unrecovered H2 is combusted to preheat the reactor feeds. If a higher hydrogen recovery were used, additional natural gas would be required to supplement the burner duty. This process configuration is presented in Section 6.3. The base case simulation output appears reasonable. The oxidation zone was calculated to be 18°C hotter than the inlet temperature of the reactor core (600°C). The oxidation temperature is a function of the solids circulation rate, the preheat temperature of the reactor feeds and the specified maximum membrane temperature. The flow of oxidation air was adjusted to satisfy the reactor energy balance, resulting in an 02:CH4 molar ratio of 0.39. Approximately 10% of the gases in the upper reactor were estimated to recycle to the reactor core with the circulating solids. Due to hydrogen removal, partially countered by the increase in molar flow due to reaction, the number of moles of gas leaving the reactor core (excluding permeate H2) is approximately two-thirds of the gas entering the core. 149 Table 6.1: ICFBMR base case simulation Variable description Base case Notes Reactor pressure 2,500 kPa Max 1,500 kPa in ICFBMR pilot plant <D Total permeate H 2 flow 30 NmJ/h Lower side for distributed H 2 market JO .2 Steam-to-carbon molar ratio of feed 3.0 Typical industrial value a > Permeate FbjCFL, ratio 2.5 Reasonable objective c <D T 3 C Superficial core velocity (Ucore) 0.25 m/s Based on cold model tests u D. CU TD Max membrane operating temperature 600°C Temperature of stream 4 in Figure 6.1 C Feed preheat temperature 600°C Assures energy integration Reactor heat losses 2kW -10% of endothermic reaction heat iriables C H 4 flow 12Nm7h Set by H 2 flow and H 2 recovery iriables Steam flow 36NmVh Set by CFL, flow and H20:CFL| ratio TO > £3 GvCFL, molar ratio 0.39 Adjusted by reactor heat balance 1) -a c Solids circulation 1,356 kg/h Set by correlation with Ucore (Eq. 6.2) a, Q Oxidation zone temperature 618°C Adjusted to match maximum membrane operating temperature Membrane thickness (r) 50E-6 m Readily commercially available <8 c Permeate sweep gas None Reduces connections and ensures product impurities Meml Permeate pressure 101 kPa No vacuum equipment required Meml Membrane efficiency 100% Area predicted by Eq. 6.3 150 Table 6.2: Stream summary for 30 Nm Th reactor base case simulation (Note: Stream numbers refer to Figure 6.1) Stream 1 2 3 4 5 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 6.10 8.2 Mass Flow [kg/h] 37.55 1558.98 5.33 42.88 1559 42.61 42.34 42.07 41.80 41.53 41.26 40.99 40.72 40.45 40.18 0.27 Volumetric flow [m3/h] 6.18 1.14 0.65 7.61 1.14 7.21 6.93 6.65 6.37 6.10 5.83 5.57 5.32 5.07 4.82 9.66 Molar Flow [kmol/h] 2.14 38.90 0.22 2.62 38.90 2.49 2.40 2.31 2.22 2.14 2.05 1.97 1.89 1.80 1.72 0.13 Temperature [C] 600 618 618 600 600 600 597 594 592 588 585 582 579 575 572 603 Pressure [kPa] 2500 2500 2500 2500 2500 2500 2500 2500 2500 2500 2500 2500 2500 2500 2500 101 Heat Flow [kJ/h] -3.79E+05 7.63E+06 -3.27E+04 -3.86E+05 7.61 E+06 -3.88E+05 -3.87E+05 -3.85E+05 -3.83E+05 -3.81 E+05 -3.79E+05 -3.77E+05 -3.75E+05 -3.72E+05 -3.70E+05 2.27E+03 Mole fraction - CH4 0.25 0 0.0131 0.1551 0 0.1634 0.1607 0.1572 0.1530 0.1480 0.1421 0.1352 0.1274 0.1187 0.1086 0 Mole fraction - CO 0 0 0.0154 0.0084 0 0.0088 0.0096 0.0103 0.0110 0.0116 0.0122 0.0128 0.0133 0.0137 0.0141 0 Mole fraction - C02 0 0 0.1778 0.0578 0 0.0609 0.0718 0.0839 0.0973 0.1122 0.1288 0.1471 0.1672 0.1895 0.2140 0 Mole fraction -H20 0.75 0 0.4095 0.5532 0 0.5830 0.5877 0.5913 0.5940 0.5959 0.5968 0.5961 0.5947 0.5925 0.5885 0 Mole fraction - N2 0 0 0.3040 0.0252 0 0.0266 0.0276 0.0286 0.0298 0.0310 0.0323 0.0336 0:0351 0.0367 0.0384 0 Mole fraction - 02 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Mole fraction - H2 0 0 0.0802 0.2003 0 0.1572 0.1428 0.1287 0.1150 0.1013 0.0878 0.0751 0.0623 0.0489 0.0365 1.000 Mole fraction - Ca 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 Sream 9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9 9.10 10.1 11 12 13 14 15 Mass flow [kg/h] 42.61 42.34 42.07 41.80 41.53 41.26 40.99 40.72 40.45 40.18 1559 28.84 - 69.02 63.69 2.70 Volumetric flow [m3/h] 7.32 7.04 6.76 6.49 6.22 5.96 5.70 5.45 5.20 4.96 1.14 2.92 - 8.35 7.71 95.5 Molar Flow [kmol/h] 2.53 2.44 2.36 2.27 2.19 2.10 2.02 1.94 1.86 1.78 38.90 1.00 - 2.81 2.60 1.34 Temperature [C] 597 594 592 588 585 582 579 575 572 569 597 600 618 618 593 Pressure [kPa] 2500 2500 2500 2500 2500 2500 2500 2500 2500 2500 2500 2500 - 2500 2500 101 Heat Flow [kJ/h] -3.85E+05 -3.83E+05 -3.81 E+05 -3.79E+05 -3.77E+05 -3.75E+05 -3.73E+05 -3.70E+05 -3.68E+05 -3.65E+05 7.61 E+06 1.76E+04 7.20E+03 -4.23E+05 -3.90E+05 2.23E+04 Mole fraction - CH4 0.1522 0.1486 0.1443 0.1392 0.1333 0.1266 0.1190 0.1105 0.1008 0.0902 0 0 - 0.0131 0.0131 0 Mole fraction - CO 0.0091 0.0097 0.0103 0.0109 0.0115 0.0120 0.0124 0.0128 0.0131 0.0133 0 0 - 0.0154 0.0154 0 Mole fraction - C02 0.0680 0.0793 0.0918 0.1056 0.1209 0.1378 0.1562 0.1763 0.1987 0.2230 0 0 - 0.1778 0.1778 0 Mole fraction - H20 0.5566 0.5589 0.5603 0.5606 0.5600 0.5582 0.5554 0.5515 0.5463 0.5400 0 0 - 0.4095 0.4095 0 Mole fraction - N2 0.0261 0.0271 0.0281 0.0291 0.0303 0.0315 0.0328 0.0341 0.0356 0.0372 0 0.79 - 0.3040 0.3040 0 Mole fraction - 02 0 0 0 0 0 0 0 0 0 0 0 0.21 - 0 0 0 Mole fraction - H2 0.1881 0.1764 0.1652 0.1545 0.1441 0.1340 0.1243 0.1148 0.1055 0.0963 0 0 - 0.0802 0.0802 1 Mole fraction - Ca 0 0 0 0 0 0 0 0 0 0 1 0 - 0 0 0 Figure 6.2 presents the temperatures and average hydrogen partial pressures for the ten membrane permeation stages for the base case (stage 1 is at the bottom, 10 at the top). A significant temperature drop (~30°C) is predicted over the reactor core. Another item of note is the reduction in the hydrogen partial pressure as gases flow up the reactor core, corresponding to a reduced permeation driving force. The reactor-side H 2 partial pressure in the final permeation stage is -250 kPa, still well above the base case permeate-side pressure (101 kPa). 600 600 200 1 8 10 2 3 4 5 6 7 Hydrogen removal stage Figure 6.2: Permeation stage temperature and reactor hydrogen partial pressure (base case) Each membrane stage in the simulation is assumed to permeate an equal flow of hydrogen. However, as the temperature and hydrogen partial pressure are reduced as the gases pass upward through the reactor core, the permeation area required for each stage increases (see "adiabatic" curve in Figure 6.3). The permeation area in the last stage is approximately twice that required for the first H 2 stage, indicating that membrane panels at the bottom of the reactor would produce more H 2 than panels of equal area located higher in the reactor core. The total permeation area (for the assumed 50 um thick membrane) is the sum of the areas of all permeation stages, calculated to be 2.34 m2. As the calculated permeation area (equation 6.3) is proportional to the membrane thickness, 25 urn thick membranes would halve the required membrane. If the internal solids circulation were very high, the reactor core would approach isothermal operation. The effect of an isothermal core (600°C) on membrane area is also presented in Figure 6.3 ("isothermal" curve). The total membrane area is calculated to be 2.01 m2 for the isothermal case, 14% less than calculated if the core operated adiabatically. 152 0.4 0 -I , , , 1 , 1 1 1 1 1 2 3 4 5 6 7 8 9 10 Hydrogen removal stage Figure 6.3: Calculated permeation stage area (base case, 50 um Pd7 5Ag25 membrane) 6.2.2 Effect of Reactor Variables The influence of the major process variables on the ICFBMR design was investigated by parametric studies using the HYSYS simulation. Perhaps the most important ICFBMR reactor design characteristic is the required membrane area. As the ICFBMR reactor geometry is governed by the membrane layout, the total reactor volume can be estimated based on the calculated membrane area. One of the disadvantages of the ICFBMR design is that it tends to have a low overall space velocity relative to fixed bed configurations. It is therefore important to reduce the membrane area in order to produce a compact reactor design. In addition, the palladium membranes represent a significant system cost. • Reactor pressure: Figure 6.4 demonstrates the strong effect of reactor operating pressure on the predicted membrane area. Increasing reactor pressure above 2,500 kPa decreases the area only marginally. The calculated membrane area is normalized to the base case (2.34 m2). The solids circulation rate was assumed to be only a function of the superficial core velocity (equation 6.1, above). If the core superficial velocity is held constant, increasing the reactor pressure reduces the solids circulation rate as it decreases the flow area in the reactor core (Acore). Decreasing the solids circulation increases the temperature drop in the reactor core, increasing the membrane area. The data in Figure 6.4 account for this effect. • Steam-to-carbon ratio: Also shown on Figure 6.4 is the effect of the steam-to-carbon ratio on the normalized membrane area. Given the weak effect for the limited range considered, other issues are likely to govern the preferred steam-to-carbon ratio, such as the potential for carbon formation on the catalyst at low molar ratios. 153 2.5 1,000 1,500 2,000 2,500 3,000 Reactor pressure (kPa) Figure 6.4: Effect of reactor pressure and steam-to-carbon ratio on normalized base case membrane area (50 um Pd 7 5Ag 2 5 membranes) • Hydrogen recovery: Figure 6.5 presents the influence of hydrogen recovery (moles of permeate H 2 per mole of CFL, fed to the reactor) on the normalized membrane area at reactor pressures of 1,500 and 2,500 kPa. As expected, the membrane area requirement increases with increasing hydrogen recovery. Hydrogen recoveries beyond a H2:CHt molar ratio of 3.0 become progressively more difficult, as the hydrogen partial pressure in the reactor becomes very low, and reaction kinetics may become limiting. Also shown in Figure 6.5 is the required 0 2 to CFL, molar ratio as a function of hydrogen recovery. As more hydrogen is removed from the reactor, the reaction duty increases, as more oxidation air is required to generate the heat needed for autothermal operation. H2 product:CH4 feed molar ratio (-) Figure 6.5: Effect of hydrogen recovery on normalized base case membrane area 154 Membrane temperature: The base case assumes a maximum operating temperature of 600°C. The ICFBMR pilot reactor membranes operated at temperatures up to 650°C. Increasing the membrane operating temperature strongly decreases the membrane area as it increases the specific membrane flux (see equation 6.3) and, more importantly, favourably shifts the SMR equilibrium. However, operation at temperatures below 600°C, and perhaps as low as 550°C, is likely needed for successful long-term operation of palladium-based membranes. Figure 6.6 clearly illustrates the incentive for the development of high-temperature membranes. 2.0 0.5 I , , 1 1 550 575 600 625 650 Maximum membrane operating temperature (°C) Figure 6.6: Effect of membrane operating temperature on normalized base case membrane area Feed preheat temperature: The base case assumes that the feed streams (steam, methane and air) enter the reactor at 600°C. Figure 6.7 demonstrates that the normalized membrane area is relatively insensitive to the feed preheat temperature, despite the significant change in the required flow of oxidation air. This highlights how the ICFBMR facilitates circulation of heat within the reactor, but not nitrogen from the oxidation air, which would otherwise reduce the H 2 flux through the membrane. 155 1.25 0.50 0.75 400 700 500 600 Feed preheat temperature (°C) Figure 6.7: Effect of feed preheat temperature on normalized membrane area Solids circulation rate: Increasing the solids circulation rate decreases the temperature drop in the reactor core (Figure 6.8). An isothermal core will result in higher fluxes and improved reactor equilibrium in the upper reactor core. If solids circulation is lower than predicted in the simulation, there is a penalty on the predicted membrane area. Increasing solids circulation beyond the base case has only a modest effect on membrane area. 1.25 CU k-ro 0) c CO I . n E cu E TJ CU N ~m E i_ o z 1.00 0.75 -•-Area -o-Temperature 0 1 2 3 4 Normalized solids circulation (-) Figure 6.8: Effect of normalized solids circulation rate on normalized base case membrane area Increased solids circulation transports more gas from the top of the reactor, diluting the hydrogen in the reactor core, thus modulating the benefits of a more isothermal core. As noted above, it was assumed here that the gas circulation is volumetrically equal to that of the catalyst circulation. 156 Hence, as solids circulation increases, more nitrogen is circulated to the core, thereby reducing the hydrogen partial pressure. At very high solids circulation, the benefits of an increasingly isothermal reactor core are slightly outweighed by the nitrogen dilution effect, and the area curve turns slightly upward. • Permeate pressure: Reducing the permeate H 2 partial pressure increases the driving force for permeation, thus reducing the required membrane area. This may be achieved in membrane reactors by utilizing permeate-side sweep gas. If a high-purity H 2 product were desired, the only sweep gas that may be practical would be steam. However, the heat duty to generate sweep steam reduces the overall thermal efficiency of the process, produces a wet hydrogen product, and introduces impurities unless the water used to make the steam were to be subjected to rigorous purification. A more practical method of enhancing the hydrogen flux may be to expose the permeate stream to vacuum. However, there is a risk that vacuum operation could cause ingress of air into the permeate H 2 , adding impurities to the hydrogen product and creating safety and fiammabilify issues. Vacuum operation would also add to hydrogen compression costs. However, as shown in Figure 6.9, there can be a significant reduction in membrane area with vacuum operation. X 2.5 -i RS CO U.O " E z 0.0 -! ! 1 ' 1 0 50 100 150 200 Permeate H2 partial pressure (kPa) Figure 6.9: Effect of permeate H 2 partial pressure on normalized base case membrane area 6.3 Design of a 30 Nm 3/h ICFBMR To assess the costs and performance of a commercial ICFBMR system, a design for a 30 Nm3/h hydrogen production unit was developed with the HYSYS process simulator. This capacity is five times larger than the nominal design rate of the ICFBMR pilot reactor, and would be sufficient to support a fleet of about 150 fuel cell cars. The design assumes autothermal SMR operation using 157 direct air addition, with no external heat transfer to the reactor. At this capacity, the endothennic SMR duty is approximately 20 kW, a reasonably large value for indirect transfer in a conventional SMR design, but one that could be readily handled in an autothermal ICFBMR design. 6.3.1 Design Basis The main feeds are natural gas, city water and atmospheric air (Table A6.1.3, Appendix 6.1). Based on recent MRT operating experience (Li, 2005a), 25 um thick Pd75Ag25 membranes are chosen, but are derated by 50% to account for membrane and reactor inefficiencies and thus modeled as 50 um thick. The design assumes most of the reactor base case conditions presented previously. • Autothermal operation: The reactor is assumed to operate adiabatically through direct air addition. Electrically powered heaters are installed for reactor start-up, but are not used for normal operation. Preheating of steam and feed is accomplished through heat exchange with combusted reactor off-gas. • Reactor conditions: o 2,500 kPa reactor pressure. o 3.0 steam-to-carbon molar feed ratio. o 2/4 moles of hydrogen product per mole of feed methane. o 2 kW reactor heat loss. • Membranes: o Maximum 600°C operating temperature. o 25 urn thick, palladium silver (25%) membranes, derated by 50%. o Pure hydrogen product (i.e. no sweep gas). o Maximum operating temperature of membranes of 600°C. o Permeate pressure is 101 kPa. • Hydrogen product: o The hydrogen permeate stream is compressed to 3,000 kPa. o Hydrogen storage and dispensing is outside the scope of the process design. • Reactor geometry and hydrodynamics: o Superficial gas velocity within the range of 0.1 to 0.3 m/s, with the final velocity to be determined by layout considerations, o Membranes arranged with vertical communication slots between adjacent flow channels. 158 • Burner. The non-permeate reactor off-gas has fuel value from residual CH4, CO and H 2 . It is combusted to generate the high-temperature flue gas needed to raise steam and preheat the reactor feed. The burner duty can be supplemented by fresh natural gas if necessary. 6.3.2 Process Simula t ion Several process options, mainly focused around heat integration, were considered in developing the flowsheet. The process flow diagram and associated stream table are shown in Figure 6.10 and Table 6.3 respectively. Figure A6.1.4 in Appendix 6.1 presents the process flow schematic output from the HYSYS simulation. The major unit operations included in the 30 NnvVh PFD include: • Water treatment: City water is treated in a mixed ion exchange bed to remove trace hardness and organic compounds. Treated city water is combined with water recovered from the reactor off-gas, then pumped to the process. • Natural gas treatment: Natural gas is desulfurized in two parallel beds of ZnO adsorbent. • Gas compression: Natural gas and oxidation are compressed to 100 kPa above the reactor pressure. An adiabatic efficiency of 75% is assumed for the compressors. Inter-stage compressor cooling is assumed to reject heat to ambient air. • Burner: Non-permeate reactor off-gas is combusted in a forced draft burner. Excess air is adjusted to achieve the desired burner operating temperature. Increasing the burner temperatures (i.e. minimizing excess air) results in higher overall system efficiencies. However, a maximum burner operating temperature of 1,000°C is imposed to allow standard materials of construction to be used. A heat loss of 2 kW from the burner to the environment is assumed to account for non-reactor heat losses. The bulk of the burner duty is provided by the reactor off-gas. A small amount of natural gas is also combusted to represent the burner pilot flame. • Heat exchange: Steam is raised and the reactor feeds preheated through heat exchange with reactor off-gas and burner gas. • ICFBMR: Preheated natural gas and steam react to form hydrogen, most of which is extracted by the membranes to produce a high-purity H 2 stream. Sufficient air is added to allow the reactor to operate adiabatically (with a 2 kW ambient heat loss allowance). After the non-permeate reactor off-gas (ROG) leaves the reactor, it is cooled and condensed. Condensate is recycled back to the feed water tank. The residual heating value in the cooled ROG is recovered in the burner. • H2 compression: The hydrogen permeate is compressed in a multistage compressor. 159 Table 6.3: Stream table for 30 Nm 3/h flowsheet Stream 1 2 3 4 5 7 8 9 10 11 13 14 15 16 20 21 Mass Flow [kg/h] 8.69 . 8.59 8.59 8.59 11.12 28.93 28.93 28.93 37.52 37.52 2.69 • 2.69 2.69 2.69 63.78 63.78 Actual Volume Flow [m3/h] 10.78 10.66 0.73 1.44 0.011 0.030 0.14 0.94 2.23 6.18 95.90 62.88 34.36 1.21 7.73 6.53 Molar Flow [kgmole/h] 0.541 0.535 0.535 0.535 0.617 1.606 1.606 1.606 2.142 2.142 1.332 1.332 1.332 1.332 2.603 2.603 Temperature [C] 15 15 160 564 15 68 226 226 203 600 601 300 40 50 618 482 Pressure [kPa] 120 120 2600 2600 300 2600 2600 2600 2600 2500 101 101 101 3000 2500 2500 Vapour Fraction 1 1 1 1 0 0 0.05 0.40 0.72 1 1 1 1 1 1 1 Molecular Weight 16.04 16.04 16.04 16.04 18.02 18.02 18.02 18.02 17.52 17.52 2.02 2.02 2.02 2.02 24.50 24.50 Mass Density [kg/m3] 0.81 0.81 11.71 5.96 1015 975 200 30.90 16.80 6.07 0.03 0.04 0.08 2.23 8.26 9.78 Viscosity [cP] 1.09E-02 1.09E-02 1.56E-02 2.48E-02 1.14E+00 4.13E-01 - - - 2.66E-02 1.97E-02 1.52E-02 9.16E-03 9.48E-03 3.27E-02 2.91E-02 Thermal ConductivityJW/m-K] 3.23E-02 3.23E-02 5.66E-02 1.33E-01 5.95E-01 6.61E-01 - - - 9.24E-02 3.79E-01 2.84E-01 1.82E-01 1.88E-01 8.03E-02 6.87E-02 Mass Heat Capacity [kJ/kg-C] 2.23 2.23 2.71 4.08 4.05 4.05 4.83 3.87 3.02 2.70 14.73 14.60 14.12 14.26 1.62 1.56 Heat Flow [kJ/h] -4.08E+04 -4.03E+04 -3.74E+04 -2.56E+04 -1.76E+05 -4.52E+05 -4.30E+05 -4.11E+05 -4.36E+05 -3.78E+05 2.25E+04 1.07E+04 5.67E+02 9.54E+02 -3.90E+05 -4.04E+05 Comp Mole Frac (Methane) 1 1 1 1 0 0 0 0 0.25 0.25 0 0 0 0 0.013 0.013 Comp Mole Frac (CO) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.015 0.015 Comp Mole Frac (C02) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.177 0.177 Comp Mole Frac (H20) 0 0 0 0 1 1 1 1 0.75 0.75 0 0 0 0 0.409 0.409 Comp Mole Frac (Nitrogen) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.304 0.304 Comp Mole Frac (Oxygen) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Comp Mole Frac (Hydrogen) 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0.081 0.081 Stream 22 23 24 26 27 30 31 32 35. 36 37 38 39 40 41 Mass Flow [kg/h] 63.78 63.78 63.78 45.94 17.84 28.95 28.95 28.95 0.10 43.28 43.28 43.28 89.31 89.31 89.31 Actual Volume Flow [m3/h] 4.43 3.70 1.99 40.84 0.019 23.71 1.40 2.82 0.12 35.44 34.90 56.21 302.05 173.99 127.44 Molar Flow [kgmole/h] 2.603 2.603 2.603 1.614 0.990 1.003 1.003 1.003 0.006 1.500 1.500 1.500 2.995 2.995 2.995 Temperature [C] 250 177 100 92 100 15 160 600 15 15 21 200 952 432 245 Pressure [kPa] 2500 2500 2500 120 2500 101.3 2600 2600 120 101.325 105 105 101 101 101.3 Vapour Fraction 1 0.99 0.62 1 0 1 1 1 1 1 1 1 1 1 1 Molecular Weight 24.50 24.50 24.50 28.47 18.03 28.85 28.85 28.85 16.04 28.85 28.85 28.85 29.82 29.82 29.82 Mass Density [kg/m3] 14.41 17.22 32.06 1:12 949 1.22 20.73 10.25 0.81 1.22 1.24 0.77 0.30 0.51 0.70 Viscosity [cP] 2.13E-02 1.96E-02 2.79E-01 1.83E-02 2.49E-02 4.29E-02 1.09E-02 1.83E-02 1.86E-02 2.63E-02 4.99E-02 3.20E-02 2.57E-02 Thermal Conductivity [W/m-K] 4.81 E-02 - 3.72E-02 6.81E-01 2.52E-02 3.55E-02 6.06E-02 3.23E-02 2.52E-02 2.56E-02 3.71 E-02 8.01 E-02 5.04E-02 3.82E-02 Mass Heat Capacity [kJ/kg-C] 1.49 1.50 2.01 1.15 4.10 1.01 1.05 1.13 2.23 1.01 1.01 1.03 1.32 1.17 1.11 Heat Flow [kJ/h] -4.27E+05 -4.35E+05 -4.80E+05 -2.03E+05 -2.76E+05 -3.01E+02 3.89E+03 1.76E+04 -4.52E+02 -4.50E+02 -1.92E+02 7.71 E+03 -2.03E+05 -2.61 E+05 -2.80E+05 Comp Mole Frac (Methane) • 0.013 0.013 0.013 0.021 0.000 0 0 0 1 0 0 0 0 0 0 Comp Mole Frac (CO) 0.015 0.015 0.015 0.025 0.000 0 0 0 0 0 0 0 0 0 0 Comp Mole Frac (C02) 0.177 0.177 0.177 0.285 0.001 0 0 0 0 0 0 0 0.181 0.181 0.181 Comp Mole Frac (H20) 0.409 0.409 0.409 0.048 0.999 0 0 0 0 0 0 0 0.123 0.123 0.123 Comp Mole Frac (Nitrogen) 0.304 0.304 0.304 0.491 0.000 0.79 0.79 0.79 0 0.79 0.79 0.79 0.660 0.660 0.660 Comp Mole Frac (Oxygen) 0 0 0 0 0 0.21 0.21 0.21 0 0.21 0.21 0.21 0.036 0.036 0.036 Comp Mole Frac (Hydrogen) 0.081 0.081 0.081 0.130 0.000 0 0 0 0 0 0 0 0 0 0 Heat stream H1 H2 H3 H4 H5 H6 H7 Heat Flow [kJ/h] 7,200 7,200 21,388 4,524 39,405 10,104 45,142 A pinch analysis was performed to look at possible heat recovery schemes (see Figure A6.1.3 and Table A6.1.4, Appendix 6.1). The pinch temperature is 225°C, the feed steam boiling point. The reactor pressure has a significant impact on the pinch analysis, as it sets the boiling temperature of the feed steam. The burner gases are vented to atmosphere at ~250°C, ensuring a positive temperature driving force to raise steam. Table 6.4 summarizes the PFD feed and utility consumptions. The molar ratio of permeate (pure) hydrogen to total natural gas consumed is 2.48. Table 6.4: PFD stream summary for 30 Nm3/h design Description PFD value Notes Natural gas 12.1 Nm7h 0.1 NmJ/h to burner, balance to reformer T3 <U <U Air 56 NmJ/h Atmospheric air, 60% to burner, balance to reformer lu Water 11 kg/h Water recovery from ROG reduces consumption <*> Cooling water ~2m J /h Air coolers could eliminate use of cooling water '-*-» Nitrogen 0 N m 7 h Used to purge system for start-up and shutdown 5 Electricity 21 kW Compressors If electrical power is ignored, the energy efficiency of conversion of the methane to hydrogen is 80%, based on the higher heating values of hydrogen and methane. The overall energy efficiency O7) is calculated to be 69%, based on the electrical power consumed and higher heating values of methane and hydrogen. HHVH2 (134kmol I h)(2S6MJ I kmol) _ Q 6 9 . V~ HHVCH4 + power ~ (0.536kmol / h)(S9lMJ / kmol) + (Q.021MJ/s)(3600s/h) ~ Table A6.1.5 (Appendix 6.1) summarizes the process losses. Waste heat from the compressors and condensation of the reactor off-gas accounts for the bulk of the thermal inefficiency. It is difficult to recover energy from these low-grade sources within the process, though most of the rejected heat could be captured for domestic hot water and space heating. 6.3.3 R e a c t o r L a y o u t Experience gained in design and fabrication of the ICFBMR pilot reactor (Chapter 4) showed that the layout of the reactor internals poses a number of mechanical challenges. One of the biggest issues was the numerous membrane tubes and instruments that had to be neatly fixed within the reactor and sealed at the reactor shell. Minimizing the number of vessel connections is clearly important. 162 Another consideration is the size of commercially available palladium foil. It is difficult for suppliers to cold-roll wide strips of very thin palladium. Foils of area 6" x 12" (0.15 x 0.30 m) are available down to 25 urn thickness. Foils of 3" x 12" (0.076 x 0.30 m) are also now available down to 15 urn. In order to minimize wastage of palladium, membrane panels based on a foil widthof 3" (0.076 m) or 6" (0.15 m) are likely to be preferred. The base case simulation assumes a superficial core velocity of 0.25 m/s, from which the core flow area is then calculated to be 0.0068 m2. Figure 6.11 presents one possible membrane layout, based on a nearly-square core draft box and five staggered membrane panels of the same dimensions as in the ICFBMR pilot reactor (83 mm wide, 6 mm thick). In order to reduce the core box dimensions, the space between the membrane panels was set at 13 mm, slightly smaller than the panel spacing in the pilot reactor and cold model (16 mm). The core box can fit within a 6" (0.15 m) reactor shell. The area available in the core for flow of this layout is 0.0076 m2, corresponding to a superficial core velocity of 0.22 m/s. 6" shell (0.15 m)-^ v All dimensions in mm Membrane panel Core box Figure 6.11: Plan view of option 1 layout - 30 Nm3/h reactor (Ucore = 0.22 m/s) 64 membrane panels would be required to match the membrane area calculated for the base case (2.34 m2) for the Option 1 layout. This would require 13 membrane panel stacks, making the core box almost 4 m tall, an impractical dimension. Thinner membranes would reduce the membrane area, reducing the reactor height. However, basing a design on 50 urn (or 25 um membranes working at 163 50% efficiency) is currently more realistic. Alternatively, the superficial gas velocity in the core could be reduced from the original specification of 0.25 m/s. As the solids circulation flux varies approximately linearly with superficial velocity, decreasing the core velocity will not likely penalize the design significantly. An alternative reactor layout is presented in Figure 6.12. In this layout, 6% " x 12" (0.16 x 0.30 m) membrane panels are installed in a staggered pattern, 13 mm apart within the central zone of the reactor. The area open to flow in the core in this layout is 0.014 m2, corresponding to a core velocity of 0.12 m/s. The circulating solids travel in two opposing segments rather than in an annular region. 30 membrane panels in six stacks provide the membrane area. The membrane area can be accommodated in an 8" (0.20 m) diameter, ~2.6 m tall reactor (Figure 6.13). This layout has a lower proportion of the area for return of solids than in the pilot plant reactor, and additional cold model testing would be required to verify that sufficient solids circulation could be achieved with this layout. Figure 6.13 shows the permeate H 2 connecting through the top flange, but the side flange configuration presented in Figure 4.21 (Chapter 4) could also be used. 164 Off-gas CH 4 /H 2 0 Wind box i L Filter Disengagement zone Settled catalyst bed Oxidant distributor Core box Membrane panels (30) 8" (0.31 m) shell Figure 6.13: Elevation drawing for Option 2-30 NmVh reactor (Ucore = 0.12 m/s) As the reactor is not heated externally, conventional stainless steel can be used for the reactor shell. In order to accelerate heat-up of the reactor, several small tubular heaters could be installed in both solids return zones. 165 6.3.4 Economics A simplified economic analysis was performed to estimate costs to build and operate the 30 NmVh ICFBMR unit. It should be noted that the values presented here are not optimized and are for a one-off system. Costs to produce multiple units of a production-type design would be significantly less than values presented, as engineering, capital and fabrication costs could be reduced and amortized over many units. However, the economic analysis gives a starting point for comparison of the ICFBMR technology against other emerging small-scale H 2 technologies. Unless otherwise noted, all costs are in 2006 Canadian dollars. Equipment Costs A simplified piping and instrumentation drawing (PID) is presented in Figures A6.2.1 and A6.2.2 of Appendix 6.2. An equipment list, which includes nominal sizing, materials of construction and estimated capital costs, is presented in Table A6.2.1, Appendix 6.2. In Chapter 2, the cost to produce the 50 urn thick Pd7sAg25 membranes used in the ICFBMR pilot plant was estimated at $40,700/m2. Given the larger number of membranes (30), a larger palladium area and thinner Pd material (25 um) compared to the pilot ICFBMR, the capital cost estimate for the 30 NmVh design assumes a fabricated membrane cost of $10,000/m2 CAD. Note that the palladium value in the 25 um membranes is approximately $2,000/m2 at the average Pd price for the last five years ($279US/oz, average 2002-2007). This implies that the fabricated membrane cost should be relatively insensitive to small variations in the spot palladium market. Bare equipment cost is estimated at $220,000 CAD, the breakdown of which is presented in Figure 6.14. Membranes represent -10% of the estimated bare equipment cost. The three gas compressors are the largest component of the equipment capital. By applying a Lang factor of 3.5 to the bare equipment cost, the total capital cost to supply the first 30 NmVh ICFBMR unit is estimated to be $770,000. Use of a simple Lang factor is not appropriate for estimating costs of multiple production units, as it overestimates engineering and installation costs, but can be justified for the first ICFBMR skid. 166 Burner Miscellaneous Reactor (excluding membranes) Membranes Figure 6.14: Estimated 30 Nm /h bare equipment cost components (values listed in Table A6.2.1) Operating Costs The list of assumptions (Table A6.2.2, Appendix 6.2) of the economic analysis of the 30 NmVh base case includes: • 8,000 hours of operation/year, at an average production rate of 85%. • The membranes are replaced annually. No credit is given for recycling palladium, as this represents only a portion (likely less than 10%) of the membrane replacement costs. • Catalyst and desulfurizer adsorbent are replaced annually. • Inflation rate of 2% applied to consumables and utilities. • Annual maintenance costs are estimated at 5% of bare equipment cost, plus an additional $5,000 for each membrane changeover. • Other miscellaneous expenses, such as insurance and annual taxes, are assumed to be 5% of bare equipment cost. Based on the above assumptions, the variable operating costs are estimated to be $5.00/kg of Fb, the breakdown of which is presented in Figure 6.15 (see data in Table A6.2.3, Appendix 6.2). When gas and electricity are excluded, the variable operating costs are estimated to be $2.88/kg H 2 . Membrane replacement represents the largest operating cost component, slightly larger than the cost of natural gas. Clearly there is a significant economic incentive to reduce operating costs by extending membrane service life and reducing fabricated membrane costs. 167 Natural gas (base case, values listed in Table A6.2.3) Cost Recovery A cash flow analysis was performed to calculate the hydrogen selling price required to recover the capital and operating costs. Assumptions included: • Cost of capital 10%. • Capital depreciated (double-declining) over the 10-year service life. • Margin tax rate of 38%. • No scrap value for equipment. The cash flow analysis is presented in Tables A6.2.3 and A6.2.4, Appendix 6.2. Capital recovery was estimated at $7.00/kg H 2 , which when combined with operating costs, gave a selling price of $ 12.7/kg H 2 . Given the one-off nature of this design, it is not surprising that the break-even H 2 price is very high (see Figure 1.2 in Chapter 1 for current H 2 delivery costs). Government subsidies or other financial incentives would likely be required to make the first ICFBMR unit economically viable. An alternate economic analysis was performed to compare the ICFBMR technology with conventional small-scale reformers on a similar footing. Costs for the 30 NmVh ICFBMR unit where compared to estimates for a 222 NmVh small-scale conventional SMR filling station using the same cost scaling parameters from a recent hydrogen study (Table E-35 in "The Hydrogen Economy", National Academy of Engineering, 2004). The basis and output from this analysis are presented in Appendix 6.3. Using the NAE methodology, the cost of hydrogen production, which included capital recovery, was calculated to be $4.12 and $6.73/kg H 2 for small-scale conventional SMR and the 30 NmVh ICFBMR 168 systems respectively. This estimated cost for the ICFBMR system is approximately half that estimated using the NPV analysis presented above, with the reduction mostly a result of aggressive capital cost factors appropriate for a multiple production unit. A portion of the estimated price difference between the conventional SMR (222 Nm3/h) and ICFBMR (30 Nm3/h) systems is due to relatively higher capital costs of a lower capacity system. However, increased variable costs, mainly attributable to annual membrane replacement, represent the main extra expense of the ICFBMR system. It should be noted that the analysis assumes that the ICFBMR uses approximately 25% less natural gas than the conventional SMR, but this saving is more than offset by membrane replacement costs. 6.4 Discussion The membrane flux characteristics (i.e. foil thickness) and the reactor operating conditions determine the required membrane area. The membrane area is a key design parameter of the ICFBMR configuration, as this controls reactor sizing and strongly influences both capital and operating costs. Simulation shows that high reactor pressure (> -2,500 kPa) and low permeate pressure (< -100 kPa) are the major reactor parameters that can be used to reduce membrane requirements. Increasing the reactor temperature significantly reduces membrane area, but operating conditions are likely to be governed by maximum allowable membrane temperatures (<600°C). There is a clear incentive to develop high-flux (i.e. thin) membranes that can operate at higher temperatures and larger pressure differentials. Modelling suggests that the ICFBMR design is relatively tolerant to variations in feed preheat temperatures and solids circulation. The amount of nitrogen internally circulated within the ICFBMR does not significantly influence membrane requirements, indicating that there is little economic benefit of using oxygen in place of oxidation air. This finding also tends to discount the financial merits of incorporating oxygen-permeable membranes within the reactor, unless they were cheaper to operate than an air compressor. However, if the non-permeate reactor off-gas was to be treated to remove C0 2 for sequestration, oxygen addition might be attractive. The 30 NmVh flowsheet predicts that relatively high system efficiencies (-69%, based on HHV) can likely be achieved for small-scale H 2 production with the ICFBMR design. As the required membrane area and superficial gas velocity tend to make the reactor tall and narrow, minimization of the membrane area is needed to obtain a practical design. In order to reduce the reactor diameter, it may 169 be desirable to reduce the downflow (annular) area of the ICFBMR. Study of alternate reactor membrane layouts may be useful to determine the effect on solids circulation. Table 6.5 presents several figures of merit for the 30 Nm3/h ICFBMR unit compared to a large-scale and small-scale conventional SMR system. The ICFBMR system efficiency compares favourably with conventional SMR systems. The relative reactor volume of the ICFBMR is significantly larger than for a conventional SMR primary reformer. However for a fairer comparison, the volume of the shift reactors and PSA should also be included in the conventional SMR values. Table 6.5: Figures of merit for 30 Nm3/h H 2 ICFBMR and conventional SMR production units Figure of merit Large-scale conventional SMR Small-scale conventional SMR (NAE, 2004) Proposed 30 NmVh ICFBMR design H2:CH4 molar ratio 2.5 1.91 2.48 Overall thermal efficiency 70-80% (steam export) -60% (no steam export) 56% (incl. power, no steam export) 80% excluding power 69% including power Catalyst productivity -3,000 NmVh H 2/m 3 (primary reformer only) 500 NmVh H2/m3 Reactor productivity -1,000 Nm J/hH 2/m J (primary reformer furnace only) 375 NmVh H2/mJ Membrane H 2 productivity N/A N/A 13 NmVh m2 Estimated H 2 cost with capital recovery ($CAD) $1.62/kg (267,000 NmVh, Table E-19, NAE, 2004) $4.12/kg (222 NmVh, see Appendix 6.3) $6.73/kg (30 NmVh, see Appendix 6.3) The economic analyses clearly demonstrate that the capital and operating costs of membranes are significant. Reducing the initial capital cost of the membranes is important, but membrane longevity may be a more important cost factor. If membranes can be shown to operate for two or more years, the ICFBMR unit economics improve significantly. Finally, it should be noted that the hydrogen produced in the ICFBMR has the potential to be of higher purity than H 2 produced from a conventional SMR system. A cost premium for the membrane hydrogen may be warranted. 170 Chapter 7. Conclusions and Recommendations 7.1 Conclusions The main objective of this study was to evaluate the ICFBMR concept for production of high-purity H 2 to serve the emerging distributed H 2 market. Although the results from the pilot plant work were somewhat below expectations due to catalyst performance and the limited installed membrane area, the potential for the ICFBMR to supply fuel-cell quality H 2 from a small-scale ATR process has been demonstrated. Some specific achievements and conclusions from this study: • Double-sided planar membranes using 50-um thick palladium alloy foil (25% silver) were successfully prepared using proprietary techniques of Membrane Reactor Technologies Ltd (MRT). Sievert's Law could be used to characterize the panel membrane flux. The calculated active permeation area of the membrane panels was a fraction (from VA to VA) of the panel permeation geometry. Post-run analysis suggest that the reduction in permeation area was likely due to inter-diffusion of the palladium foil with the stainless steel substrate. MRT has subsequently improved their process for depositing the inter-diffusion barrier. • There were no significant membrane failures during the 14 days of hot operation in the fluidized bed pilot reactor. High-purity hydrogen was produced (>99.999% H 2 , excluding N2) from the in-situ membranes for the first 180 hours of pilot testing. Although several small leaks developed, all of the membranes survived the reactor testing intact (up to 650°C with no protective covers), and relatively high quality hydrogen (-99.8% dry basis, excluding N2) was produced for the remainder of the pilot testing. Traces of steam were also likely present in the permeate H 2 at the end of the testing, but this was not measured. No decrease in permeation was seen in the membranes, despite heavy deposits of catalyst on the membrane surface. • Catalyst solids were proven to internally circulate up the reformer core and down the annulus in both the cold model and hot pilot reactor. Results from cold model testing indicate that the solids flux increases proportionally to the main gas feed flow the core to a superficial velocity (Ucore) of -0.3 m/s, and that the solids flux levelled off at superficial velocities above -0.5 m/s. It is postulated that this levelling was due to a gradual transition from bubbling to turbulent 171 fluidization. The addition of a small amount of secondary air to the bottom of the annulus increased the solids circulation significantly. Cold model testing showed that communication slots are required between the flow channels in the membrane core in order to prevent maldistribution of gas and solids. Solids flux correlations obtained from cold modelling closely matched the measured solids flux in the pilot reactor, providing some confidence that dimensionless scaling of cold model data to reformer conditions is valid. The estimate of the effective membrane area during the pilot reforming tests with no N2-sweep was close to the actual permeation area. This suggests that the "effectiveness" of the membrane area was close to 100%. The N 2 sweep enhanced H 2 permeation, but was not as fully effective as predicted from modelling, The reformer was successfully operated under both SMR (external heating only) and ATR (direct air addition) conditions. The reactor operation was very stable. Six membrane panels were installed in the pilot reactor, less than half the full complement, and as a result, the reactor performance was very much flux-limited. The maximum measured permeate hydrogen production was 1.06 NmVh at 589°C. The highest measured H 2 recovery was 1.17 Nm3/Nm3 of natural gas feed. The reactor was operated up to 650°C, but low catalyst activity hindered reactor performance. As expected, H 2 production increased with increasing temperature and lower permeate H 2 partial pressure. The overall reactor performance could be adequately simulated with a simple equilibrium model coupled with a permeation equation. Helium tracer studies in both the cold model and hot pilot reformer indicated that only a portion of the gas in the upper reactor (-10%) is circulated to the reactor core with the returning solids. This ensures that only a portion of the nitrogen in the oxidation air reaches the permeation zone. This was confirmed in ATR testing, where it was found that air addition to the top of the reactor had little effect on the hydrogen flux, indicating that the ICFBMR system requires less membrane area than a well-mixed ATR reactor using air. It also suggests that the ICFBMR system operating with air can have similar membrane area as ATR reactors using oxygen or oxygen-permeable membranes. The deactivation of the NiO on alumina SMR catalyst initially used in the pilot reactor was not unexpected, as this material is known to be intolerant of the cyclical oxidizing conditions present 172 when air is fed to the reformer. However, the poor performance of the novel ATR was a disappointment, and curtailed the pilot plant program. The cause of the low ATR catalyst activity is not known, as there was a non-analysis agreement in place for this material. • A commercial simulation program (HYSYS) was used to create a model of the ICFBMR based on a series of equilibrium reactors. The effect of the main process variables on reactor performance was investigated. When coupled to cold model solids flux correlations, the model shows that the ICFBMR geometry permits enough solids circulation to match the thermal demand of the reforming reaction with direct air addition (ATR). • Simulation of a 30 NmVh H 2 ICFBMR production system indicated that high thermal efficiencies could potentially be achieved. Economic analysis suggests membrane cost and membrane longevity are key areas for the viability of this process. 7.2 Alternative Reactor Applications Membrane reactors tend to be flux-limited, and thus the required membrane area governs the reactor geometry. This results in a larger solids volume than required for catalysis. Higher flux membranes would reduce the reactor volume, lowering reactor, metal and catalyst costs. Distributed reforming using a natural gas feedstock may not meet long-term environmental goals. However, it is likely that novel autothermal reforming technology, of which ICFBMR is an example, will play a role in the transition to a hydrogen economy. In addition, this technology can likely be applied to alternative methane sources that are "greener" than natural gas, such as anaerobic digester gas. It is also possible that higher hydrocarbons (C3+) can be successfully reformed in this reactor configuration. The ICFBMR continually moves catalyst between reforming and oxidation zones, which may alleviate coking problems that can occur when reforming higher hydrocarbons in conventional fixed bed reactors. Another possible variation on the ICFBMR reformer is sorption-enhanced reforming (SER), where calcium-based solids, such as limestone or dolomite, are used to capture C 0 2 from the reforming zone. Recent experimental work at UBC (Johnsen et al., 2006) showed experimentally the potential for SER to be carried out in a bubbling bed reactor to produce relatively high-purity hydrogen. This concept could be extended to a fluidized bed membrane reformer, where both C 0 2 and H 2 removal from the reforming zone would shift the thermodynamic equilibrium. C 0 2 could then be separately removed from the calcium solids in a separate calciner or calcining zone. A schematic of a possible 173 configuration is presented in Figure 7.1. Absorption of C 0 2 is exothermic, which would help supply most or some of the reforming heat. Heat, and perhaps stripping steam, would be added to the calciner to desorb C0 2 . The bed solids, a mix of SMR catalyst and C02-sorbent, is continually circulated between the beds. A C02-rich stream is produced for sequestration. A similar configuration using an external solids circulation loop was proposed by Prasad and Elnashaie (2004). This SER concept highlights the unique ability of fluidized beds to move heat and solids from separate reaction zones. Calciner (C02 desorption) | Heat -|CQ2/Stearrj> -| Off-gas > / P " CaC03, • Catalyst 1 1 i , , , H 2 f • Membranes ' llllilli CaO, Catalyst Bed surface Reformer (SMR reaction, H2 separation and C02 absorption) H2 > j Steam ;> |CH4/Steam> Figure 7.1: Sorption-enhanced ICFBMR concept In this study, the ICFBMR has been applied to high-purity hydrogen production from steam methane reforming. One could also envision applying this fluidized bed membrane system to a number of other high-temperature catalytic processes, such as propane dehydrogenation, where hydrogen needs to be removed and heat transfer is important. 7.3 Recommendations The follow areas merit further study: • Reactor Geometry: 174 o Additional cold modelling to determine whether reactor sizing can be reduced (space-time yield increased) by using different layouts (e.g. 2 chord returns instead of an annulus). o Optimization of feed distributors and catalyst return geometry, o Side-mounted membrane flanges for ease of reactor assembly. o Investigate the effect of other reactor geometries on the rate of solids circulation, including reactor height and membrane spacing. • Membranes: o Determine longevity of Pd membranes in a fluidized bed environment at varying temperature and fluidization velocities (both covered and uncovered). o Find long-term flux stability of membranes to ensure that palladium - substrate inter-diffusion does not occur. o Investigate whether membrane efficiency remains high when H 2 recovery is increased to economically practical values. • Catalysis: o Long-term testing of a fluidizable ATR catalyst. • C0 2 sequestration: o Evaluate the application of ICFBMR to sorbent enhanced reforming. Finally, it is re-emphasized that membranes remain the key to the successful development of any membrane reactor. Although it is believed that the ICFBMR concept greatly facilitates the use of H 2 -permeable membranes, without development of robust, high-flux, cost-effective membranes that can tolerate the high temperatures and pressures in the reforming process, the ICFBMR concept will fail to compete with conventional reforming processes. 175 References Aasberg-Petersen, K., Nielsen, C.S., Jorgensen, S.L., "Membrane reforming for hydrogen", Catalysis Today, Vol. 46, 193-201 (1998). Abashar, M.E.E., Alhumaizi, K.I., Adris, A.M., "Investigation of methane-steam reforming in fluidized bed membrane reactors", Trans. IChemE, Vol. 81, Part A, No. 11, 251-258 (2003). Abba, I.A., "A generalized fluidized fed reactor model across the flow regimes", Ph.D. Thesis, University of British Columbia, Vancouver, Canada (2001). Abba, I.A., Grace, J.R., Bi, H.T., "Application of the generic fluidized-bed reactor model to the fluidized-bed membrane reactor process for steam methane reforming with oxygen input", Industrial and Engineering Chemistry Research, Vol. 42, No. 12, 2736-2745 (2003). Adris, A.M., "Fluidized-bed membrane reactor for steam-methane reforming: Experimental verification and model validation", Ph.D. Thesis, University of British Columbia, Vancouver (1994). Adris, A.M., Grace, J.R., Lim. C.J., Elnashaie, S.S., "Fluidized bed reaction system for stream/hydrocarbon gas reforming to produce hydrogen", US Patent 5,326,550 (1994). Adris, A.M., Pruden, B.B., Grace, J.R., Lim, C.J., "On the reported attempts to radically improve the performance of the steam methane reforming reactor", Canadian Journal of Chemical Engineering, Vol. 74, 177-186(1996). Adris, A.M., Grace, J.R., "Characteristics of fluidized-bed membrane reactors: scale-up and practical issues", Industrial and Engineering Chemistry Research, Vol. 36, No. 11, 4549-4556 (1997a). Adris, A.M., Lim, C.J., Grace, J.R., "Fluidized-bed membrane reactor for steam methane reforming: Model verification and parametric study", Chemical Engineering Science, Vol. 52, No. 10, 1609-1622 (1997b). Adris, A.M., Grace, J.R., Lim, C.J., Elnashaie, S.S.E., "Fluidized bed reaction system for steam/hydrocarbon gas reforming to produce hydrogen", Canadian patent CA 2,081,170 (2002). Amandusson, H., Ekedahl, L.-G., Dannetun, H., "Hydrogen permeation through surface modified Pd and PdAg membranes", Journal of Membrane Science, Vol. 193, 35-47 (2001). Armor, J.N, "The multiple roles for catalysis in the production of H 2", Applied Catalysis A: General, Vol. 176, No. 2, 159-176(1999). Armor, J.N, "Catalysis and the hydrogen economy", Catalysis Letters, Vol. 101, No. 3-4, 131-135 (2005). Ayabe, S., Omoto, H., Utaka, T., Kikuchi, R. Sasaki, K., Teraoaka, Y., Eguchi, K., "Catalytic autothermal reforming of methane and propane over supported metal catalysts", Applied Catalysis A: General, Vol. 241 261-269 (2003). Balachandran, U., Kleefisch, M.S., Kobylinski, T.P., Morissette, S.L., Pei, S., "Oxygen ion-conducting dense membrane ceramic membranes", US Patent 5,639,437, assigned to Amoco (1997). 176 Benes, N.E., Verweij, H., "Comparison of macro- and microscopic theories describing multicomponent mass transport in microporous media", Langmuir, Vol. 15, No. 23, 8292-8299 (1999). Bi, H.T., Ellis, N., Abba, A., Grace, J.R., "A state-of-the-art review of gas-solid turbulent fluidization", Chem. Eng. Sci., Vol. 55, No. 1, 1-37 (2000). Biegler, L. T., Jiang, L., Fox, V.G., "Recent advances in simulation and optimal design of pressure swing adsorption systems", Separation and Purification Reviews, Vol. 33, No. 1, 1-39 (2004). Bolthrunis, CO. , Silverman, R.W., Ferrari, D., "The rocky road to commercialization: Breakthroughs and challenges in the commercialization of fluidized bed reactors", Fluidization XI, eds. Arena, U., Chirone, R., Miccio M., Salatino, P., Engineering Conferences International, NY, (2004). Boyd, T., Grace, J., Lim, C.J., Adris, A.M., "Hydrogen from an internally circulating fluidized bed membrane reactor", International Journal of Chemical Reactor Engineering, Vol. 3, A58 (2005). Buxbaum, R.E., Kinney, A.B., "Hydrogen transport through tubular membranes of palladium-coated tantalum and niobium", Industrial and Engineering Chemistry Research, Vol. 32, No.2, 530-537 (1996). Chen, X., Honda, K., Zhang, Z., "A comprehensive comparison of CH4-CO2 reforming activities of NiO/Al 20 3 catalysts under fixed- and fluidized-bed operations", Applied Catalysis A: General, Vol. 288, No. 1-2, 86-97, (2005). Chen, Z., Elnashaie, S.S.E.H., "Autothermal CFB membrane reformer for hydrogen production from heptane", Chemical Engineering Research and Design, Vol. 83, No. 7, 893-899 (2005). Chu, C.Y., Hwang, S.J., "Attrition and sulfation of calcium sorbent and solids circulation rate in an internally circulating fluidized bed", Powder Technology, Vol. 127, 185-195 (2002). Corella, J., Orio, A., Aznar, P., "Biomass gasification with air in fluidized bed: Reforming of the gas composition with commercial steam reforming catalysts", Industrial & Engineering Chemistry Research, Vol. 37, No. 12, 4617-4624 (1998). Coronas, J., Santamaria, J., "Catalytic reactors based on porous ceramic membranes", Catalysis Today, Vol. 51, 377-389 (1999). Cromarty, B.J, Hooper, C.W., "Increasing the throughput of an existing hydrogen plant", International Journal of Hydrogen Energy, Vol. 22, No. 1, 17-22 (1997). Davidson, J.F., Clift, R., Harrison, D., "Fluidization", 2nd edition, Academic Press, London (1985). Deshmukh, S.A.R.K., "Membrane assisted fluidized bed membrane reactor: experimental demonstration for partial oxidation of methanol", Ph.D. Thesis, University of Twente, Netherlands (2004). Deshmukh, S.A.R.K., Laverman, A.H.G., van Sint Annaland, M., Kuipers, J.A.M., "Development of a membrane-assisted fluidized bed membrane reactor. 1. Gas phase back-mixing and bubble-to-emulsion phase mass transfer using tracer injection and ultrasound experiments", Industrial and Engineering Chemistry Research, Vol. 44, No. 16, 5955-5965 (2005). 177 Deshmukh, S.A.R.K., Heinrich, S., Mori, L., van Sint Annaland, M., Kuipers, J.A.M., "Membrane assisted fluidized bed reactors: Potential and hurdles", Chemical Engineering Science, Vol. 62, 416-436 (2007). Dogan, M., Posarac, D., Grace, J., Adris, A.M., Lim, C.J., "Modeling of autothermal steam methane reforming in a fluidized bed membrane reactor", International Journal of Chemical Reaction Engineering, Vol. 1, Article A2 (2003). Dixon, A.G., "Recent research in catalytic inorganic membrane reactors", International Journal of Chemical Reactor Engineering, 1, Review R6. Dyer, P.N., Richards, R.E., Russek, S.L., Taylor, D.M., "Ion transport membrane technology for oxygen separation and syngas production", Solid State Ionics, Vol. 134, 21-33 (2000). Edlund, D., "Hydrogen separation and purification using dense metallic membranes", IdaTech Corp., presented at DOE Hydrogen Separations Workshop (2004). Edlund, D., "A versatile, low-cost, and compact flue processor for low-temperature fuel cells", IdaTech Corp. website: (1999). Ellis, N., "Hydrodynamics of Gas-Solid Turbulent Fluidized Beds", Ph.D. Thesis, University of British Columbia, Vancouver, Canada (2003). Ewan, B.C.R., Allen, R.W.K, "A figure of merit assessment of the routes to hydrogen", International Journal of Hydrogen Energy, Vol. 30, No.8, 809-819 (2005). Galuszka, J., Fouda, S., Pandey, R., Ahmed, S., "Process for producing syngas and hydrogen from natural gas using a membrane reactor", US Patent 5,637,259, assigned to Natural Resources Canada (1997). Gavalas, G.R., Megiris, C.E., Nam, S.W., "Deposition on H2-permselective Si0 2 films", Chem. Eng. Sci., Vol. 44, No. 9, 1829-1835 (1989). Glicksman, L.R., "Scaling relationships for fluidized beds", Chemical Engineering Science, Vol. 39, No. 9, 1373-1379(1984). Glicksman, L.R., Hyre, M., Woloshun, K., "Simplified scaling relationships for fluidized beds", Powder Technology, Vol. 77 (2), 177-199 (1993). Grace, J.R., "Fluidized bed hydrodynamics", Chap 8.1 in Handbook of Multiphase Systems, ed. Hetsroni, G., Hemisphere, Washington (1982). Grace, J.R., "Contacting modes and behaviour classification of gas-solid and other two-phase suspensions", Canadian Journal of Chemical Engineering, Vol. 64, 353-362 (1986). Grace, J.R., Elnashaie, S.S.E.H., Lim, C.J., "Hydrogen production in fluidized beds with in-situ membranes", International Journal of Chemical Reactor Engineering, Vol. 3, A41 (2005). Grace, J.R., Lim, C.J., Adris, A.M., Xie, D., Boyd, D.A., Wolfs, W . M , Brereton, C.M.H., "Internally circulating fluidized bed membrane reactor system", US Patent 7,142,231, assigned to Membrane Reactor Technologies Ltd. (2006). 178 Gryaznov, V., "Metal containing membranes for the production of ultrapure hydrogen and the recovery of hydrogen isotopes", Separation and Purification Methods, Vol. 29, No. 2, 171-187 (2000). Haider, A., Levenspiel, O., "Drag coefficient and terminal velocity of spherical and nonspherical particles", Powder Technology, Vol. 58, No. 1, 63-70 (1989). He, Y.L., Lim, J.L., Grace, J.R., "Scale-up studies of spouted beds", Chemical Engineering Science, Vol. 52, No. 2, 329-339 (1997). Heydorn, B., Schwendener, H., Mori, S., "Hydrogen", Chemical Economics Handbook, SRI International, Menlo Park, USA (1994). Holleck, G.L., "Diffusion and solubility of hydrogen in palladium and palladium-silver alloys", Journal of Physical Chemistry, Vol. 74, No. 3, 503-511 (1970). Horio, M., Nonaka, A., Sawa, Y., Muchi, I., "A new similarity rule for fluidized bed scale-up", AIChE Journal, Vol. 150, No. 9, 1466-1481 (1986). Hurlbert, R.C, Konecny, J.O., "Diffusion of hydrogen through palladium", Journal of Chemical Physics, Vol. 34, No.2, 655-658 (1961). Hseih, H.P., Bhave, R.R., Fleming, L.L, "Microporous alumina membranes", Journal of Membrane Science, Vol. 39, Issue 3, 221-241 (1988). Imai, H., Morimoto, H., Tominaga, A., Hirashima, H., "Structural changes in sol-gel derived Si0 2 and Ti02 films by exposure to water vapour", "Journal of Sol-Gel Science and Technology, Vol. 10, No.l, 45-54 (1997). Islam, M.T., "Palladium coated high-flux tubular membranes for hydrogen separation at high temperatures and differential pressures", M.Sc. Thesis, University of Calgary, Canada (1997). Jarosch, K., de Lasa, H.I., "Novel riser simulator for methane reforming using high temperature membranes", Chemical Engineering Science, Vol. 54, No. 10, 1455-1460 (1999). Jarosch, K., de Lasa, H.I., "Permeability, selectivity, and testing of hydrogen diffusion membranes suitable for use in steam reforming" Industrial and Engineering Chemistry Research, Vol. 40, No. 23, 5391-5397 (2001). Johnsen, K., Ryu, H.J., Grace, J.R., Lim, C.J., "Sorption-enhanced steam reforming of methane in a fluidized bed reactor with dolomite as C02-acceptor", Chemical Engineering Science, Vol. 61, 1195-1202 (2006). Juda, W., Krueger, C. W., Bombard, R. T., "Diffusion-bonded palladium-copper alloy framed membrane for pure hydrogen generators and the like and method of preparing the same", US patent 5,904,754, assigned to Walter Juda Associates (1999). Judd, M.R., Dixon, P.D., AIChE Symp. Ser. 74 (176) 38 (1978). Jung, D.S., Ron, S.A., Kim, S.D., Guy, C , "Regeneration characteristics of hydrodesulfurization catalyst by combustion in an internally circulating fluidized bed", Fluidization XI, eds. Arena, U., Chirone, R., Miccio M., Salatino, P., Engineering Conferences International, NY, (2004). 179 Kast, W., Hohenthanner, C.R., "Mass transfer within the gas-phase of porous media", International Journal of Heat and Mass Transfer, Vol. 43, 807-823 (2000). Kelkar, V.V., Ng, K.A., "Development of fluidized catalytic reactors: Screening and scale-up", AIChE J., Vol. 48, No. 7, 1498-1518 (2002). Klassen, A., Membrane Reactor Technologies Ltd., personal communication (2005). Knapton, A.G., "Palladium alloys for hydrogen diffusion membranes - A review of high permeability materials", Plat. Met. Rev., Vol. 21, 44-50 (1977). Knowlton, T.M., Karri, S.B.R., Issangya, A., "Scale up of fluidized-bed hydrodynamics", Powder Technology, Vol. 150, No. 2, 72-77 (2005). Kunii, D., Levenspiel, O., "Fluidization Engineering", 2nd edition, Butterworth-Heinemann, Newton, MA (1991). Kurungot, S., Yamaguchi, T, "Stability of Rh-Y-Al203 catalyst layer by ceria doping for steam reforming in an integrated catalytic membrane reactor system", Catalyst Letters, Vol. 92, No. 3-4, 181-187(2004). Lewis, F.A., "The palladium hydrogen system", Academic Press, London (1967). Li, A., Membrane Reactor Technologies Ltd., personal communication (2003). Li, A., Membrane Reactor Technologies Ltd., personal communication (2005a). Li, A., "Diffusion bonding for metallic membrane joining with metallic module", US Patent Application, 20050109821 (2005b). Lin, Y.S., "Microporous and dense inorganic membranes: current status and prospects", Separation and Purification Technology, Vol., 25, 39-55 (2001). Lin, Y.S., Kumakiri, I., Nair, B.N., Alsyouri, H., "Microporous inorganic membranes", Separation and Purification Methods, Vol. 31, No. 2, 229-379 (2002). Marschall, K.J., Mleczko, L., "CFD modeling of an internally circulating fluidized-bed reactor", Chemical Engineering Science, 51, 2085-2093 (1999). Matsen, J.M., "Design and scale-up of CFB catalytic reactors" in "Circulating Fluidized Beds", Grace, J.R., Avidan, A.A., Knowlton, T.M., (Eds.), Chapman and Hall, London, 489-503 (1997). Milne, B.J., Berruti, F., Behre, L.A., de Bruijn, T.J.W., "The internally circulating fluidized bed (ICFB): A novel solution to gas bypassing in spouted beds", Canadian Journal of Chemical Engineering, Vol. 70, 910-915 (1992). Minet, R.G.; Vasileiadis, S.P.; Tsotsis, T.T., "Experimental studies of a ceramic membrane reactor for the steam/methane reaction at moderate temperatures (400-700°C)", Preprints - Division of Petroleum Chemistry, American Chemical Society, Vol. 37, No. 1, 245-251 (1992). Mleczko, L., Ostrowski, T., Wurzel, T., "A fluidized bed membrane reactor for the catalytic partial oxidation of methane to synthesis gas", Chemical Engineering Science, Vol. 51, No. 11, 3187-3192 (1996). 180 Morreale, B.D., Ciocco, M.V., Enick, R.M., Morsi, B.L, Howard, B.H., Cugini, A.V., Rothenberger, K.S., "The permeability of hydrogen in bulk palladium at elevated temperatures and pressures", Journal of Membrane Science, Vol. 212, 87-97 (2003). Morreale, B., Chen, L., Killmeyer, R.P., Dorris, S.E., Lee, T.H.; Balachandran, U. "Dense cermet membranes for hydrogen separation" 2005 AIChE Spring National Meeting, Conference Proceedings, 2005 AIChE Spring National Meeting, Conference Proceedings, p. 1645 (2005). Mukainakano, Y., Li, B., Kado, S., Miyazawa, T., Okumura, K., Miyao, T., Naito, S., Kunimori, K., Tomishage, K., " Surface modification of Ni catalysts with trace Pd and Rh for oxidative steam reforming of methane", Applied Catalysis A, 318, 252-264 (2007). Nam, S.W., Gavalas, G.R "Stability of H2-permselective Si0 2 films formed by chemical vapour deposition", AIChE Symp. Series, 85 (268), 68-74 (1989). Namkung, W., Guy, C , Boisselle, F., Legros, R., "Effect of temperature on gas bypassing and solids circulation rate in an internally circulating fluidized bed", Canadian Journal of Chemical Engineering, Vol. 78, 1025- 1031 (2000). National Academy of Engineering, "The hydrogen economy: opportunities, costs, barriers, and R&D needs", National Academies Press (2004). Natural Resources Canada, "Canadian hydrogen survey - 2004/2005", prepared by Dalcor Consultants Ltd and Camford Information Services Inc. (2005). Nijmeijer, A., "Hydrogen-selective silica membranes for use in membrane steam reforming", Ph.D. thesis, University of Twente, Netherlands (1999). Ogden, J.M., "Developing an infrastructure for hydrogen vehicles: a Southern California case study", International Journal of Hydrogen Energy, Vol. 24, Issue 8, 709-730 (1999). Ogden, J.M., "Review of small stationary reformers for hydrogen production", Center for Energy and Environmental Studies, Princeton University (2001). Ohashi, H., Ohya, H., Aihara, M., Youichi, N., Semenova, S., "Hydrogen production from hydrogen sulphide using membrane reactor integrated with porous membrane having thermal and corrosion resistance", Journal of Membrane Science, Vol. 146, 39-52 (1998). Onstot, W.J., Minet, R.G., Tsotsis, T.T., "Design aspects of membrane reactors for dry reforming of methane for the production of hydrogen", Industrial and Engineering Chemistry Research, Vol. 40; No. 1,242-251 (2001). Paglieri, S.N., Way, J.D., "Innovations in palladium membrane research", Separation and Purification Methods, Vol. 31, No. 1, 1-169 (2002). Pal, V., Singh, M., Gupta, B., "Analysis of thermal expansion coefficients under the effect of high temperature for minerals", Journal of Physics and Chemistry of Solids, Vol. 60, 1895-1896 (1999). Patil, C.S., "Membrane reactor technology for ultrapure hydrogen production", Ph.D. thesis, University of Twente, Netherlands (2005). Patil, C.S., van Sint Annaland, M., Kuipers, J.A.M., "Design of a novel autothermal membrane-assisted fluidized-bed reactor for the production of ultrapure hydrogen from methane", Industrial Engineering Chemical Research, Vol. 44, 9502-9512 (2005). 181 Pena, M.A., Gomez, J.P., Fierro, J.L.G., "New catalytic routes for syngas and hydrogen production", Applied Catalysis A: General, Vol. 144, No. 1-2, 7-57 (1996). Prasad, P., Elnashaie, S.S.E.H., "Novel circulating fluidized-bed membrane reformer for the efficient production of ultraclean fuels from hydrocarbons", Industrial and Engineering Chemistry Research, Vol. 41, No. 25, 6518-6527(2002). Prasad, P., Elnashaie, S.S.E.H., "Novel circulating fluidized-bed membrane reformer using carbon dioxide sequestration", Industrial and Engineering Chemistry Research, Vol. 43, 494-501 (2004). Qi, A., Wang, S., Fu, G., Ni, C , Wu, D., "La-Ce-Ni-O monolithic perovskite catalysts potential for gasoline autothermal reforming system", Applied Catalysis A, Vol. 281, No. 1-2, 233-246 (2005). Rakib, M.A., Alhumaizi, K.I., "Modelling of a fluidized bed membrane reactor for the steam reforming of methane: advantages of oxygen addition for favourable hydrogen production", Energy and Fuels, Vol. 19, 2129-2239 (2005). Ridler, D.E., Twigg, M.V., "Steam Reforming" in Catalyst Handbook, edited by Twigg, D.E., 2nd edition, Manson publishing, London (1996). Romm, J.J., "The hype about hydrogen", Issues in Science & Technology, Vol. 20, No. 3, 74-81 (2004). Rostrup-Nielsen, J.R., "Catalytic Steam Reforming", Catalysis Science and Technology, Vol. 5, Springer, Verlag, Berlin (1984). Rostrup-Nielsen, J.R., "Production of synthesis gas", Catalysis Today, Vol. 4, 305-324 (1993). Rostrup-Nielsen, J.R., "Syngas in perspective", Catalysis Today, Vol. 71, 243-247 (2002). Rothenberger, K.S., Cugini, A.V., Howard, B.H., Killmeyer, R.P., Ciocco, M.V., Morreale, B.D., Enick, R.M., Bustamante, F., Mardilivich, I.O., Ma, Y.H., "High pressure hydrogen permeance of porous stainless steel coated with a thin palladium film via electroless plating", Journal of Membrane Science, Vol. 244, 55-68 (2004). Roy, R., Davidson, J.F., "Similarity between gas-fluidized beds at elevated temperature and pressure", Fluidization VI, edited by Grace, J.R., Shemilt, L.W. Bergougnou, M.A., Engineering Foundation, New York, 293-300 (1989). Roy, S., "Fluidized bed steam methane reforming membrane with high-flux membranes and oxygen input", Ph.D. Thesis, University of Calgary, Canada (1998). Roy, S., Pruden, B., Adris, A.M., Grace, J.R., Lim, C.J., "Fluidized bed steam methane reforming with oxygen input", Chem. Eng. Sci., Vol. 54, 2095-2102 (1999). Roy, S., Cox, B.G., Adris, A.M., Pruden, B.B., "Economics and simulation of fluidized bed membrane reforming", International Journal of Hydrogen Energy, Vol. 23, No. 9, 745-752 (1998). Rowe, P.N., Partridge, B.A., "An x-ray study of bubbles in fluidised beds", Chemical Engineering Research and Design, Vol. 43a, 157-175 (1965). Schwartz, M., White, J.H., Sammels, A.F., "Solid state oxygen anion and electron mediating membrane and catalytic membrane reactor containing them", Assigned to Eltron Research Inc., US Patent 6,033,632 (2000). 182 Shih, H.H., Chu, C.Y., Hwang, S.J., "Solids circulation and attrition rates and gas bypassing in an internally circulating fluidized bed", Industrial and Engineering Chemistry Research, Vol. 42, No. 23, 5915-5923 (2003). Shu, J., Grandjean, B.P.A., Van Neste, A., Kaliaguine, S., "Catalytic palladium-based membrane reactors: a review", Canadian Journal of Chemical Engineering, Vol. 69, 1036-1060 (1991). Shu, J., Bongondo, B.E.W., Grandjean, B.P.A., Adnot, A., Kaliaguine, S., "Surface segregation of Pd-Ag membranes upon hydrogen permeation", Surface Science, Vol. 291, No. 1, 129-138 (1993). Soliman, M.A., Elnashaie, S.S.E.H., Al-Ubaid, A.S., Adris, A.M., "Simulation of steam reformers with methane", Chemical Engineering Science, Vol. 43, No. 8, 1801-1806 (1988). Song, B.H., Kim, Y.T., Sang, D.K., "Circulation of solids and gas bypassing in an internally circulating fluidized bed with a draft tube", Chemical Engineering Journal, Vol. 68, 115-122 (1997). Spath, P.L., Mann, M.K., "Life cycle assessment of hydrogen production via natural gas steam reforming", National Renewable Energy Laboratory, report TP-570-27637 (2001). Sperling, D., Ogden, J., "The hope for hydrogen", Issues in Science and Technology, Vol. 20, No. 3, 82-86 (2004). SRI Consulting, Suresh, B., Inoguchi, Y., Schlag, S., "CEH Report: Hydrogen", Menlo Park, California (2004). Stitt, E.H., Abbott, P.E.J., Cromarty, B.J., Crewdson, B.J., "Emerging trends in syngas and hydrogen", CatCon 2000, Houston, TX (2000). Stoll, R.E., von Linde, F., "Hydrogen - what are the costs?", Hydrocarbon Processing, Vol. 79, No. 12,42-46(2000). Thomas, C.E., Kuhn, I. F., James, B., Lomax, F., Baum, G., "Affordable hydrogen supply pathways for fuel cell vehicles", International Journal of Hydrogen Energy, Vol. 23, No. 6, 507-516 (1998). Tomishige, K., Nurunnabi, M., Maruyama, K., Kunimori, K., "Effect of oxygen addition to steam and dry reforming of methane on bed temperature profile over Pt and Ni catalysts", Fuel Processing Technology, 85, 1103-1120, 2004. Tong, H.D., Gielens, J.G., Gardeniers, J.G.E., Jansen, H.V., Berenschot, J.W., de Boer, M.J., van Rijn, C , Elwenspoek, M., "Microsieve supporting palladium-silver alloy membrane and applications to hydrogen separation", Journal of Microelectromechanical Systems, Vol. 14, No.l, 113-124 (2005). Tosti, S., Bettinali, L., Violante, V., "Rolled thin Pd and Pd-Ag membranes for hydrogen separation", International Journal of Hydrogen Energy, Vol. 25, 319-325 (2000). Tosti, S., "Supported and laminated Pd-based metallic membranes", International Journal of Hydrogen Energy, Vol. 28, 1445-1454 (2003). Touloukian, Y.S., Kirby, R.K., Taylor, R.E., Desai, P.D., "Thermal expansion - metallic elements and alloys" in Vol.12 of "Thermophysical properties of matter", IFI/Plenum, New York (1970). Tsuru, T., Yamaguchi, K., Yoshioka, T., Asaeda, M., "Methane steam reforming by microporous catalytic membrane reactors", AIChE Journal, Vol. 54, No. 11, 2794-2805 (2004). 183 Uemiya, S., Matsuda, E., Kikuchi, E., "Hydrogen permeable palladium-alloy membrane supported on porous ceramics", Journal of Membrane Science, Vol. 56, No. 3, 315-325 (1991). Uemiya, S., "Brief review of steam reforming using a metal membrane reactor", Topics in Catalysis, Vol. 29, No. 1-2, 79-84 (2004). Usami, Y., Fukusako, S., Yamada, M., "Heat and mass transfer in a reforming catalyst bed: Analytical prediction of distributions in the catalyst bed", Heat Transfer - Asian Research, Vol. 32, No. 4,367-380 (2003). Van Beurden, P., van Dijk, E., van Delft, Y., van der Brink, R., Jansen, D., "Catalysts for hydrogen production in membrane and sorbent reformers", Energy Research Centre of the Netherlands, Vankelecom, I.F.J., Jacobs, P.A., "Dense organic catalytic membranes for fine chemical synthesis", Catalysis Today, Vol. 56, 147-157 (2000). Ward, T.L., Dao, T., "Model of hydrogen permeation behaviour in palladium membranes", Journal of Membrane Science, Vol. 153, 211-231 (1999). Watson, J., Daly, F., "Steam reforming catalysts for microchannel reactor", Velocys Inc., Xie, D., "Experimental study of gas and particle recycling in an internally circulating fluidized bed membrane reactor, Report No.5", Internal report, Membrane Reactor Technologies Ltd. (2002). Xu, J, Froment, G.F., "Methane steam reforming, methanation and water gas shift: I. Intrinsic kinetics", AIChE Journal, Vol. 35 (1), 88-96 (1989). Xu, J, Froment, G.F., "Methane steam reforming, methanation and water gas shift: II. Diffusional limitations and reactor simulation", AIChE Journal, Vol. 35 (1), 97-103 (1989). Yasuda, I., Shirasaki, Y., Mori, T., Maeda, K., Furuta, H., Kondo, T., "Development and demonstration of highly-efficient hydrogen production from natural gas for fuel cell vehicles", presented at GASEX 2004, Singapore (2004). Yates, J.G., "Effects of temperature and pressure on gas-solid fluidization", Chemical Engineering Science, Vol. 51, No. 2, 167-205 (1996). Yi, K. B., Harrison, D.P., "Low-pressure sorption-enhanced hydrogen production", Industrial & Engineering Chemistry Research, Vol. 44, No. 6, 1665 -1669 (2005). Zaika, Z.D., Minet, R.G., Tsotsis, T.T., "A high temperature catalytic membrane reactor for propane dehydrogenation", Journal of Membrane Science, Vol. 77, 221-232 (1993). Zakkay, V., Sellakumar, K. M.; Radhakrishnan, R.; McClung, J. D., "Vertical heat exchanger for high pressure fluidized bed coal combustors", Energy Progress, Vol. 6, No. 4, 248-253 (1986). 184 Appendix 1 - Introduction Details Steam:carbon molar ratio Temperature (°C) Figure A l . 1: Equilibrium methane conversion in SMR as a function of steam-to-carbon ratio and temperature (pressure = 1,000 kPa, graph contours are lines of constant conversion) Table A 1.1: Steps employing catalysts in SMR (Armor, 1999) Operation Temperature (°C) 1 Sulfur conversion (HDS) 290-370 2 H2S removal (ZnO) 340-390 3 Chloride removal (Alumina) 25-400 4 Pre-reforming 300-525 5 Steam methane reforming 850 6 High temperature water gas shift 340-360 7 Low temperature water gas shift 200 8 Methanation 320 9 NOx removal (ammonia, SCR) 350 Table A 1.2: SMR reactions with oxygen input (Abba et al., 2003 after Xu and Froment, 1990) SMR reaction with oxygen addition Rate RI CH 4 + H zO <-» CO + 3H2 hi R2 CO + H 2 0 C0 2 + H 2 R3 CH4 + 2H20 <-> C0 2 + 4H2 kr3 R4 CFL, + 202 *+ C0 2 + 2H2 R5 CH 4 + 1.5H 20«-»CO + 2H20 kr5 185 Table A 1.3: SMR reaction rates (Abba et al., 2003 after Xu and Froment, 1990) Reaction rate equation RI r \ = ^ri (PCH4PHIO 1 PHI ~ PHI PCO 1 KEQL) / K R2 r i = kri(PQOPHIO 1 PHI ~ PHIPCOI 1 Keq2)/ K R3 r _ t / p p2 p 3 5 _ p 0 5 n IV- X( \ , 2 '3 — A >3V J CHA1 HIO1 H2 1 H2 1 C02 '  1^ eq,\ 1^ eq,2 ) 1 N~ R4 '4 — N-R4ACH4A02 R5 r -k r 0 ' 5 Y 0 7 5 '5 — A ' r 5 A C / / 4 A 0 2 where K = 1 + KcoPco + KH2PH2 + KCH4PCH4 + KH20PH20 1PH2 Table A1.4: SMR reaction rate parameters (Abba et al., 2003 after Xu and Froment, 1990) Rate/ adsorption constant Pre-exponent factor Activation energy / heat of adsorption (kJ/mol) Ki 8.34E17(molPaus/kgs) 240.1 kr2 1.22E1 (mol/kg s Pa) 67.1 kr3 2.01E17(mol Pa6'/kg s) 243.9 kr4 1.26E12(mol/kg s) 170.3 krS 3.08E10(mol/kg s) 139.7 K-co 8.23E-10 (Pa1) -70.65 K-CH4 6.65E-9 (Pa1) -38.28 KH20 1.77E5 (-) -88.68 KH2 6.12E-14(Pa"') -82.90 Keq.l 1.2E23(TV) 223.10 Keq.2 1.77E-2 (-) -36.58 186 Appendix 2 - Membrane Details JM Base Prices Palladium From: Jul 1992 US$ Monthly Average To: Jan 2007 in*, ; ; 1992 I 1993 I 1994 I 199S I 1996 I 1997 I 1998 I 1999 I 2000 I 2001 I 2002 I 2003 I 2044 I 200S I 2006 I Figure A2.1: Palladium pricing 1992-2007 (source: Johnson Matthey website) 187 SideB Side A Side A Panel #1 3/16" tube 3/16" tube Side B Side A lots of small creases, especially on LHS Side B Side A Panel #5 Panel #3 3/16" tube 3/16" tube J \ 1 faint crack numerous surface bumps, due to alumina? Side B small ' crease small number of surface bumps, due to alumina? Panel #4 ladder-like |marks faintly visible from substate Side B Side A Panel #6 Side B Panel#8 Panel A is not shown, but was clear both sided Figure A2.2: Crease defects in membrane panels installed in ICFBMR (after bonding, but before service in the ICFBMR pilot reactor) 188 Table A2.1: Permeation rig components Tag Description Size/ Capacity Model / Manufacturer V-10 316SS pressure vessel 10" IDx25" Axton Manufacturing H-10 Semi-cylindrical ceramic fiber heaters (2) 3.5 kW/ 240VAC Watlow, model VS112A18 FIC-01 Feed mass flow controller 0-6 SLM H 2 Brooks, model 5850S FIC-02 Sweep mass flow controller 0-2 SLM H 2 Brooks, model 5850S FI-03 Permeate mass flow meter 0-1 SLM N 2 Omega, model FMA-A2306SS FI-04 Rotameter (CHE3495A) Calibration in figure below Fisher-Porter VP-10 Vacuum pump lA HP Fisher Technical, model LAV-3 Table A2.2: Gas chromatograph details GC#1 GC#2 Application • Permeation rig testing • Permeate from ICFBMR • ICFBMR ROG testing pilot plant Model Shimadzu GC-8AIT Shimadzu GC-8AIT (CHE4116A) (CHE3577A) Detector type TCD TCD Detector current 60 mA 60 mA Sample introduction Automatic sampling valve Syringe injection (0.1 m) Column Hayesep DB 30' x 1/8" OD, Supelco 15' x 1/8" OD, Carboxen 100/120 mesh, Lot 154 1000, Lot 080797 Injector temperature 150°C 110°C Column temperature 120°C 100°C Carrier / reference gas Argon Argon Carrier / reference flow 22 / 33 ml/min Integrator Shimadzu CR601 (CHE4116B) Shimadzu C-R3A (CHE4116B) Integrator settings Width 5 5 Slope 29.76 15 Drift 0 0 Chart speed 4 mm/min 6 mm/min Min. area 1 100 Tvpical gas factors H2 3.9min/3.643E-05 3.86 min/3.9302E-5 N2 4.52 min / 4.820E-04 4.52 min / 6.859E-4 CO 4.78 min/3.567E-04 4.78 min/4.706E-4 CH4 6.29 min / 1.326E-04 6.28 min / 1.441E-4 C02 8.95 min / 4.632E-04 8.84 min / 5.036E-4 189 Table A2.3: Hydrogen permeation data for Panel A - effect pf hydrogen partial pressure Data point Permeation vessel conditions Sweep argon (SLM) Permeate pressure (kPa) H2flux (SLM) [H2] in permeate (vol%) Permeate H2 partial pressure (kPa) Sq. root H2 partial pressure (PaA0.5) H2 feed (SLM) Temp (°C) Pressure (kPa) H2 partial pressure (kPa) No sweep 1 3.0 549 134 134 0 101.3 0.78 100% 101.3 47.5 No sweep 2 6.0 543 170 170 0 101.3 1.41 100% 101.3 94.5 No sweep ,3 6.0 549 205 205 0 101.3 1.91 100% 101.3 134.4 Sweep gas 4 6.0 543 170 170 0.40 101.3 1.88 83% 83.6 123.6 Sweep gas 5 6.0 543 170 170 0.67 101.3 2.03 75% 76.2 136.7 Sweep gas 6 6.0 543 170 170 0.93 101.3 2.10 69% 70.1 148.1 Sweep gas 7 6.0 543 170 170 1.20 101.3 2.15 64% 65.0 157.9 Vacuum 8 6.0 548 170 170 0.0 67.5 2.26 100% 67.5 153.1 Vacuum 9 6.0 548 170 170 ' 0.0 50.5 2.70 100% 50.5 188.0 Vacuum 10 6.0 548 170 170 0.0 33.6 3.18 100% 33.6 229.5 Table A2.4: Effect of temperature on permeation for Panel A (pure hydrogen in permeation vessel, no sweep gas to permeate) Temperature ( ° C ) Measured H2 flux (SLM) Permeate pressure (kPa) Vessel pressure 238 kPa 307 kPa 376 kPa 455 1.39 2.06 2.64 101 502 1.76 2.42 3.03 101 538 1.85 2.61 3.32 101 564 1.91 2.71 3.53 101 Table A2.5: Effect of temperature on permeation for Panels 1-8 (pure hydrogen in permeation vessel, no sweep gas to permeate, vessel at 300 kPa, permeate at 101 kpa) Panel 1 Panel 3 Panel 6 Temp ( °C ) Flux (SLM) Temp ( °C) Flux (SLM) Temp ( °C ) Flux (SLM) 463 1.44 463 1.14 463 0.93 483 1.53 485 1.20 532 0.96 505 1.74 503 1.39 545 1.03 536 1.87 536 1.53 570.5 1.17 555 2.02 554 1.69 Panel 4 Panel 8 Panel 5 Temp ( °C ) Flux (SLM) Temp ( °C ) Flux (SLM) Temp ( °C ) Flux (SLM) 463 0.69 450 1.14 463 1.69 515 1.07 450 0.99 482 1.77 537 1.13 474 1.21 507 2.13 547.5 1.25 505 1.42 536 2.23 568.5 1.45 530 1.52 555 2.42 544 1.51 191 Appendix 3.1 Cold Model Installation Instruction Figure A3.1.1 presents a schematic of the cold model and include part numbers listed in the instructions below. Wear appropriate personal safety protection. 1. Disassemble full column: • Drain catalyst from column a nozzle on the lower column section • Close return standpipe valves <1> • Disconnect return standpipe from column <2> • Remove main 6" feed pipe <3> • Unbolt and remove windbox <4> • Remove lower 4' column section 2. Install internals: • Assemble empty core box o Remove all panels <5> from the core box <9> o Insert VS" location tube <11> through top of box. This tube is about 11" long so as to fit in the column and will act as the guide to the Vi" rod <10>, which secures the core box. o Install lA" filter impulse connections <23> onto core box channels o Locate box <9> on the floor directly under the 12" column • Feed rope <17> from 2" angled nozzle <8> at the top of column through to the open bottom and tie to middle of VS" box support tube <11> • Feed Vi" plastic impulse connection lines <from upper column ports through open bottom of 12" column, ensuring that they are not tangled • Connect impulse lines to the appropriate impulse filters <23> on the core box <9> • Tie wrap impulse feed lines <18> as needed • Lift / raise box into the column in increments: o Always tie off support rope <17> o Pull slack of Vi" impulse tubes <18> through the connection ports on column wall o Ensure impulse lines <17> are not tangled • Secure box support by inserting 3/16" hanging tube <10> through VS" support tube <11> and through the external Plexiglas bosses. Secure 3/16" tube with Swagelok port connections <12> to NPT drilled bosses • Center and level core box with Vi" lateral tubes <21>, both east-west and north-south and secure with Swagelok fittings <22> • Pull final slack of Vi" impulse lines <18> and seal Swagelok fittings on column wall • Pull rope <17> tight and secure to the inside of the column nozzle <8> so as to permit the internals to be lowered during disassembly • Slide in the core box panels <5> from the bottom and secure in place 3. Install oxidant feed line • Insert a Vi" steel guide through the column with the opposing port connections • Push the 3/16" oxidant distributor <6> over one end of the Vi" guide and feed the distributor through the column to the opposing port 192 • Secure the distributor with Swagelok port connections <7> to the Plexiglas bosses on the column wall • Attach distributor to oxidant feed line 4. Install secondary feed distributor: • Center, and secure distributor <16> in each of the four quadrants • Secure lA" distributor feed lines <15> with Swagelok port connections <14> to Plexiglas bosses • Attach both distributor feed lines <15> to building air supply 5. Install windbox: • Last check of internals • Make sure the two distributor plates <20> are oriented so the drilled holes in the plates match • Lift and bolt the windbox <4> to the column, ensuring that the square feed hole pattern matches the orientation of the core box <9> • Reattach the 6" feed line <3> to the windbox <4> and to the flexible connection on the building air supply 6. Solids addition: • Add a small flow of gas to distributors to reduce solids back sifting and aid bed settling • Remove the steel exit elbow <13> on top of the column • Using a funnel, fill the column with FCC until the settled bed level is 25 mm above the core box <9> • Reattach steel exit elbow <13> 7. Instrumentation: • Connect V" impulse lines <18> to appropriate pressure transmitters <19> • With no gas flow to column, check the calibration of the pressure transmitters <19> on the data acquisition computer and zero if necessary 8. Column disassembly: • Drain solids from column via nozzle near bottom of column • Remove main 6" feed pipe <3> • Unbolt windbox <4> and remove remaining solids - this can be messy • Remove secondary distributor < 14-16> • Remove oxidant distributor <6-7> • Remove core box panels <5> by unbolting setscrews at the bottom of the box <9> assembly • Disconnect impulse lines <18> connected to core box <9> • Open 2" angled nozzle <8> near top of column and tie off the rope <17> attached to the core box • Remove the 3/8" tube <10> securing the core box to the column • Using the rope <17>, slowly lower the core bolt out the bottom of the column 193 <13> exit nozzle elbow Cyclones <17> rope to raise core box <8> angled nozzle (2") <7> oxidant fitting (2) <6> 0.375" oxidant distributor 12> core box support fitting (2) <10> 0.375" core box support rod <11> 14" core box location tube <16> secondary distributor (4) <14> 2° distributor fitting (2) <15> % " distributor feed <20> distributor plates <1> cyclone return valves V V <2> standpipe return ^ P T ^ <19> pressure transmitters <18> V*" plastic impulse tubing <9> Plexiglas core box <23> pressure filter fitting (many) <5> core panels (8) <21 > VA" centering tubes (6) <22> centering tube fitting (6) <3> main feed pipe (6") Figure A3.1.1: Cold model assembly schematic 194 Appendix 3.2 - Cold Model Testing Details 6.35 /Top distr typ' "I T / plate 5 8 _ ^ L_ / ibutor ^ . 38 pm II i i II M II l i , I " 6 4 _ J L_ \ Bottom 6.35 ? .1 I distributor W Section A-A plate (all dimensions in mm) Figure A3.2.3: Cold model main distributor plate Oxidant feed -182-Settled bed depth of FCC particles 75 50 25 r Oxidant distributor (12 mmOD, 2x13 holes, 1 mmdiam) Oxidant I feed Top of core box Column shell (285 mm ID) Figure A3.2.4: Elevation view of cold model oxidant distributor (all dimensions in mm) 196 Table A3.2.1: Cold model feed air rotameters Tag Service Scale Model Material FI-OT Main air feed (1) 0-600 SLM air Omega FL-1658-SS Glass body, SS float FI-02 Main air feed (2) 0-236 SLM Omega FL-1656-SS Glass body, SS float FI-03 Secondary air feed 3-30 LPM air Cole Parmer U-32461-56 Acrylic body, SS float FI-04 Oxidant air 30-280 LPM air Cole Parmer U-3 2461-64 Acrylic body, SS float re Q. OL 3 Si •- 1 •a  1 c "55 0 A -0 100 200 300 400 500 600 700 Air Flow (SLM) Figure A3.2.5: Pressure drop over cold model main distributors 800 o Secondary (per tube) • Oxidant 0 <!>-20 80 Fi) 40 60 Air flow (SLM) ;ure A3.2.6: Pressure drop over cold model secondary and oxidant distributors 100 197 Table A3.2.2: Cold model pressure transducer data Description Type Tag Channel Calibrated full scale range (kPa) Over main distributor plate Pd 18 20 14.8 Bottom of North Quadrant Pg 5 11 15.6 Freeboard (c/w atmosphere) Pd 1 6 3.2 Core bottom channel 4 (c/w atmosphere) Pd 6 0 14.9 Core bottom channel 5 (c/w atmosphere) Pd 13 21 96.9 Core mid channel 1 Pg 12 1 96.5 Core mid channel 2 Pg 23 3 16.0 Core mid channel 3 Pg 14 15 15.9 Core mid channel 4 Pg 9 13 15.9 Core mid channel 5 Pg 8 7 14.8 Core mid channel 6 Pg 7 9 15.8 Core mid channel 7 Pg 22 2 15.6 Core mid channel 8 Pg 4 5 16.4 Core top channel 4 (c/w freeboard) Pg 0 4 3.2 Core top channel 5 (c/w freeboard) Pd 17 17 3.1 Core top channel 6 (c/w freeboard) Pd 16 12 3.2 Note: Pd = differential pressure, Pg = absolute pressure, c/w = connected with routed up south quadrant Figure A3.2.7: Plan view of cold model.pressure tap connections with vertically slotted panels 198 Freeboard ^@ (Elev 2090) Top core channel (5) (^)-Mid core channels (1,3,5,7) (&^> Bottom core channel (5) (@^« Annular sample points (4) (elev = 325) North quadrant ^gg (elev = 70) Windbox Core box Windbox Elevation view looking East All dimensions in mm •(^) Top core channels (4,6) Mid core channels (2,4,6,8) • { ® ) Bottom core channel (4) Top core pressure taps (elev= 1494) Mid core pressure taps (elev = 862) Bottom core pressure taps (elev = 230) Distributor (elev = 0) Figure A3.2.8: Elevation view of cold model pressure tap connections (looking east) 199 0 0.0025 0.005 0.0075 0.01 0.0125 0.015 Velocity (m/s) Figure A3.2.9: Cold model pressure drop vs. superficial air velocity for SMR (NiO) catalyst 2.0 1.5 Q. Q. O •D 1.0 0 3 co Q) 0.5 A A / • A A A A A • • A / A f A / 7 A increasing flow • decreasing flow r Umf = 0.0030 0.0 0.0000 0.0025 0.0050 0.0075 0.0100 Superficial velocity (m/s) Figure A3.2.10: Cold model pressure drop vs. superficial air velocity for ATR catalyst 200 120 0 0 20 40 60 80 100 Rotameter Reading (bottom ball) Figure A3.2.11: Calibration of FI-06 used for helium injection in cold model testing at 207 kPag as measured against a mass flow meter operating with helium 12 0.35 0.40 0.45 0.50 0.55 Signal (volts) Figure A3.2.12: Typical calibration curve of thermal conductivity detector (TCD) with helium used in cold model system (TCD #2 on 050426, xlOO amplification) 201 9 ° 7 O ' > c s * (0 3 cr c ? 3 x re 1 0 Figure x x x x x x x x X X X X X X X X X M X X X X X X x x x *x &X X X X X X *x X 100 200 600 700 300 400 500 Main air flow (SLM) A3.2.13: Ratio of maximum to minimum quadrant solids velocity for cold model data 800 (0 Q . J £ Q . O <u 3 (A (A d) Q . - 7 0) c c ra xz O horizontally slotted panels y = -0.0027X + 8.77 R2 = 0.994 A • 30 SLM annular air 20 SLM annular air O 10 SLM annular air - Linear (20 SLM annular air) 100 200 300 400 500 Main Feed (SLM) 600 700 800 Figure A3.2.14: Pressure drop over channel 4 in cold model core with horizontally slotted panels (95% confidence interval) 202 94 SLM main air Frequency (Hz) 0 2 4 6 8 10 Frequency (Hz) 236 SLM main air 0 2 4 6 8 10 Frequency (Hz) 236 SLM main air 0 2 4 Frequency (Hz) 480 SLM main air 0 2 4 Frequency (Hz) 480 SLM main air 0 2 4 6 8 10 Frequency (Hz) 2 4 6 8 10 Frequency (Hz) 800 SLM main air 0 2 4 6 Frequency (Hz) 713 SLM main air 0 2 4 6 8 10 Frequency(Hz) (a) Solid panels (b) Vertically slotted panels Figure A3.2.15: Gage pressure at mid4ieight of cold model channel 4, 20 SLM air to annulus 203 12.0 7 10.0 -! 1 . 1 : ! 10.0 10.5 11.0 11.5 12.0 12.5 Corrected pressure in bottom of channel 4 (kPa) Figure A3.2.16: Pressure of annulus vs. channel 4 pressure in cold model with vertically slotted panels (corrected for elevation difference) 204 Figure A4.1.1: ICFBMR main distributor drilling pattern 200 CL ^ 150 Q. O •o 100 a) i _ 3 tn 50 o i_ CL 0 i • A A A oxidant • annular • main 0 20 40 60 80 Air flow at ambient conditions (SLM) Figure A4.1.2: ICFBMR pressure drops for three gas distributors 100 205 206 NOZZLE SCHCDULC Figure A4.1.4: ICFBMR external pressure vessel drawing (1) Figure A4.1.5: ICFBMR external pressure vessel drawing (2) w . f t * a s t a « u ; s AX row 7 . ft J / L L Of MATERIALS i\ )/<• I ii iff V r w r t ft) ?/r* 'ma to tmT~ * 7542A03 j f t ' - t f t n c w r i n w o t . )/ S i n BAI w a o s ra > y ' OA sure K H c«ip dt w u w w ; yuam. I 1 I ) B L O C K T H I C K N E S S A N D C A P F R O M S A * B L A D E T H I C K N E S S . I 1 2 ) B O R E D t A U E T I R T O W A T C H 0 . 0 . O F 3 * N P S P F E . I 1 3 ) R E M O V E A L L 8 U R R S A N D S H A R P E D G E S . R O U N D O F F S H A R P C O R N E R S . I 1 4) F O U R (4) C A S T E R S e / « M O U N T I N G H A R D W A R E S U P P U E D B T N O R A M . CD 5 ) A L L F I L L E T W E L D S T O B E V M I N I M U M U N L E S S O T H E R W I S E N O T E D . CD 6 ) Aa I T E M S T O B E P R E P A R E D A N D P A I N T E D . I N S T R U C T I O N S T O F O L L O W . I I 7 ) M A X I M U M D E S I G N L O A D - 7 0 0 0 l b s . • 8 ) T R W . m V E S S E L T O F R A M E P R I O R T O S H I P P I N G . I ' S f fit. 5X' M 1 C L E A N I N G k P A W T W C : S U R F A C E P f t F P . - S U R F A C E S M U S T B E C L E A N A N D F R E E O F L O O S E R U S T , M I L L S C A L E . D E T E R I O R A T E D P R E V I O U S C O A T I N G S . D I R T , O R , G R E A S E A N D C H E M I C A L C O N T A M I N A N T S . A B R A S I V E B L A S T I N G S R £ C D M U E N D E D . P R I M E R - ( 1 ) C O A T 7 0 6 9 R E D P R I M E R . F J N J S t J - I N D U S T R I A L E N A M E L . G R E Y C O L O U R . F O L L O W M A N U F . I N S T R U C T I O N S F O R C O V E R A G E . < f Q 0 T | r C A W - E71T-1 ( l -D)GMAW => ER70S-6 iVWvfACTUKD 1: N O R A M E N C W E E R 1 N G A N D C O N S T R U C T O R S L T D . f o r U B C I C F B M R I N T E R N A L C I R C U L A T I N G F L U l D I Z l N G B E D M E M B R A N E R E F O R M E R M O B I L E S U P P O R T F R A M E D E T A I L S Figure A4.1.6: ICFBMR support steel drawing Figure A4.1.7: ICFBMR internal vessel drawing (1) Figure A4.1.8: ICFBMR internal vessel drawing (2) Figure A4.1.9: ICFBMR main distributor mechanical drawing V COIL A (9 TURNS) H' COIL 8 (9 TURNS) H" COIL C (9 TURNS) NOZZLE SCHEDULE KSM MPS SCM IfWWC DESIGN CODES AND SPECIFICATIONS VESSEL aw. BE OESICWD AM FABROIED IN ACCORDANCE fir* - ASK Ml D(V 1. 7001 EDITIOH. - U" STA* REOD. DESIGN CONDITIONS CONTEXTS: PROCESS CAS k WIROCEN SwfEP SPECK CRAW. 0.01 UTERKAL OESCN PRESSURE 0 TEltfEJWWtE: 50) PStt I WOT ECTDBW DESKN PRESSURE 4 TEMPERATURE 15 PSK O MOT ommc PRESSURE • TEKPGUTIME: •CO P9C O 4C0T UMHJM C O O H P * TEMPERATUR : 3J T • SCO rac RADWOWItt: N01 JKT0 AS PER UH*-J3(B) MC W-11(c) OMRCr M « T TESTMCi NOT KE0TJ AS PER U*-5l(BHlXl>) AND W3-M(a)(lXb) POST «QJ> HEAT IKEArUENT: MOT urns AS pa uai UCS-M NOTE p) w m-iotooXd) CORROSOI mcnfHS: NONE 0E9CN «M) SPEED: NOT A OESCN ffiOUREMENT SEKWCr NOT A OtSCN REOUWEVENT MATERIALS OF CONSTRUCTION SfCLL: SAt« yon B. SEAMLESS PIPE KAft SAW GRADE iPB PFE CAP NOZZLE: SAiDJ SEAMLESS Ftft (PER SOCDULE) FIANCES.' SA-105 (EXCEPT AS NOTED) REPADS: N/A SUPPORTS N/A MIFJQW. ATOMS; SA KI CRADE A EITERWL FTTTMS: SA 202 GRADE A VAMMr OC: SA-105 STUDS / NUTS SA-1H-B7 / SA-1M-JH CASKETS: CtLON JStO WEIGHT, CAPACITY & AREA GENERAL NOTES i NORAM PROirct MD ?(MM DATE INTERNAL CIRCULATING FLUIDIZED BED MEMBRANE REFORMER UBC ICFBMR 3 COIL CAST PREHEATER BLOCK 2 rwrcufn APPtKWEO EOT TEW [t-W « IIIV-N.T.S. 1 1 OF 1 | ZO845-40OJ | A Figure A4.1.10: ICFBMR internal preheater All dimensions in mm Material: 304L SS VA Swagelok nuts welded to the box to hold support guide from upper flange r 2 . 7 - » CD OO 83 88 V V CN Plan view Elevation Figure A4.1.11: ICFBMR draft core box 214 1407 305 Dummy 305 Membrane 22-63.50 305 Membrane 64 305 Membrane View 1: Membrane assembly before sealing strips Membrane connector (part 2) 1524 1422 r Dummy Membrane Membrane Membrane Core draft box View 2: Membrane assembly with sealing strips Six communication slots (57 x 22 mm) Part C: Membrane backing strip (graphite gasket underneath) Figure A4.1.12: Membrane assembly general arrangement (dimensions in mm) 0.50-H Part 3 - 4 Req'd 304/316SS (%" Thick) Note: Faces do not need to be machined Figure A4.1.13: Dummy membrane panel (all dimensions in inches) 3.25-1/8" thick 304 SS 8x3/16" thru' C - 8 backing strips '/«• thick 304 SS 16x3/16" thru' 0.1875-»-j-»- 0.1875-»| D -1 dummy panel Notes: • All SS materials supplied • All dimensions in inches • Dims to hole centres 2.875 Figure A4.1.14: Membrane backing strips and full-length dummy panel (all dimensions in inches) 217 CO 82.55 | Part 2 - Connector Piece (4 Required) o> Notes: 1. All dims are inches 1. Man 316 or 304 SS 3. 3W x ZV* sheet supplied UBC Chemical & Biological Engineering I C F B M R Membrane Connector P i ece Designed / Drawn By: T.Boyd SIZE A FSCM NO DWG NO ICFBMR-CNT-001 REV B Date: Nov. 18. 2003 S C A L E 1:1 | [SHEET 3 OF 16 Figure A4.1.15: Membrane connector piece (all dimensions in mm) 218 P U R O C W I R O C C N S W E E P U r m O C E N t 5 M O T E S : 1 . E S D - E M E R G E N C Y S h V T D O W N . NORAM MEMBRANE REACTOR Technologies U.B.C PILOT PLANT CYLINDER SHED / GAS COMPRESSOR KTS 1 10F3 1 D-Z0MS-H11 1 A Figure A4.1.16: P&ID for pilot plant (gas cylinders) N O T E S : 1 . ' C C " D E N O T E S C O N N E C T I O N T O C A S C H R O M A T O G R A P H 2 . A L L T U B * I S I / * " S S S W A C O . O K U N L E S S O T H E R W I S E N O T E D 3 . R U P T U R E D G K R A T E D 3 5 0 P S B O 2TC. 2 5 0 P S l C O 1 W C NORAM MEMBRANE REACTOR Technologies U.B.C PILOT PLANT FEED SYSTEMS Figure A4.1.17: P&ID for pilot plant (feeds and condensers) Figure A4.1.18: P&ID for pilot plant (external preheater) 222 Table A4.1.1: ICFBMR thermocouples Tag Location Measurement location A B C D TE-114 Annulus (quadrant 1) 25 mm below core box 508 mm above point A 508 mm above point B 584 mm above point C TE-115 Annulus (quadrant 2) 25 mm below core box 508 mm above point A 508 mm above point B 584 mm above point C TE-116 Annulus (quadrant 3) 25 mm below core box 508 mm above point A 508 mm above point B 584 mm above point C TE-117 Core (channel 1) 25 mm below core box 508 mm above point A 508 mm above point B 584 mm above point C TE-118 Core (channel 4) 25 mm below core box 508 mm above point A 508 mm above point B 584 mm above point C TE-119 Reactor freeboard 25mm above core box 152 mm above point A 152 mm above point B X Table A4.1.2: ICFBMR pressure transducers Tag Service Model Range DP-01 Differential pressure in reactor (various points) Omega PX750 0-2.54 m WC DP-02 Differential pressure in reactor (various points) Omega PX750 0-2.54 m WC PT-121 Outer reactor (V-01) pressure (gauge) Omega PX182 0-3,449 kPa(g) PT-272 Inner reactor (V-02) pressure (gauge) Omega PX503 0-2,069kPa(g) Table A4.1.3: ICFBMR pressure tap and sample points Tag Service / Location Filter on tip? PT-A Freeboard, 150mm below top flange No PT-B Splash zone, 140 mm above core box Yes PT-C Adjacent to oxidant distributor Yes PT-D Top of center flow channel Yes PT-E Mid4ieight of center flow channel No PT-F Bottom of center flow channel No PT-G Annular quadrant 3 top - 127 mm below top of core box No PT-H Annular quadrant 3 bottom - 127 mm above bottom of core box No PT-I Annular quadrant 1 top - 127 mm below top of core box Yes PT-J Annular quadrant 1 bottom - 127 mm above bottom of core box Yes 223 4 point annular thermocouples (3) TE-114/115/116 3 point splash zone thermocouple rm 4 point core thermocouples (2) TE-117/118 TC point C TC point B TC point A TC point D PT-A TC point C TC point B H TC point A — Figure A4.1.20: ICFBMR instrument locations 224 Table A4.1.4: Equipment list for I C F B M R pilot plant Tag Name Size Materials Description C-1 Gas compressor 50 SLM at 20 MPa - Fuelmaker model P30 DSU Desulfurizer vertical cylindrical vessel 0.65m x0.1m OD SS vessel Cylindrical vessel filled with sulfur adsorbant (CalgonSulfasorb-8 activated carbon pellets impregnated with 8% copper oxide, mesh size 4x10) DIU Deionizer 3 cylinders Fibreglass US Filter (activated carbon bed followed by 2 cation exchange resin beds) V302A/B Water pump 0.71mx0.11mOD 316 SS N2 padded vessels E-1 Interchanger Shell: 1.22m x 0.22 OD 316 SS Rating 1.35 MPa Tubes: 6.2mm OD x 2.4m 316 SS Area = 0.047 m2 E-2 Condenser Shell: 1.22m x 0.22 OD 316 SS Rating 1.35 MPa Tubes: 6.2mm OD x 1.2m 316 SS Area = 0.024 m2 E-3 External preheater Outer shell: 0.29m ID x 1.52m CS Rating 1.64 MPa Inner shell: 0.1m ID x 1.3m 316 SS Rating 1.28 MPa Heater: 2 x 6 kW (240V) -H-1 Internal preheater Block: Lead See equipment drawing Tubes: 3 tube coils 316 SS See equipment drawing Heater: 2 x 6 kW (240V) ceramic fibre 2xWatlowVS108A18T H-2 Reactor heater A oriqinal: 2 x 4.3 kW (240V) ceramic fibre 2x Watlow VS108A30S revised: 2 x 5 kW (240V) Incoloy tubes 2x WatlowRDN134J10S H-3 Reactor heater B oriqinal: 2 x 4.3 kW (240V) ceramic fibre 2 x Watlow VS108A30S revised: 2 x 5 kW (240V) Incoloy tubes 2xWatlowRDN134J10S V-02 Internal reactor O.135/0.256m ID x 2.31m 304/310 SS See equipment drawings V-01 External reactor 0.41 m OD x 3.64m overall CS See equipment drawings tijl, Overview.cim 0 © ® File View Help to Figure A4.1.21: ICFBMR pilot plant control screen #1 Figure A4.1.22: ICFBMR pilot plant control screen #2 Figure A4.1.23: I C F B M R pilot plant control screen #3 to to so % FBMR.cim File View Help m • 0 FEEDS r~~~F9MR OFF-OAS V-01 Pressure [3453 Wag PI-121 F B M R P r e s s u r e B a l a n c e 13964kpa T i . i i 3 h 372.0C Oxidation Zone TMio|i372.QtC V-01 : | j V-02 | ; 1372.0 C 1-511 kpag 12(01/200514:58:29 HYDROGEN X V - 1 3 1 -|1372.QC 1372.0oCTic-iQ9 P R E H E A T E D F E E D T I -111 1372.0C 1372.0C NITROGEN FLUID BED ANNULUS 1372.0C 1372.0 C 1372.0 C 1372.0 C T M 1 4 T I - 1 1 S T l - 1 1 6 T I -117 T E M P E R A T U R E P R O F I L E S P R E S S U R E T A P S [3<Tl kpa DPI-001 35.4 k P a DPI-002 [23.5 kpa s m a i i D P 0.S SLPM 1372.0C T i c - n o O A S / S T E A M Figure A4.1.24: I C F B M R pilot plant control screen #4 % , TRIPS, t-ue view Help to O j tYAWCOl* ["""FSR*""" TRIPS TUNING, ESDI TRIPS PAHH305 1340.1 | -504.6 PAHH272 1349.9 | -4BS.9 PAHH202 1340.1 j -479.3 TAHH108 800.0 | 1372.0 TAHH109 i 600.0 1372.0 TAHH114A 6S0.0 [ 1372.0 TAHH114B 750.0 | 1372.0 TAHH114C S50.0 | 1372.0 TAHH1UD [ 6500 | 1372.0 TAHH115A | 650.0 | 1372.0 TAHH115B | 650,0 1372.0 TAHH115C | 725.0 | 1372.0 TAHH1150 675.0 1372.0 TAHH116A | 675.0 | 1372.0 TAHH116B | 675.0 1372.0 TAHH116C 650.0 1372.0 TAHH116D I 650.0 | 1372.0 ESD4 TRIPS TAHH110 | 750.0 | .1372.0 TAHH220 TAHH225 TAHH230 TAHH235 E S D 2 T R I P S - TAHH117A TAHH117B TAHH117C TAHH117D TAHH118A TAHH118B TAHH118C TAHH118D TAHH119A TAHH119B TAHH119C TAHH111 TAHH112 TAHH113 ESD5 TRIPS 1373.0 | 1372.0 400.0 21.5 650 0 23.0 650.0 | 21.5, 675.0 1372.0 675.0 1372.0 675.0 1372.0 675.0 1372.0 675.0 1372.0 650.0 1372.0 675.0 1372.0 675.0 1372.0 675.0 1372.0 650.0 1372.0 650.0 1372.0 ESD '3 TRIPS 650.0 650.0 1372.0 1372.0 12/0112005 14:58:41 TRIP STATUS R E S E T S E S D 2 E S D 3 E S D 4 | 650.0 | 1372.0 Note: The resets refer only to Trip "Causes" and not "effects" If a one of the trips is active, the effects of that and all subsequent trips are activated. Figure A4.1.25: ICFBMR pilot plant control screen #5 118" chamfer on each corner 3" x 3/8" sheet (typ' 5) equally spaced { 3/8" * | 4 x 'A" OD rings on box centreline 4 x 1/8" thick tab 1/8" hole, on box centrelines Connection Tabs All dimensions in inches All materials carbon steel Figure A4.1.26: ICFBMR dummy internals for commissioning 231 200 0 50 100 150 200 250 300 Nitrogen flow (SLM) Figure A4.1.27 ICFBMR off-gas filter pressure drop (Pure N 2 flow to reactor, 20°C, ambient pressure downstream of filter, no catalyst in reactor) Table A4.1.5: Natural gas analysis Sample Location: BC Gas, Tilbury Gate Station, Delta BC (2002) Analyzer: On-line Daniel gas chromatograph Component Mole Percent (Volume Percent) Nitrogen 0.7 Carbon Dioxide , 0 . 2 Methane 95.5 Ethane 2.9 Propane 0.5 iso-Butane 0.05 n-Butane 0.1 Pentanes 0.04 Hexanes 0.01 Total 100.00 Gross Heating Value (MJ/m3) 38.60 @ 15.0°C and 101.325 kPa Relative Density 0.580 @ 15.0°C and 101.325 kPa Average Molecular Weight 16.78 *Total Sulphur (mg S/m3) Approx. 15 (11 ppm) Hydrogen Sulphide (mg/m3) Approx. 2 (1.5 ppm) Water Content (mg/m3) Approx. 30 (40 ppm) *Note: The total sulphur includes naturally occurring sulphur compounds plus added odorant (approx. 5 mg S/m3 or 3.5 ppm) plus hydrogen sulphide. 232 Appendix 4.2 - ICFBMR Assembly Procedure 1. General Notes It is important that appropriate safety gear be worn during assembly. As a minimum, steel-toed boots, eye protection, hardhat, work gloves and work clothes should be worn. Dust masks must be worn when using insulation and other dusty materials. Kaowool and ceramic blanket and paper insulation were used to insulate the inside of the carbon steel outer reactor. Installers should be aware that the paper insulation has organic binders that release vapour on heat-up. Care should be taken to bake these binders before closing up the reactor as the oils can accumulate, leading to heater failure and condensation in vent tubing. All flange bolts should be tightened in the appropriate "star" pattern. 2. Reactor to Horizonta l For the following instructions, it assumed that the reactor cradle swings north-south and that the reactor is lowered due south. • Two people are required to lower the ICFBMR reactor to the horizontal position safely. • The main reactor support frame must be anchored to the ground to prevent movement. This can be achieved by drilling screwed Hilti anchors into the cement floor. For the Gas Gun installation, the reactor must be moved away from its location adjacent to the pilot plant, as there is insufficient overhead and horizontal clearance. • Two block and tackles (or high quality chain pullers) are required to lower and raise the reactor. See Figure A4.2.1. The reactor is balanced on the main support frame on the four horizontal pipes. In order to lower, the reactor must first be pulled off vertical (Tackle #1) before the weight is supported by the lowering tackle (Tackle #2). • Tackle #1: This is used to pull the reactor off vertical. Attach the fixed end of a block and tackle (min 5 tonne capacity) to the horizontal bar opposite the side the reactor is being dropped. An appropriately rated chain or lifting strap can be used. The hook of the tackle is then attached to the eyebolt on the main lower flange at the due south position. Make sure that the tackle does not bind when raised or lowered. Exert slight tension. • Tackle #2: This supports the weight of the reactor once it is pulled off vertical. Attach the fixed end of a block and tackle (min 5 tonne capacity) to the horizontal bar side on which the reactor is being lowered. An appropriately rated chain or lifting strap can be used. The hook of the tackle is 233 then attached to the lowest eyebolt located on the preheater wye pipe support guide. Make sure that the tackle does not bind when raised or lowered. Apply slight tension. • Remove the two northern U-clamps that hold the main column supports. Loosen (do not remove!) the two southern U-clamps so that the supports can pivot. • Put an operator on each tackle. Make sure that there is a clear escape route for each operator if needed. • Slowly pull the reactor the reactor off vertical with Tackle #1. Tackle #2 MUST be tightened at the same time to ensure that there is no slack. • At about 5°, the reactor will be teeter as the weight shifts onto Tackle #2. Lower the reactor slowly with Tackle #2. Take in the slack with Tackle #1. • A wheeled support leg is strapped to the upper reactor. Keep lowering the reactor onto the support leg and then install the two horizontal guides to anchor the support leg to the main frame. 3. Main Flange Removal • Top flange (weight 145 kg): Attach a moveable crane to the flange eyebolt. Be careful to protect the top flange gasket surface. • Remove the bottom flat flange (weight 145 kg). Attach a moveable crane to the flange eyebolt with a tie rope to the bottom of the flange to prevent the flange from swinging out when it is unbolted. Be careful to protect the top flange gasket surface. • Remove the preheater support section (wye) with an appropriately sized crane. The wye weighs 185 kg when empty, about 250 kg if the preheater is installed. Make sure the reactor is anchored when removing to prevent the reactor shifting. • The reactor anchors can now be removed and the reactor assembly wheeled to the desired location. 4. Internal Preheater Assembly (HI) The preheater assembly consists of three separate stainless steel coils (two 3/8" tubes, one V" tube) embedded into a cast lead block shaped into an annular block. One of the coils melted during the block pour, so a separate coil for the oxidant supply was inserted tightly inside the preheater core. The procedure below was used for the preheater assembly, but should not have to be repeated unless there is a failure of some of the components. Figure A4.2.2 is a photograph of the complete preheater assembly ready to be attached to the main reactor wye. 234 • Place the two hemi-spherical Watlow ceramic heaters adjacent to the pre-heater block. Attach to the block with metal strapping on the ceramic shoulders. • Attach pigtail tubing to upstream end o f pre-heater block. The tubing should have several corners between the wye end-cap and preheater to allow for thermal expansion between these two fixed points. • Support wye end cap in the vertical position. • Install blanket insulation into bottom of wye end cap and around heaters. • Place the preheater block assembly into wye piece end cap and rest onto the end support tabs. The three tube ends should protrude through the wye end. Bol t preheater in. • Place heater leads next to the wye shell and bring out lead ends through the main flange connection. Insulate between heaters and the vessel. . • From outside, slip N P T tube fittings over preheater inlets and tighten to shell. Put tube fitting and ferrules over tubes and tighten to seal. Teflon ferrules may be used to make the system removable. • Lif t and bolt the wye assembly to the main reactor flange 5. Main Column Insulation Liner and Power Lead Pass-Throughs Two insulated stainless sheaths are installed in the 16" (0.41 m) diameter column. The following procedure outlines how the sheaths and power leads were installed. Barring a failure, this should not need to be repeated. • Drop the big wye assembly. • Wrap and tape the two 13" (0.33 m) diameter stainless steel liners separately with blanket / paper insulation until diameter o f the 16" (0.41 m) column is achieved. • Insert the longer liner from the bottom flange top until the top is about 23" (0.58 m) from the top flange. The two lower nozzles (B and C) should be accessible from the inside. There are six 240 V A C heater elements installed, each with two leads. The leads are brought through three N P T - Conax fittings on Nozz le B . Each fitting holds four insulated wire leads. • Place additional insulation, for example Teflon tubing, on the Conax leads that w i l l be inside the reactor. • Make sure each end o f the pass-through lead is labelled and then install the three N P T - Conax fittings on the flange. • The 12 leads from the Conax fitting should be secured to the vessel wal l with aluminium tape and directed towards the bottom open flange (M2) . 235 There are a number of pass-through thermocouples on Nozzle C: • Place additional insulation, for example Teflon tubing, on the thermocouple leads inside the reactor. • Make sure each end of the thermocouple is labelled and then install the Conax thermocouple fittings on nozzle C. • The thermocouples should be secured to the vessel wall with aluminium tape and the wire sent towards the bottom open flange (M2). The shorter insulated stainless sheath is now installed in the 16" (0.41 m) column: • Insert the shorter insulated liner from the bottom flange top until it mates the upper liner • Take care that the power and thermocouple leads remain in place against the column wall • The thermocouple for the preheater block can now be placed (TE-110) into the drilled hole in the end of the cast block • Raise and install the wye. 6. Internal Reactor Assembly The internal reactor is made from 304 / 310 stainless steel. The system is designed for the main body of the reactor to remain in the outer reactor, while internals are changed after removal via the top flange. • Position the four Watlow tubular heaters on the internal reactor. Try to have the heater terminals insulated and as far away as possible from the reactor. Attach the heaters with stainless steel strapping. Figure A4.2.4 is a photograph of the four heaters strapped to the reactor, prior to insulation wrapping. • Place the reactor wall thermocouple into the well (TE-108 top, TE-109 bottom). Bring the thermocouple leads down past the bottom flange. • Attach the high temperature N i power leads to the heater terminals. Ensure that the terminals are well insulated. • Insulate top of the heaters and upper reactor with ceramic paper and blanket until the diameter of the insulation is about 12!4" (0.32 m) so as to slip into the stainless liner already installed in the pressure vessel. Ensure that the heater and thermocouple leads are outside the insulation and are taped down with aluminium tape. The heater leads are brought to the bottom of the assembly. • Screw in the stainless threaded rod supports into the 4 tabs on top of the internal reactor. The rods should project about 19" (0.48 m) above the inner reactor flange. 236 • With the top and bottom flanges off, insert the J-shaped lifting jig into the internal reactor from the top. Lift the reactor with the jig with a crane. • Before inserting the reactor, place a 10' (3 m) long 3/8" (9.5 mm) OD tube in the sheath. This will become line transferring oxidant gas from the preheater to the top of the internal reactor. • Insert the internal vessel into the pressure vessel via the top flange. If the insulation is tight, it may be worthwhile to attach a rope or cable puller to the bottom flange to help insert the reactor. Take care not to rip or detach the heater and thermocouple leads. • Take care not to rip the insulation on the inner reactor on the internal pressure vessel tabs. The internal reactor has to be twisted slightly several times to allow the top flange to pass the tabs on the pressure vessel. • Insert the internal reactor until it meets the top of the stainless sheath. Secure the four support rods onto the internal taps with nuts. • When the distributor assembly is ready to be installed, place the gasket and bottom flange via the bottom manway (M2). Use anti-seize on the bottom bolts. The flange bolt heads have been shaved to lock on the backside of the flange when pulled, but might require a little playing while bolting. 7. Heater Connections • All heater and power leads are now in the bottom wye. Connect the appropriate heater leads with the Conax leads with heavy-duty connectors. Insulate the connections well and secure the wire assemblies to the uninsulated shell of the wye. • Drill out the back of the electrical box to match the positions of the three Conax lead fittings on nozzle B. Mount box to nozzle • Attach Conax leads to the appropriate slot on the terminal strip. • Check the resistances of each heater set. Check resistance to ground and the other heaters. • Connect the two reactor wall thermocouples (TE-108, TE-109) to the Conax plugs from nozzle C. 8. Preheater Connections The three outlets from the preheater are connected to the distributors. Remember to use anti-seize on all tubing connections. • Connect the appropriate thermocouple into a tubing tee to measure the preheat temperature (TE-111 main feed, TE-112 secondary feed, TE-113 oxidant). • Make sure that the connections have a couple of bends to allow for thermal expansion. 237 • Pressure test and insulate lines. 9. Insertion of Reactor Internals • Insertion o f the reactor internals can be tricky due the space constraints in the vessel headspace. As many connections should be made as possible before inserting the internals. • Place the top inner gasket, and insert the reactor internals. Make sure the orientation is correct. (See Figure A4.2.3). • Place and tighten top flange o f the inner reactor. The large diameter tubing in the top headspace should be tackled first: • Rel ief / pressure line from the internal reactor (3/8", 9.5 mm) out through nozzle D • Reactor outlet from internal reactor filters (3/8", 9.5 mm) out through nozzle D • Rel ief line from the external reactor (1", 25 mm): cap with screen to prevent plugging and bring out through nozzle D • Filter backflush connections, i f required QA", 6 mm through nozzle D) The reactor internals may consist o f the following and are connected through the inner reactor top flange: • Oxidant distributor: Connect to the preheat tube extending up the internal cladding. • Membrane panels with inlet (1/8", 3 mm) and outlet (Vi", 6 mm). Connect with mating tubing brought through nozzles E and D . • Pressure taps: Most easily brought out by the 1/8" (6 mm) Conax fittings. The taps are then connected in the head space with 1/8" (6 mm) tubes • Thermocouples (1/8", 6 mm): Multipoint thermocouples are passed through the Conax split fitting on nozzle G . Assemble fitting once all thermocouples are in place. Final button-up: • Insulate branch nozzles once all o f the connections are made. • Pressure test the internal reactor and fittings. Check top and bottom connections. 10. Raise Reactor If not already done, the reactor needs to be raised to the vertical. This is done in the reverse order of how it is lowered (described previously, see Figure A4.2.5). I f no overhead crane is available, it may be preferable to attach the top flange ( M l ) before lifting. It can be swung open again once vertical. 238 The reactor can now be rolled to its final location and the four corners of the support structure anchored. The internal reactor needs to be level, note that this may not be the same as levelling the external vessel. Use shims under the main supports or the jackscrews. The main reactor and frame needs to be connected to ground. Once catalyst is filled, the top and bottom headspaces can be insulated and the top (Ml) and bottom (M2) flanges placed. Pressure test the external vessel prior to operation. 11. Final External Connections • Make sure the 240 VAC heater controllers are locked out. Attach the 240 VAC leads and grounds to the heater terminal box. • Connect three feed lines to preheater. • Connect nitrogen purge to preheater wye cap. • Plug in all thermocouples to the PLC • Connect pressure taps to transmitters and purge meters • Connect membrane sweep and permeate lines • Connect ROG (non-permeate reactor off-gas) outlet • Connect rupture disks and pipe to outdoors • If required, connect filter backslash piping. Figure A4.2.1: ICFBMR block positions - Tackle #1 is right (north), Tackle # 2 is left (south) 239 Figure A4.2.2: ICFBMR preheater (HI) - note the three exit connections and heater leads Figure A4.2.4: ICFBMR internal reactor shell and heaters (H2 and H3) Appendix 4.3 - ICFBMR Operating Procedures Pre-Operation 1 Gas chromatography (GC) a. Ensure gas GC is on and calibrated. b. Run gas standard. 2 Start PLC and control system: a. Turn on 120 V A C power switch in heater control cabinet. The PLC indication lights should go on. b. Turn on 120 V A C power switch to solenoids in heater control cabinet. c. Turn on 24 V D C breaker in PLC cabinet. Green light should go on the 24 V D C power supply. d. Turn on 120 V A C breaker in PLC cabinet to power the solenoids (push red button). e. Power up the Human Machine Interface (HMI) computer. f. Start HMI I C F B M R control program by clicking "Pilot Plant- ICFMBR" icon. g. Check that inputs and outputs are connected on working. h. If desired, shutdown and restart PLC data logger. 3 Ensure all cylinders in the outside shed are full and connected. 4 Ensure that fuel lines (hydrogen and methane) are purged to expel any oxygen. 5 Open flow of cooling water to the condenser E2 and check drain to confirm the flow. 6 Line up permeate: a. Ensure that all permeate outlet valves are fully open. b. Close all sweep rotameter valves. c. Close all sweep inlet valves. d. Ensure combined permeate line is vented. 7 Open nitrogen supply and adjust regulators: a. Process (200 psig). b. Purge (200 psig). c. Sweep (20 psig). 8 F i l l water tanks: a. Open vent solenoid valve of desired tank and throttle flow with manual valve. b. When pressure is vented, open manual and solenoid water supply valves. c. Monitor HMI and close water supply solenoid valve when tank is full. d. Close vent valves. 242 9 Check level in E2 - should be small heel of water present for condensation of ROG. 10 Ensure that reactor off-gas (ROG) pressure control valve is fully open in manual (PCV-272) and that valve bypass is open. Route ROG to vent. 11 External preheater: a. Ensure preheater output is turned off in PLC. b. Hook 120VAC power to the preheater magnetic switch in electrical cabinet and energize switch. c. Turn on 220 VAC preheater power supply in electrical cabinet. 12 ICFBMR heaters: a. Ensure the ICFBMR heater controllers are in manual at 0%. b. Ensure the supply switches to the three heaters are off on the ICFBMR heater control box. c. With the 480 VAC supply isolated, measure the resistance of the six ICFBMR heaters at the supply box. (HI ~ 14 Q., H2 and H3 ~ 12 Q.). d. Measure the resistance to ground for each heater set. It should ideally be greater than 0.2 MQ. e. Turn on the 480 VAC supply to the transformer. Supply light should illuminate on the transformer switch box. f. Turn on the 240 - 480 VAC transformer via the transformer switch box. Power supply should be indicated on the transformer switch box and the ICFBMR heater control box. Operation Start-up 1 Purge nitrogen: a. Open flow to purge meters. b. Slowly pressurize the external preheater to 75 psig (520 kPag) with the vent valve closed. c. Pressurize the two water tanks using manual and solenoid valves. d. Make sure that the manual vent from the external ICFBMR vessel is closed. e. Open nitrogen to the external ICFBMR pressure regulator, which should maintain the pressure in the vessel at about 15 psig (100 kPag) over the internal reactor pressure. 2 Introduce process nitrogen: a. Ensure that secondary feed line is lined up to branch from main feed line. b. Ensure process nitrogen 3-way valve is routed to main feed line. 243 c. Ramp process nitrogen flow to 15- 20 S L M . d. Open nitrogen supply to oxidant flowmeter and maintain small flow through the oxidant distributor. Purge permeate lines: a. Open all sweep isolation valves to membranes. b. Crack open sweep rotameters to trickle nitrogen through the membranes. It is important to keep the permeate pressure very low to prevent reverse pressure on the membranes. Start heat-up: a. Turn on external preheater control to 300°C. b. Turn on I C F B M R heater power supply (HI, H2 and H3) and put on low fire (5-15%). c. Heat up the reactor at a rate of <2°C/min, paying special attention to the core membrane temperatures. Fi l l natural gas tanks: a. Check pressure of the two natural gas tanks. The first tank (fill tank) should be at about 2,000 psig and the second (feed tank) at about 1,000 psig. b. If not to pressure, fill each tank to the desired pressure with the natural gas compressor: i . Open inlet valves to compressor. i i . Line up compressor discharge to correct cylinder. i i i . Start compressor and fill to desired pressure. iv. Stop compressor and isolate tank and compressor. c. Depending on the gas usage, the gas cylinders wil l need periodic filling while running. The Fuelmaker compressor has a capacity of about 53 S L M . Introduce hydrogen: a. Open the hydrogen supply and set regulator at 200 psig (1,400 kPag). b. Open the hydrogen supply solenoid. c. Once the core reactor has stabilised at 250°C, gradually introduce hydrogen up to a composition of 15-20% hydrogen. Process nitrogen can be cut back as the reactor heats up to conserve supplies. Raise temperature: a. Heat the reactor at a rate of <3°C/min, paying special attention to the core membrane temperatures. 244 b. Once the core has reached about 500°C, increase the reactor pressure to about 30 psig (200 kPag). Watch the external I C F B M R pressure to see if it follows automatically. If it does not, investigate the operation of the tracking regulator. 8 Introduce steam: a. Check the ROG temperature exit E l , which wil l likely start to show increased temperature. b. Ensure water control valve is closed. c. Open water feed manual / automatic valves on one of the tanks. d. Crack open the water control valve to feed water at a low rate (2-3 kg/h). e. Monitor the feed line temperatures and pressures to ensure that the feed to the reactor is fully vaporized. f. Supply steam for 30 minutes and check ROG composition for any carbon reduction. 9 Introduce methane: a. Line up the second (fill tank) gas cylinder (fill tank at 500 (3,450 kPag) to 1,000 (6,900 kPag) psig) and isolate the connection between the two cylinders. b. Set the gas regulator in the cylinder shed at 200 psig (1,400 kPag) c. Open the methane supply valve in the cylinder shed. d. Open the methane isolation valve upstream of the gas controller. e. Open the methane supply solenoid via the HMI. f. Crack the methane flow and gradually increase to match desired steam-to-carbon ratio. 10 Stabilize operations: a. Maintain temperature uniformity in bed by adjusting heater settings. b. Take samples of the ROG using the direct sample line to GC#1. Once CO or C 0 2 shows up in the ROG to confirm reaction, the hydrogen supply can be closed: i . Shut hydrogen controller via the HMI. i i . Close hydrogen supply solenoid via the HMI . i i i . Close isolation valve in cylinder shed and isolate hydrogen cylinders. c. Take GC analyses of the ROG until reproducible values are achieved. d. Watch the I C F B M R pressure and stabilise operations. e. Sweep nitrogen flows can be gradually increased, but always ensure that the permeate pressure is well below the reactor pressure. 245 Shutoff heaters (4): a. Turn off the three ICFBMR heaters on the HMI. b. Shut the three ICFBMR heaters switches on the control box. c. Turn off ICFBMR transformer and 480 VAC supply. d. Turn off external process heater on HMI. e. Turn off 120 VAC supply to external preheater magnetic switch. f. Turn off 240 VAC supply to external preheater. Reduce NG and steam flows in tandem to a minimum value. The endothermic reforming reaction can be used to help reduce the reactor temperature. Introduce hydrogen to the feed with the H2 mass flow controller. Reduce pressure: a. Slowly reduce the reactor pressure to 30 - 45 psig (200 - 300 kPag). This can be done in conjunction with feed flow reduction. b. As preheater feed pressure drops, manually vent pressure from external preheater shell while reducing nitrogen padding regulator. c. TR-123, the ICFBMR tracking regulator, should vent pressure in the outer ICFBMR vessel to follow the internal pressure, but this has been known not to work. Watch the pressure balance and manually vent pressure if necessary. Note that the rupture disk between the two ICFBMR vessels is designed to relieve at 100 psi (690 kPa) differential pressure. Reduce sweep nitrogen flows to minimum value to conserve nitrogen. Shutoff NG and steam: a. Open process nitrogen to about 10 SLM (ensure it is connected to the main feed line) and adjust hydrogen flow to give desired concentration. b. Shut NG flow via controller: 1. Shut NG feed solenoid via the HMI. 2. Close manual valve just upstream of the NG controller. 3. Close NG delivery valve in the cylinder shed. 4. If not done already, stop gas compressor and isolate suction and discharge. 5. Close and isolate NG delivery tanks. c. Stop steam flow several minutes after NG: 1. Close water feed throttling valves. 2. Close manual and solenoid valve on feed delivery from water tanks. 246 3. Close steam feed isolation valve. 4. If shutting down for a lengthy period, isolate nitrogen pad from both water tanks and vent pressure. d. It is important that all steam be purged from the preheater and reactor. Maintain gas flow for at least 2 hours after steam shutoff. 7 Once the reactor pressure has been reduced to less than 60 psig (400 kPag), fully open the vent from the outer I C F B M R vessel and external preheater. 8 Reduce the reactor core temperature to 250°C. Depending on the process flows this can take many hours. It is important that a significant partial hydrogen of hydrogen is maintained in the reactor during this time to prevent membrane damage. It is also preferable to maintain a small positive pressure (15-30 psig, 100-200 kPag) to ensure that the membrane is pressed onto its support. 9 Stop hydrogen once the membrane temperature has reached 250°C: a. Shutoff hydrogen flow controller via the HMI. b. Close hydrogen supply solenoid valve via the HMI. c. Close manual valve just upstream of the hydrogen controller. d. Close hydrogen delivery valve in the cylinder shed and isolate cylinders. 10 Maintain process and purge hydrogen for several hours to purge hydrogen from the membranes and reactor. 11 Stop sweep nitrogen: a. Keep membrane outlets open. b. Close membrane sweep inlets. c. Isolate nitrogen sweep cylinders and feed valve in cylinder shed. 12 Shutoff process nitrogen: a. Fully open ROG pressure valve (PCV-272) and open manual bypass. b. Reduce and stop process nitrogen via HMI controller. Close process nitrogen supply valve and isolate cylinders in shed. 13 Close purge nitrogen suply valve and isolate cylinders in shed. 14 Shutoff cooling water to E2. 247 Operation Running Checklist Below is a list of operating tasks required to keep the pilot plant running. These tasks are in addition to any experimental activities. Table A4.3.1: Operator task checklist Task Frequency 1 Check cylinder pressures, especially nitrogen: • Once new cylinder is on, check on post-regulator pressure as small differences can make differences, for example the water flow from the tanks Every 30 mins 2 Check for low levels in water tanks on computer and fill when switching: • Make sure second tank is full, close vent (manual and solenoid valve) and open nitrogen pad (manual and solenoid valve) • Open discharge solenoid on second tank • Open manual discharge valve on the second tank and at the same time close the discharge on the first tank • Close water discharge solenoid valve on first (empty) tank • Vent empty tank via manual and solenoid valves • When empty, open manual and solenoid water fill valves • When HMI indicates as full, open manual and solenoid water fill valves • Close vent on filled tank (manual and solenoid) and open nitrogen pad. Tank is now ready to be placed on-line When low level alarm shows on HMI 3 Drain E2 - do not drain fully empty as heel of water is used for condensation. Take care of the hot water under pressure. Every time new water tank is introduced 4 Check permeate line temperatures to ensure no large leaks Hourly 5 Check preheater shell pressure. Manually pad or vent if necessary to maintain a shell overpressure of 15-45 psi 30 mins or more frequently when changing reactor pressure 6 Check external ICFBMR pressure and manually vent if necessary 30 mins or more frequently when changing reactor pressure 248 Appendix 4.4 - Catalyst Microreactor System A microreactor test stand owned by Membrane Reactor Technologies Ltd. was used to evaluate the catalysts used in the ICFBMR pilot reactor. Rigorous catalyst testing was not undertaken; rather simple, standard experiments were conducted to check the reactivity of catalyst samples before and after pilot plant runs. Both the nickel-based SMR and noble metal ATR catalysts were tested. Figure A4.4.1 presents a flow schematic of the microreactor. Mass flow controllers set the gas and water flows. The feed was heated and water vaporized in an electrical preheater prior to being introduced into the heated microreactor. The system could run unattended and was connected to a supervisory computer, which logged data and performed safety shutdowns. Reactor exit gases were analyzed in a gas chromatograph. Figure A4.4.2 presents details of the downflow microreactor. Vent G C Figure A4.4.1: Microreactor flow schematic 2 4 9 F l o w i n Watlow ceramic cylindrical heater (550 W, 372" OD, ID, 12" long) Thermocouple Thermocouple (1/16" OD, moveable) Thermocouple sheath (1/8" OD) Reactor tube (3/8" OD, %" ID) Ceramic microspheres (1 mm diam) Quartz wool Catalyst sample Alumina powder Quartz wool Ceramic microspheres F l o w o u t Figure A4.4.2: Microreactor geometry 2 5 0 Appendix 4.5 - Commissioning Plan -133-u VA" (6 mm) ring heater Section 18 Single 0.6 mm orifice at each of the 4 distribution points VJ" (6 mm) feed line Figure A4.5.1: Original pilot ICFBMR secondary distributor (dimensions in mm) VA" (6 mm) SS tube fitting for inlet Heavy wall %" (15 mm) SS pipe 14-1 mm diameter orifices (7 per side) Figure A4.5.2: Original pilot ICFBMR air distributor 251 4 £ 2 y = 4.77E-04X2 + 2.67E-02x + 2.52E-01 R2 = 0.9998 A . A A — " 0 10 20 30 40 50 60 Rotameter Reading Figure A4.5.3: Calibration of helium rotameter (CHE 4251C) at 1,380 kPag 252 Appendix 4.6 - ICFBMR Pilot Plant Data Table A4.6.1: Gas chromatograph analyses from ICFBMR helium tracer testing (reactor at 1,100 kPa, core at ~595°C, natural gas feed = 42 S L M , steam-to-carbon ratio = 3.0, no air, N 2 content due to purges to pressure impulse lines) Sample location GC analysis (dry) CH4 conversion H2 N2 CO CH4 C02 He Reactor off-gas 45.76 7.29 0.89 29.84 15.61 0.610 35.6% Point 1 (annulus) 45.01 20.75 6.53 10.36 17.06 0.295 69.5% Point E (bottom slot 2) 51.66 10.10 3.11 21.41 13.69 0.030 44.0% Point D (top slot 2) 39.61 30.81 3.46 13.00 13.07 0.043 56.0% Point H (top slot 3) 48.24 11.58 3.39 20.04 16.73 0.031 50.1% Table A4.6.2: Membrane flux testing during run #5 (Pure H2 feed to reactor, no sweep to permeate) Reactor Permeate H2 feed to Measured Predicted flux for Effective Effective / Reactor pressure pressure reactor H2 flux area = 0.0258 m2 permeation theoretical Membrane Date Time temp. (°C) (kPa a) (kPa a) (SLM) (SLM) (SLM) area (m2) area #5 July 11 05 2023 555 206 101 19.1 1.91 2.70 0.0183 70.8% All 6 July 11 05 2049 555 266 101 19.1 11.29 3.93 0.0742 47.9% All 6 July 11 05 2053 555 250 101 19.1 10.59 3.62 0.0756 48.8% #3 July 11 05 2103 555 250 101 19.1 1.81 3.62 0.0129 50.2% #4 July 11 05 2102 555 250 101 19.1 1.38 3.62 0.0099 38.2% #1 July 11 05 2109 555 250 101 19.1 2.02 3.62 0.0144 55.8% #8 July 11 05 2114 555 250 101 19.1 1.74 3.62 0.0124 48.0% #6 July 11 05 2115 555 250 101 19.1 1.12 3.62 0.0080 30.9% #5 July 11 05 2119 555 250 101 19.1 2.47 3.62 0.0176 68.4% All 6 July 11 05 2127 555 150 101 19.1 4.94 1.38 0.0924 59.7% #3 July 11 05 2154 555 150 101 19.1 0.64 1.38 0.0119 46.2% #4 July 11 05 2150 555 150 101 19.1 0.57 1.38 0.0106 41.2% #1 July 11 05 2157 555 150 101 19.1 0.87 1.38 0.0163 63.3% #8 July 11 05 2204 555 150 101 19.1 0.55 1.38 0.0103 39.8% #6 July 11 05 2206 555 150 101 19.1 0.37 1.38 0.0070 27.1% #5 July 11 05 2210 555 150 101 19.1 1.03 1.38 0.0193 74.7% Membrane number #1 #3 #4 #5 #6 #8 Average 0.0154 0.0124 0.0102 0.0184 0.0075 0.0113 Standard deviation 0.0014 0.0007 0.0005 0.0008 0.0007 0.0015 253 Table A4.6.3: Membrane flux testing during run #6 (Pure H 2 feed to reactor, no sweep to permeate) Membrane Date Time Reactor temp. ("C) Reactor pressure (kPa a) Permeate pressure (kPa a) H2 feed to reactor (SLM) Measured H2 flux (SLM) Predicted flux for area = 0.0258 m2 (SLM) Effective permeation area (m2) Effective / theoretical area #4 Aug 18 05 1300 551 164 101 10.0 0.59 1.72 0.0089 34.6% #4 1306 553 174 101 10.0 0.70 1.96 0.0092 35.6% #3 1310 553 180 101 10.0 0.96 2.11 0.0117 45.4% #3 1315 553 182 101 10.0 0.96 2.15 0.0115 44.5% #1 1322 560 180 101 10.0 1.16 2.13 0.0141 54.7% #1 1322 560 180 101 10.0 1.17 2.13 0.0141 54.8% #8 1326 562 178 101 10.0 0.48 2.09 0.0059 22.9% #8 1328 562 178 101 10.0 0.49 2.09 0.0061 23.6% #8 1346 570 208 101 13.6 0.81 2.81 0.0075 28.9% #8 1348 570 213 101 13.6 0.92 2.92 0.0082 31.7% #8 1350 570 213 101 13.0 0.90 2.92 0.0079 30.7% #6 1353 570 217 101 13.0 0.90 3.01 0.0077 29.8% #6 1355 570 217 101 13.0 0.96 3.01 0.0083 32.0% #5 1357 569 219 101 13.0 1.95 3.05 0.0165 64.0% #5 1357 569 219 101 13.0 1.80 3.05 0.0152 58.9% #3 1410 565 199 101 13.0 1.26 2.59 0.0125 48.5% #3 1411 565- 199 101 13.0 1.31 2.59 0.0130 50.5% #4 1416 565 201 101 13.0 1.05 2.63 0.0103 40.0% #4 1417 567 201 101 13.0 1.01 2.64 0.0099 38.4% #1 1419 565 201 101 13.0 1.52 2.63 0.0149 57.8% #1 1420 567 201 101 13.0 1.54 2.64 0.0150 58.2% #8 Aug 19 05 1219 576 204 101 12.0 0.67 2.75 0.0063 24.4% #8 1??5 576 204 101 12.0 0.69 2.75 0.0065 25.2% #6 1227 574 204 101 12.0 0.56 2.74 0.0052 20.3% #6 1228 573 204 101 12.0 0.52 2.73 0.0049 19.0% #5 1231 572 201 101 12.0 1.54 2.66 0.0149 57.8% #5 1233 572 201 101 12.0 1.59 2.66 0.0154 59.8% #3 1237 570 204 101 13.0 1.26 2.72 0.0120 46.4% #3 1238 570 204 101 13.0 1.26 2.72 0.0120 46.4% #4 1242 570 201 101 13.0 1.07 2.65 0.0104 40.3% #4 1243 570 201 101 13.0 1.07 2.65 0.0104 40.3% #1 1246 570 201 101 13.0 1.32 2.65 0.0128 49.7% #1 1247 570 201 101 13.0 1.33 2.65 0.0130 50.2% #6 repeat 1251 570 201 101 13.0 0.52 2.65 0.0051 19.6% #6 1252 570 201 101 13.0 0.51 2.65 0.0050 19.4% #8 Aug 20 05 1130 585 196 101 13.0 1.09 2.60 0.0108 41.8% #8 1131 585 196 101 13.0 1.08 2.60 0.0107 41.6% #8 1135 585 196 101 13.0 0.97 2.60 0.0097 37.5% #6 1139 585 199 101 13.0 0.78 2.67 0.0076 29.4% #6 1140 585 199 101 13.0 0.77 2.67 0.0075 29.0% #5 1143 585 199 101 13.0 1.93 2.67 0.0187 72.4% #5 1144 585 199 101 13.0 1.88 2.67 0.0182 70.6% #3 1150 585 203 101 13.0 1.16 2.76 0.0108 41.9% #3 1151 585 203 101 13.0 1.14 2.76 0.0107 41.4% #4 1146 587 197 101 13.0 1.03 2.63 0.0101 39.3% #4 1147 587 197 101 13.0 0.97 . 2.63 0.0096 37.1% #1 1155 586 203 101 13.0 1.58 2.76 0.0148 57.2% #1 1156 586 I 203 101 13.0 1.56 2.76 0.014E 56.3% Membrane number #1 #3 #4 #5 #6 #8 Average 0.0142 0.