UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Model experiments of autoxidation reaction fouling Wilson, David Ian 1994

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

Item Metadata


831-ubc_1994-893591.pdf [ 7.48MB ]
JSON: 831-1.0058579.json
JSON-LD: 831-1.0058579-ld.json
RDF/XML (Pretty): 831-1.0058579-rdf.xml
RDF/JSON: 831-1.0058579-rdf.json
Turtle: 831-1.0058579-turtle.txt
N-Triples: 831-1.0058579-rdf-ntriples.txt
Original Record: 831-1.0058579-source.json
Full Text

Full Text

MODEL EXPERIMENTS OF AUTOXIDATION REACTION FOULINGbyDavid Ian WilsonM.A. (1988) Jesus College, CambridgeM.Eng. (1989) Jesus College, CambridgeA THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDepartment of Chemical EngineeringWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAMarch, 1994© D.I.Wilson, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.Department of____________________The University of British ColumbiaVancouver, CanadaDate fl t9-DE-6 (2188)AbstractChemical reaction fouling of heat exchangers is a severe problem in thepetrochemicals industry, where deposits can be formed by a wide range of undesirablereactions. Autoxidation has been identified as a prime source of deposit formation inoxygenated process streams and fuel storage systems but the fouling mechanism has notbeen fuily investigated.The fouling of heat exchangers subject to autoxidative fouling was studied usingmodel solutions of an active alkene, indene, in inert solvents saturated with air. The heatexchangers were operated at moderate surface temperatures (180-250°C) and at turbulentflow velocities. The experiments featured air pressures of 342-397 kPa and heat fluxes of90-280 kW/m2.The effects of chemical reaction rate, surface temperature and flow velocitywere investigated and compared with existing chemical reaction fouling models. Chemicalinitiators were used to eliminate the chemical induction periods observed under ‘natural’thermal initiation and permitted the study of the chemical reaction rate and surfacetemperature as separate variables. The chemistry of the physical system and the complexreaction mechanism prevented extensive model development.Two fouling probes were used: an annular probe which allowed visual inspectionof deposit formation and a novel tubular heat exchanger constructed during this workwhich allowed inspection of the deposit in situ after an experiment. The same foulingmechanism was found to generate deposit in both probes.Chemical analyses were developed to monitor the autoxidation reaction during thebatch fouling experiments. The results confirmed that fouling was caused by the depositionof insoluble polyperoxide gums generated by the reaction of indene and oxygen. The gumsilaged on the heat exchanger surface to form complex oxygenated solids which were noteasily removed. These results confirmed the hypotheses of Asomaning and Watkinson(1992).The kinetics of indene autoxidation was studied in a separate series of semi-batchstirred tank experiments. The rate of formation of polyperoxides was influenced by thesolvent nature, temperature, oxygen concentration and mode of initiation. The aromaticpolyperoxides exhibited limited solubility in aliphatic solvents. The kinetics of indeneautoxidation could not be described by the schemes reported in the literature and werefound to be subject to oxygen mass transfer effects.The fouling resistance behaviour was controlled by conditions in the bulk fluid. Theinitial, linear fouling rate decreased with increasing flow velocity and increased with bulkreaction rate and surface temperature. A simple fouling model, involving generation ofdeposit in the reaction zone next to the heat transfer surface and an attachment factor relatedto the mean fluid residence time, was fitted to the experimental data. Once the solubilitylimit was reached, the fouling resistance showed increasing rate behaviour, caused by thedeposition of globules of insoluble gum.The effect of an antioxidant on the fouling process was studied. The efficiency ofthe antioxidant, di-t-butyl-4-methylphenol, was found to be severely reduced under theenhanced thermal conditions in the heat exchanger.Simulated ageing experiments were performed to investigate the fate ofpolyperoxides exposed to the enhanced temperatures on the heat exchanger surface. Thestudies confirmed that the insoluble polyperoxide gums undergo ageing processes afterdeposition.U’Table of ContentsAbstract iiTable of Contents ivList of Tables viiList of Figures xAcknowledgement xviINTRODUCTION 12. LITERATURE REVIEW 72.1 The Role of Autoxidation in Chemical Reaction Fouling 72.2 Experimental Studies in Autoxidation Reaction Fouling 112.2.1 Fuel Stability Studies 132.2.2 Thermal Fouling Studies 182.3 Autoxidation Chemistry 212.3.1 Autoxidation Mechanisms and Kinetics 212.3.2 Solvent and Additive Effects in Autoxidation 252.3.3 The Autoxidation of Indene, C9H8 282.3.4 Antioxidation 302.4 Mechanistic Modelling of Fouling 312.4.1 Transport and Adhesion Models 332.4.2 Transport and Reaction Models 362.4.3 Reaction Engineering 432.5 Objective 443. EXPERIMENTAL MATERIALS AND METHODS 463.1 Materials and Physical Properties 463.1.1 Model Solutions 463.1.2 Physical Properties 503.1.2.1 Kerosene Properties 503.1.2.2 Paraflex Properties 513.1.2.3 Concentration of Dissolved Gases 553.2 Portable Fouling Research Unit (PFRU) Apparatus 563.2.1 PFRU Heat Transfer 623.2.2 PFRU Experimental Procedure 643.3 Stirred Cell Reactor (SCR) 663.4 Tube Fouling Unit (TFU) 683.4.1 Tube Fouling Unit Apparatus 693.4.2 Surface Temperature Measurement 743.4.3 Data Collection and Processing 78iv3.4.4 TFUHeatTransfer 823.4.5 TFU Operating Procedures 863.5 GumAgeingOven 883.6 Chemical Analysis 903.6.1 Hydroperoxide Analysis- Peroxide Number (POx) 913.6.2 Polyperoxide Analysis (Gum in Solution Assay) 923.6.3 Indene Concentration- Gas Liquid Chromatography 953.6.4 Further Analysis- FTIR, SEM 994. AUTOXIDATION OF MODEL SOLUTIONS 1004.1 Solvent Effects in Indene Autoxidation 1004.2 Autoxidation Kinetics 1064.2.1 Mass Transfer Effects in Autoxidation Kinetics 1084.2.2 Kinetics of Autoxidation in Mass Transfer 1124.2.3 Gas Phase Resistance Effects in Mass Transfer 1174.2.4 Oxygen Effects in Autoxidation 1214.3 Chemically Initiated Autoxidation of Indene 1244.4 Temperature Effects in Autoxidation 1274.5 Autoxidation in the Presence of Antioxidants 1324.6 A Kinetic Model of Indene Autoxidation 1354.7 Ageing of Polyperoxide Gums 1464.8 Summary of Autoxidation Studies 1515. INITIAL FOULING EXPERIMENTS 1535.1 Model Solution Selection 1535.1.1 Tetralin as Solvent 1545.1.2 Hexadec-1-ene as Dopant 1565.1.3 IndeneasDopant 1605.1.4 Deposit Characterisation 1625.1.5 Initial Mechanistic Insights 1645.2 Effects of Dopant Concentration 1675.3 Fouling in Model Solutions with Two Dopants 1735.4 Temperature Effects in Thermally Initiated Fouling 1775.5 Velocity and Surface Temperature Effects in ChemicallyInitiated Fouling 1865.6 Stages in Autoxidation Fouling 1915.7 Fouling in the Presence of Antioxidants 1975.8 Summary of Initial Fouling Experiments 2036. FOULING EXPERIMENTS IN THE TUBE FOULING UNIT 2066.1 Autoxidation of Indene in the TFU 2066.2 Initial Fouling Studies 2096.2.1 TubePressureDrop 2116.2.2 Local Fouling Behaviour 2146.2.3 Deposit Distribution and Morphology 218v6.3 Surface Temperature Effects in TFU Fouling 2266.4 Velocity Effects in TFU Fouling 2296.5 Comparison of TFU and PFRU Fouling Probes 2346.6 Summary of TFU Fouling Studies 2377. MODELS OF ASPECTS OF AUTOXIDATION FOULING 2407.1 Autoxidation Kinetics in the Fouling Apparatus 2427.1.1 Model Development 2427.1.2 Model Performance 2467.2 Fouling Mechanisms in the Initial Fouling Regime 2497.2.1 Particulate Fouling Models 2497.2.2 Chemical Reaction Fouling Models 2547.3 Analysis of Fouling Rate and Behaviour 2577.3.1 Formulation of a Lumped Parameter Fouling Model 2587.3.2 Analysis of Experimental Fouling Data 2608. CONCLUSIONS 2709. RECOMIVIENDATIONS FOR FURTHER STUDY 275Abbreviations 277Nomenclature 278References 283APPENDICESA. 1 Experimental Apparatus and Configuration 294A.2 Experimental Summaries 301A.3 Sample Calculations 307A.4 Fouling Model Calculations 312B.1 DataSummary 316viLIST OF TABLES2.1 Examples of Chemical Reaction Fouling in Refineries 82.2 Summary of Autoxidation Related Fouling Studies 122.3 Fouling Results of Taylor (1969b) and Asomaning and Watkinson (1992) 142.4 Activation Energies Reported in Chemical Reaction Fouling 162.5 Velocity Effects Reported in Chemical Reaction Fouling 172.6 Summary of Chemical Reaction Fouling Models 343.1 Alkenes Used in Model Solutions 483.2 Properties of Solvents Used in Model Solutions 493.3 Physical Properties of Initiators and Antioxidants 483.4 Estimates of Dissolved Oxygen Concentration in Paraflex and Kerosene 553.5 Comparison of Nusselt Numbers in PFRU Heat Transfer 633.6 TFUAlarmMatrix 724.1 Solvent Effects in the Autoxidation of Model Solutions of Indene 1044.2 Mass Transfer Effects in Batch Autoxidation of Indene 1104.3 Regression of Thermally Initiated Autoxidation Data to Mass TransferModels 1184.4 Gas Phase Resistance Effects in Autoxidation Kinetics 1204.5 Comparison of Autoxidation Kinetics in the SCR and PFRU 1204.6 Oxygen Effects in the Initiated Autoxidation of Indene in Paraflex 1234.7 Chemically Initiated Autoxidation of Indene in Paraflex 1264.8 Activation Energies in the Autoxidation of Indene in Model Solutions 1294.9 Autoxidation of Indene in Paraflex in the Presence of Antioxidants 1344.10 Elemental Analysis of Aged Polyperoxide Gums 150vii5.1 Summary of Initial Fouling Experiments 1555.2 Chemical Analysis of Deposits From Initial Fouling Runs 1655.3 Effects of Indene Concentration in Fouling Experiments 1695.4 Fouling from Solutions of Indene and Dicyclopentadiene in Paraflex 1745.5 Temperature Effects in Thermally Initiated Fouling Experiments 1785.6 Elemental Analysis of Fouling Deposits 1855.7 Effects of Bulk Chemical Parameters on Fouling and Reaction Behaviour 1855.8 Reaction Diagnostics from Chemically Initiated PFRU Fouling Experiments 1895.9 Effect of Flow Rate on Initial Fouling Rate in Initiated PFRU Experiments 1925.10 Reaction Diagnostics and Fouling Summary of Interrupted Fouling Runs 1945.11 Fouling in the Presence of an Antioxidant (2,6,di-t-butyl 4-methylphenol) 1996.1 Comparison of Indene Autoxidation Kinetics Under TFU Conditions 2086.2 Initial Conditions and Tube Pressure Drops in TFU Fouling Runs 2156.3 Local Heat Transfer and Fouling Behaviour in Run 504 2166.4 Comparison of Sand Roughness Criteria with Gum Globule Dimensionsin TFU Fouling Run 503 2246.5 Surface Temperature and Flow Velocity Effects on Initial Fouling Rate 2276.6 Elemental Analysis of Deposits from TFU Runs 2307.1 PFRU Reaction Coupling Model Parameters 2437.2 Experimental Data from Thermally Initiated Fouling Experiments0.41M Indene in Paraflex; 79 kPa Oxygen Saturation 2487.3 Particle Relaxation Times and Mass Transfer Coefficients in ParticulateFouling Calculated at Conditions Used in Fouling Runs 2527.4 Comparison of Initial Fouling Rate Activation Energies 261vm7.5 Comparison of Activation Energies of Adjusted Fouling Rate (Rf7V) 262A. 1.1 Experiment Nomenclature 294A. 1.2 TFU Equipment Specifications 295A.2. 1 Comparison of Indene Autoxidation Kinetics under TFU Conditions 301A.4 Fouling Model Calculation Spreadsheet 314B. 1 Summary of PFRU Fouling Experiments #001-016, 030 317B .2 Summary of Thermally Initiated PFRU Experiments #017-036 318B .3 Summary of Chemically Initiated PFRU Fouling Experiments 319B .4 Summary of Final PFRU Fouling ExperimentsAntioxidant, Interrupted and Comparison Studies 320B.5 Summary of SCR Experiments Performed by Lai and Wilson 320B .6 Summary of SCR Experiments 321B .7 Summary of Tube Fouling Unit Experiments 322lxLIST OF FIGURES1.1 Heat Transfer in the Presence of Fouling Deposits 31.2 Types of Fouling Behaviour 52.1 Autoxidation of Indene 292.2 Antioxidant Action of 2,6,di-t-butyl 4-methylphenol (BMP) 292.3 Schematic of Processes Involved in Chemical Reaction Fouling 322.4 Comparison of the Experimental Fouling Data of Crittenden et al. from thePolymerization Fouling of Styrene in Kerosene and the Predictions fromEpstein’s Fouling Model (1993b) 392.5 Panchal and Watkinson Autoxidation Fouling Models 423.1 Kinematic Viscosities of Solvents Used in Model Solutions 523.2 Prandtl Number of Solvents Used on Model Solutions 523.3 Viscosity of Mixtures of Indene in Paraflex at 20°C; Comparison ofCorrelation Predictions and Experimental Data 543.4 Schematic Diagram of PFRU Apparatus 573.5 PFRU Orifice Plate Calibration 583.6 Schematic Diagram of PFRU Fouling Probe 603.7 Schematic Diagram of Stirred Cell Reactor (SCR) Apparatus 673.8 Schematic Diagram of Tube Fouling Unit Apparatus 703.9 Schematic Diagram of Heated Section Construction 753.10 Details of TFU Surface Thermocouple and Mounting Designs 773.11 Comparison of Predicted TFU Nusselt Number with Experimental Data 833.12 Thermal Entry Length Effects in TFU Heat Transfer 833.13 Comparison of Experimental Fanning Friction Factor Data with theCorrelations of Blasius and Gnielinski 85x3.14 Schematic Diagram of Gum Ageing Oven Apparatus 893.15 Photograph of Gum Assay Filters Showing Change in Gum Appearance 943.16 Gas Chromatogram of Solution of Indene in Paraflex 963.17 Gas Chromatograms of Indene in Kerosene from Flame Tonisation andPhoto-lonisation Detectors in Series 984. la-c Solvent Effects in Indene Autoxidation: Analysis Results 1024.2 Kinetic Plots of Indene Concentration Data: Comparison of Kinetic Models 1074.3 Comparison of Mass Transfer Kinetic Models of the Autoxidation ofIndene in Paraflex 1184.4 Temperature Effects on the Gum Solubility Limit, g*, in Paraflex 1284.5 Variation of the Gum Solubility Limit, g*, with Indene Concentration 1284.6 Temperature Effects in the Thermally Initiated Autoxidation of lOwt%Indene in Kerosene: Peroxide Number Behaviour 1294.7 Kinetic Model of Indene Autoxidation: Effect of K on Mass Concentrationof Gum and Comparison with Data from Runs 140, 141 (e = 1) 1404.8 Kinetic Model of Indene Autoxidation : Effect of K on Peroxide Numberand Comparison with Data from Runs 140, 141 (e = 1) 1404.9 Kinetic Model of Indene Autoxidation: Effect of e on Mass Concentrations(K = 0.025m3/mol.min: Data from SCR Runs 140, 141) 1414.10 Kinetic Model of Indene Autoxidation : Effect of e on Peroxide Number(K = 0.025m3/mol.min: Data from SCR Runs 140, 141) 1414.11 Kinetic Model of Indene Autoxidation : Revised Mass Concentration Curvesat K 0.025m3/mol.min with Data from SCR Runs 140, 141 1424.12 Kinetic Model of Indene Autoxidation: Revised Peroxide Concentrations atK = 0.025m3/mol.min with Data from SCR Runs 140, 141 1424.13 Kinetic Model of Indene Autoxidation: Comparison of Revised MassConcentration Curves and Data from SCR Run 142 at Tb = 120°C 1444.14 Kinetic Model of Indene Autoxidation: Comparison of Revised MassConcentration Curves and Data from SCR Run 142 at Tb = 120°C 144)4.15 Kinetic Model of Indene Autoxidation: Comparison of Revised MassConcentration Curves with Data from TFU Runs 501, 503, 504 1454.16 Kinetic Model of Indene Autoxidation : Comparison of Revised PeroxideConcentrations with Data from TFU Runs 501, 503, 504 1454.17 FTIR Spectra of Samples from Gum Ageing Experiments at 200°C 1474.18 Reduction in Gum Mass During Ageing Oven Experiments 1494.19 Reduction in Gum Mass During TGA Ageing Experiments 1495.1 Peroxide Number Analyses from Initial Fouling Experiments using ModelSolutions with Tetralin as Solvent 1575.2 Fouling Resistance and Hydroperoxide Concentration Profiles fromFouling of Hexadecene in Kerosene in Run 014 1595.3 Fouling Resistance and Hydroperoxide Concentration Profiles fromFouling of Indene in Kerosene in Run 013 1615.4 Photograph of Fouled PFRU Probe Following Fouling Run 014.Thermally Initiated Autoxidation of 0.388 M Hexadecene in Kerosene. 1635.5 Deposit Thickness Profile of PFRU Probe Fouled by the Autoxidationof 0.388 M Hexadecene in Kerosene 1635.6 FTIR Spectra of PFRU Deposit and Soluble Gum from AutoxidationFouling of Indene in Kerosene in Run 013 1665.7 Fouling Resistance and Peroxide Number Profiles from Indene in Paraflex 1695.8 Kinetic Fit of Peroxide Data from Initial Indene Fouling Runs to Eqn.[4.36] 1725.9 Fouling Resistance Profiles in Indene/DCP Studies 1755.10 Peroxide Number versus Time in Indene/DCP Studies 1755.11 Temperature Effects in Thermally Initiated Fouling Experiments in Paraflex 1805.12 Effects of Surface Temperature on Fouling Resistance Profiles in ThermallyInitiated Solutions of Indene in Paraflex Runs 025, 028, 029, 031, 032 1805.13 Gum Concentration and Fouling resistance in Thermally Initiated Fouling 182x5.14 Surface Temperature Effects on Initial Fouling Rate in Thermally andChemically Initiated Fouling Runs 1895.15 Log-log Plot of Velocity Effects in Chemically Initiated Fouling 1925.16 Photomicrographs of PFRU Surface Deposit from Interrupted Runs 1955.17 Autoxidation Fouling in the Presence of Antioxidant: Gum Concentrations 1965.18 Fouling Resistance Profiles in the Presence of an Antioxidant 1965.19 Effect of an Antioxidant (BMP) on Reaction and Fouling Induction Periods 2026. la-c Comparison of Indene Autoxidation in SCR, TFU and PFRU 2076.2 Fouling Resistance Profiles Showing TFU Reproducibility 2106.3 Overall Fouling Resistance and Soluble Gum Concentration in Run 503 2126.4 Tube Pressure Drop and Soluble Gum Concentration in Run 505 2136.5 Local Fouling Resistances in Run 504 2176.6 Photograph of Fouled TFU Tube from Run 507 after Division into Sections 2196.7 TFU Test Section Deposition Regions 2196.8 Photograph of Thermal Entry Length Deposition in Run 507 2196.9 Optical and Scanning Electron Micrographs of Deposit from Run 507 2236.10 Variation in Deposit Coverage with Axial Position in TFU Runs 502-504 2246.11 Arrhenius Plot of Initial and Final Fouling Rates in TFU Experiments 2276.12 Plot of Final Local Fouling Resistance Against TFU Deposit Coverage forEstimation of Deposit Thermal Conductivity 2306.13 Tube Pressure Drop and Overall Fouling Resistance in TFU Run 510 2326.14 Comparison of Fouling Resistance Profiles in TFU and PFRU 2356.15 Fouling Resistance Profiles From Initial PFRU Runs at High IndeneConcentrations showing Unusual Fouling Behaviour 2387.1 Compartmental Kinetic Model of PFRU System 243xiii7.2 Regression of First Order Indene Rate Constant from Data fromSCR Experiment 122 2477.3 PFRU Reaction Coupling Model Predictions of Volume Effects in KineticParameters in the PFRU System 2477.4 Comparison of Reaction Coupling Model Predictions and ExperimentalInduction Periods in Thermally Initiated Fouling Runs in Paraflex 2487.5 Schematic of Fouling Model Mechanism 2597.6 Arrhenius Plot of Initial Fouling Rates 2617.7 Arrhenius Plot of (RfYV) Against Surface Temperature 2627.8 Arrhenius Plot of 0 versus Inverse of Surface Temperature for PFRU Runs 2657.9 Fouling Model Analysis: Variation of 0(210°C) with Friction Velocity 2677.10 Paterson and Fryer Fouling Model Analysis: Variation of 0(210°C)ISthwith Friction Velocity 269A. 1.1 TFU Orifice and Rotameter Flow Calibrations 296A. 1.2 Schematic Diagram of TFU Power Circuitry 297A. 1.3 Schematic Diagram of TFU Signal Processing System: Meter Outputs 298A. 1.4 Schematic Diagram of TFU Heated Section Voltage Processing Circuit 299A. 1.5 Schematic Representation of Mass Transfer Effects in Gas/Liquid Reactions 300A.2. 1 Effects of Surface Temperature on Chemically Initiated Fouling Solutionsof Indene in Paraflex 302A.2.2 Effect of Flow Velocity on Initiated FoulingTSurf 222°C a. Fouling Resistance Profiles; b. Gum Analysis Results 303Tsurf 248°C c. Fouling Resistance Profiles; d. Gum Analysis Results 304A.2.3 Peroxide Number, Soluble Gum and Indene Concentration Results inTFU Fouling Runs- Tsurf Varied 305xivA.2.4 Peroxide Number, Soluble Gum and Indene Concentration Results inTFU Fouling Runs - Flow Velocity Varied 306A4. 1 Estimation of Friction Factor Velocity Dependence using the j-Factor Heatand Mass Transfer Analogy: Variation of Nusselt Number with Re 313xvAcknowledgementMy sincere thanks are due to all members of the Department of ChemicalEngineering for their encouragement and company in the course of this study. I amparticularly indebted to my supervisor, Prof. Paul Watkinson, for his advice, patience andsupport over the last five years.This study is a testament to the skill and experience of the workshop, stores andtechnical staff of the department. I am profoundly grateful for their long suffering patience,humour and general humanity. The contributions of Dr. K.C. Teo, Ronald Lai, SamuelAsomaning and Dr. Guohong Zhang are gratefully acknowledged.The financial support of the University of British Columbia and the NaturalSciences and Engineering Research Council of Canada are gratefully acknowledged, as isthe support of the Royal Ulster Constabulary Benevolent Fund.This volume is dedicated to William James Wilson, the man who encouragedme to seek but did not live to see me find.xvi1. Introduction1. IntroductionHeat exchanger fouling is generally defined as the unwanted accumulation ofdeposit on heat transfer surfaces (Taborek et at., 1972). Deposit accumulation increases theoperating costs of process plant heat exchangers as fouling reduces the efficiency of heattransfer and increases frictional energy losses. Fouled heat exchangers have to be cleanedor replaced, which involves significant capital and operating costs associated with cleaningand plant down time. The economics of heat exchanger fouling are discussed in detail byVan Nostrand et at. (1981), Garret-Price et at. (1985) and Pritchard (1990), who estimatedfouling related costs in UK industry as 0.3% of GNP in 1990.Taborek et at. described fouling as ‘the major unresolved problem in heat transfer’in 1972, a view repeated by Chenoweth in 1990. The improved design of heat exchangersafforded by advances in heat transfer science and technology can be negated in industrialpractice by fouling, which is a widespread and poorly understood phenomenon. The goalof engineering fouling research is to understand the mechanisms and rates of fouling so asto either prevent fouling or to design robust, fouling-resistant heat exchangers (Fryer et al.,1990). The design of fouling resistant heat exchanger networks has been discussed byKotjabasakis and Linnhoff (1987) and Fryer (1987), as the effect of fouling in networks isoften observed later than desirable.Heat exchanger designers currently compensate for fouling by overdesign or byusing specialised antifouling equipment. The latter devices are usually expensive andinvolve cleaning-in-place strategies, such as the Taprogge system (Bott, 1988) or thefluidised bed heat exchanger (Kiaren, 1983). The use of turbulence promoters inconventional shell and tube units (Gough and Rogers, 1987; Crittenden et at., 1993) andother surface geometries (Rabas and Chenoweth, 1991) is increasing as the increased11. Introductionpressure drop is offset by enhanced heat transfer capacity and superior antifoulingperformance under certain fouling conditions. These devices feature enhanced surface shearrates to retard deposition and are often used without a complete understanding of thefouling mechanism involved. Zhang et at. (1993) related the antifouling behaviour to thelower heat transfer surface temperature necessary to achieve a given heating duty in onesuch device.Heat exchangers for fouling service are designed with excess heat transfer capacityin order to offset the losses in efficiency caused by fouling. Equation [1.11 is the generaldesign equation for a simple heat exchanger transferring Q Watts over an area A withtemperature driving force im AT as shown in Figure 1.1.Q = UAImAT [1.1]The overall heat transfer coefficient, U, is related to individual heat transfer contributionsby1/U = 1/h + ôf,1/?\.f,1 + ômetlXmet + óf,2!?f,2+ 1/h2 [1.21when A1 = A2 and subscripts are defined in Figure 1.1. As deposit thickness Of] or 0f2increases over time, U decreases. The usual design strategy is to compensate for theeventual reduction in the overall heat transfer coefficient, U, by increasing the heat transfersurface area, A. The decrease in heat transfer performance is often described by a foulingresistance, Rj, defined asRf = 1/U(t) - 1JU(t=O) [1.3]Since Rf is defined in terms of heat transfer it includes the effects of deposit thickness,surface roughness and deposit composition variations. The value of Rf chosen for a set ofconditions (fluid, velocity, temperature) is usually a correction factor based on previousexperience published by the Tubular Exchangers Manufacturers’ Association (TEMA).This value is often used in design (erroneously) as an asymptotic value of foulingresistance which assumes that fouling will reach a steady state after which no further21. IntroductionFigure 1.1 Heat Transfer in the Presence of Fouling DepositsHeating MediumT11 IFlow1 refers to the processed fluid, 2 refers to the heating medium, ômet is the tube thickness,and of2 the foulant thicknesses. The heat transfer driving force is the log meantemperature difference, lmz\T, defined as(T2, - T1,11)- (T2,0 - Ti,0)1mAT = (T2,- T1 in)in (T2,0- T1,0)for a cocurrent heat exchanger shown hereTi0,T20oil6met6f231. Introduciiondeposit is formed. In many industrial cases such ‘asymptotic fouling’ is never observed.The TEMA approach has been criticised as encouraging a steady state solution to a dynamicproblem (Bott and Walker, 1971) which ignores the role of optimal operating conditions indecreasing fouling rates. The commonly observed types of fouling resistance behaviourand associated terms are shown in Figure 1.2. The improper use of the TEMA standardscan exacerbate fouling, and some examples are given by Bott (1990) in The FoulingNotebook, a nomograph intended to improve engineering practice. The scope for improvedheat exchanger design is significant and requires reliable understanding of the mechanismsinvolved in deposit formation, removal and cleaning.Fouling occurs in diverse media and so Epstein (1983) classified fouling by themechanism responsible for deposit generation. Epstein identified five classes of fouling asbiofouling, particulate fouling, crystallisation fouling, corrosion fouling and chemicalreaction fouling. Chemical reaction fouling was defined as the formation of deposit at theheat transfer surface by chemical reaction in which the surface itself is not a reactant;surface participation is included in the corrosion fouling category. Chemical reactionfouling is an extremely widespread phenomenon as many process fluids undergo foulantgenerating reactions when heated. Examples of chemical reaction fouling are common inthe chemical, nuclear and food industries. Bohnet (1987) identified organic fluid fouling asone of the areas of fouling lacking conclusive study as the diversity of possible reactionspresents a complex matrix of possible or concurrent mechanisms. This study addresses onearea of chemical reaction fouling in the processing of organic fluids, where deposits areformed by the autocatalytic oxidation (autoxidation) of compounds in the processed stream.The first engineering analysis of refinery fouling was performed by Nelson in 1934but the level of understanding is still relatively sparse. Industry attention to chemicalreaction fouling has focussed on fouling mitigation through the overdesign of process plantand the use of chemical additives to inhibit reaction, disperse foulant precursors or weaken4FoulingResistanceRfRf*0asymptoticfoulingincreasingratefoulingsawtoothfoulinglinearfoulingfalling ratefoulingTime51. Introductiondeposits for efficient cleaning (Cowan and Weintritt, 1978; Nathan 1970). The mechanisticunderstanding of chemical reaction fouling remains relatively unclear despite the significantcosts of chemical reaction fouling and the operating expense of antifouling programmes.This work describes an investigation of chemical reaction fouling caused byautoxidation reactions in oxygenated organic solutions being heated in conventional heatexchanger geometries operating at moderate surface temperatures (100-250°C) in theturbulent flow regime. A known chemical reaction system was chosen using modelsolutions of an active component, indene, in relatively inert solvents. Characteristics of thechemical reaction which generated fouling precursors were studied separately in a series ofbatch reactor experiments and used to develop a model of indene autoxidation and toexplain the fouling resistance behaviour observed in the fouling experiments.Section 2 is a summary of relevant published material describing autoxidation andchemical reaction fouling. Section 3 describes the experimental methodologies used. Theautoxidation chemistry of the model solution is described in Section 4. Initial foulingexperiments performed to study the effect of operating parameters are summarised inSection 5; Section 6 describes model fouling experiments performed in a novel fouling unitconstructed for this research (the Tube Fouling Unit, abbreviated to TFIJ). Section 7 is asummary of fouling model investigations.62. Literature Review2. Literature Review2.1 The Role of Autoxidation in Chemical Reaction FoulingThe study of organic fluid fouling is complicated by the diversity of the mediuminvolved. Refinery feedstocks and process streams contain species which can undergo arange of chemical reactions depending on composition and the operating conditionsinvolved. Lawler (1979) and Scarborough et at. (1979) concluded that any fluid’s foulingbehaviour under given conditions of temperature and pressure is dominated by its chemicalcomposition. Organic fluid fouling often involves chemical reaction fouling in conjunctionwith other fouling mechanisms, particularly corrosion fouling, particulate fouling andprecipitation fouling. Heat exchangers downstream of crude oil desalters are subject to allfour processes so that deposits from these units often contain rusts, sand and organicmaterial. Table 2.1 shows the operating conditions and feedstocks involved in variousrefinery heat exchangers subject to señous fouling and their likely deposition mechanisms.Alkenes and asphaltenes occur most often and these chemical groups have been identifiedas the primary sources of fouling deposits in petroleum refining (Dickakian and Seay,1988, Eaton and Lux 1984, Lambourn and Durrieu 1983) due to their tendency to forminsoluble particulates in solution. Studies of fuel storage stability which focus on chemicalreactivity rather than heat transfer, such as that by Schwartz et at. (1964), identified alkenesas the major source of gums during fuel storage.Watkinson’s extensive review of organic reaction fouling (1988) confirmed theprimacy of asphaltenes and alkenes in organic reaction fouling. Asphaltenes are large,complex ring structure molecules found in heavier crudes and bitumens which formdeposits both by reaction to form coke-like deposits and by precipitation. Asphaltenefouling is discussed by Dickakian and Seay (1988), Scarborough et at. (1979) andCrittenden et at. (1993). The present work is concerned primarily with alkene fouling,72. Literature ReviewExamples of Chemical Reaction Fouling in RefineriesProcess Stage Heater Fluid Stream Maximum Wall Deposition Source Reference__________________Temperatures [°C]Crude Preheat Crude/residues 374 Asphaltenes and Lambourn & DurrieuFeS particulates (1983)Crude Preheat Crude/residues 400 Organics, rust N.A.C.E. (1970)Crude Preheat Crude/residues 310 Asphaltenes Weiland et at. (1949)Crude Preheat Crude/residues 250 Salt, FeS, Coking Crittenden et al. (1992)Catalytic Cracker Recirculation 357 Oxygenated alkenes Butler et a!. (1949)reflux/gas oilCatalytic Reformer Gas oil 350 Viscous sludge Weiland et at. (1949)/catalyst slurry from alkenesCatalytic Reformer NaphthaJeffluent 510 Unsaturated species Dugan et at. (1978)Hydrodesulphunzer Naphtha/effluent 440 Organics, FeS, rust N.A.C.E. (1970)Hydrodesuiphurizer Naphthalkerosene 440 Unsaturated species Dugan et al. (1978)Naphtha Boiler Naphtha 340 Unsaturated species Butler et a!. (1949)Reformer Gas Oil 510 Organics,rust, FeS N.A.C.E. (1970)NH4C1Coker-Visbreaker Residues 540 Organics NA.C.E. (1970)Alkylation Unit 240 Organics/rust N.A.C.E. (1970)Hydroformer Naphtha 180 Oxygenated alkenes Weiland et at. (1949)Desulfurizer Naphtha 360 Metal ions, FeS Crawford & Miller (1963)Table 2.182. Literature Reviewwhich Watkinson identified as forming fouling deposits via three reaction mechanisms;pyrolysis/condensation, polymerisation and autoxidation.At higher temperatures, such as those found in cracking furnaces and decokers, thepyrolysis of hydrocarbons generate a range of free radicals which combine to form cokesand long chain polymers. Condensation of alkenes via Diels-Alder reactions can alsoproduce longer chain compounds. At temperatures greater than 700°C, alkanes undergothermal pyrolysis to form alkenes and other products which foul the cracking reactorsextensively. Alkane pyrolysis reactor coking has been studied extensively by Froment andco-workers (1981) so that simulations can be used to predict reactor operating behaviour.At lower temperatures, usually in the liquid phase, the nature of the alkene free radicalreaction is determined by the operating conditions. In the absence of dissolved oxygen,alkenes can undergo addition polymerisation to form medium to long chain polymericspecies which are insoluble in solution or undergo decomposition on heated surfaces. Freeradical initiation occurs via metal ions or heteroatomic species present in solution, as wellas thermal generation of free radicals in solution. Polymerisation fouling has been studiedby Palen and Westwater (1966), Fetissoff (1982), Crittenden et al. (1987a) and Oufer(1990).In the presence of dissolved oxygen, alkyl and allyl radicals are oxidised to alkoxyor peroxy radicals which form peroxides and other oxygenated products via autoxidationreactions. Alkenes undergo autoxidation readily and can form polymeric peroxides whichdecompose to form insoluble deposits in heat exchangers. Section 2.3 is a review ofautoxidation chemistry. Watkinson’s review (1988) and that of Crittenden (1988b)identified autoxidation as a mechanism meriting further study owing to the relative lack ofheat transfer studies involving autoxidation fouling and its significance in industrialapplications.The key role of oxygen and autoxidation in hydrocarbon plant fouling has beenreported by Canapary (1961) and Butler et al. (1949), who achieved lower fouling rates by92. Literature Reviewstripping the process feedstocks of air. Braun and Hausler (1976) reported reduced foulingrates from crude oil feedstocks at reduced oxygen concentrations and showed that thedeposit morphology also changed with oxygen concentration. Eaton and Lux (1984)reported complicated oxygen effects in their crude oil fouling studies. Alkenes are found inhigher concentrations in cracked feedstocks: Gillies (1979), Canapary (1961) andCrawford and Miller (1963) all traced the heavy fouling in these alkene-rich streams to thealkene components. The effect of autoxidation initiators such as metal ions wasdemonstrated by Crawford and Miller, who observed higher fouling rates when copperions were added to their feedstocks. The deposits formed in plant fouling indicate anautoxidative source; Vranos etal. (1981) found oxygen concentrations of 10-20 wt% in jetfuel deposits and infra-red analysis indicated that the solids contained significant levels ofhydroxyl groups and ketones generated by hydroperoxide decomposition.The formation of insoluble gums in hydrocarbon fuels during storage has beenstudied due to the undesirable effects of such materials on end use performance. Schwartzet al. (1964) identified the autoxidation of alkenes as a primary source of gums andsubsequent studies have confirmed this (Mayo and Lan 1986). Reaction mechanisms andfuel composition effects have been studied extensively, as well as the role of oxygen,initiators (metal ions, ultra violet light), heteronuclear species and inhibitors (antioxidantsand chelating agents). Recent experiments investigating specific aspects of fuel stabilityhave been performed using ‘model’ solutions of compounds in relatively inert base stocksused as a control. Morris et al. (1991) have thus used solutions of indene (a reactivealkene) in JP5 jet fuel to assess the effect of various thiols in jet fuel storage. Mayo et al.(1988) related gum formation tendencies to the ability of dopant species to form polymericperoxides. Although not strictly fouling studies, these low temperature experiments (20-120°C) contain valuable information about hydrocarbon reactivity and autoxidation.Watkinson (1988) commented that if the extensive library of information available fromfuel stability studies were to be connected to fouling behaviour, it would greatly reduce the102. Literature Reviewexperimentation required to understand the interaction of autoxidation, solutioncomposition and fouling. Asomaning and Watkinson (1992) studied the autoxidation ofselected alkenes in kerosene from this basis and found considerable agreement withTaylor’s jet fuel stability studies (1969b).Autoxidation has thus been identified as a significant source of deposits in industrialheaters and of gums in fuel storage. There is a wealth of previous basic chemical researchin autoxidation. The mechanism of autoxidation fouling is still poorly understood,however, so most industrial research has focused on the mitigation of such fouling bychemical treatment rather than by optimising design and operation. Antioxidant technologyis valuable commercial property and so the literature is seldom current.2.2 Experimental Studies in Autoxidation Reaction FoulingFew studies extending the understanding of fuel storage stabilities to thermalfouling behaviour have been performed, although suggested in reviews (Watkinson, 1988;Crittenden, 1988b). Similarly, autoxidation has been suggested as the primary source ofdeposition in oxygenated feedstock fouling but few results have been published linking theeffects involved quantitatively. This is needed for the design of robust antifouling heatexchangers, such as described by Bradley and Fryer (1992) using an existing foulingmodel to assess the scope of various design and operating strategies.Table 2.2 is a summary of experimental studies of fouling where autoxidation hasbeen identified as a primary source of deposition. These studies can be divided into reactionmechanism studies and thermal fouling studies. Little work has been performed toquantitatively link autoxidation chemistry to foulant generation and thermal effects, whichis a goal of the current study.112. Literature ReviewTable 2.2 Summary of Autoxidation Related Fouling StudiesFeedstock Fouling Probe, Temperature Pressure Experiment Reference(Measurements) Range Period(°C) (kPa) (hours)Jet fuels, doped HeatedTube 150 <Tb <260 20.6 t < 4 Taylor eta!.alkanes [oxygenated] . (1967, 68a,b, 69a,b)(Mass Deposition)Jet fuels, doped Heated Tube 150 <Tb < 560 6990 t < 4 Taylor et a!.alkanes [deoxygenated] (1974, 76, 78, 80)(Mass Deposition)Oil Refinery Hot Wire Probe 200 < Tw < 340 1480 t < 50 Hausler and ThalmayerFeedstocks . (1975), Braun and(Thermal Fouling) Tb < 100 Hausler (1976)Crude oil, kerosene, Hot Wire Probe 175 <T <400 Pressurised t < 3 Latos and Frankeshale oils. (1982)(Thermal Fouling) Tb 90Crude Oil Refinery Heated Probe, T <287 2000 t 20 Eaton and Lux (1984)cuts with pitch, resin Rotating Cylinder 71 <Tb < 287(Thermal Fouling)Jet Fuel Heated Tube 120 <Tb < 355 6990 t < 16 Vranos eta!. (1981)(Mass Deposition)Aerated n-dodecane Modified JFT’OT 190 < Tw < 538 5597 t < 2.5 Hazlett eta!. (1977)(Mass Depsoition)Jet fuels, Stabilizers Modified JFTOT T <310 5597 t < 2.5 Morris et al. (1988,(Mass Deposition) 1989)Model Fuels, Modified JFTOT 190 <T <538 5597 t <2.5 Morris and MushrushAdditives. . (1991)(Mass Deposition)Aerated kerosene Tubeside 160 < T < 380 106 - 253 t 100 Crittenden and Khater(1984)Vapounsation(Thermal Fouling)Alkenes in aerated Hot Wire Probe, 140 < Tw <204 480 t < 50 Asomaning andkerosene Watkinson (1992)Annular Probe Tb 80(Thermal Fouling)Styrene in heptane Annular Probe 151 <T < 190 1034 t < 68 Oufer (1990),(Thermal Fouling) 87 < Tb < 100Tb - bulk temperature (°C); T - maximum wall temperature (°C)122. Literature Review2.2.1 Fuel Stability StudiesThe fuel stability studies of Taylor et al. (1967 - 1980), Hazlett et at. (1977) andHausler and Thalmayer (1975) involved passing hydrocarbon streams over heated surfacesunder evaporative conditions and measuring the formation of deposits and of oxidationproducts in solution. The effect of surface temperature is reported in terms of activationenergies for deposit formation rates based on an Arrhenius type expression, i.e.Deposition rate dmf/dt = B exp [- Eact!R Tsurt] [2.11or as a threshold temperature below which little deposition occurs. The effect of liquid flowrate and thus hydrodynamics was not reported. Taylor initially studied deposition from jetfuel blends doped with various additives under oxygenated and deoxygenated conditions.Deposition rates were significantly higher in the presence of dissolved oxygen, andsolution and solid analyses were consistent with an autoxidation mechanism. Depositionfrom blends doped with organic oxygen, nitrogen and sulphur species varied withcompound structure; peroxides and their analogues increased deposition while phenolanalogues reduced or maintained deposition rates. Hazlett et al. (1977) reported similarresults for sulphur derivatives using the Jet Fuel Thermal Oxidation Tester (JFTOT), acommercial device used to assess the fouling tendency of process liquids. Taylor (1968b)studied deposition on various metal surfaces and found enhancement with metals that couldinitiate radical formation. Steels were relatively inert compared to copper, the metal used byCrawford and Miller (1963) in their crude oil tests.Taylor (1968a,b, 1969b) used ‘model solutions’ of 10 wt% alkanes, alkenes andaromatics in n-dodecane to study the effect of jet fuel component structure on depositionrates and found heavy deposition from alkenes. Table 2.3 is reproduced from Asomaningand Watkinson (1992) and shows Taylor’s results and the measured thermal fouling resultsfrom their thermal fouling experiments. Hexadec- 1 -ene, dicyclopentadiene and indene areknown to form polymeric peroxides and these show very large deposition rates. This132. Literature ReviewTable 2.3 Fouling Results of Taylor (1969b) and Asomaning and Watkinson (1992)Alkene Structural Formula Molecular Boiling Ratio of Mass Rf, Fouling Ratio of Rf toFormula Point Deposition to Resistance t Rf (indene) t(101 kPa) Deposition of after 40 hrs,indene [m2.KJW]oct-1-ene Me(CH2)5CH= C8H16 121.6°C 0.0 0.0 0.0dec-l-ene Me(CH7CH= C10H2 170.9°C 0.1 0.015* 0.002*hexadec-1-ene Me(CH2)13CH=CH2 C1 6-32 284.4°C - 0.25 0.334-vinylcyclohexene C8H12 127.0°C 0.55 0.16 0.211,5-cyclooctadiene C8H12 150.8°C- 0.0 0.0indene C9H8 182.6°C 1.0 0.75 1.0dicyclopentadiene C10H2 170.0°C - 0.45** 2.2**Reproduced from Taylor (1969b) [] and Asomaning and Watkinson (1992) [tI.- 10 wt% alkene in dodecane evaporated at 135°C under 20.6 kPa oxygen saturation.t - 10 wt% alkene in aerated kerosene at 40.6 kPa oxygen saturation, Re 9600, Tbulk 80°C,T5 180°C, Heat flux 300 kW/m2. * - Heat Flux 250 kW/m2; ** - Heat Flux 350 kW/m2142. Literature Reviewvariation with compound structure confirms Canapary’s hypothesis on the complex natureof organic reaction fouling. Hazlett et al. (1977) analysed the fuels for hydroperoxides, theinitial product of autoxidation, and detected significantly higher levels post heating. Theyalso employed on-line gas chromatography (GC) to measure dissolved oxygen levels insolution and found that deposition rates and deposit patterns were determined by oxygenconcentration. Taylor and Wallace (1967) obtained an activation energy of 42 id/mo! for jetfuel deposit formation at temperatures <300°C; above this temperature the activation energydecreased, indicating that different mechanisms were involved (or that mass transfer wassignificant). This work confirms that it is important to identify the range of any foulingstudies before comparisons are made. Table 2.4 shows the activation energies reported inautoxidation fouling studies. These range from 30-120 kJ/mol (excluding Oufer’s results),indicating that chemical reaction steps play a significant role in deposit formation.The effects of flow rate and temperature were studied by Vranos et al. (1981) intheir study of jet fuel coking. Table 2.5 lists the velocity effects reported in the literatureand shows conflicting trends. Many early studies did not report whether constant interfacetemperatures were used, which would decouple temperature effects from any velocityeffect. Epstein (1993a,b) proposed a model for chemical reaction fouling which includedcomplex velocity effects as reported by Crittenden et al. (1987b). Vranos et al. found thecoking rate to be proportional to Re°6; this is one of the few instances of fouling rateincreasing with flow rate under autoxidative conditions and suggested that coking wascontrolled by mass transfer. Deposit studies using FTIR and elemental analysis indicatedthat the deposits were formed by liquid phase oxidation reactions, while chemical analysisof solution samples detected hydroperoxides and other liquid phase oxidation products.Transmission electron microscope studies showed that the foulant consisted of particulatesof 15A diameter; Vranos et al. concluded that solubility processes as well as oxidationreactions were involved in the formation of deposits. This reaction and precipitation152. Literature ReviewTable 2.4 Activation Energies Reported in Chemical Reaction FoulingFeedstock Temperature Range Activation Energy Eact Reference(°C) (kJ/mol)Crude Oil (heavy) 160 - 280 21 Crittenden et at. (1992)(light) 160 - 280 33Refinery Feedstocks 200 - 340 40 - 120 Braun and Hausler (1976)Gas Oil 146 - 204 120 Watkinson and Epstein(1968)Crude Oil Feedstocks 365- 447 53 Scarborough etal. (1979)Used Lubricating Oils 343 - 455 74 - 97 Steele et al. (1981)Jet Fuels 93 - 260 40 Taylor and Wallace (1967)Crude Oil Feedstocks 71 - 287 36 Eaton and Lux (1984)Shale Oil 175 - 400 15 - 20 Latos and Franke (1982)Jet Fuel 149 - 260 42 Vranos etal . (1981)Kerosene 160 - 380 70 Crittenden and Khater(Vapourisation) (1984)Dilute Polymerisation of 80-150 56 Crittenden etal. (1987a)styrene in keroseneStyrene in heptane 164 - 201.1 179-3 19 Oufer (1990)162. Literature ReviewTable 2.5 Velocity Effects Reported in Chemical Reaction FoulingReference Feeds tock Flow Velocity Effect of Flow CommentsRate Increase on(mis) Fouling RateNelson (1958)Chantry and Church(1958)Charlesworth (1963)Parkins (1961)Canapary(1961)Hillyer (1963)Smith (1969)Watkinson and Epstein(1969)*Scarborough et at.(1979)Vranos et al. (1981)*Dickakian and Seay(1988)Fetissoff (1982)*Oufer (1990)Cnttenden et al.(1987a)*Crittenden et at. (1992)Desalted, wet andcorrosive crude oilsForced CirculationReboilersOrganic Coolants -TerphenylsGas OilTerphenylsJet Fuel (kerosene)Sour Gas OilCrude oilJet FuelsCrude OilStyrene in heptaneStyrene in heptaneStyrene in keroseneCrude Oil Feeds0.3 <Urn <2.13 <Urn < 101 Urn 34500 Re 10 0009800 Re 41 900600 < Re < 10 0008600 < Re < 28 8001 <Urn < 2.51100 <Re <5200decreasesdecreasesdecreasesdecreasesdecreasesincreasesdecreasesdecreasesincreasesincreasesno effectdecreasesincreases ordecreasesinconclusiveIndustrial plant dataRe1Re 1.77Re 0.6boiling studiesRe(35 5.8)Maximum in RerelationshipIndustrial plant dataurn mean fluid velocity: * indicates that wall temperatures were fixed172. Literature Reviewhypothesis arises in other cases of chemical reaction fouling, notably milk fouling andpolymerisation fouling.2.2.2 Thermal Fouling StudiesThermal fouling studies involve the measurement of heat transfer coefficients andare performed to assess the significance of fouling on heat transfer. Many of the foulingstudies listed in Table 2.2 did not include the effects of deposition on heat transfer. Most ofthe cases listed in Table 2.2 were studies of specific feedstocks rather than of autoxidation,the mechanism responsible for fouling.Previous research has shown that dissolved oxygen is an essential feature ofautoxidation fouling. Eaton and Lux (1984) found the fouling behaviour of theirhydrodesulfurizer feedstock to depend on the feedstock nature. Fouling rates increasedsignificantly with dissolved oxygen concentration, whereas Braun and Hausler (1976),who used a series of petroleum feedstocks, found oxygen important in some cases and notin others. The reactants or fouling precursors were not identified. Hausler and Thalmayerused feedstocks doped with suiphides and found that in the presence of oxygen, thesuiphides initiated fouling whereas in its absence the sulphides instead initiated corrosion.The latter result confirms the need to maintain a well defined reaction system in foulingstudies. Oufer (1990) reported shorter induction periods in oxygenated solutions of styrenein heptane and generally larger rates, though no conclusive trend was established. Ouferalso investigated the effect of oxygen on thiol-doped styrene solutions and again reportedshorter induction periods.Crittenden and Khater (1984) studied fouling in a horizontal, single tube kerosenevapouriser under oxygenated conditions and attributed fouling to a liquid phaseautoxidation mechanism. Their results showed strong variation of fouling with axial and182. Literature Reviewcircumferential location due to the complicated distributions of oxygen concentration andbubble density in an evaporative system.The complexity of chemical reactions in chemical reaction fouling has spurred theuse of model solutions in experimental studies so as to simplify the system and tracedeposit formation back to a known kinetic scheme. The effects of flow velocity,temperature and surface geometry can then be interpreted in terms of a known foulantgeneration process. More reliable physical property and kinetic parameters can then be usedin testing fouling models. The first study of chemical reaction fouling using modelsolutions was performed by Palen and Westwater (1966) using solutions of styrene intoluene. Thermally initiated solutions of 35-100 v/v% styrene were used to study foulingunder pool boiling conditions. Fouling was reported to be extremely rapid after aninduction period which decreased with increases in styrene concentration and heat flux.Crittenden et al. (1987a,b) reported the first systematic study of chemical reaction foulingwith associated reaction kinetic studies in their study of polymerisation fouling usingsolutions of 1 v!v% styrene in kerosene. Chemical reaction parameters were obtainedseparately from the fouling experiments and used in a numerical model to compare thefouling data with the model predictions. This study did not involve autoxidation andreported complex velocity effects which are discussed further in Section 2.4. The sameapproach was used by Oufer (1990) to study fouling in boiling solutions of styrene inheptane. Oufer varied styrene and oxygen concentrations, surface temperature, flowvelocity and the effect of sulphur additives. No chemical analysis was performed toestablish solution reaction chemistry; the large activation energies (>179 kJ/mol) and strongvelocity effects (dR1/ t ceRe’ where n ranged from -3.5 to -5.78) reported do not correlatewell with previous results from Tables 2.4 and 2.5, many of which are for non-boilingsolutions.Asomaning and Watkinson’s (1992) study of fouling from oxygenated modelsolutions of alkenes in kerosene has been mentioned as a thermal fouling extension of192. Literature ReviewTaylor’s jet fuel work (1969b). Their results reproduced in Table 2.3 show that the alkeneswhich Taylor found to produce most gum also generated the heaviest fouling deposits.Deposit analyses indicated that the foulant was derived from polymeric peroxides andindene polyperoxide was obtained as a precipitate from the resultant solutions of indene inkerosene. No solution chemistry analysis was performed to establish the controllingreaction. Fouling runs of indene and dicyclopentadiene under deoxygenated conditionsgave reduced fouling consistent with differences in autoxidation and polymerisation rates,while runs at higher heat fluxes produced heavier fouling. No activation energies or flowrate effects were reported.Asomaning and Watkinson’s study identified indene, dicyclopentadiene andhexadec-1-ene as alkenes for further model solution studies in autoxidation fouling. Thecurrent work is part of an extended project using such solutions described in part byWilson and Watkinson (1992), Zhang et al. (1992), Lai (1992), Lai (1993) and Zhang etal. (1993).The experimental study of autoxidation fouling has thus progressed from explainingresults in terms of a general n-th order chemical reaction with unidentified reactants tomodel solution studies investigating the processes in the fouling mechanism. Therelationship between reaction rates and deposition still remains unclear, however.The apparatus used in autoxidation fouling studies consists of compact heatexchangers with well defined geometries and flow patterns. Electrical heating is commonlyused as the power and heat fluxes are readily controlled and monitored. In most of the fueldeposition studies the fluid is heated using a once-through arrangement at hightemperatures in order to accelerate the fouling rate. In thermal fouling studies recirculationof the process liquid is generally employed in order to use reasonable quantities of fluid asexperimental periods are frequently long. The selection and design of fouling probes hasbeen reviewed by Fetissoff (1982) and Chenoweth (1988). The latter also describescommercially available fuel oxidation testers. Researchers have tended recently to use202. Literature Reviewsimpler flow geometries (annulus, tube) for thermal fouling in order to facilitate modellingstudies. Many of the analytical techniques available to the modern organic chemist havebeen used in solution analysis. Determination of dissolved oxygen concentrations iscomplex but has been performed by Hazlett et al. (1977). Modern surface analysis methodshave similarly been used to investigate deposit structure (Belmar-Beimy and Fryer, 1992)and a general review of analytical techniques is given by Watkinson (1988).2.3 Autoxidation Chemistry2.3.1 Autoxidation Mechanisms and KineticsAutoxidation is the general term for reactions involving the autocatalytic oxidationof hydrocarbons by a free radical mechanism where the dominant radical species is theperoxy radical, ROO. Autoxidation occurs in paint drying, fuel storage and rubberdegradation, with the latter phenomenon spurring the original research into autoxidationmechanisms. A review of the extensive material describing autoxidation and itsmechanisms is given by Reich and Stivala (1969). The discussion which follows is asummary of autoxidation mechanisms related to alkanes and alkenes under the conditionsused in the current study.Hydrocarbons can undergo addition polymerisation in the presence of alkylradicals, R, as in the chain (vinyl) polymerisation of styreneR + R - R-R + R — R-R-R [2.21Alkenes usually undergo addition polymerisation more rapidly than alkanes due to the easeof addition of R to the C=C double bond. In the presence of significant concentrations ofdissolved oxygen, however, the alkyl or allyl radical combines with molecular oxygen toform the peroxy radical, R02 via the fast stepR + 02 — RO2 [2.3]212. Literature ReviewAutoxidation chemistry is determined by peroxy radical behaviour and the mechanism ofautoxidation was summarised in the ‘modified basic autoxidation scheme’ below by VanSickle et al. (1965a, 65b, 67) as follows. The labels are those found in the literature.Initiation (R1)— R [1]Propagation R + 02 — R02 [2]R02 + RH— RO2H + R (H abstraction)[3a]R02 + RH— RO2H (E R) (addition) [3b]R02H + 02 - R02 ( RO2) [2’]R02— RO + MO (epoxidation) [71R0 + RH- ROH + R [8]RO + RH- ROM (E R•) [9]Termination 2 R02— inert products [6]2 R0— inert products [5]2 R— inert products [4]where MO denotes epoxide, M denotes a monomer unit, RH, and R0 is the alkoxyradical. The reaction scheme is quite complex and the values of the respective kinetic rateconstants are determined by hydrocarbon structure and solvent nature. In the absence ofsignificant steric crowding, the rate constants k2 and Ic2’ are equal.Steps [2,2’] are relatively rapid in significant oxygen concentrations so thedominant radical at high oxygen concentrations is the peroxy radical and steps [4, 51 can beignored. Steps [3a,3b] are thus rate determining and the reaction rate can be written asd[02]/dt = d[RH]Idt =- (k3a+ 3b) [R02] [RH] [2.4]The steady state approximation is applied to the peroxy radical concentration. Equating thefree radical initiation rate, R1, to termination via step [6] gives the general result for longkinetic chain lengths:d[02]/dt = d[RH]/dt =- (k3a+kb) (RI2.k6)°5[RH] [2.5]The reaction rate is thus determined by the mode and rate of free radical initiation.222. Literature ReviewAs the oxygen concentration decreases, the alkoxy radical and eventually the alkylradical become dominant, altering the mechanism appropriately. The conditions at whichautoxidation (via the peroxy radical) becomes an alkoxy radical process are determined bythe monomer structure. Mayo et al. (1956a,b) thus observed a minimum in styrene reactionrate at 50°C as the oxygen pressure was increased from 0 kPa to 100 kPa, corresponding tothe region between addition polymerisation and autoxidation where R& is the dominantradical species. The transition in mechanisms explains the complex effects reported foroxygen concentration in chemical reaction fouling. The oxygen concentrations involved inthe transition between radical mechanisms are determined by the alkene structure and arethus difficult to predict for mixtures. Nicholson (1991) thus reported the use of airblanketing of methylacrylic acid in order to inhibit polymerisation and did not reportsignificant autoxidation problems. Asomaning and Watkinson (1992) found their aeratedand deoxygenated fouling resistance data to mirror the expected polymerisation rates forindene polymerisation at the experimental conditions employed.The peroxy radical can react with a hydrocarbon molecule via hydrogen abstraction[3aj to form the hydroperoxide or by direct addition [3bj to form a dimeric or polymericperoxy radical. The competition between these steps is determined by the hydrocarbon’sstructure. Alkanes generally favour hydrogen abstraction and the reaction rate is governedby the energetics of hydrogen abstraction and the ability to stabilise the resulting radical.The tendency of alkenes to undergo addition (oxidative copolymerisation) is alsodetermined by radical stability, with aromatic alkenes and terminally double-bonded longchain alkenes reported as favouring this pathway over abstraction. The addition stepgenerates polymeric peroxides, which Mayo and Lan (1986) identified as the primarysource of gum in fuel storage and which Asomaning and Watkinson (1992) considered tobe the source of deposits in their fouling studies. Heavy deposition was observed inindene, hexadec- 1-ene and dicyclopentadiene, the alkenes most likely to favour addition,while oct-1-ene, dec-1-ene, which tend to form hydroperoxides, showed little fouling. Van232. Literature ReviewSickle et at. (196Th) studied the tendency of alkanes and alkenes to form epoxides andalkols by radical rearrangement (steps [7], [8]) and reported values ofk21k8> 1000 foralkenes which favour addition at high oxygen pressures. They explained this in terms ofresonance energy losses and emphasised the role of ring strain in reducing the tendency toform epoxides inS- ring alkenes such as cyclopentene. Steps [7,81 are thus unlikely to besignificant in the autoxidation of compounds such as indene. These results mirror thefouling tendencies observed by Asomaning and Watkinson, who concluded thatautoxidation fouling was caused by the formation of insoluble polymeric peroxides whichsubsequently decomposed on the hot heat exchanger surface.The propagation step products can undergo further reaction as peroxides andpolyperoxides are known to be unstable at high temperatures and decompose to formcarbonyls, alkols and various free radical species. Tert-butyl-hydroperoxide, (Me)3COOH,is thus a common free radical initiator. The formation of these oxygenated secondaryreaction products from hydroperoxide decomposition was demonstrated by Mazius (1965)and this explains the significant concentrations of carbonyls reported by Hazlett et at.(1977) and Vranos et at. (1981) in their fouling studies. The decomposition ofhydroperoxides has been reported to be unimolecular at low concentrations but bimolecularat high concentrations, especially where hydrogen bonding is likely to generate peroxidedimers in non-polar solvents.Equation [2.5] indicates that autoxidation rates are controlled by the initiation rate,R1. Hydroperoxide decomposition has been reported to be the primary initiation source inmany studies, which would involve a kinetic expression of the form;R1 = {k11, [RO2H] + [RO2H]} [2.6]Proposed reaction schemes are;Unimolecular RO2H + RH—‘ RO + R + H20 (Neiman, 1964) [2.7]Bimolecular 2 ROH— RO2 + R + H20 (Toboisky et al. 1950) [2.8]242. Literature ReviewToboisky et al. (1950) modelled the autoxidation of 1,4-dimethylcyclohexane using step[2.8] as the initiation source but did not consider an initiation step as complex as [2.6],which would give a complex rate expression when substituted into equation [2.5]. Mostchemical kinetic studies have considered the initial stages of autoxidation, at low extents ofconversion and consequently small peroxide and side product concentrations, where R1 isrelatively well defined. An alternative approach to complex reaction kinetics was used byNorton and Drayer (1968) in their kinetic model of hexadec-1-ene autoxidation to formhydro- and poly-peroxides. The data were fitted to an empirically based scheme whichdescribed the distribution of products observed in the experimental datahexadec-1-ene -4 hydroperoxide— polyperoxide —* insoluble peroxides [2.9]The model did not reflect the mechanism of the free radical chemistry involved but did offersome rationalisation of the processes involved. The rate constants obtained were thuslumped kinetic constants; such an empirical approach may be more appropriate in caseswith complex initiation sources.2.3.2 Solvent and Additive Effects in AutoxidationAutoxidation kinetics and product distributions are influenced by temperature,product reactivity, oxygen concentration and solvent nature. Increases in temperaturegenerally increase reaction rates but can introduce side reactions of labile products.Reaction products such as acids can inhibit further autoxidation. The oxygen dependence ofthe dominant free radical has been discussed previously but analysis of Eqns. [l]-[9] hasbeen shown to predict the higher epoxide yields at lower oxygen concentrations reportedfrom experiments (Reich and Stivala, 1969).Solvent effects are observed in rate constants, hydroperoxide behaviour andproduct solubility. Polar solvents tend to enhance the propagation rate constants (k3a, k3b)252. Literature Reviewwhile non-polar solvents have retarding effects. Solvents with abstractable protonsavailable can interrupt the solute reaction via chain transfer to the solvent (5).Chain Transfer R + S-H —* R-H + S [2.101The solvent can undergo autoxidation itself (co-oxidation) or act as a sink for radicals.Since the solvent is usually in excess, chain transfer will reduce the solute reaction rate andcan also affect the product distribution. Solvent effects in hydroperoxide decompositioninvolve hydrogen bonding of the polarised hydroperoxide molecules to form dimers innon-polar solvents, while enhanced hydroperoxide decomposition rates have been reportedin polar solvents. Cage mechanisms, where solvent viscosity is significant, have also beenpostulated for peroxy radical termination. Product solubility depends on solvent nature,where the maxim ‘like dissolves like’ applies. Most autoxidation products are soluble in theoriginal hydrocarbon but particular exceptions are polymeric peroxides, such as reported byNorton and Drayer (1968). Polymeric indene peroxide was soluble in pure indene but notin solutions with more aliphatic, less aromatic and less polar solvents so Russell usedsolvent precipitation to identify the reaction products of indene autoxidation (1956a). Thesolubility of reaction products in solutions depends on the solvent nature, often expressedin terms of the Hildebrand solubility parameter, The solubility parameter is related tothe free energy of solution and is used in the estimation of the activity coefficient of adissolved solute in equilibrium with the solvent. The value of aH is usually found byexperiment, but Small (1953) proposed an additive approach and estimation methods aregiven in ASTM D3827-86. The solubility of a solute (2) containing no solvent (1) is givenbymy2 2 = - [AH/RTJ (1 - T/T) + F (Cp(1,2)) [2.11]where lny2 = [V2t12!RTI (ãH,1-8H,2)2 [2.121where c2 is the mol fraction of solute in solution and V2 is the molar volume of solute.This approach is not valid for long chain polymers where Flory-Huggins solutionthermodynamics apply. The solubility of foulant precursors is an important yet poorly262. Literature Reviewunderstood aspect of chemical reaction fouling as many researchers have reported thatdeposits originated as insoluble materials precipitated from solution rather than generatedby simple wall reactions.The roles of sulphur, nitrogen and oxygen heteroatomic species in autoxidationhave been discussed extensively in the autoxidation literature. The structure of a compounddetermines the C-X bond energy, which dictates its reactivity towards the peroxy radicalsand hydroperoxides formed during autoxidation. Changes in mechanisms are reported atenhanced temperatures due to fission of the C-X bond to form free radicals or condensationproducts. Disuiphides and diazo compounds contain the weakest C-X-X bonds and react insimilar fashion to their oxygen analogues, the peroxides, which are recognised free radicalinitiators. The thermal disociation of disuiphide, for example, is given byR-C-S-S-C-R - 2 R-C-S [2.13]Thiols similarly resemble phenols in acting as antioxidants or inhibitors at lowtemperatures, above which they contribute to deposit formation. White et al. (1983) foundphenols to be a source of deposits via phenolic coupling reactions, while Taylor (1976)reported enhanced deposition rates of sulphur-doped fuels above threshold temperatureswhich he linked to C-S bond strengths. Taylor (1970) observed enhanced deposition rateswhen various oxygenated dopants were added to aerated jet fuels at 100 ppm, dependingon dopant structure. Watkinson (1988) reviewed the role of oxygen, sulphur and nitrogenheteroatomic species in fuel stability studies and fouling work and found similar patterns toexist in both. The significance of these species in fouling is shown by the increasedconcentrations of sulphur, oxygen and nitrogen reported in deposit analyses whencompared to the fluid (Watkinson and Epstein, 1969). Dissolved metal ions are well knownsources of radical initiation and can also function as hydroperoxide decomposers.Crawford and Miller’s study of metal ions in oxygen uptake showed significant increases atconcentrations as low as 1 ppm (1963). The role of metal ions in autoxidation wasreviewed by Reich and Stivala (1969).272. Literature Review2.3.3 The Autoxidation of Indene, C9H8The present work concentrates on model solutions of the aromatic alkene, indene,in solvents which are relatively inert to autoxidation. Indene autoxidation was studied byRussell (1956a,b), who reported significant polymeric peroxide formation and anaddition/abstraction ratio [k3a/k3b] of 4:1 at 50°C. The major products were polymericindene peroxides ((C9H8OO), where n ranged from 2-10), carbonyls, and indenehydroperoxide but not epoxide. Studies were performed in pure indene or in solutions inbromobenzene. Ueno et al. (1974) analysed the products from the extended autoxidation ofindene and reported a product slate of various carbonyls and aldehydes consistent withpolyperoxide formation and degradation. Morris et al. (1988) and Morris and Mushrush(1991) have also used indene in model solution studies of sulphur compounds in fuelstorage stability. Figure 2.1 is a summary of the indene autoxidation mechanism.Russell also investigated the mechanism of thermal initiation in indene and reportedit as a bimolecular reaction with oxygen to form the peroxy radical;C9H8 + 02 - C9H800• [2.14]This mechanism has been disputed on energetic grounds in favour of one involving theunimolecular dissociation of hydroperoxide. Carisson and Robb (1966) investigated theautoxidation of indene in excess antioxidant at 70-95°C and proposed a termolecularinitiation step involving two indene molecules and an oxygen molecule with an activationenergy of 78.7 kJ/mol. They reported that traces of hydroperoxide increased autoxidationrates significantly.Indene polyperoxides were formed by the cooxidation of indene with oxygen viathe indene peroxy radical. Russell (1956b) studied thermal and AJEN initiated oxidationand concluded that the reaction mechanism did not change under a different initiationsource. Howard and Ingold (1962) measured the rate constants for the initial propagation282. Literature ReviewFigure 2.1 Autoxidation of Indene00indeneinitiation indene peroxy radical0 7 0abstractioVadditionocrHhydroperoxideabstractionhydroperoxidepolyperoxyradical _0.Z additionpolymeric peroxidesFigure 2.2 Antioxidant Action of 2,6 t-butyl,4 methyl phenol (BMP)ROO +R00+tBiROOH +tBu tBu292. Literature Reviewand termination steps in indene autoxidation at 50°C but, like Russell, did not report anyactivation energies. The activation energy of indene disappearance given by Equation [2.5]is a composite figure, Eact,overaii;Eact, overall = 0.5 Eact,init + Eact,propagation - 0.5 Eact,termination [2.151Howard and Ingold reported an indene additionlabstraction ratio of 9:1 at 30°C.2.3.4 AntioxidationAdditives are often used to prevent the formation of permanent deposits on heattransfer surfaces. Stephenson and Rowe (1993) discussed the mitigation strategiesemployed in ethene plants and the factors which determine the choice of additive;dispersants and detergents are used to minimise the deposition of foulant precursors,whose formation is inhibited by adding antioxidants, or chelating agents which formcomplexes with the metal ions present in solution. Antioxidants are primarily preventativein action and Scott (1965) describes the two modes of antioxidant action as hydroperoxidedecomposition and radical deactivation. In the former, hydroperoxides are decomposed toform stable products without generating new radicals; this can be achieved by selectedmetal ions, sulphur compounds or strong organic acids. In the latter, radical scavengers(AR), react with available peroxy radicals to form stable species, thus interrupting the chainreaction; examples of radical scavengers are hindered phenols, amines and certainthiophenes.RO2 + A-H — RO2H + A (stable) [2.16]RO2 + A —> RO2A (stable) [2.17]Figure 2.2 shows the reaction of a common gasoline antioxidant, 2,4,di-butyl-4-methylphenol (BMP), with peroxy radicals to form stable products. The efficiency ofhindered phenols as antioxidants is structure dependent. The chemistry of antioxidation hasbeen reviewed by Reich and Stivala (1969). Antioxidation remains an important research302. Literature Reviewarea as antioxidant efficiency also depends on operating conditions; Morris et al. (1988)reported how the antioxidation efficiency of a metal ion, thiophene, B1VIP and an aminevaried with operating temperature. BMP was relatively ineffective at temperatures above a‘ceiling temperature’ of 100°C. Ceiling temperatures refer to the temperature above whichan antioxidant rapidly loses efficiency.2.4 Mechanistic Modelling of FoulingThe aim of mechanistic modelling is to be able to predict and explain the behaviourof a fouling system when parameters such as hydrodynamics, temperature and chemistryare varied. A reliable and robust fouling model could be used both in the design ofantifouling heat exchangers (Fryer et al., 1990) or in the development of control strategiesfor heat exchangers in fouling service (Fryer and Slater, 1985). The accuracy of amechanistic model is determined by the understanding of the physical phenomena involved.The complexity and number of processes involved in chemical reaction fouling hasprevented the development of a universal model, so that certain cases, such as coking, arerelatively well understood and modelled while others, such as autoxidation, are not.Figure 2.3 is a schematic of the processes involved in chemical reaction fouling.The foulant precursor can be generated in situ or transported into the unit with the bulkfluid. In situ generation can occur at the surface itself, in the ‘reaction zone’ near thesurface where conditions are favourable, or in the bulk fluid (i.e., conditions are favourableeverywhere). The foulants andlor reactants may have to be convected to the reaction zoneand the surface, and likewise reaction products or the foulant (if still mobile) may bereturned to the bulk fluid and may cause deposition downstream. The foulant precursor canform deposit either by surface layer growth or by the precipitation and/or adhesion ofinsoluble agglomerates. Froment (1981) reported the morphology of coke deposits inethane crackers as a mixture of amorphous coke and regular, dendritic regions similar to312. Literature ReviewFigure 23 Schematic of Processes Involved in Chemical Reaction FoulingHot WallAged DepositIBulkFluidFlowAdapted from Crittenden ci al. (1987b)New DepositfoulantprecursorsREACTIONZONE4products(includingfoulant)-Ii,Heat Flow322. Literature Reviewmetal crystal growth. Most studies of deposit morphology in the medium temperature rangedescribe the deposit as originating from insoluble materials. Belmar-Beimy and Fryer(1992) and Crittenden et al. (1987b) both emphasise the role of (in)solubility in foulinginvolving two different materials - milk proteins and polystyrene, respectively. Dickakian(1990) has also emphasised the role of asphaltene insolubility in crude oil fouling. Themechanism of forming solid deposits from the liquid and gaseous phase is the controllingstep in chemical reaction fouling models. Once formed, the deposit may undergo furtherreaction as described by Atkins (1962) or be removed by shear forces. Existing modelsinvolve various combinations of these steps and Table 2.6 is a summary of chemicalreaction fouling models in the literature.2.4.1 Transport and Adhesion ModelsMany fouling models describe the phenomenon as a series of transport, reaction,adhesion and removal steps. Kern and Seaton (1959) formulated an empirical model tocorrelate plant data of particulate fouling, which fitted an expression of the formRf(t) = Rf*( 1 - exp[-b.t]) [2.18]where Rf* is the asymptotic fouling resistance and 1/b a time constant. Kern and Seatonmodelled fouling as a competition between a deposition flux, dep proportional to theconcentration of precursors C and the mass flow rate W, and a removal flux, remproportional to the deposit thickness Xf and the wall shear stress, ‘r.dRf/dt = [1I2] dxf/dt = Pdep - ørem [2.19]This model gives the initial fouling rate as(dRf/dt) t=O = ødep = Rf*.b = kdep Cp W [2.20]Taborek et al. (1972) included a deposit shear strength factor, iyD, in the removal termwhich Pinhero (1981) suggested should be a function of velocity. The increase in lrem withdeposit thickness has been rationalised by Loo and Bridgewater (1981) in a theory of33Modelsolutionevaporation-styreneAutoxidationFoulinginModelSolutionsHydrocarbonsinseneralRateisdirectlydependentuponthermalboundarylayerthicknessConstantmonthlyincreaseincokeresistanceforvariousrefinerystreamsHydrodynamicboundarylayeranddiffusionpartialdifferentialequations(1)instantaneousfirstorderreactioninzoneclosetothewall(2)veryrapidcrystallisationathotsurfaceMasstransferandadhesionofsuspendedparticles(1)stickingprobabilityproportionaltoexp(-E/RT)(2)stickingprobabilityinverselyproportionaltoshearstressatwallKineticcontrol-tworeactions(1)FirstorderdissociationofAintoproducts(2)ZerothordercokeformationKineticsand/ormasstransfercontrolwithfirstorderreactionKineticscontrol(1)atsurfacetemperature(2)firstorderinconcentrationofcrackingproductsBoundarytayerasdifferentialchemicalreactor-kinetic/adhesioncontrol(I)Foulantgenerationinboundarylayer(2)DepositionviastickingprobabilityasdescribedbyWatkinsonandEpsteinKineticsandmasstransfercontrolwithsecondorderreactionlbMassTransferfrombulkliquidIaMassTransferwithChemicalReaction2BoundaryLayerGenerationofFoulanlKineticsandmasstransfercontrolwithnthorderreactionandanattachmentfactorinvolvingmeanresidencetimeatthesurfaceNoneconsidered.FryerandSlater(1985)simulationusesfirstorderKernandSeatonshearremovalterm(1)FirstorderKernandSestonshearremovalterm(2)NegligiblefoulantbackdiffusionFoulantBackDiffusioninvokedinlb,2FoulingRatecanbereducedbyincreasingfluidvelocityTwolayerconcept-porouscokelayeradjacenttomidandhardcokenexttowall(I)Solutionwithdiffusioncontrolfitsplantdata.Foulingratepredictedtoincreasewithvelocity(2)Extendedtoconsidercolloidaltransfertohot surface(1)Correctpredictionofinitialratedependenceonvelocity(2)Incorrectpredictionofasymptoticresistanceonvelocity(1)Quasi-steadystateassumption(2)UntestedSolutionwithmasstransfercontrolfitsplantrun-timedata(1)Quasi-stesdystateassumption(2)GoodagreementbetweenindustrialdataandnumericalsimulationAuthorsApplicationDepositionTermRemovalTermRemarksa n —I 0 C <.5 CsOilrefiningFiredrefineryheatersOrganiccoolantsinnuclearreactorsLiquidphasegasoilfoulingVapourphasepyrolsisVapourphasepyrolysisUltrapyrolysinofethaneNelson(1934)Atkins(1962)Nijsing(1964)WatkinsonandEpstein(1970)JackmanandAris(1971)Fernandez-BsujinandSolomon(1976)SsndaramandFroment(1979)Crittenden,KolaczkowskiandHout(1979)PatersonandFryer(1988)Oufer,BrebcrandKnudsen(1993)PanchalandWatkinson(1993)Epstein(1993a,b)NoneconsideredNoneconsideredProductdiffusionbacktotheliquidisanintegralpartofthedifferentialequationsFirstorderKernandSeatonshearremovaltermNoneconsideredNoneconsideredNoneconsidered(1)Diffusionoffoulantbackintobulkfluid(2)FirstorderKernandSeatonshearremovaltermHydrocarbonsinKineticsand/ormasstransfercontrolwithfirstordergeneralreactionLID I ri C C C,,MilkFouling(1)Limitedtesting(2)Complex-manyparameters(3)Extendedtotwo-layerconceptproposedbyAtkinsCorrectpredictionofinitialratedependenceonvelocityUnknownparametersinboilingmasstransfer.Largedeviationsfromobservedresults.Fittedtoeachsetofexperimentaldatala,2showedtrendsobservedinRfprofilesNostickingprobability;batchkineticsPredictedvelocityandtemperatureeffectsinstyrenepolymerisationfoulingnoneconsidered2. Literature Reviewdeposit shattering by thermal stresses. The removal flux has been assumed to beproportional to ‘r in many cases and some of the possible interpretations are reviewed byEpstein (1988). For asymptotic fouling, Equation [2.19] gives(dRfldt) = 0 = ødep - Prem = kciep Cp W - Rf* tw/14JD [2.2 1]and thusRf*= kdep Cp W NJD Itw [2.22]Rf* should thus be proportional to irjy’Win turbulent flow. Watkinson and Epstein (1968)modeled gas oil fouling as due to suspended particulates but observed very differentbehaviour to that predicted by Kern and Seaton’ s model; the initial fouling rate decreasedasW increased and Rf* did not obey [2.22]. They proposed a model featuring masstransfer of particulates to the surface followed by adhesion to form deposit, and removal;[1I2’] dxfldt = a1 S Jdep - a2 tw Xf [2.23]where J dep = km (Cp,bulk - Cp,surf [ 0]) [2.24]and S, the ‘sticking probability’ was used as defined by Parkins (1961) asdeposition rate = S oc exp [2.25]particle fluxWatkinson and Epstein’s Arrhenius type term represented the formation of an adhesivebond, while the shear stress dependence correlates the removal of material by shear forceprocesses. Equation [2.23] was successfully fitted to their initial rate data but not theasymptotic fouling resistance.The ‘sticking probability’ concept describes fouling as a competition betweenadhesion and removal phenomena in the process of deposit formation rather than removalof foulant material once it has been incorporated in the deposit surface. Asymptotic foulingis rarely reported in cases of chemical reaction fouling, so removal would seem to be lesssignificant than in cases of particulate fouling. Removal processes in particulate foulinghave been investigated by Yung et al. (1989). The sticking probability has also beeninterpreted by Epstein (1981) and Paterson and Fryer (1988) as a competition between352. Literature Reviewcharacteristic residence times, between that necessary for bond formation and that for theperiodic renewal of fluid at the surface. Cleaver and Yates (1975,1976) obtained a similarexpression for particle deposition based on the observed phenomenon of turbulent bursts.Small bursts of local eddy activity from the turbulent bulk occur with periods of order [100v/u*2] disrupting the laminar sublayer and thus renew the fluid at the surface. u’ is thefriction velocity, defined as = Um/(f/2). Vatistas (1989) considered the process in termsof removal probabilities, defining a dimensionless adhesion time based on removal andadhesion processes in particulate fouling. Paterson and Fryer (1988) similarly described Sas proportional to the time for which the surface and foulant precursor are exposed to eachother, when adhesive processes operate. Most work on sticking probability has involvedparticulate fouling rather than chemical reaction fouling. Turner (1993) reviewed thepublished data and models for particulate sticking probabilities. He suggested that thecommonly used Arrhenius term in S for particulates arose from the activation energy of thedouble layer potential rather than from some adhesive chemical reaction stage. Turner alsodiscussed the forces and processes affecting particle deposition at a solid surface.2.4.2 Transport and Reaction ModelsNeither Kern and Seaton nor Watkinson and Epstein incorporated chemical reactioneffects to describe the common case where foulant precursors are generated in situ.Crittenden et al. (1987b) reviewed chemical reaction fouling and described a generalfouling model which incorporated most of the cases reported in the literature. The flux offouling precursor to the reaction zone is given by= km (C,bu1k - Cp,suri) [2.261The flux is balanced by the consumption of precursor in the deposit generating reaction,rxn = kr Cp,surf ‘ [2.27]where kr = k° exp ( - Eact IR Tsurf) [2.281362. Literature ReviewThe fouling rate is obtained by equating [2.19], [2.26] and [2.27], then rearranging to givedRf/dt= l/ d (Xf)/dt = {l/pf Xf }[ Cp,bullc /{1/km + 1/kr Cp,surt’’}1 [2.291Equation [2.33] shows that surface reaction controls the fouling rate at high mass transferrates (large k) and conversely mass transfer is the controlling step at high surfacetemperatures. The latter case corresponds to conditions in pyrolysis reactors modelled byFemandez-Baujin and Solomon (1976), where fouling increases with mass flow rate.Froment (1981) describes successful simulations of pyrolysis reactors using simpledeposition models and a detailed kinetic model to predict C,b14lk for each component.Huntrods et al. (1990) similarly simulated a propane pyrolysis condenser where crackedproducts are cooled and described the form of the ‘deposit factor’ in some detail.Autoxidation reaction rates in the range under study are more likely to involve mixedreactionlmass transfer control.The comprehensive model of Crittenden et al (1987b) was used to explain theresults from an experimental study of polymerisation fouling from solutions of 1 v/v%styrene in kerosene. The model can be summarised asdRf/dt= ødep-4rem P ageing [2.30]The deposition term, dep was based on [2.29] but included a ‘back convection’ flux,back, of foulant generated in the reaction zone transported back into the bulk solution ratherthan forming a deposit;ødep = (1Jpj) (J- iback) [2.31]where Jm km (Cp,bulk - Cp,surf) = krCp,surt deposition [2.32]Tback = kt (Cf,f - Cf,b1k [ 0]) removal offoulant [2.33]This gives an effective interfacial foulant concentration, Cfsurf. which was found to bestrongly dependent on temperature and was interpreted as the solubility limit of the foulantin solution. Direct calculation of Cfsurf involved too many parameters and in its absence themodel overpredicted fouling rates at higher temperatures and did not show the complexvelocity effects observed.372. Literature ReviewThe rem term took the form suggested by Kern and Seaton and the çbageing termincorporated a deposit ageing correction first suggested by Atkins (1962). The deposit wasmodelled as two solid phases of different thermal conductivities. Recently generated foulantwas observed as an amorphous, tarry material with lower thermal conductivity than thecoke-like material found after extended exposure to surface temperatures. 4ageing thusincludes the reduction in Rf associated with deposit ageing.Epstein (1993a,b) included a sticking factor in the reaction term in equation [2.291to explain the velocity effects rather than invoke foulant back convection; the reaction termwas written as the product of an Arrhenius term and the fluid residence time near thesurface. This yielded a reaction plus attachment term of the formkr = k34 v exp [- EactfR Tsuff] / u*2 [2.34]This term introduced a velocity dependence into the reaction term in Equation [2.29] whichpredicted the observed extrema in fouling rates at given surface temperatures as the flowrate increased. Epstein (1993a) initially assumed that the transport term, km. could beestimated using the simplified equation of Metzner and Friend (1958) in calculating k34.km = u*I1l.8 Sc213 [2.35]The calculated values of k34 obtained from the experimental data increased withtemperature, which Epstein suggested could be due to thermophoresis, the diffusion ofparticles down a temperature gradient, or another force resisting attachment which increaseswith surface temperature. Epstein (1993b) later relaxed the form of the transport term, km,and calculated the model constants from one point within Crittenden et al. ‘s data. Themodel predictions are shown along with the data in Figure 2.4 and show excellentagreement with the data with an average absolute deviation of 14.2%, which was within thereported experimental error. The denominator in Equation [2.35] was calculated to be502.3 rather than 11.8; Epstein attributed this to the non-isothermal conditions present, the382. Literature ReviewFigure 2.4 Comparison of the Experimental Fouling Data of Crittenden et al. from thePolymerisation Fouling of Styrene in Kerosene and the Predictions fromEpstein’s Fouling Model (1993b)(preprinted with permission)2100 200 300 400 500G/(kg /m2s)600 700-)2‘F’392. Literature Reviewrelatively low Reynolds numbers involved and the fact that was based on fluidproperties evaluated at the surface rather than the bulk fluid temperature.Oufer (1990) formulated a transport and reaction model to describe fouling frommodel solutions of styrene in heptane under boiling conditions. Back convection of thepolystyrene product was considered to be negligible. The polymerisation of styrene wasfound to be second order with respect to styrene from independent kinetic studies soEquation [2.29] was solved by setting Equation [2.26] equal to [2.271. The model did notgive good agreement with the observed initial fouling rate, which was attributed to errors inestimating the mass transfer coefficient. This model and that of Crittenden et al. (1987b)emphasised the need to understand the order and mechanism of the chemical reaction beforeattempting any mechanistic modelling.Zhang and Watkinson (1991) used the reaction kinetics reported by Russell (1956b)to model liquid phase fouling from the autoxidation of indene. They used a transport andreaction model similar to Equation [2.29] with n = 1.5 but found that there were too manyunknown parameters involved to make valid conclusions.This work was extended by Panchal and Watkinson (1993) to compare threefouling models with fouling data obtained from annular and tubular fouling monitors,operating under identical hydrodynamic and chemical conditions. These devices operated inbatch reactor mode so that bulk concentrations of reactant changed over time. The foulingmodels used an indene autoxidation model similar to that described by Norton and Drayer(1968) to simulate the bulk and surface reactions;ReactantsA1,2 ‘ Precursor B Foulant C [2.36](soluble) (sparingly soluble) (insoluble)The foulant precursor, B, was assumed to consist of 2-6 polyperoxide units and theinsoluble foulant, C, was described as a 16-polyperoxide unit material. The first reactionused Russell’s kinetics,402. Literature ReviewRA1,A2B = k37 CAl 1.5 0.5 [2.37]while the formation of foulant was taken to be first order in foulant precursor,RB_,C = k38 CB [2.38]where k38 was regressed from fouling data. This autoxidation model was used to introducethe time dependent bulk concentrations into the respective fouling models shown in Figure2.5. The models differ in the foulant generation step. Case 1 describes the scenario whereprecursor generation occurs primarily in the bulk solution; precursor formation near the hotsurface is ignored so that bulk reaction control applies. Case lb corresponds to masstransfer control, where foulant generated in the bulk is transported to the wall where it isassumed to adhere completely. Case la involves mass transfer of precursor from the bulkto react at the wall to form foulant. Any foulant found at the wall was assumed to formdeposit, ignoring any sticking probability or attachment factor arguments. Case lb wasrejected as it did not predict the increase in deposition with surface temperature found in theexperimental data. The Case la model followed the trends observed in the data forvariations in surface and bulk temperature, though less successfully than Case 2. Thesecond model described foulant generation as occurring primarily in the thermal boundarylayer. Back diffusion of precursor as descibed by Crittenden et al. (1987b) was includedand the equations solved numerically. This model was found to give the best agreementwith the experimental data. The Case 3 model is similar to that described by Crittenden etal. (1987b) but seemed to ignore reaction in the bulk, which was an intrinsic part of theCase 1 model. This model did not predict the effects of bulk temperature, unlike Cases laand 2.No flow velocity effects were reported and the differences in fouling data from thetwo fouling monitors were not discussed. Each model included an unknown parameterwhich was calculated by fitting the model predictions to the data at a selected point. This isa common procedure in fouling modelling and the values obtained were reasonable. The412. Literature ReviewFigure 2.5 Panchal and Watkinson Autoxidation Fouling ModelsCase 1: Precursor Generation in Bulk Solutionmolecular transferla A > B > B ---> C on wallparticulate trtEmSpOrtlb A --> B --> C > C sticks on wallCase 2: Precursor Generation in Boundary LayerA > A > Bmolecular transferB < B > C C sticks on wallCase 3 : Precursor Generation on Wall Surfacemolecular transferA > A ---> B on wallB < B ---> C on wallbulk solution thermal boundary layer> < >Reaction MechanismReactant A —--> Precursor B ----> Foulant CFrom Panchal and Watkinson (1993)422. Literature Reviewsuccess of the boundary layer formulation concurs with the reaction engineering approachof Paterson and Fryer (1988).2.4.3 Reaction EngineeringPaterson and Fryer (1988) adopted a ‘reaction engineering’ analysis to explain theform of the results observed in milk fouling. This approach echoed that of Nelson (1934)and concentrated on the ‘reaction zone’ next to the heat transfer surface where the enhancedtemperatures give rise to the maximum foulant generation rates. In the absence ofsignificant bulk concentrations of foulant precursor, the conditions in this region of thefluid boundary layer dominate the fouling process. The zone is modelled as a differentialchemical reactor of volume related to the laminar sublayer thickness with generation rate N,.Nr = ar exp [- E IR Tsurt] { 1//tw} [2.39]The initial fouling rate is the product of the generation rate and a sticking probability basedon residence time argumentsdRf/dt a Ar exp [- E1. IR T] { 1I’/t} As exp [- Es IR Tsurf] { 1N’t} [2.40]dRf/dt a A0 exp [- Ef/R Tff] { l/-r} [2.41]Paterson and Fryer correlated their fouling data in the form of a fouling Biot number, BifRf U0 , and thus found the initial rate to vary as i/urn. This approach is a simplification ofthe surface/fluid processes involved but gives some insight into the relationship betweenthe primary factors involved.Fryer et at. (1990) also discussed the result of large concentrations of foulantprecursor in the bulk fluid, when mass transfer of material from the bulk swamps thegeneration of foulant in the reaction zone. The reaction engineering analysis is then invalidand fouling is best described as a generation-agglomeration-adhesion process.432. Literature ReviewMechanistic modeling of chemical reaction fouling in autoxidative conditions is lessdeveloped than that of pyrolysis reactors, for example, due to(a) The range and complexity of liquid phase free radical reactions, which involvecomplex kinetics;(b) The difficulty in performing reliable autoxidation fouling experiments;(c) The operating conditions involved feature mixed reactionlconvection control of thefouling process;(d) The precise mechanism of deposit generation has not been identified.A study of autoxidation fouling must thus address the chemical behaviour of the system aswell as its thermal performance in order that reliable mechanistic models can be developed.2.5 ObjectiveAutoxidation has been identified as a major source of fouling deposit in heatingoxygenated hydrocarbon streams from ambient temperatures to 400°C, above whichcondensation and pyrolysis reactions become dominant. Fuel stability studies haveidentified the effect of chemical structure, conditions and initiating species on the formationof polymeric peroxide gums in fuel storage. Alkenes, particularly conjugated alkenes, tendto form such gums more readily. Asomaning and Watkinson (1992) proved the hypothesisdrawn from plant observations that autoxidation fouling is linked to the formation of suchgums and also found that fouling was determined by the alkene structure, as described infuel stability studies.Autoxidation fouling is still poorly understood and no mechanistic model exists toexplain the interaction of temperature, hydrodynamics and chemical reaction in suchsystems. Studies have identified the significance of dissolved oxygen, temperature andfeedstock composition but reported velocity effects contradict, partly because of thediversity of experimental methods. Crittenden et al. (1987b) reported a fouling model for442. Literature Reviewpolymerisation fouling using a well defined chemical system and this approach was used toinvestigate autoxidation fouling.The objective of this study is therefore to investigate the mechanisms involved inautoxidation fouling and the links between the series of reaction and deposition steps whichgenerate foulant. The process is poorly understood and requires experimental verificationof the mechanisms involved. This knowledge could then be used to formulate and verify amechanistic model for autoxidation fouling at moderate temperatures in turbulent flow,single phase heat transfer to hydrocarbon liquids.Intermediate objectives were1. Identify a suitable mOdel solution for studies of autoxidative fouling;2. Determine the model solution reaction mechanism and significant kineticparameters;3. Determine the effects of temperature and velocity in autoxidation fouling;4. Test numerical fouling models against observed results.The use of model solutions of active alkenes in inert solvents reduces the reactioncomplexity so that the formation of deposit and deposit structure can be compared with thealkene reaction kinetics and products. Model solutions also facilitate the development ofmathematical models of fouling as physical and chemical parameters can be estimatedreliably to test the reliability of such models.453. Experimental Methods and Materials3. Experimental Materials and MethodsThis chapter describes the experimental materials, apparatus and procedures used inthe current study of autoxidation fouling. The model solution used in the fouling studieswas selected after a search of potential alkenes and solvents described in Sections 4.1 and5.1. The experimental apparatus and chemical analyses were developed during theexperimental program on the basis of operating experience. Experiments were numberedaccording to type as described in Table A. Materials and Physical Properties3.1.1 Model SolutionsThe model solutions used in the fouling and kinetic studies of indene autoxidationconsisted of an alkene which undergoes autoxidation readily under the experimentalconditions and a solvent which should remain relatively inert. Candidates for solvent liquidwere selected on the basis of their chemical inertness to autoxidation, chemical nature(aromatic, polar, aliphatic, miscibility with the alkene), safety in handling, boiling point (tomaintain single phase heat transfer) and cost. Alkenes were selected on the basis of cost,availability, toxicity and existing knowledge in the fouling and chemical literature.The alkene used was primarily indene (C9H8), based on the experience ofAsomaning and Watkinson (1992). This was obtained as a technical grade liquid (92wt%+) from Aldrich Ltd. and was stored frozen until required. No attempt was made topurify the indene due to the large quantities involved. Gas Chromatography - MassSpectrometry (GCMS) analysis indicated that the impurities present were those expectedfrom indene manufacture. Peroxide analysis showed that relatively little autoxidation hadoccurred during production and storage. Different batches of indene did cause variations in463. Experimental Methods and Materialsthe rate of autoxidation and so series of experiments were performed using indene from thesame manufacturer’s batch. The physical properties of indene are summarised in Table 3.1.Table 3.1 also lists the properties of the other alkenes investigated; hexadec-1-ene,C16H32,and dicyclopentadiene, (DCP), C10H8the dimer of the cyclic diene pentadiene.Hexadec-1-ene was obtained as a technical grade liquid from Aldrich Ltd. (94 wt%) andused as obtained. DCP was also obtained as a technical grade liquid from Aldrich Ltd. (95wt%) but contained an oxidation inhibitor, p-t-butylcatechol, which had to be removed bydistillation. This was performed by G. Zhang using a Penn State column and the distillatewas stored frozen under nitrogen. DCP was used in two fouling experiments to study theinteraction of two active alkenes in a model solution. DCP was not used subsequently dueto problems in the peroxide analysis and its bad odour.Table 3.2 summarises the properties of the various solvents used in the modelsolutions. Trichlorobenzene (Aldrich Ltd.), toluene (BDH) and n-octane (Fisher) wereobtained as high purity liquids. The kerosene used was a commercial kerosene supplied byImperial Oil Ltd. and contained an unknown oxidation inhibitor. 13C nuclear magneticresonance (n.m.r.) spectroscopy showed that the kerosene was mainly paraffinic in naturebut also contained some alkene and aromatic compounds. Paraflex HT1OTM is a process oilmarketed by Petro-Canada Ltd. It consists of a lubricating oil basestock hydrotreated twiceto saturate all components to their alkane equivalents, giving a very stable, inert liquid. 13Cn.m.r. did not show any unsaturated species.Two free radical initiators were used to study the effects of chemical initiation onthe autoxidation of indene. Benzoyl peroxide (bP) (Aldrich Ltd.) and 2-2’-azo-bis-2-methylpropionitrile (ABN) (Pfaltz and Bauer) are both thermally activated initiators whichdecompose on heating to from free radical fragments. These materials were obtained as473. Experimental Methods and MaterialsTable 3.1Table 3.3Alkenes used in Model SolutionsPhysical Properties of Initiators and AntioxidantsCompound Benzoyl Azobis-2-methyl 2,6-t-butyl-4- Tributylamineperoxide propionitrile methyiphenol (TBA)(bP) (ABN) (BMP)Formula (C6H5CO)20 [(CH3)2C CN)N] [(CR3)Cl2.CR3. [CH3(2)jN6H2OHFormula Mass 242.2 164.0 220.36 185.4Activity Initiator Initiator Antioxidant AntioxidantMelting Point 104-106°C 102-103°C 69-70°C -70°CDensity (20°C) solid solid solid 778 kg/rn3Boiling Point - -- 216°CStructure-oCompound Indene Hexadec- 1 -ene Dicyclopentadiene(DCP)FormulaFormula MassActivityMelting PointDensity (20°C)Boiling PointStructureC9H8116.1aromatic alkene- 2.0 °C996 kg/rn3181.6 °C(C5H)2130.21diene dirner-1.0°C986 kg/rn3170.0 °CC16H32224.3aliphatic alkene4.0 °C783 kg/rn3274°C483. Experimental Methods and MaterialsTable 3.2 Properties of Solvents used in Model SolutionsSolvent n-octane toluene Paraflex kerosene trichioro- tetralinProperty HT 10 benzeneSupplier Fisher BDH Petro Canada Imperial Oil Aldrich Ltd. Aldrich Ltd.Ltd Co.Density 705 kg/rn3 873.4 kg/rn3 855.4 kg/rn3 808.5 kg/rn3 1459 kg/rn3 973 kg/rn3(15.6°C)Colour [D1500j + 30 ÷ 30Flash Point 13.3 °C 4.4°C 166°C 45°C 105.58 °C 71.13 °C[D92]Viscosity 0.57 rnPa.s 0.62 mPa.s 23.9 mPa.s 2.26 mPa.s 3.20 mPa.s 2.3 rnPa.s(15.6°F)Pour Point -21°C[D97]Aromatics 0 wt% 100 wt% 0.06 wt% 1-2 wt% 100 wt% 100 wt%[PCM 435]Prandtl Number(100°C) 5.97 4.57 37.3 12.4 10.1 10.5(80°C) 6.33 4.98 53 14.1 12.3 12.2Chemical aliphatic aromatic saturated ahphatic, chlorinated aromaticActivity compounds some aromaticunsaturatesFormula Mass 114.2 92.1 310 145.5 132GC Distillation[D28871lwt% 87°C 190°C10 314°C30 333 °C50 347°C 230°C90 369°C 250°C99 464°CBoiling Pt. 125.6°C 110.6°C 213.0°C 207.6°CStructure/VVReferences in parentheses refer to ASTM or API testing methods493. Experimental Methods and Materialspure solids from the manufacturers and stored frozen after opening. No attempt was madeto purify the initiators. The effect of antioxidants on indene autoxidation was studied usingdi-t-butyl-4-methylphenol (BMP) and tri-t-butylamine (TBA), obtained as 99 wt% purecompounds from Aldrich Ltd. Table 3.3 summarises the relevant properties of thesematerials. All but TBA exist as solids at room temperature and were added to the solutionby dissolving the powder in the indene before this was added to the reactor or holding tankto start an experiment.3.1.2 Physical PropertiesThe physical properties of trichlorobenzene, indene, tetralin, n-octane and toluenewere readily available from the literature (Coulson and Richardson 1983; Daubert andDanner 1989), whereas the Paraflex and kerosene were obtained in 205L bulk shipmentswith only general property data. The boiling point, viscosity and density were measured inorder to obtain accurate data for use in the general correlations for specific heat capacity andthermal conductivity. The kerosene data was collected and analysed by G. Zhang (1990).The Paraflex analysis was performed in conjunction with R. Lai. Kerosene PropertiesKerosene viscosity was measured using a Fenske viscometer immersed in aconstant temperature bath; kerosene density was measured using a hydrometer and the dataregressed to give the following functions of temperature. The thermal conductivity and heatcapacity were then estimated by Zhang using an API method. The data were collected overthe range 45-100°C.Density p (kg/rn3) = 820.15-0.7 T(°C) [3.1]Viscosity i. (Pa.s) = in {1I(T(°C) + 44.737) + 8.5142}- 2.1419 [3.2]503. Experimental Methods and MaterialsHeat Capacity Cp (J/kg.K) = 1874.7 + 3.7559 T(°C) [3.3]Thermal Conductivity ?. (W/m.K) = 0.13 11 - 1.4123 x 10-4T(°C) [3.41Figures 3.1 and 3.2 show the kinematic viscosities and Prandtl numbers of the solventsused in the model solutions along with that of indene, the alkene selected for foulingstudies. The difference in physical properties between indene and Paraflex is much greaterthan that between indene and the other solvents, so the density and viscosity of solutions ofindene in Paraflex were also measured and compared with existing correlations. Paraflex PropertiesParaflex viscosity was measured using a Haake VT500 rotary viscometer equippedwith an NV coaxial cylinder sensor system for liquids of low viscosity. The NV system ismounted in a controlled temperature bath and the unit is interfaced with a computer for dataanalysis. The results were fitted to a number of viscosity-temperature expressions and thebest correlation was obtained using the form derivd by Puttagunta et al. (1992) forconventional petroleum liquids;Kinematic Viscosity logy0(v (m2/s) = KVb + KVc[1 + (T(°C) - 37.78)!310.93]s [3.5]where KVb = 1.95; KVc = -0.8696; KVs = 2.1642. The densities of both Paraflex andsolutions of indene in Paraflex were measured using specific gravity bottles. Petro Canadarecommended the use of the Fisher Dens fly/Temperature Tables to calculate Paraflexdensity at other temperatures. The values obtained for the range 15-100°C were then fittedto the expressionDensity p (kg/rn3) = 864.04 - 0.588 T(°C) [3.6]Petro Canada also supplied expressions from a Shell Thermia Oils Technical Bulletin forthe thermal conductivity and heat capacity, based on the liquid density at 60°F (15.6°C).These expressions were used to give the following property functions.513. Experimental Methods and Materials1000 20Figure 3.1 Kinematic Viscosities of Solvents used in Model Solutionsicr70 20 40 60 80 100Temperature(°C)Figure 3.2 Prandtl Number of Solvents used in Model Solutions120io2101 -I • I I I —— Iindene40 8060Temperature(°C)100 120523. Experimental Methods and MaterialsThermal Conductivity (W/m.K) = 0.13711 - 7.40524x iO T(°C) [3.7]Heat Capacity Cp (J/kg.K) = 1821.6 + 3.6676T(°C) [3.8]The heat capacity calculated at 20°C compared favourably with the predictions of the Chuehand Swanson method (see Coulson et al. 1983) and the results of some simple calorimetryexperiments using hexane and warm Paraflex in a vacuum flask. The mean molecular massof the Paraflex oil was estimated at 310, which was confirmed by the viscosity estimationprocedure described in ASTM D-2502-87.The densities of solutions of 0.0-1.0 mol/L indene in Paraflex were measured at20°C and fitted a linear function of indene concentration;Density p (kg/rn3) = 152.28 + 14.96 [indene] (mol/L) [3.91The viscosities of 0.0 - 1.0 mol/L solutions of indene in Paraflex were measured at 20°Cusing a Cannon Manning Semi-Micro Viscometer following ASTM Standard Test D445.This method gave more reproducable results at 20°C than the Haake rotary viscometer. Thevalues of Paraflex and indene viscosity reported at 20°C agreed within the limits ofexperimental error. Figure 3.3 shows the data and the predictions of the viscosity mixingrules of Kern (1950), Chhabra (1992) and Souders (see Coulson et al., 1983).Kern = w1/pt + w2Iu [3.10]Chhabra 1/Jtmix = {iI(jt12) + (1)2/(1422) }O.5 [3.11]Souders logio(log1(10 ptmjx))= Prnix Is,mix/Mmix X i0 2.9 [3.12]where w1 is the mass fraction of component i, ct is the mole fraction of component i;‘S,i?xand M11 the mole averaged Souders viscosity constant and molecular mass respectively.None of the mixing rules exhibits the observed viscosity behaviour so the viscosity of asolution was estimated using the viscosity of Paraflex and a ratio obtained from theexperimental data (e.g. 0.79 for 0.405 M indene in Paraflex). The change in viscosity wasevident in the TFU fouling experiments as the heat transfer coefficients reported forsolutions of indene in Paraflex were larger than for Paraflex alone at the same mass flowrate. The viscosity dependence in Re (—0.8) is larger than that in Pr (— -0.33), so a larger533. Experimental Methods and MaterialsFigure 3.3 Viscosity of Mixtures of Indene in Paraflex at 20°C; Comparison ofCorrelation Predictions and Experimental Data30251 .0200.0 0.2 0.4 0.6 0.8[indene](mol/L)543. Experimental Methods and Materials3.1.2.3 Concentration of Dissolved GasesAutoxidation does not occur in the absence of oxygen, the solubility of which inorganic liquids is thus extremely important. A direct method for determining dissolvedoxygen concentration was not available during the current work; the on-line gaschromatography method described by Hazlett et at. (1977) was beyond the scope of thisstudy. The saturated concentration of a dissolved gas can be calculated using ASTMmethods D2779-86 and D3827-86, which use the liquid density and solubility parameter,aH, respectively to calculate the Bunsen coefficient of the dissolved gas. Table 3.4 showsthe values calculated for the solubility of oxygen in kerosene and Paraflex at 377 kPa airsaturation. The values for kerosene vary markedly and neither method shows the expecteddecrease in oxygen solubility with liquid temperature. It is evident , however, that oxygenis sparingly soluble in the solvents used in the current study. The ASTM D2779-86 methodwas used in subsequent calculations; this requires an estimate of the Ostwald coefficient foroxygen, L0, obtained from a plot. The graphical data were fitted to Equation [3.13] with acoefficient of variation (R2) 0.996.= 0.15876 + 5.9342x104T(°C) - 7.3367x10T2 [3.13]The calculation method is described in detail in the ASTM procedure.Table 3.4 Estimates of Dissolved Oxygen Concentration in Paraflex and kerosene[ppmw at Pair = 377 kP]value of Nu is expected. The difference in thermal conductivities was negligible, so wasignored.Solvent 80°CParaflexkeroseneMethod100°C21431080°C215314100°C186213ASTM D2779-86 ASTM D3827-86188213553. Experimental Methods and Materials3.2 Portable Fouling Research Unit (PFRU) ApparatusThis apparatus was constructed by Fetissoff (1982) and modified by subsequentworkers. The device features two fouling monitors; a Hot Wire Probe (HWP) designed tostudy fouling in laminar regime liquid flows, and the Portable Fouling Research Unit(PFRU) probe designed for more turbulent liquid flows. This apparatus was used in theinitial fouling experiments and in the studies of fouling in the presence of antioxidants. TheHWP proved difficult to operate and was not used after the first fouling studies.Figure 3.4 is a schematic drawing of the PFRU apparatus. All surfaces in contactwith the liquid were constructed from stainless steel as this was not thought to be asignificant source of metal ions. Seals were made from Teflon where possible as thisoffered greatest resistance to the fouling solutions and cleaning liquids. Rubber-based sealsand components were found to degrade slowly on exposure to the autoxidation solutions.The apparatus was located inside a partial fume hood.Liquid is pumped by a 3 h.p. centrifugal pump from a 9.45 L holding tank througha flow control valve, a set of orifice plates, the fouling probe and a rotameter before beingreturned to the tank. The orifice plate = 15.8 mm, dor 7.94 mm, Cor = 0.639(100°C)) and rotameter were calibrated at the liquid operating temperature using a bucketand a stopwatch. The pressure drop was recorded using a mercury manometer. Therotameter was later removed in order to increase the maximum flow rate when investigatingflow velocity effects. Figure 3.5 shows the orifice plate calibration for Paraflex at 80°C and100°C using the expressionW = CorAt {2 p APorI[(dt,ildor)4- 1j}0.5 [3.14]where W is the mass flow rate, A is the tube cross sectional area (= 3tdt2/4), APor is thepressure drop across the orifice and Cor the office discharge coefficient. Similar calibrationswere performed for tetralin and kerosene.563. Experimental Methods and MaterialsFigure 3.4 Schematic Diagram of PFRU ApparatusAir or CoolingWaterPumpOrifice Meters573. Experimental Methods and MaterialsFigure 3.5 PFRU Orifice Plate Calibration for Paraflex0.400.350.300.250.201140. 50 100 150 200 250 300(a)583. Experimental Methods and MaterialsLiquid samples were taken downstream of the fouling probe using a sequence ofball valves arranged so that sampling did not disturb the flow. Samples had been initiallytaken from the holding tank but this did disturb the flow. Air or gas mixtures were suppliedto the airspace in the holding tank, providing a gas blanket above the solution. Mixing inthe holding tank was provided by recirculation of the liquid. Relief valves were fitted toavoid overpressure conditions at the pump and in the tank. The system was insulatedthroughout and cooling was supplied by mains water passing through a coil in the holdingtank, controlled by a needle valve. Heating tapes on the tank exterior and lagging were usedto bring the liquid to its operating temperature and were usually left on during anexperiment. Temperatures were measured using ‘J’ type thermocouples located at variouspoints in the system. Pressure gauges and mercury manometers were used to monitorabsolute and differential pressures respectively. Temperatures and pressures weredisplayed on the instrument panel.Figure 3.6 shows the design of the PFRU probe. Liquid flows up the annulusbetween the metal core and the outside wall. An electrically heated section 102mm in lengthis located 20 equivalent diameters downstream of the probe entrance. The outer wall of theannulus is constructed of glass at this point to allow visual inspection of the heated surfaceduring the experiments. The PFRU probe was originally supplied by Heat TransferResearch Inc. (HTRI, College Station, TX) and had developed an unrepairable bend whichprevented true alignment of the inner core. This prevented its use in heat transfer studies.The heated section consisted of an electrical heater embedded in a ceramic matrix andsheathed with a stainless steel tube. Imbedded in the sheath are four ‘J’ type thermocoupleswhich measure the temperature at a distance x below the surface; the surface temperature isthen calculated using calibrations supplied by 1-ITRIT surf = T measured - (Xs/?net) q [3.15]593. Experimental Methods and MaterialsFigure 3.6 Schematic Diagram of PFRU Fouling ProbeI I294 mm10.7 mm* -- 25.4 mmtPLOW60to Digitrend DisplayandDoric235Dataloggerto Variac150 mmA78mm .102 mm3. Experimental Methods and MaterialsThe thermocouples are located 80 mm from the start of the heated section. A high surfacetemperature trip was installed to prevent the probe burning itself out at temperatures greaterthan 350°C.The heater is powered from a 110 Vac supply via a variac; the voltage is measuredand used to calculate the power and thus the heat flux at the surface. The voltage wasrectified and stepped down for the datalogger using circuitry described by Fetissoff (1982).The heat flux calibration was performed by G. Zhang and was fitted to a fourth orderpolynomialq (kW/m2) = 135.67- 1129.2 v (volts) + 4143.7 v2 - 4914.6 v3 + 2438 v4 [3.16]where v is the datalogger voltage reading. The PFRU heating voltage and systemtemperatures were recorded remotely on paper tape using a Done 235 datalogger. Thesedata were transferred to a Microsoft ExcelTM spreadsheet for further processing andpresentation.The variac setting is kept constant during an experiment so that the device operatesat a constant heat flux. If the deposit does not modify the solid surface/liquid heat transfercoefficient then this condition also gives a constant depositlliquid interface temperature.Any increase in metal surface temperature is then due to the thermal resistance of thedeposit. The heat transfer coefficient was calculated from;q = U(t) (Tsurf - Tb) [3.17]The fouling resistance could then be calculated using the expressionRf(t) = 1/U(t) - i/u0 [3.18]The clean heat transfer coefficient, U0, was based on the average value calculated at thestart of the experiment.613. Experimental Methods and Materials3.2.1 PFRU Heat TransferThe heat transfer characteristics of the PFRU were studied using clean solventsunder experimental conditions. One of the thermocouples was inoperative and the skewedalignment of the annulus core produced non-uniform surface temperatures. The heattransfer coefficient, U, in Equation [3.171 was calculated using a mean surface temperatureand expressed as a Nusselt number, Nu = Udh/A, where d11 is the hydrodynamic diameter,= da,o - da,j = 14.3 mm. Heat transfer tests were performed for kerosene and Paraflex,the most frequently used solvents. Table 3.5 compares some results with the predictionsfrom the following heat transfer correlations.Monrad and Pelton (1942) Nu = 0.020 Re08 Pr (da,o/da,i)°53 [3.19)Wiegand (1945) Nu = 0.023 Re08 Pr 0.4 (dao/dai)°5 [3.20]Gnielinski (1986) Nu = (f18) (Re - 1000) Pr [3.21]1 + 12.7 ./(f/8) [Pr213-1]where the Moody friction factor,f is given by Knudsen’s expression for annuli (1958);f = 0.304 Re-025 [3.221The Sieder-Tate (1936) surface temperature correction, (pc/ji)°’4was also used incalculating the final values. The experimental results in Table 3.5 showed some agreementwith the annuli correlations. Asomaning (1990) reported similar findings in the PFRUusing kerosene alone. Systematic error sources lie in the reliability of the physical propertydata and the fouling probe itself. The geometry of the PFRU probe is such that thermalentry length effects will still be significant at the thermocouple location, which correspondsto x/dj = 5.6. This is particularly true for Paraflex, where the high solvent viscosity resultsin a low Reynolds number. Kay and Lambourn (see Fetissoff, 1982) calculated thethermal entry length effects for constant heat flux and fully developed flow between parallelplates at Re = 7906, Pr = 10. Interpolating their results for x/dh = 5.6 gave Nu(x)!Nu =1.074, which is significant although the large value of Pr for Paraflex would tend to623. Experimental Methods and MaterialsTable 3.5 Comparison of Nusselt Numbers in PFRU Heat TransferSolvent kerosene tetralin trichioro Paraflex ParaflexbenzeneBulk Temperature 80°C 80°C 80°C 80°C 100°CPr 14.1 12.2 12.3 53.0 37.3Re 10690 10720 12090 2550 5145Nu(PFRU) 163 155 172 97-106 137Nu(Wiegand) 181 144 166 109 166Nu (Monrad-Pelton) 142 134 154 79 122Nu (Gnielinski) 119 113 132 41 90633. Experimental Methods aiwl Materialsreduce this figure. Heat transfer data from the initial fouling experiments showed that Nuce Re065’which does not agree with the correlations. The PFRU was thus concluded to beunsuitable for heat transfer studies but still useful for initial fouling studies.This work is concerned with single phase heat transfer so the variation of Nu withheat flux was studied in order to determine the onset of subcooled boiling. The maximumoperating surface temperature for kerosene at 378 kPabs and Re = 10690 was found to bearound 210°C, whilst the maximum surface temperature in tetralin at 378 kPa abs, Re =7671 was 270°C. Subcooled boiling was not observed in Paraflex below 300°C.Subcooled heat transfer in kerosene in the PFRU is discussed by Zhang et al. (1993). TheNusselt number for Paraflex at Re = 2550 and Tblk = 80°C increased as q0’3; thiscorresponds to a surface temperature correction such as that of Sieder and Tate.3.2.2 PFRU Experimental ProcedureThe annular probe and the apparatus were cleaned thoroughly before eachexperiment. The tank was filled with the charge of solvent and run for several hours at therequired bulk temperature and air overpressure in order to ensure that the liquid wassaturated with air and the system warmed up. Compressed air was usually drawn from thebuilding supply although this often fluctuated due to demand from other users. Thedatalogger was started before adding any alkene as it was located in an adjoining room. Tostart a fouling experiment, the system was quickly depressurised and momentarily shutdown so that the alkene could be added and rinsed in with some of the solvent charge. Thepump was turned on, air pressure restored and the filling line converted to the sampling linewhile the unit returned to the required operating conditions. The PFRU heat flux and flowrate were set and the first sample was then taken. This starting procedure rarely took morethan twenty minutes. Attention was paid to adjust the cooling water flow rate to maintain asteady bulk liquid temperature.643. Experimental Methods and MaterialsSolution samples were taken as required during the experiment and visible changesin solution nature and heated surface appearance were recorded. The following aliquotswere quickly pipetted into suitable sample containers - 2 mL (indene concentration), 10 mL(gum assay), 2 x 2OmL (Peroxide Number). The shut-down procedure involveddepressurising the system whilst gradually reducing the PFRU power level in order toavoid disrupting any deposit by a thermal shock. Once the pressure had reduced toatmospheric and the PFRU power to zero, the pump was turned off and the probe quicklyremoved. The probe was left to cool then washed in hexane to remove any residual solvent.The reaction liquor was drained from the system and a charge of solvent rinse added with asmall amount of acetone, primarily to cool the apparatus down. The probe was replaced bya dummy probe, the coolant flow increased and the rinse circulated until the apparatus wascool enough to allow safe loading of acetone. Two acetone rinses in succession were usedto remove any residual indene polyperoxide material; methyl-iso-butyl-ketone (MIBK) andtetrohydrofuran (THF) were also used to clean the system where the unit showed signs ofcontamination. The acetone was removed from the system by prolonged purging with air.Any further acetone was removed by two rinses with the next solvent. The rinse volumeswere coordinated in such a way as to recycle as much solvent as possible without affectingthe efficiency of the cleaning program.The cleaned probe was inspected by eye and using an optical microscope.Photographs of the fouled probe were taken for reference. The deposit thickness wasmeasured using vernier calipers but this was complicated by the softness or brittle nature ofthe deposit. Photographs and diapositives of the magnified deposit surface were obtainedusing either a 35 mm camera or a Polaroid camera fitted to a Nikon HFX-II opticalmicroscope system. The magnification was calibrated using an etched vernier scale. Thedeposit was often scraped off the probe and kept for further analysis, taking care not todamage the metal surface. The probe was then cleaned using acetone and household cleanerbefore the next experiment.653. Experimental Methods and Materials3.3 Stirred Cell Reactor (SCR)The autoxidation of indene in solution was studied in a stirred batch reactorconstructed by G. Zhang. Figure 3.7 is a schematic drawing of the apparatus. The 3Lcylindrical reactor and all fittings in contact with the liquid were constructed from type 316stainless steel; Teflon seals were used where possible. Air was admitted via a gas linewhich could be set to sparge gas into the bulk liquid or into the gas space above the liquid.The latter configuration mirrors the PFRU holding tank geometry whereas the formercorresponds to the TFU arrangement. Gas flows were monitored using rotameters andcalibrated using graduated cylinders immersed in a water bath. Air was supplied from thebuilding supply or from gas cylinders; a one way stop valve was fitted to the buildingsupply line to prevent suckback when the feed pressure cycled. The system pressure wasread from an Omega 0-100 psig pressure guage. A pressure relief valve was used to avoidoverpressure as Alexander (1990) describes various explosions caused by runawayautoxidations at higher temperature and pressure.The reactor was mounted in a stirred, heated oil bath located inside a fume hood.The oil bath heaters were controlled by an Omega Model 49 temperature controllerconnected to the reactor’s ‘K’ type thermocouple. The oil was circulated by a GKH 1/40h.p. electric motor, controlled by a GKH GT-21 speed controller. The reaction liquor wasagitated by an impeller driven by a top-mounted electric motor (GKH, 1/8 h.p.) throughmechanical seals. The impeller speed was controlled by a GKH S-12 motor controller.Few problems arose in control of the reactor.Samples of liquid were withdrawn through a 1/8 inch sample line and a needlevalve. The sample volume was collected in a 100 mL jar and the aliquots described inSection 3.2.2 pipetted into the respective containers. The sample line and pipettes tended toget blocked by insoluble orange/red gum towards the end of an experiment; no solutionwas found to this problem. This gum also settled out rapidly in the collection jar and so66Figure 3.7 Schematic Diagram of Stirred Cell Reactor (SCR) Apparatusto vapourtrap pot3. Experimental Methods and MaterialsMotorI —ReliefThermocouple IvalveI —RotameterGasSupplyOne WayValvestiffer..\ SS SS’.\ S\\\\ S”. \Heating Oil Bath- / / / / // / / // / / // / / / / / / / / / / / / / / / / 7, / / /7, /673. Eperimenta1 Methods and Materialsgum analysis during this period showed considerable scatter. The pipettes were rinsed withacetone and hexane and left to dry before re-use.An experiment was started by charging the reactor with a volume of solvent andrunning the reactor at the required conditions for several hours. The reactor was thenquickly depressurised, a known volume of indene added and the system re-pressurised.The air flow rate, liquid temperature and motor speeds were checked before the first samplewas taken, ensuring that the solution was fully mixed. Samples were taken at regularintervals until the experiment was completed, at which time the reactor was depressurisedand the heater turned off. The reactor was removed from the oil bath and the remainingliquor kept for analysis or disposal. Any gum formed on the reactor walls was noted,washed in hexane and removed for analysis. The reactor was then stripped down andcleaned with acetone to remove any residual products. Severe contamination occurred in thetrichlorobenzene run and more aggressive cleaning was required.3.4 Tube Fouling Unit (TFU)The Tube Fouling Unit was constructed in order to overcome the shortcomings inthe design of the PFRU fouling monitor; examination of the PFRU deposit is limited andthermal entry length effects are significant along its relatively short heated section. TheTFU provided a longer heated length with a known axial temperature profile, and allowedthe deposit to be examined in situ as the fouled sections were designed to be removed afteruse. Operational experience from the PFRU proved invaluable in designing and operatingthe TFU.683. Experimental Methods and Materials3.4.1 Tube Fouling Unit ApparatusThe apparatus is a modified version of Watkinson’s design (1968) adapted to useremovable test sections. The test section consisted of a 1.83m long 304L stainless steeltube (d,0 9.525 mm, 9.017 mm) heated across its central section by alternatingelectric current. The power to the heated section was kept constant during a run so that thedevice operated at constant heat flux, as in the PFRU.Figure 3.8 is a schematic diagram of the apparatus. All materials in contact with theliquid were constructed from stainless steel or Teflon. Liquid was recirculated from a 65 Lholding tank through a pump, an orifice plate, the heated section, a series of coolers and arotameter, then returned to the tank as a jet to promote mixing. Hoke globe valves wereused to split the flow between the test section and a return line, which ensured good mixingwhen lower flow rates were used in the heated section. A mesh strainer could be switchedon line as desired. The specifications and origins of the equipment are listed in Table A.1.2in Appendix 1. The tank included filling and drain ports, a low level alarm float device anda 3 kW low heat flux heater to warm the liquid to the operating temperature before anexperiment. The tank heater was controlled by an Omega CN 911 controller and was notrequired during an experiment. The system was pressurised by gas which was bubbledinto the liquid from the base of the tank and vented through the top. Gas flows could bedrawn from the building compressed air supply or a cylinder source and were controlled byrotameters on the control panel. The rotameters were calibrated as described previously.The tank and all system piping was insulated by fibreglass insulation. The piping sectionbetween the pump and orifice plate was also fitted with heating tapes connected to avariable transformer in case insulation losses were significant but the pump heat output wasfound to counter any such losses.Liquid flow rates were monitored using a rotameter and a set of orifice platesconnected to a differential pressure transducer. The orifice plates (d,1 = 12.2 7mm, d0,. =693. Experimental Methods and MaterialsFigure 3.8 Schematic Diagram of Tube Fouling Unit ApparatusPRV primarygas to ventgas linePAR - Pressure Alarm High; LAL - Level Alarm Low; TAR - Temperature Alarm High; PRV - PressureRelief Valve; T - Thermocouple; P - Pressure Transducer; AP - Differential Pressure TransducerTAHsample line/drainsample line/drainheating tape lines703. Experimental Methods and Materials9.525 mm, Cor = 0.571 ) were calibrated using pure Paraflex at 100°C following theprocedure described in Section 3.2. The calibration plots for the rotameter and orifice platesare given in Figure A.1.l. The orifice plate pressure drop was recorded for data processingwhile the rotameter was used to monitor the flow rate and provide a visual record of theliquid.The temperature of the bulk liquid entering and leaving the heated section wasmonitored by thermocouples mounted in T-pieces and positioned in the centre of the flow.The fluid temperatures were displayed on the control panel for reference and controlpurposes. A series of alarms were built in to the system wiring to minimise risks and aredescribed in Table 3.6. The pressure drop across the heated section was monitored by adifferential pressure transducer connected to T-pieces at the inlet and exit of the test section;absolute pressure transducers were also fitted in case the drop exceeded the working rangeof the differential unit. The differential pressure transducers were calibrated using a Marsh0-15 psig test gauge.The coolers removed the heat supplied in the heated section. The primary coolerconsisted of an 2.44 m long x 1 inch diameter copper tube around the 1/2 inch stainlesssteel tube carrying the process liquid. Mains cold water passed countercurrent to theprocess fluid and was controlled using a rotameter and globe valve on the control panel.The primary cooler was overdesigned for the heating duties and flow rates involved andmaintained the bulk temperature constant during all Paraflex heat transfer tests. The primarycooler capacity decreased during fouling experiments due to fouling so two auxiliarycoolers were added downstream of the primary unit. These were of similar construction tothe primary cooler but were shorter (0.5 m, 0.4 m) and used cooling water from a separatetap supply, so that they could be operated independently of the primary unit. Cooler foulingproved to be a serious problem in operating the TFU at high heat fluxes and imposed limitson the maximum surface temperature which could be maintained at a given flow rate over a713. Exverimentai Methods and MaterialsOverpressureFluid Leak/Runs DryOverheated LiquidTube Burn OutThermocouple DamagePower SurgesPAR: PRVLALTAH on cooler outletTAR on tube surafceTAR on tube surfaceReset relaysSystem Shut DownSystem Shut DownSystem Shut DownHeating OffHeating OffPower stays offTable 3.6 TFU Alarm MatrixHazani Cause Trip ActionBlockage/ExplosionLeakage/RuptureCooilng Water FailureFlow StoppedTube too hotPower Failure/Restored723. Experimental Methods and Materialsgiven heated length. Cooler fouling was not observed in the PFRU due to theproportionately smaller heating rates and the design of cooler, i.e. an immersed coil.Liquid samples could be withdrawn at various points in the process fluid circuit.The main sampling position was located before the orifice plate and consisted of a needlevalve and a ball valve in series. Similar lines were located at the primary cooler exit and thepump. The sample line after the test section consisted of two electrically operated valvesconnected by a 100 mL volume of 3/8 inch piping. This safety arrangement was used asthe process fluid line was beyond normal reach and ran close to the heated section. Themanual sample points were used along with the tank drain valve in cleaning the system.The heated section was heated by the direct passage of alternating current. The tubelengths were constructed of drawn T304 stainless steel (ASTM A269-80A) and weresupplied by Greenville Tube Corp, Clarksville, Arkansas. The tubes had a nominalthickness of 10 thousands of an inch, but tube thickness measurements using a ballmicrometer and calculations from tube masses gave values ranging between 11 and 14thou. The tube was connected to the fouling loop by Teflon SwagelokTM fittings whichensured that the test section was electrically isolated from the rest of the apparatus. Currentwas supplied to the test section from a mains 208 Vac supply via a power variac and a 208-19V step-down transformer. The step-down transformer was connected to the test sectionby pairs of #3 welding cables bolted to 10mm thick copper busbars. These busbars wereisolated from the frame by Tufnol holders and were designed in two parts which screwedtogether to clamp the tube in place. The first busbar was located 500 mm (55 tubediameters) from the upstream fitting to ensure well developed duct flow in the heatedsection. The second busbar marked the end of the heated section and was positioned asdictated by the power requirement. The resistance of the stainless steel tube wasunderestimated at the design stage and the step-down transformer oversized for long test733. Eperiinental Methods and Materialslengths; high heat fluxes required larger currents and thus a shorter test section. Figure 3.9shows the arrangement of the test section in schematic form.Figure A.1.2 describes the electrical network in detail. The voltage across the testsection, v, was measured using an a.c. panel meter and displayed on the control panel. Thecurrent, I, was measured using a current transformer on one set of welding cables. Thecurrent transformer output was measured using an ac ammeter and this was calibrated usingan Amprobe ACD-9 Current meter:I (Amps) = 0.06733 + 142.85 (ammeter reading) R2 = 1.000 [3.23]The power to the heated section was calculated assuming a power factor of unity (afterWatkinson, 1968) as equal to vI, then converted to a heat flux using an estimated surfacearea based on a tubing thickness of 10 thou;Heat Flux q (kW/m2) = vi / [L ru (0.009017)] = 35.3 vI/L 13.24]where L is the length of the heated section. The maximum error in local heat flux wasestimated using a tube thickness of 14 thou; this increased the value of q by 1.7%. Tubethicknesses varied along each tube and the means were closer to 11 thou than 14 thou; thenominal thickness was thus considered a reasonable figure to use in the calculations.3.4.2 Surface Temperature MeasurementWatkinson (1968) and Hopkins (1973) used similar electrically heated test sectionswhere the tube was reused after each experiment. These workers used silver soldered andglued thermocouples respectively to measure the tube’s outer surface temperature; thesemethods were not appropriate for the current system of readily removable test sections. Analternative was sought which would provide accurate surface temperature measurementwith electrical isolation and durability. The method selected involved thin film temperaturesensors which were compressed against the side of the tube using clamping blocks743. Experimental Methods and MaterialsFigure 3.9 Schematic Diagram of Heated Section ConstructionTufnol Steel CopperLava Ceramic PTFE Fibreglass insulationTube [] Ceramic Fibre%%%%%‘s F-s,,_,_ F5%%‘.%\% ,sF-’%%%%%% F-s,,_,_ F-..,,_,,‘‘5%’’’’‘‘‘F.,,‘5%’’’’F.,,’,,‘‘‘‘‘‘5-‘‘F.,,.‘‘5%’’’‘‘1,,,‘‘‘‘5’’,,,,,.‘‘‘‘‘‘5‘‘F.,,.‘‘5’’’’‘‘‘‘F..‘‘‘‘‘5’‘‘‘‘,.‘‘‘‘‘‘5F.,’,’,‘‘‘‘‘‘5‘‘‘‘‘F.‘‘‘5’’’‘‘‘‘F..‘‘‘‘‘‘5‘‘‘‘‘F.‘‘‘F.,.‘‘‘‘‘‘5‘‘‘‘F.,‘‘‘‘‘‘5‘‘F.’,,‘‘‘5’’’‘‘F.,,,—F. F / / / f fF. / F. F.“5‘‘‘‘‘‘5,,,,,.‘‘‘‘‘5’‘‘‘‘‘F.‘‘‘‘‘‘5‘‘‘‘‘‘5‘‘‘‘‘F.‘‘5’’’’‘5’’’’’‘‘‘‘‘F.‘‘‘‘‘‘5‘‘‘‘‘‘5‘‘‘F.,‘‘‘‘‘‘5‘‘‘F.,‘‘‘‘‘‘5‘f_f,‘‘‘‘5’’,,‘,‘‘‘‘‘‘‘5F.,,,,,‘5,%,%ø,%,%, Ns;s;;-s’5’’’’’ ——% (5f_f,,,,, F‘5,,,,,,,F.‘%\\%%%%%% ‘5,,,,,,,,F.‘%%%\%\\\% ‘5.,_,f,,,F. 3\\%\%%\%% “5f__f,,,, F.‘5f_f,,,,, 1\\\\\%%\% ‘5,,,,,,,, F%%\\\%\\\ ‘5,,_,,,F., FS.,,,,f,F., F‘5,,,,fF.F., F.\\\%%%%%% ‘5f_f__F.,, F.\%\%\%\%\ S.,,,,,,,, F.%\\%\%\%% 5.,,,_,F.,F. >3\\%%%%\\% S_,,,_,F.,F. >3\%\\\%‘5%% 5..a‘‘‘‘‘F.,,‘55,5,Thermocouple wire753. Experimental Methods and Materialsfabricated from heat resistant materials. The method was developed during the current workand updated regularly so that some of the earlier data is less reliable.The thin film sensors used were essentially ribbon thermocouples mounted in a thinpolyimide (KaptonTM)carrier which provided electrical resistance up to 270°C, abovewhich temperature the polyimide quickly degraded. The sensors used were ‘K’ type‘Stikon’ 20112 Foil thermocouples manufactured by RdF Corporation of Hudson, NH.Figure 3.10 shows the design of a sensor and the two designs of sensor clamping blockused in the fouling experiments. Both designs consist of solid blocks of material whichcould be screwed or fastened together to compress the sensor against the tube. The sensorand its copper overbraided lead had to be mounted normal to the length of the tube as thealternating current’s magnetic field could cause large fluctuations in the thermocouplesignal. This is discussed further in Section 3.4.3. The original design was constructedentirely of Tufnol but this polymeric composite was found to char after prolonged exposureto temperatures > 200°C. Design I was developed where the Tufnol mount was retainedfor its strength and flexibility but featured a compression section fabricated from LavaTM,an insulating ceramic material kindly donated by AlSiMag Ltd. This material is machinableuntil it is fired at 1050°C, when it becomes a hard, insulating ceramic. The Lava saddlesappeared to function well but the Tufnol holders charred after continual high temperatureexposure, so Design II was developed where the clamping block was machined entirelyfrom Lava. Design I featured a clip/screw fastening system whereas Design II used offsetscrews to provide the clamping action.Ten thermocouple mountings were located along the heated section and could bemoved as required. The intervening sections were insulated as shown in Figure 3.9; a Lavacollar was placed around the tube and held in place by a wrap of ceramic fibre woven tape.The heated section was then wrapped tightly with 3 layers of aluminum backed fibreglassblanket which were secured by removable straps. These straps and the screw tightenedthermocouple mountings made removal of the tube straightforward. The heated section and763. Experimental Methods and MaterialsFigure 3.10 Details of TFU Surface Thermocouple and Mounting DesignsMATRIX‘2‘I’i-IV1,c-I6mm19 mm 23mm60 mm10mm 17 mm(not to scale)Design I Design IIthermocouple glued into a grooved channel in lower clamping blockStikon 20112 Thin Foil Thermocouple (Dimensions in inches)BUTT BONDEDJUNCTION——.50.75IIr°°773. Experimental Methods and Materialsthe electrical contacts were enclosed by a steel cover which prevented accidental contactwith any live surfaces.The unheated sections of the tube were insulated using a wrap of ceramic fibrewoven tape and fibreglass piping insulation fastened in place by removable straps. Asurface sensor was taped on to the outside of the insulation at the heated section to providean estimate of heat losses to the surroundings.A high temperature alarm sensor, consisting of a thermocouple mounting similar toDesign I, was positioned at the hotter end of the heated section. The Tufnol support wassquare in section and did not extend to the apparatus frame. This sensor was dedicated tothe alarm only so that PC failure would not create a hazard.3.4.3 Data Collection and ProcessingExperimental data were collected using the LabTechTM datalogging softwarepackage on a PC equipped with a DAS-8 analog/digital interface board. Data were saved toa floppy disk and then transferred to a faster PC for data processing using spreadsheetingand graphical software packages.Fluid and surface temperatures were measured using a multiplexer with coldjunction compensation calibrated at 0°C and 100°C. The error in the multiplexer temperatureoutput was given as ±0.7°C. The thermocouples’ accuracy was checked up to 300°C usinga Hewlett Packard Model 2801 quartz thermometer and an oil bath. The difference inrecorded temperatures did not exceed the manufacturer’s specifications. The tank bulkliquid temperature was obtained as a 1 mV/°C analog output from the tank heater controller;this was amplified by an operating amplifier and the output fed to the DAS-8. Tank heatertemperature errors were ±0.5°C. Similar error limits applied to the TAH alarm sensors butthese signals were isolated from the datalogging system and not recorded.783. Experimental Methods and MaterialsThe orifice plate and tube section pressure drops were obtained as analog outputsfrom the DP-350 pressure sensors. These signals were filtered through operationalamplifiers to remove background noise then connected to the DAS-8. The absolute pressuretransducer outputs ranged from 0-50 mV so were amplified and filtered before connectionto the DAS-8; the DAS-8 is sensitive to noise when input signals are less than 0.5 V. Allamplified and filtered signals were calibrated within the LabTech software to account foroffsets caused by signal processing.The current through the heated section was measured by adapting the current meterto give an analog output as well as a digital reading. The meter LED control signal wasrectified, amplified and filtered then fed to the DAS-8. The PC and meter readings werethen both calibrated using the Amprobe current meter. The voltage across the heated sectionwas measured using a 1:1 isolating transformer to isolate the DAS-8 from the heatingcurrent; the voltage from the transformer was then rectified and ratioed using two resistorsin series (10 Q, 20 Q). The DAS-8 output was calibrated using the Novatron panel meterafter loading effects had been eliminated. The signal processing arrangements are shownschematically in Figures A.1.3 and A.1.4.All system data were logged and displayed on the PC screen so that the systemcould be easily monitored. The surface thermocouples were subject to interference from theheating current magnetic field, which caused waves in the temperature readouts ofamplitude ± 4°C at maximum power. An oscilloscope showed that the noise had afrequency of 50 Hz. This noise was filtered out by sampling every six seconds andcalculating the moving average over the previous minute. The value obtained each minutewas then logged against the midpoint of the averaging period. This procedure reducedthermocouple fluctuations to ±1.5°C. A sampling period of one minute was consideredreasonable during experiments lasting five hours or longer. All process variables weresmoothed in a similar fashion. Data collection was often interrupted to save the data in casea power failure occurred.793. Experimental Methods and MaterialsThe heat flux in the heated section was assumed to be uniform and calculated usingEquation [3.241. The heat transferred to the liquid (Qe vi) was compared to the enthalpygain of the fluid (Qi);Q (Watts) = W Cp (Tb,out - Tb,j0) [3.25]where Cp was evaluated at the mean bulk temperature, Tb (Tb,ot + Tb,jfl)/2. The heattransfer coefficient h(x) measured by a thermocouple at axial location x m from the firstbusbar was calculated as:h(x) (W/m2.K) = q![T,(x) - Tb(x)] [3.261The increase in bulk temperature along the heated section was assumed to be linear underconditions of uniform heat flux and small increases in bulk fluid temperature;Tb(x) = Tb(x=O) + [Tb(X=L) - Tb(x=O)] xIL [3.27]where L is the length of the heated section. The temperature on the inside surface of thetube was assumed to be given by the analytical solution of the steady state heat conductionequation for a long hollow cylinder with uniform internal heat generation and an adiabaticouter wall;(x) = T,0(x) + Qe i. - r in (s-)27lL?.met 2 r - r2 r1 [3.281This correction ranged from 1 to 2°C and was around 1% of the temperature differencebetween the wall and the fluid. The value of Amej used was calculated using T,(x).Watkinson (1968) used Qirather than Qe in Equation [3.28] as insulation losses amountedto 3% Qe. Insulation losses calculated using Qi were less than 2% Qe and were subject toerrors in flow rate calculation, temperature measurement and heat capacity estimation.Insulation losses were also estimated by considering free convection from the insulationsurface to the ambient atmosphere.q11 = Uconv - Tanib) [3.29]where T5 and Tomb were the insulation surface and ambient temperatures respectively.The free convection heat transfer coefficient to the quiescent air was estimated at 10803. Exverirnental Methods and MaterialsW/m2.K. This value was a conservative estimate in order to account for radiative heattransfer losses. The values of Q (calculated using an insulation diameter of 100 mm)obtained were smaller than those given by Qe - Q. This correction was included in thecalculations for completeness and Qe- Q was thus used in Equation 3.28.Theovera1l heat transfer coefficient was calculated by a simple integration of theheat transfer coefficients along the heated section after the thermal entry length. The surfaceentry length was judged to have ended when surface temperatures increased in step with thebulk fluid temperature. Any thermocouples registering unusual temperatures were alsoomitted from the calculation. Watkinson (1968) calculated the overall heat transfercoefficient, U, (based on the inside of the tube) by fitting a quadratic through 15 data pointsand calculating the integralU=[3.30]The method used here was to integrate the following integral using the trapezoid rule;=fXL(Qe - dxL J 2trL(T,(x) - Tb(X)) [3.31]This approach was deemed more suitable for the fewer data points with non-uniformspacing. The two methods were compared in a number of the heat transfer tests andshowed good agreement. The simpler method was maintained as it was easier toincorporate into a spreadsheet and also generated the local heat transfer coefficients whichwere required to study local fouling behaviour. The fouling resistance was then calculatedusing Equation [1.3].813. Experimental Methods and Materials3.4.4 TFU Heat TransferThe performance of the TFU as a heat exchanger was studied using pure Paraflex atatmospheric pressure and a bulk liquid temperature of 100°C. A test section length of 770mm was used in these trials. The Paraflex flow rate, cooling water flow and heating powersetting were set at the required conditions and the system left to reach thermal equilibriumfor 15 minutes before data collection was started. The data were then processed asdescribed in Section 3.4.3.The insulation heat losses were studied in the absence of heating across thecomplete range of flow rates. No discernible drop in fluid temperature was recorded acrossthe tube in these tests and surface temperatures ranged from 1-5°C below the fluidtemperature. Heat transfer tests were performed with Tb1k = 100°C, flow rates from 1.0 -4.5 mIs and the variac set at 80% and 90% capacity (185±5 kW/m2 and 228±3 kW/m2),respectively. These tests also provided operating characteristics for future foulingexperiments. Figure 3.11 shows the correlation between the experimental data and Nupredicted by the Gnielinski correlation (Equation [3.21]) using Gnielinski’s expression forthe Moody friction factor in smooth tubes;f = (0.79 In (Re) - 1.64)2 [3.32]The Figure also shows the clean Nu data from the fouling experiments described in Section7. The agreement between the sets of data and with the correlation was good and thedeviation lay within the estimates of experimental error for the apparatus (5 % at Re1 =7760). The heat transfer coefficients from runs at the same Re1 increased slightly with heatflux, as reported in Section 3.2.1.Figure 3.12 shows the axial variation in the local heat transfer coefficient at fourvalues of Re17. The local heat transfer coefficient decreases from a high inlet value andincreases towards the exit of the tube as the bulk temperature rises. The thermal entrylength is significant at lower values of Re1, but almost insignificant at Re1> 10 000, in821.301 . Experimental Methods and MaterialsRe = 5454 • Re = 7703 I Re = 10805 • Re 14127Heat Flux = 183 kW/m2;Paraflex at Tbulk,jn = 100°C; Re calculated at inlet83Comparison of Predicted TFU Nusselt Number with Experimental DataFigure 3.11300 -2502001: 150100500-0Figure 3.12.. . . .• Fouling Runso Paraflex TestsI . . . . I .50 100 150 200 250 300Nu(Gnielinski)Thermal Entry Length Effects in TFU Heat TransferAA Aa - —...— .- ......I A0 10 20 30 40 50 60xID70 803.Experiinental Methods and Materialsagreement with the literature. Various correlations were tested against these data but mostfailed in the first few tube diameters where the heat transfer coefficient approaches infinitevalues. The profiles could be generated by a numerical solution of the governing enthalpyand momentum equations but this was beyond the scope of the current study. A furtherunknown is the heat loss by conduction through the copper terminals and cables. The heatloss to the terminals is thought to eliminate conduction along the tube as the busbars presenta surface area 40 times greater than the stainless steel tube cross section.The pressure drop across the tube section, AP1 in the insulation and heat transfertests was compared with the pressure drop calculated using Equation [3.33] and anestimated tube thickness of 12 thou. TFU conditions were Tblk,t 100°C, Pair = 101 and377 kPa, Re 5000 - 16000. End effects in the fittings (Ke, K) were estimated usingvelocity losses given by Kay and Nedderman (1985) and were small relative to the tubelength contribution.= + Ke + 4f12 2r1) [3.331where the p and Urn are the average fluid density and flow velocity along the test section, Lis the separation between the pressure taps, and f is the Fanning friction factor. Thecalculated values of the friction factor from the heat transfer and fouling runs were plottedagainst literature correlations in Figure 3.13. Blasius’ friction factor correlation applies tosmooth tubes at Re > 4000;f = 0.079 Re025 [3.34]The figure shows that the experimental values of f tend usually to be larger than thecorrelations’ predictions. The discrepancy was considered to be within reasonableexperimental error given the uncertainties in tube thicknesses, variations between tubes,tube roughness and fluid properties. The agreement in pressure drop and heat transfer datawith existing correlations established the reliability of the TFU as an experimental heattransfer section.843. Experimental Methods and MaterialsFigure 3.13 Comparison of Experimental Fanning Friction Factor Data with theCorrelations of Blasius and Gnielinski0.01 10.0100.0090.0080.0070.0060.005 -0.005 0.006 0.007 0.008 0.009 0.010 0.01111411.40 Gniehnski• e18 -• Blasius.•• 9,.3 -•o•oCalculated Friction Factor853. Experimental Methods and Materials3.4.5 TFU Operating ProceduresThe TFU operating procedures were similar to the PFRU and benefittedsignificantly from the experience gained from this apparatus. The experimental methodsdiffered primarily in the handling of the removable tube test sections.A test section tube was first cleaned with acetone and hexane using tubing .brushes,then dried and weighed. The tube surfaces to be in contact with the copper terminals andthe terminals themselves were thoroughly cleaned with contact cleaning solvent and left todry. The tube was then installed and the unit pressurised to the operating pressure with airin order to detect any leaks. The tank was then charged with solvent and the solventcirculated at the operating pressure to check for leaks again. If no leaks were detected, theterminal and thermocouple locations were marked using a permanent marker before theclamps were tightened in place and the test section insulated. The data collection systemwas then activated and the solvent circulated at atmospheric pressure. The thermocouplereadings were monitored at a variac setting of 50% in order to check that they had beenproperly installed. Any other adjustments were performed at this stage.To start a run, the system was pressurised and the solvent circulated while the tankheater brought the liquid to the required temperature. Once thermal equilibrium was reachedat the given flow rate, data were collected for ten minutes to provide insulation performanceinformation. The variac was then turned to the required setting and the system left toreestablish thermal equilibrium. Adjustments in variac setting were often necessary toachieve the required surface temperatures as the change in viscosity on adding indene toParaflex caused the surface temperature to drop. Air bubbles were cleared from all coolerand pressure lines at this time. Pure solvent heat transfer data was collected for 5-10minutes and the system was then depressurised. Indene and any additives were addedwhilst running via the tank filling port. The system was pressurised, flow conditionschecked and data collection started before the first sample was taken to start the experiment.863. Experimental Methods and MaterialsExperimental procedures (sampling and analysis) during a run were similar to thosein the PFRU except that the ëooling water flow had to be monitored once gum formed andcaused fouling in the primary cooler. Once this became significant the auxiliary coolerswere used to maIntaIn the desired bulk fluid temperature.Experiments were run for 8 hours unless heavy fouling required an earlier stop.Run #502 was curtailed by a power failure. To terminate an experiment, the data loggerwas stopped before simultaneously reducing the heated section power and depressurisingthe system. The pump was then stopped and all process liquid drained from the system. Itwas important to maintain the gas flow at this stage in order to avoid gum blockage of thegas feed line. The test section was carefully removed and replaced by a dummy sectIon forcleaning. The cleaning procedures described in Section 3.2 were used with scaled up rinsequantities. The test section was rinsed with hexane and left to dry.The terminal and thermocouple locations were recorded from the drIed test sectionbefore the tube was cut up into labelled 50.8 mm sections using a pair of steel snips. Thesnips dislodged a minimum of material without generating any steel dust. Significant losseswere noticeable with heavy, brittle deposits. The cut sections were rinsed with hexane,dried in a vacuum oven for an hour and weighed. The sections could then be analysed asdesired. After analysis, any remaining foulant was scraped off gently and the metal cleanedusing acetone, chromic acid solution and distilled water. The dried sections were thenwashed with hexane, dried under vacuum and weighed again. The mass of foulant wasexpressed as mg/cm2 by assuming a tubing thickness of 10 thou and calculating a meansection length from its mass and the mass of the clean tube. The thin metal tubing was idealfor optical and electron microscope analysis.873. Experimental Methods and Materials3.5 Gum Ageing OvenThe chemical literature describes the homolysis of peroxides in solution but there isscant coverage of the degradation of peroxide gums following deposition on a solidsurface. A series of gum ageing experiments was performed to investigate the fate andkinetics of the pyrolytic degradation of gums generated during indene autoxidation. Theexperiments were performed in an inert (nitrogen) atmosphere under isothermal conditionsand were intended to trace physical and chemical changes in the gum during the ageingpIu.Figure 3.14 is a schematic of the apparatus, which consists of an adapted Varianenes 14(31) UC with a top opening tic!. Ihe oven is purged with nitrogen and all otherconnections are blanked off. The temperature distribution inside the oven was checked witha manouvreable K type thermocouple and found to be close to isothermal. The unit waslocated inside a fume hood to remove any noxious fumes generated.A glass weighing crucible was cleaned and dried at 350°C before being weighed.Approximately 200 mg. of finely ground sample was weighed into the crucible and themass recorded. The crucible was quickly placed in the centre of the oven and heated for therequired period then removed and left to cool before weighing. The product was easilyremoved for further analysis and testing. Blank samples showed zero weight loss andrepeat samples showed good reproducibility.The gum ageing experiments were also duplicated using Thermal GravimetricAnalysis (TGA). Approximately 10 mg of finely ground gum was equilibrated at 30°C thenheated at 100°C/mm to the ageing temperature in a nitrogen atmosphere. The experimentswere performed using a Perkin-Elmer TGS-2 device interfaced to a Perkin Elmer TAGSdata station.88Figure 3 14 Schematic Diagram of Gum Ageing Oven Apparanitrogen SupplyRotajner893. Experimental Methods and Materials3.6 Chemical AnalysisThe difficulties in analysis of autoxidation mixtures have been discussed in detail byLink and Formo (1962). Analysis of the model solutions was significantly simpler thanindustrial fluids but it cannot be described as a complete monitor of the reaction system.The products of indene autoxidation range from hydroperoxides and polyperoxides toepoxides, carbonyls and rearrangement products which are not available for use asreference compounds or standards. The chemical analyses developed for solution samplesdescribed here represent a set of reasonably accurate, simple methods which monitor thesignificant steps involved in the generation of fouling precursors. Methods such as HighPerformance Liquid Chromatography could ostensibly have been used to separateindividual polyperoxide fractions but the costs were deemed unreasonable. The solidanalysis methods represent a similarly pragmatic approach to the study of a complexorganic material.The solution analyses were designed around an autoxidation scheme adapted fromthat of Norton and Drayer (1968):A ---> B ---> C ---> D [3.35]indene hydroperoxide polyperoxide insoluble peroxidesMethod GLC Peroxide Number Gum Assay -This scheme is not based on the chemical mechanisms described in Section 2 andrepresents a simplification of the reaction system to a form which can be described by theavailable analytical methods. No method was found to determine the amount of insolubleperoxides formed during an experiment. The amount of insoluble gum remaining after anSCR experiment depended on the number and timing of samples so that a simple kineticanalysis is not feasible.903. Experimental Methods and Materials3.6.1 Hydroperoxide Analysis - Peroxide Number (POx)A titration method for hydroperoxide determination was chosen over precipitationfrom base or polarography due to the simplicity and reliability of the iodine titration forautoxidative systems. The method employed is based on ASTM E298. Hydroperoxide isreduced by iodide, yielding iodine which is then titrated against standardised sodiumthiosuiphate. The redox reactions are;R-OOH + 21- + 2H ----> R-OH + 12 + H20 [3.36]12 + 2NaSO3 ---> 21- + Na2S4O6 + 2Na [3.37]The Peroxide Number is expressed as the number of milliequivalents of oxygen per litre(meq!L) solution and it can be seen that one mol of hydroperoxide reacts with 2 mols ofthiosuiphate. The titration is specific to the hydroperoxide end group and thus the peroxidenumber includes contributions from both monomeric and polymeric hydroperoxides.Polyperoxide linkages could be included in the Peroxide Number if stronger oxidants suchas boiling hydriodic acid (Russell, 1956a) were used.A volume of sample corresponding of 1-4 milliequivalents of oxygen was dissolvedin 20 mL glacial acetic acid in a stoppered Erylenmeyer flask which had been deaeratedusing a carbon dioxide purge. The exclusion of air prevents air oxidation of iodide insolution. The mixture was deaerated, 5 mL saturated aqueous sodium iodide solutionadded, deaerated again and then immersed in a water bath at 60°C for one hour. The flaskwas then cooled, 50 mL deaerated distilled water added and titrated against 0. 1M sodiumthiosulphate. 1-2 mL starch solution (5 gIL) was used as indicator, which lost its dark bluecolour at the end point. The sodium thiosulphate was standardised against potassiumpermanganate and stored under carbon dioxide. The Peroxide Number was calculated asPOx (meqlL) = 1000 x 0.1 x (volume Na2S2O3)(volume sample) [3.38]913. Experimental Methods and MaterialsSamples were run as duplicates with a control to monitor reagent quality. The method wastested using t-butyl hydroperoxide and found to be satisfactory. The analysis did not workin the DCP experiments as the DCP seemed to react directly with the iodide or iodinegenerated. Hydroperoxides are acidic and alkali precipitation was tried using aqueoussodium hydroxide on a sample from an SCR experiment. A brown solid was obtained butthe technique was not developed further.3.6.2 Polyperoxide Analysis (Gum in Solution Assay)The early PFRU fouling studies showed that the Peroxide Number was not a directmeasure of foulant precursor concentration. The work of Russell (1956a), Mayo and Lam(1986) and the fuel stability literature emphasised the importance of gum species in solutionso a gum assay was developed, based on RusseWs separation of indene polyperoxidesusing hexane.10 mL of solution was pipetted into a 50 mL Erylenmeyer flask, stoppered andstored frozen to halt the reaction and precipitate gum from the solvent. 20 mL hexane(BDH, Fisher) was added to the flask before analysis, shaken and cooled again. 20 mLwas found to be sufficient excess for kerosene, Paraflex and n-octane, but 30 mL wasneeded for toluene and trichlorobenzene. A Whatman #1 filter paper was dried undervacuum and weighed. The filter paper was placed in a Buchner funnel, rinsed with hexaneand the sample solution filtered through it. lOmL cold hexane was used to rinse the flaskand paper before both were dried under vacuum for 30-60 minutes. The flask and filterpaper were weighed, the paper discarded and the flask cleaned with acetone, dried andweighed. The gum content was calculated by difference and expressed as g/L. Initial trialsshowed reasonable accuracy (±0.2 g/L) and repeatability but later experiments indicatedthat duplicate sampling was preferrable.923. Experimental Methods and MaterialsThe gum obtained varied with the solvent and the stage of the reaction. The gumrecovered from the aromatic solvents was a darker orange colour than the yellow or ambergum produced by Paraflex and n-octane. The insoluble gum found in the SCR unit after aParaflex or kerosene experiment was always darker in colour, confirming this hypothesis.Figure 3.15 is a photograph of the filter papers from the assay after two Paraflexexperiments and shows that the gum changes form during the reaction. Section 4.1describes how a gum solubility limit exists in aliphatic solvents; before this concentration isreached, the gum is suspended in solution and is collected on the filter paper. The filterpaper from subsequent samples are clean as the gum, which coalesces readily on freezing,sticks to the flask as a translucent material. The assay could not provide information aboutgum agglomerate sizes.The mean molecular mass of the gum produced during selected indene in keroseneexperiments was estimated using the benzene melting point method. The gum precipitatedby hexane was dissolved in 2 mL pure benzene (Fisher, 99.99mo1%) and frozen in aboiling point tube fitted with a stirrer and a ‘J’ type thermocouple. This approach was usedbecause the sticky gum is difficult to weigh out accurately. The freezing point was notedand the molecular mass calculated via:M = [ATe pt m(gum)]/[z.Tf.,0m(benzene)] [3.39]where is the molal freezing point depression constant (5.1 K for benzene) andthe freezing point depression (K). The estimate is subject to errors in mass andtemperature measurement, gum hexane content and temperature accuracy, but wasconsidered a helpful guide in the analysis programme. Estimates of M for the gumproduced from indene in kerosene varied from 350 to 450, corresponding to 2-4 indenepolyperoxide units of mass (116.1 + 32 = 148.1).933. Experimental Methocts’ and MaterialsFigure 3.15 Photograph of Gum Assay Filters showing Change in Gum Appearance943. Experimental Methods and Materials3.6.3 Indene Concentration - Gas Liquid ChromatographyGas-Liquid Chromatography (GLC or GC) is the standard method for thequantitative analysis of organic liquid solutions. The Bromine Number Test for unsaturates(ASTM Dii 59-57T) and ultraviolet spectroscopy were also evaluated but neither of thesetechniques proved to be as specific or as accurate as GC methods. The disadvantages ofGC methods are the relativley long run times associated with the solvents used and thedestruction of any peroxides in the injector, preventing their detection.2 mL samples of solution were stored frozen until analysis, when they were mixedwith 5 mL of GC solvent. The GC solution used varied with each solvent but was based ondichloromethane, CH21 which dissolved all samples and is quickly eluted from thecolumn. Paraflex, n-octane, toluene and trichlorobenzene used a 200:1 v/v solution ofMeC12and an internal standard, n-decane. The majority of the samples were anlaysed usingone of two machines, using identical columns and temperature programming. A VarianVista 6000 GC fitted with a thermal conductivity detector and manual injector was usedinitially, with helium as the gas phase. A Varian CDS 401 data station performed allchromatogram acquisition and integration. Subsequent analyses were performed on aVarian 3400 Star GC fitted with a liquid autosampler, flame ionisation (FID) andphotionisation (PID) detectors. Nitrogen was used as the carrier gas. Separation employeda 1.83 m x 1/8 inch stainless steel packed column with 80/100 Supelcoport as the stationaryphase and 10 wt% OV-lOl as the liquid phase. These columns were repacked and reconditionned as required. Some early analyses were also performed using a HewlettPackard Series 2 GC fitted with a Mass Spectrometer detector which allowed peakidentification. These analyses used a 12 m HP-i capillary column with helium as the gasphase.The pure solvents eluted as single peaks while Paraflex and kerosene eluted as aseries of peaks. Figure 3.16 is a chromatogram generated by the Varian 3400 unit from a953. Experimental Methods and MaterialsFigure 3.16 Gas Chromatogram of Solution of Indene in ParaflexIf)0GoIf)0iI I IC)C.)Column Temperature C)250°C‘40°C100°C40I I I II II II I I II I II I II I I I TimeI I I0 6 95 29.5 32(minutes)963. Experimental Methods and Materials0.41M solution of indene in Paraflex and shows the temperature program used. TheParaflex components elute after indene and the internal standard so that quantitative analysisis straightforward. The indene concentration is calculated using the internal standard as areference. The ratio of indene peak area to that of the n-decane is calculated and comparedwith a calibration curve prepared using standard solutions. The sample/solvent ratio waschosen to ensure a linear calibration curve.Indene analysis in kerosene solutions was complicated as indene elutes among thekerosene components. An added standard method was used with the Varian 6000 unit,where a known amount of indene is added to the sample via the GC solution to ‘boost’ theindene peak. The GC solution was a 200:1:1 v/v mixture of CH21,indene and octadec- 1-ene, a reference which elutes after the kerosene components. Peak height was used in thecalculations and an accuracy of ±5% was observed in the calibration. Added standardanalysis usually employs a series of samples with different amounts of added standard butthis approach was considered to be too time intensive in the absence of an autosampler. Adifferent approach was used later with the Varian 3400 system. This unit featured a PIDwith a 10 meV UV lamp, making the detector more sensitive to aromatics and alkenes thanalkanes. Figure 3.17 is a comparison of PID and FID chromatograms from a solution of 10wt% indene in kerosene; the detectors are arranged in series and the selectivity for indene isapparent. Internal standard analysis at 1:20 v/v dilution in CH21 with bromobenzene asreference yielded ±2% accuracy over the calibration range. The PID selectivity foraromatics and alkenes over alkanes was also used in the studies of antioxidation describedin Section 5.7. The antioxidant, BMP, eluted among the Paraflex peaks at lowconcentration (ppm). The BMP peak was visible in the PID chromarogram but highbackground noise prevented quantitative study and the qualitative results were deemedsufficient. Solvent extraction was not considered.973. Experimental Methods and MaterialsFigure 3.17 Gas Chromatograms of Indene in Kerosene from Rame lonisation andPhoto-lonisation Detectors in SeriesI I I I I I I I I I I I I I I I I IW cvc ci)() rC) $‘dI I I I I I I I I I I I IPDN00• a NNmnJ 0co10coN ClRDaD0‘0rco983. Experimental Methods and Materials3.6.4 Further Analysis - FTIR, SEMA variety of analytical methods was used to provide further information about thechemical and physical nature of liquid and solid samples.Fourier Transform Infra-Red Spectroscopy (FTIR) is used to identify whichchemical functional groups are present in gas, liquid or solid samples. Two microprocessorcontrolled devices were used; a Perkin Elmer Model 1710 device and a Midac FTIR with anMCT detector. Early solids analysis used mulls of the solid sample spread between twoKBr disks. Subsequent solids analysis used the KBr pellet technique. Approximately 5 mgsample was added to 100 mg dry KBr powder and mixed together, dried and ground up.The powder was then compressed in a pellet press into a solid, transparent 11 mm disc foranalysis. Liquid analysis was performed using a,0. 1 mm pathlength liquid cell fitted withKBr windows.Elemental analysis (C, H, Cl, 0 by difference) was performed by Mr. P. Borda ofUBC Chemistry and by commercial laboratories (Canadian Microanalytical, GuelphAnalytical Laboratories).TFU tube sections were examined using a Hitachi S-2300 Scanning ElectronMicroscope located in the UBC Department of Metals and Materials Engineering. Thisdevice was equipped with a Polaroid Camera and a Quartz PCI Image Capture System forphotographic and digital recording of images. A coupon approximately 8 mm x 8 mm wascut from the tube section and gold sputtered before mounting in the SEM. Gold sputteringwas necessary to improve the thermal conductivity of the sample as the 20 kV beam quicklydestroyed the unprotected surface at higher magnifications.994. Auto xidation of Model Solutions4. Autoxidation of Model Solutions of IndeneModel solutions of hexadecene or indene in four solvents (Paraflex, kerosene,tetralin and tiichlorobenzene) were tested in the PFRU system to establish which would begood candidates for fouling experiments. These preliminary fouling studies are described inSections 5.1-3. The chemical analysis results from these runs indicated that theautoxidation of model solutions of alkenes in the PFRU system differed from themechanisms and kinetic forms reported in the chemical literature. These initial studies alsoconfirmed that the bulk chemical reaction played a significant role in the fouling process.The autoxidation of indene in model solutions was therefore studied further in the stirredcell reactor in order to identify and understand the important reaction parameters. Indenewas studied as this was the alkene selected for further fouling studies. The work wasperformed in parallel with the fouling studies and is described first in order to introduce thereaction analysis used in the fouling runs.4.1 Solvent Effects in Indene AutoxidationThe choice of solvent is important in model solutions for fouling studies as thesolvent can affect both the reaction kinetics and the physical behaviour of the reactionproducts. The initial fouling studies in Sections 5.1-3 reported both of these effects and sothe autoxidation of indene was studied in the SCR using five different solvents; Paraflex,kerosene and n-octane, ‘inert’ alkane-based solvents; toluene, an aromatic liquid whichundergoes autoxidation less quickly than indene, and trichlorobenzene, a polar aromaticsolvent. The relative inertness of the aliphatic solvents was confirmed by controlexperiments in the SCR under autoxidation conditions (Tbjlk 100°C, 150 mL/min (n.t.p.)air at 377 kPa). Minimal hydroperoxide concentrations were detected after 48 hours.Toluene did undergo autoxidation under these conditions but the concentration of1004. Autoxidation of Model Solutionshydroperoxide was low (6.8 mmol/L) after 56 hours and no soluble gum was detected.GC analysis did not detect any benzaldehyde, the secondary oxidation product.Trichlorobenzene was thought to cause contamination in the fouling experiments and thiswas also found in the kinetic experiments in the SCR. A heavy black solid was formed aswell as an orange gum which corroded the steel noticeably. This solid contained 2.05 wt%chlorine and thus the contamination source was identified.The thermally initiated autoxidation of solutions of 0.41 molJL indene was studiedin the SCR under the conditions described above. The experiments were run until thePeroxide Number showed little change or sampling was hindered by the formation ofsticky red/orange gums. Figures 4.1 a-c show the Peroxide Number, soluble gum andindene concentration results in each solvent. The Peroxide Number climbs steadily and thendecreases, except in the case of Paraflex where it maintains a ‘plateau’ value. The chemicalinduction period, ‘r, corresponds to the start of the rapid increase in POx. The rapidincrease was often observed after the POx reached a value of 5 meq/L. More precisedetermination of r was limited by the maximum sampling rate (llhr) in the analyticalmethod. The kerosene data all show a much longer delay (larger t), which is attributed tothe presence of a commercial inhibitor in the solvent. The inhibitor is presumably exhaustedafter the induction period and the kerosene data behaves similarly to the other data after timeSolvent effects are most apparent in the soluble gum concentration data. Figure4. lb shows a pronounced difference in gum behaviour between aromatic and aliphaticsolvents. In aliphatic solvents the gum concentration increased linearly with time after theinduction period until reaching a maximum, referred to as the solubility limit, g , at time t*.The value of g* is higher for kerosene, which ‘3C n.m.r. analysis showed to contain someunsaturated components. The solubility limit is thus a function of solvent nature, whichcould be explored further by using mixtures of solvents. Comparison of Figures 4. lb and1014. Autoxidation of Model SolutionszI-‘1)Figure 4. lb Gum ConcentrationC4:1Figure 4. ic indene Concentration0.500.450.400.350.300. 10 20 30 40 50 60Time(hours)0 Paraflex • kerosene A n-octane o toluene trichlorobenzene ISolutions of 0.41 M indene at Tbulk = 100CC: 79.2 kPa oxygen saturation: thermal initiationFigure 4.laFigure 4.la-c Solvent Effects in Indene Autoxidation : Analysis ResultsHydroperoxide Concentration8070 060 0 050 0040 00A3020100 1...I...0 10 20 30 40 50 605040 00o030020 0100;k,EA AiOA ••• ...0000 , I.. I0 10 20 30 40 50 60AAAoc1000*00 A 0£O1j- - - -.an...., . . .1024. Autoxidation of Model Solutions4.lc shows that indene consumption continues after t. The reactor internals were coatedwith a sticky, red insoluble gum after the experiments using aliphatic solvents. Thisinsoluble gum was not found with the aromatic solvents, suggesting that all the oxidationproducts were soluble in the solvent and that g* is a physical and not a kinetic parameter.The ratio of gum produced to indene consumed should reflect the ratio of additionto abstraction. This ratio would be exact if the gum were a primary oxidation product; thekinetic models described in Section 4.6 show that the gum formation rate is a complexfunction of the rate of indene consumption, but the yield still provides a useful parameterfor comparing sets of analytical data. The yield was calculated as the mass of gumproduced in the early stage of autoxidation (time t to t*) divided by the maximum mass ofpolyperoxide that would be generated if all the indene consumed formed equimolarquantities of indene polyperoxide linkages, of mass 148.1, i.e.Yield k3b {gum (t*)} - {gum(t)}k3b + k3a ([indene (t*)j - [indene(t)]) x 148.1 [41]where [indene] is given in (molIL) and {gumj is in (gIL). This equation assumes 1:1stoichiometry between gum and indene, which is discussed further in Section 4.6. Table4.1 summarises the calculated yields and the other kinetic parameters obtained from theseexperiments. The yield in the aromatic solvents was high but was less than 100%,indicating that other oxidation products are formed by abstraction and decomposition.Russell reported a ‘yield’ of 0.8 in indene at 50°C and 97 kPa oxygen; this is higher thanthe values in Table 4.1 due to increased chain transfer to the solvent in the model solutions.The initial rates of gum formation (between rand t*) were used to compare the rates ofautoxidation in the different solvents. The values given in Table 4.1 are similar for toluene,Paraflex, and n-octane, and lower in kerosene and trichlorobenzene. The lower initial gumrate in trichlorobenzene is thought to be linked to the contamination problem as indene isbeing consumed at similar rates (shown by kND). All the kinetic parameters from thekerosene experiment were low and prompted the study of antioxidants discussed in Section1034. Autoxidation of Model SolutionsTable 4.1 Solvent Effects in the Autoxidation of Model Solutions of IndeneSolvent Initial POxNature Rate ConstantkM, POx(‘J[mmoIIL]/hr)± 0.020.34Initial Gum Indene rateRate constantkND(g/Lbr) (1/br)± 0.1 ±0.0051.5 0.096SolubilityLimitg*(g/L)± 0.5soluble0.41 M indene under thermal initiation : Tbulk 100°C : 79.2 kPa oxygen saturationSolvent GumYield(g/g)Insoluble GumAnalysistrichlorobenzene polar, 0.715 C 65.9 wt%aromatic H 4.2 wt%toluene aromatic 0.81 2.5 0.092 soluble 0.689 C7880199n-octane aliphatic 1.04 2.23 0.139 10.0 0.529 H.582,37kerosene mainly 0.19 1.1 0.077 14.0 0.694 C7970187aliphaticParaflex aliphatic 0.85 2.2 0.118 11.0 0.567 C9H8002,61044. Autoxidation of Model Solutions4.5. Other experiments in kerosene indicated that indene reacted at similar rates to those inParaflex, which confirmed that the autoxidation reaction is not influenced strongly by inertsolvents but that the gum behaviour is strongly dependent on the solvent nature.The indene concentration decreased in a non-linear fashion with time at similar ratesin all solvents except kerosene, which featured an extended induction period and lower rateof reaction. The data in the toluene experiment, however, showed a marked reduction inrate as the reaction proceeded. The inhibition corresponded to the appearance of a new GCpeak which eluted later (i.e. more polar) than the toluene peak. This peak had the sameretention time as benzaldehyde; this secondary oxidation product of toluene is known to beself-inhibiting in autoxidation and thus appears to be inhibiting toluene autoxidation in thiscase. This result and the fouling trials in tetralin (Section 5.1) confirm the complexity ofsolvent interractions in autoxidation reaction fouling.The soluble gum formed was similar to that observed in the fouling experiments.FTIR analysis of the light yellow coloured gum formed before t showed both hydroxyland carbonyl activity. Carbonyl groups could be formed by peroxide decomposition orcondensation. Gum formed later in the experiments was orange-red in colour and containedhigher carbonyl levels. The gum was found to have a mean molecular mass of 300-400,corresponding to 2-4 peroxide linkages. More precise measurement of the gum mass ishindered by its agglomerating properties, trapped solvent and difficulty in handling thisvery tactile substance. The gums were analysed for C, H and 0 (by difference) and Table4.1 shows that the gums were similar in composition to indene polyperoxide, the expectedprimary oxidation product.1054. Autoxidation of Model Solutions42 Autoxidation KineticsThe chemical analysis results from the autoxidation experiments were fitted to a setof kinetic models and the parameters obtained were used to compare the reaction behaviourbetween individual runs. Simple models were used as no single model was found whichcould fit all the observed features of the data. The data from Section 4.1 were used here todescribe the selection of the diagnostic parameters.Figure 4.lc shows that the indene concentration decreases at similar rates in theuninhibited solvents, confirming that the reaction rate is not strongly affected by the solventnature. Van Sickle et al. (1965b) reported that the rate constant for cyclopentene oxidationincreased with the solvent polarity, but this is not reflected in the trichlorobenzene data. Theindene and hydroperoxide data were compared with the kinetic model of Russell (1956b),who reported that the thermal initiation of indene was first order in both oxygen andindene. Combining Equations [2.5] and [2.141 givesd[02]/dt = -kpR[RH]’[0]°• [4.2]where kpR is a lumped propagation rate constant. Equating the rates of oxygen and indeneconsumption, assuming constant dissolved oxygen concentration and integrating gives[pJ{]-O.5 = [j?J.{j0-O.5 + 0.5 kpR [02] t [4.3]Figure 4.2 shows the data from the Paraflex test plotted using Equation [4.3] and the firstorder model used by Norton and Drayer to describe the autoxidation of hexadecene;d[RHI/dt =- k [RH] [4.4]where k is a lumped first order rate constant. The Norton and Drayer approach isessentially a parameter fit to explain the observed experimental results. Neither equationfitted the data across the entire range but [4.3] works well until 7 hours, after which [4.4]gave a better fit. The transition corresponds to the point where the Peroxide Number stopsincreasing rapidly. This point could be interpreted as the point where hydroperoxide106Norton and Drayer (Equation 4.4) n = 1.01074. Autoxidation of Model SolutionsFigure 4.2 Kinetic Plots of Indene Concentration Data: Comparison of Kinetic Models0 20-1-1-3 104-6 00 5 10 15 20 25 30Time(hours)Thermal initiation of 0.41 mol/L indene in Paraflex, TbUlk 100°C, 79.2 kPa oxygen saturationRussell (Equation 4.3) n = 1.54. Autoxidation of Model Solutionsdecomposition overtakes the thermal initiation step. The values of kND in Table 4.1 showthat indene is being consumed at similar rates in all solvents except kerosene, confirmingthe observations in Section 4.1.These two models do not fit the data well and neither provides a satisfactorydescription of hydroperoxide behaviour. Russell did not measure hydroperoxideconcentrations, while Norton and Drayer’s model does not predict the variations in POxobserved with the solvents in these experiments.4.2.1 Mass Transfer Effects in Autoxidation KineticsRussell’s experiments were performed on a small scale and featured agitation suchthat “rates were not complicated by diffusion controlled processes”. Mass transferlimitations in reaction rate are more likely to arise at higher temperatures and on a largerscale. The oxygen for reaction with indene must be absorbed from the vapour phase and ifthe liquid reaction rate exceeds the mass transfer capacity of the reactor, the reaction ratewill show a mass transfer limitation. Oxygen is a sparingly soluble gas in petroleum oilsand mass transfer effects were consequently shown to control the reaction kinetics in theSCR and fouling experiments. A more complete treatment of mass transfer with chemicalreaction can be found in the text by Danckwerts (1970).The reaction between oxygen and indene can be represented as the reaction betweensparingly soluble reactant A and involatile reactant B:A + B Products 14.5]Assuming that no other reactions occur,- dNA/dt = - dNB/dt [4.6]where N1 is the total number of mols of component j in solution. The flux of A intosolution can be described using mass transfer coefficients;1084. Autoxidation of Model SolutionsJAAMIMA = AM k1 (CA,int CA,bulk) = AM kG(PA,g- PA,int) [4.71where AM is the surface area available for mass transfer, k1 and k0 the liquid and gas phasemass transfer coefficients, A the partial pressure of A and CA the concentration of A insolution. The subscripts g, i nt and bulk refer to the bulk gas, vapour-liquid interface andbulk solution respectively. Figure A. 1.5 is included as a schematic representation ofabsorption of A followed by chemical reaction with B. Equating the reaction in solution tothe mass transfer flux,- dNA/dt = - dNB/dt = - V dC/dt = AM k1 (CA,jflt- CA,b1k) [4.81where V is the volume of solution. Equation [4.8] holds when mass transport is byphysical (i.e. diffusional) processes alone and the chemical reaction is termed slow withrespect to mass transfer. The parameter.ki is determined by the physical properties andmixing conditions of the liquid phase alone and a maximum diffusive flux occurs whenreaction in the bulk reduces CA,bUlk to zero; the maximum flux condition can then be written- V dCB/dt = AM k1 CA,jt [4.9]The gas phase resistance to mass transfer is usually minimal in the case of sparingly solublecomponents as in this case and is ignored. Equation [4.9] was used to estimate the valuesof k1 from the data in Section 4.1 as CB was monitored and was maintained constantduring the experiments. A maximum flux argument was used as the dissolved oxygenconcentrations could not be measured in the reactors and the physical mass transfercoefficientk1was not known. The surface area Ap1 was estimated as the quiescent vapour-liquid interface area in the SCR (0.01327 m2, ignoring bubbles); a mean value of V wasused in the calculation of V dCBIdt. Table 4.2 shows the initial consumption rate of indenein the solvents above and the calculated value of k1. Typical values of k1 in thisconfiguration are O(10 mis) but the values in Table 4.2 are significantly higher. Astarita(1967) reported that the mass transfer coefficient in a well dispersed vapour/liquid systemcan be expressed as a function of the diffusivity DA and a diffusional contact time, tD, byk1 = (DjtD)°5 [4.10]1094. Autoxidation of Model SolutionsTable 4.2 Mass Transfer Effects in Batch Autoxidation of Indene-dNB/dt OxygenConcentrationx106 ASTMD2779CA,jntEstimated DiffusiontimetDx105 lW-Cl(mis) (s)DiffusiontimetD(s)t - estimated as the density of trichlorobenzene> ASTM rangeW-C Wilke & Chang; H-M Hayduk & Minhas: see Reid et al. (1987)0.41 M indene under thermal initiation : Tbulk 100°C : 79.2 kPa oxygen saturationSolvent(mol/s) (mourn3)tnchlorobenzene 7.32 5.56t 9.615 0.343 0.915toluene 11.96 4.62 10.89 0.206 0.375n-octane 14.35 12.10 8.66 1.164 1.84kerosene 4.18 7.37 4.14 5.596 5.578Paraflex 10.54 5.42 14.19 0.153 0.2891104. Autoxidation of Model SolutionsTable 4.2 includes the estimated values of tD calculated using the diffusivity correlations ofWilke and Chang (W-C) and Hayduk and Minhas (H-M) (see Reid et at., 1987). Astaritafound that 0.004 < tj < 0.040 s for physical mass transfer in well agitated systems; thecalculated values of tD all lie above this range and indicate that autoxidation is limited bymass transfer in these experiments. The main source of error is the estimated contactsurface area, AM, but this is unlikely to be an order of magnitude too small. There was nodirect method available for measuring bubble sizes (and thus surface area) at thetemperature and pressure conditions involved in the reactors. Subsequent experiments athigher gas flow rates (and presumably larger bubble surface areas) did not show anysignificant increase in the indene reaction rate. All the SCR and fouling experimentsinspected using this analysis gave a similar result.The mass transfer effect is inferred from the observed reaction rates as it cannot beconfirmed without a reliable method of measuring the dissolved oxygen concentration inthe bulk liquid. Dissolved oxygen probes used in aqueous media rely on electrochemicalprinciples and are unsuitable for use in organic solutions. A simple method was notavailable during the course of this work.The observed chemical reaction rate was markedly larger than the estimatedmaximum rate due to physical absorption alone. This result infers that the autoxidationreaction is eitherfast or instantaneous with respect to mass transfer. The reaction couldbelong to the intermediate regime (i.e. reaction rate similar to the mass transfer rate) if theestimate of AM was too conservative, but this would give very complex overall kinetics andwas not considered any further.The instantaneous regime corresponds to diffusional control of the reaction rate,where the overall reaction rate is independent of the chemical reaction rate. Laterexperiments using chemical initiators showed that the reaction rate could be increased bychemical means, which ruled out the instantaneous regime. The autoxidation reactions inthis work were thus inferred to belong to the fast regime, where most of the chemical1114. Autoxidation of Model Solutionsreaction occurs in the diffusive film next to the vapour-liquid interface and the bulkconcentration of oxygen (and thus the bulk reaction rate) is negligible.4.2.2 Kinetics of Autoxidation with Mass TransferMass transfer effects were shown to be significant in the previous section usingrelationships from the surface renewal theory of mass transfer with chemical reaction. Thefilm model was used in the following analysis of kinetics in the SCR in order to obtainsimpler results which could be compared with the experimental data. The film model isapplied to the reaction in the idealised liquid diffusion film and the overall result written interms of concentrations in a well mixed batch reactor.The absorption of gaseous reactant A (here, oxygen) into a liquid is governed byFicks law for dilute solutions-DAd2CAJdx = RA [4.11]where RA is the volumetric reaction rate of A. A similar equation can be written for reactantB (here, indene). The film model of Lewis and Whitman describes the mass transferresistance in terms of an effective diffusion film of thickness ÔD. The boundary conditionsfor mass transfer with chemical reaction, where all reaction occurs in the diffusion film, arethendCA/dx = 0 at x = ÔD [4.121CA = CA,j11t at x =0 [4.13]The solution of [4.11] depends on the form of RA. Two cases will be considered here; thekinetic scheme proposed by Russell and a diffusionlreaction model first proposed by Vande Vusse (1961) and described in further detail by Kay and Nedderman (1985).For n-th order destruction of diffusing species A, RA = - kCA”, giving [4.111 as1124. Autoxidation of Model SolutionsDAd2CA/dx = k1CA [4.14]Kay and Nedderman show that the solution of [4.14] is facilitated by multiplying across bydCA/dx and integrating;DA(d2CAIdx)(dCA/dx) = kflCA (dCAIdx)DA/2 (dCA/dx) = [k11(n+1)J CA’ + B [4.15]Applying boundary condition [4.13] gives13 — - “A,int2 \ dx x = 0 n + 1 [4.161The solution of Equation [4.15] thus involves the evaluation of the following integral(02k C’’ 213 -0.5I dx = I dC 11 A + —(n+1)DA DAJo [4.17]Numerical integration of [4.17] involves an estimate ofJ3, which includes the (unknown)concentration gradient at the gas-liquid interface.Kay and Nedderman described a simplified result for fast reactions, where thereaction is effectively complete within the diffusion film and both CA and dCAIdx are zeroat x ÔD. Writing this as one boundary conditiondCA/dx = 0 where CA = 0 [4.18]givesj3 = 0. [4.16] thus gives the concentration gradient (and hence the absorption rate) atthe gas-liquid interfacer1 ,-‘n-i-l ]0.5(dCA— V.’l’fl ‘—A,intdx= 0— L (n+1) DA i [4.19]This result was used to formulate an expression for mass transfer with Russell’s kineticscheme.Russell’s kinetic scheme is expressed by Equation [4.2], giving [4.11] as- DAd2CA/dx = - kpyCB15CA°5 [4.20]1134. Autoxidation of Model SolutionsThis can be simplified by assuming that the indene concentration in the diffusive film iseffectively constant, which is true when indene is in excess. The current experimentsinvolve oxygen solubilities of 0(5 mourn3)and initial indene concentrations of 0(410mourn3) so this assumption is justified. Equation [4.20) is thus half order in oxygenconcentration, giving [4.191 as(dCA/dx)X0 = - (4 kpyCB’5/3Dj° CA,1t°’75 [4.211The absorption rate of A and hence the reaction rate of B is then given by-- D (dC - - dN8 = - fv dCMA “ dx I, = — AN4 dt LAMJ dt [4.22]Applying [4.21] gives a kinetic model for the batch reactiondCB/dt = -(AM/V)(4kDA/3)°5CA,lt°75B° [4.23]Assuming that constant conditions apply to all quantities except CB, integrating yields theresult for Russell’s kinetics with mass transferC°25 (t) = CB°•25 (0) - (AM/V) (4 kpDA/3)°5CA,t°75 (t14) [4.24]or CB°25 (t) = CB°25 (0) - kpj.,j t [4.25)where kRAI is a lumped rate constant. The dependence on oxygen and indene concentrationsis thus significantly different from Equation [4.2] and the apparent activation energy,EactRM, will then be a composite activation energy;Eact,RM = 0.5 Ediffusion + 0.5 ( Eprop + (Ei111 - Eterm)/2) [4.26]This treatment assumes that the ratio AM/V remains constant during the experiment. Thesurface area is unknown but liquid sampling reduces the volume significantly during a longSCR experiment because of the large aliquots needed for POx analysis. This did not appearto affect the results, however, as the SCR data fitted the models as well as the data from theTFU and PFRU, where sampling losses were considerably less significant. Experimentswere not performed where the solvent volume alone was varied, but initiated runs of 0.41mol/L in Paraflex were run at 100°C using different volumes and initiator concentrations.The values of kR1obtained (0.020 (molfL)°25/hr[1 mM bP] cf. 0.0 175 [2.5 mM bP]) for1144. Autoxidation of Model Solutionsthe different volumes (1.05 L cf. 1.75 L.) show a correlation with AM/V when the effect ofthe difference in initiators is considered. The lumped kinetic constants were used incomparing reaction behaviour as complete decoupling of the individual components isimpossible without a correct estimate ofA,,1/V.Van de Vusse considered the oxidation of hydrocarbons and proposed a reactionscheme which was zeroth order in oxygen. The rate limiting step in indene autoxidation isthe reaction of the peroxy radical with an indene molecule, givingRB = RA = - (k3 + k3) CB [RO2] = - R0 [4.27]where [RO2J is the concentration of peroxy radicals and R0 the rate of disappearance ofA. This is zeroth order in oxygen concentration but oxygen is the limiting reactant so R0must obey the conditions= R0 where CA > 0 [4.28]R0 = 0 where CA =0 [4.291This approach is the basis of Norton and Drayer’s model (i.e. without mass transfer). Vande Vusse considered the slow and intermediate regimes but did not include the effects ofoxygen depletion. The diffusion equation [4.11] can be integrated assuming excess CB andconstant [R02]to giveDA CA = R0x2/2 + B1 x + 82 [4.30]Kay and Nedderman considered the general case of absorption with chemical reaction andshowed that zeroth order (i.e. simple) chemical reaction kinetics yield complex absorptionkinetics for slow or intermediate reactions.They described a simplified result for fastchemical reactions, involving the boundary conditionsCA = CA,flt at x=0 [4.13]dCA/dx = 0 where CA =0 (x = xF) [4.18]where XE 1S the point of oxygen exhaustion and in a fast reaction lies within the diffusionfilm (XE OD). The solution of Equation [4.30], where XE ôj, gives1154. Autoxidation of Model Solutions- DA (dCA/dx)o = (2 RODA)°5CA,t°•5 [4.311Applying [4.22] and [4.27] gives the kinetic model for zeroth order reaction with masstransfer;dCB/dt = - (AM/V) (2 kjRO]D)°5CA,flt°5CB°5 [4.32]Assuming constant radical concentration, this yields the resultCB°5 (t) = CB°5 (0)- (AM/V) (2 k0 [RO2jD°5CA,jflt°•5tJ2 [4.33jor CB°5 (t) = CB°5 (0) - kRt [4.341where kR is a lumped rate constant. If the oxygen concentration in the bulk is not negligible(i.e. slow or intermediate regimes) the solution contains both film and bulk contributionsand the overall kinetics become complex. In the fast regime the concentration of B isdescribed by a function in cosh(x) which does not vary significantly across the diffusivelayer.The concentration of free radicals is unlikely to be constant at the start of a thermallyinitiated experiment. If bimolecular decomposition of hydroperoxide is assumed to generatefree radicals in this phase, when the concentration of B is relatively constant (lowconversion), Equation [4.32] can be written in terms of hydroperoxide;dCB/dt = - d[RO2HI/dt =- (AM/V) (2k0[ROH]DA)°5CA,jflt°5CB°5 [4.35]This yields a hydroperoxide kinetic expression[R02H]°5(t) = [RO2H]°5(0) + kM,pox t [4.36]where kM,p05 is a lumped kinetic constant which varies as CA,tflt°5 CB°5. This form ofexpression was found to fit the thermally initiated POx data very well in the initial phaseafter the induction period and also provides an explanation for the observed hydroperoxidebehaviour. The thermally initiated experiments follow Equation [4.36] initially but as thereaction proceeds, hydroperoxide will be consumed in order to generate radicals and toform polyperoxides which the assay registers as single hydroperoxide units. The rate ofincrease will thus tail off and once the free radical reaction is fully established thehydroperoxide kinetics become too complex to model. Russell’s kinetics, by comparison,1164. Autoxidation of Model Solutionspredict that the hydroperoxide concentration will increase linearly until the consumption ofindene becomes significant. An expression similar to [4.361 can be deduced in the absenceof mass transfer by assuming that initiation occurs via unimolecular dissociation ofhydroperoxide. In this case kM,pOX would vary as CB1 .0 rather than CB°5 and the rate ofindene consumption would vary as CB [ROOHJ, i.e. would accelerate.The two kinetic models were fitted to the data from Section 4.1, where the time datawere adjusted to exclude the chemical induction period. The regressed rate constants areshown in Table 4.3 and the regression coefficients show that the models fit the datareasonably well. The values of kRM and kR are similar between solvents and show the sametrends observed in the gum rate in Section 4.1. The values of kM,poX mirror the trend in kRexcept for trichlorobenzene, where the contamination side reaction also affected the gumyield and initial gum rate. Figure 4.3 shows the regression analysis of the data from run#306 in Paraflex and the fit is plainly superior to the models without mass transfer shownin Figure 4.2. The values of the regression coefficients in Table 4.3 for the Russell-basedmodel are larger in this set of data, but this trend is reversed in most of the other thermallyinitiated experiments. The zeroth order model gave a better fit in nearly all chemicallyinitiated experiments. This model also provides a basis for the observed hydroperoxidebehaviour and was thus used to compare the indene kinetics in all SCR and foulingexperiments.4.2.3 Gas Phase Resistance Effects in Mass TransferThe gas phase resistance to mass transfer has been assumed to be negligible in theabove treatment of autoxidation kinetics. Two sets of experiments were performed toconfirm this and to study the effect of the different modes of oxygen transfer in the SCRand the PFRU reactors. In the SCR, gas is bubbled into a stirred liquid volume while in the1174. Auloxidation of Model Solutions-ISolvent Rate Constant(Russell)kRM(mol/L)025/hrR2 kRkmol/L)/hrTbUlk 100°C; Pair 377 kPa; 0.41 M indenePOx RateConstant(Eqn [4.36])(v”[meq/L]/hr) 4.3 Comparison of Mass Transfer Kinetic Models of the Autoxidation of indenein Paraflex0.8 0.80.7 .0.6‘3)0.50.4C0.30.2k0.10.0200. 5 10 15Time(hours)Table 4.3 Regression of Thermally Initiated Autoxidation Data to Mass TransferModelsRegressionFitRate Constant Regression Experiment(zeroth) Fit #R2trichlorobenzene 0.01294 0.992 0.01595 0.991 302 0.336toluene 0.01275 0.960 0.01703 0.945 303 0.812n-octane 0.01952 0.986 0.02583 0.970 304 1.045kerosene 0.00408 0.885 0.00706 0.793 305 0.192Paraflex 0.01924 0.999 0.01958 0.984 306 0.8461184. Autoxidation of Model SolutionsPFRU gas is passed over a liquid surface agitated by the recirculation of the liquid throughthe fouling loop.The effect of gas flow rate in the SCR was studied using solutions of 0.41 mol/Lindene in Paraflex at Tblk 100°C and Pm,. 377 kPa with 1 mM bP as initiator. Air at400 mLlmin and 1000 mL/min (ntp) was used, corresponding to 1.5 and 5.25 excess overthat consumed. Table 4.4 shows that the gas flow rate does not have a significant effect onthe autoxidation rate. The indene reaction rate at the higher flow rate is slightly larger,which is probably due to the increase in bubble surface area. Most SCR experiments wereperformed at gas flow rates of 150-300 mL/min so the surface area so the variation due tobubble surface area should be small. The results indicate that the gas phase resistance tomass transfer is not significant here. Kerosene was not tested at higher flow rates as the gasflows stripped the lighter components of the kerosene out of solution.The effect of gas delivery configuration was studied using thermally initiatedsolutions of 0.70 mol/L indene in kerosene at Tblk 100°C and 377 kPa. Theresults are shown in Table 4.4 and are relatively similar given the degree of experimentalerror. The indene concentration data used in calculating the kinetic constants was obtainedusing the added standard method and involves greater error margins. R2 for the kR valueshere is 0(0.95) whereas the Paraflex runs involve R2 0.998. The similarity in reactiondiagnostics despite the inherent difference in gas delivery is not completely understood butdoes indicate that gas phase resistance is not a limiting factor in autoxidation here.A set of experiments was performed using identical chemical conditions in the SCRand the PFRU in order to compare the performances of the two reactor configurations. ThePFRU was operated isothermally in these cases, as a batch reactor with mixing provided bythe pump recirculation, Thermally initiated solutions of 0.70 mol/L indene in Paraflex andkerosene were used at Tb1k 100°C and P 377 kPa. The results are summarised in1194. Autoxidation of Model SolutionsTable 4.4 Gas Phase Resistance Effects in Autoxidation KineticsConfiguration Run Solvent [indene]0 Maximum Initial Gum Gum Rate# Rate Rate Yield Constant(x105) (g/L.hr) kR(molIL) (moWs) ± 0.2 (g/g) J(molIL)/hr400 mL/min (bubbled) 140 Paraflex 0.372 1.421 2.8 0.53 0.0361l000mL/min(bubbled) 141 Paraflex 0.371 1.617 2.6 0.44 0.0399200 mL/min (blanket) 112 kerosene 0.645 0.942 2.0 0.78 0.0158200 mL/min (bubbled) 113 kerosene 0.653 0.905 2.3 0.54 0.0143200 rriL/min (bubbled) 114 kerosene 0.646 1.125 3.0 0.75 0.0149Tbulk 100°C : 79.2 kPa oxygen saturationTable 4.5 Comparison of Autoxidation Kinetics in the SCR and PFRUExperiment # Reactor [indene]o Induction Solvent Initial Gum Yield Indene RatePeriod Rate Constant, kR(mol/L) (br) (g/L.hr) (g/g) ‘I(moL(L)/br117 SCR 0.622 1.0 Paraflex 1.9 0.723 0.0161118 PFRU 0.624 1.0 Paraflex 2.8 0.778 0.0155114 SCRt 0.646 7.0 kerosene 2.0 0.750 0.0149112 SCR 0.645 8.5 kerosene 3.0 0.780 0.0158115 PFRU 0.646 7.0 kerosene 1.4 0.550 0.0113t - bubbled air in SCR: - induction period for appearance of gumThUlk = 100°C: 79 kPa oxygen saturation1204. Autoxidation of Model SolutionsTable 4.5 and show significant differences with reactor configuration. The Paraflex resultsindicate that indene is being consumed at similar rates in the SCR and PFRU but the gumformation rate is significantly larger in the PFRU. This increased gum yield was alsoobserved in the TFU and was thought to be due to the reactor configuration. The SCRfeatures dead zones below the stirrer which allow gum to precipitate out while the foulingrigs are largely free from dead zones. This hypothesis was also supported by theprecipitation of insoluble gum in the SCR and the occasional overshoots in gumconcentration above g* observed in the fouling rigs after t’. The kerosene results showthat the reaction rate in the SCR was significantly faster than in the PFRU, which isinconsistent with the other results. Autoxidation otherwise proceeds at similar rates inParaflex and kerosene. The two solvents do differ, however, in the gumlPeroxide Numberbehaviour. The gum concentration in Paraflex and the other solvents in Section 4.1increases in step with the Peroxide Number until tK. The gum concentration in kerosenelags the increase in POx and thus the induction periods in Table 4.5 are different eventhough the induction period for POx increase was equal in these cases, at zero hours.Hydroperoxide is thus formed in kerosene before soluble gum appears, which suggeststhat some oxidation products are more soluble in the slightly more aromatic kerosene.4.2.4 Oxygen Effects in AutoxidationThe kinetic models in Section 4.2.2 predict that the autoxidation rate varies withdissolved oxygen concentration. This was studied in the SCR using chemically initiatedsolutions of 0.41 mol/L indene in Paraflex and by Asomaning (1993) using thermallyinitiated solutions of 10 wt% solutions of indene in kerosene. The Paraflex runs used 2.5mM bP as initiator at Tblk 100°C and the dissolved oxygen concentration was varied byadjusting the air pressure in the reactor. The kerosene runs were performed at Tblk 85°Cand the oxygen partial pressure varied by diluting air with nitrogen.1214. Autoxidation of Model SolutionsThe reaction diagnostics from the Paraflex experiments are summarised in Table4.6. The reaction rates increase with oxygen partial pressure but the rates in run #139 arelower than expected. This run was performed using a different batch of indene andemphasised the need to use the same indene source in series of experiments. The yield ofgum increased with oxygen pressure as reported in the literature (Davies, 1961) but thevalue of g* remained constant at around 11 gIL, confirming that g* is a physical parameter.The initial gum rate was proportional to oxygen partial pressure, according toInitial gum rate = -0.660 + 0.041 P02 (kPa) [4.37]This linear relationship is the product of indene and yield both being functions of oxygenconcentration. The zeroth order rate constant, kR, varied as CAjflt where n = 0.69 (R2 =0.999) while the rate constant for mass transfer with Russell’s kinetics, kRAJ, varied as n =0.89 (R2 = 0.999). The individual zeroth order rate constants involved larger regressioncoefficients and thus better correlation to the data. The values of n are close to the models’values (0.50 and 0.75, respectively) and confirm that mass transfer is significant in theseautoxidation reactions. This result was re-confirmed by Asomaning’s study (1993), wherethe values of n for the zeroth and Russell models were found to be 0.55 (R2 = 0.995) and0.58 (R2 =0.998) respectively.Asomaning’s study did not yield sufficient gum data to perform a similar analysison gum behaviour. The Peroxide Number data was fitted to Equation [4.36) and thevalues were found to depend on CA,EflF47 which is in good agreement withEquation [4.351. A separate study of thermally initiated autoxidation of 10 wt% indene inkerosene by Zhang et al. (1992) found the initial POx data to followkR,po = 1.219 x1010P025 exp (79 300IRT) [4.38]where P02 is in kPa. This result is in good agreement with Equation [4.32). The indenedata fitted the zeroth order kinetic model reasonably well but varied as CA, ,1 .1 6 ThePeroxide Number in chemically initiated experiments did not obey Equation [4.36], asexpected.1224. Autoxidation of Mode! SolutionsTable 4.6 Oxygen Effects in the Initiated Autoxidation of Indene in ParaflexIndene RateConstantkR(1iV’(molJL).hr)0.01980.01530.00980.0185Experiment # YieldOxygen Partial Oxygen Initial GumPressure Concentration RateCA, mt_____________________(kPa(mourn3) (g/L.hr)134 79.2 5.42 2.51137 50.2 3.43 1.27138 28.4 1.94 0.49139 99.4 6.80 3.070.41 M indene in Paraflex: Tbulk = 100°C: 2.5 mM bP initiation(g/g)0.7060.63 30.3430.9621234. Autoxidation of Model SolutionsThe oxygen results confirm that the autoxidation in model solutions is subject tosignificant mass transfer effects. The experimental apparatus was not designed for arigorous study of mass transfer with chemical reaction, which was beyond the scope of thecurrent project.It is worth comparing the rates in the current study with those reported by Russell at50°C in pure indene. Russell’s kinetic expression gives an indene consumption rate of0.00384 mol!L.hr for 0.41 mol/L inclene in an inert solvent at P02 = 79.2 kPa at 50°C.Russell’s procedures did not report any surface areas which could be used in an estimate ofas above. Russell did not report an activation energy so Panchal and Watkinson (1993)used the value for styrene copolymerisation with oxygen, 96.3 kJ/mol. This yields a rate of0.0468 mol/L.hr at 100°C, which is in the same range as the value obtained in Section 4.1,0.025 mol/L.hr. Mass transfer would appear to have reduced the consumption rate.4.3 Chemically Initiated Autoxidation of indeneThe thermally initiated fouling experiments involved significant chemical inductionperiods and scatter in the kinetic parameters so chemical initiators were considered as ameans to eliminate the induction period and promote repeatable chemical reaction in thebulk liquid in fouling experiments. Two initiators, benzoyl peroxide (bP) and ABN, werestudied in the SCR using solutions of 0.40 mol/L indene in Paraflex at Tb1k 80°C andPair = 377 kPa at concentrations of 2.5, 5.0 and 7.5 mmol/L.The chemical initiators eliminated the chemical induction period completely and thePeroxide Number behaviour changed, as expected. The initial POx in bP was close to theinitial concentration of initiator and was zero in ABN. The POx then increased almostlinearly until it approached a similar plateau value as observed in the thermally initiatedexperiments, after which it remained relatively constant. The POx did increase slowly after1244. Autoxidation of Model Solutionsthis in later initiated experiments at higher temperatures, but both patterns mirror thethermally initiated trends and confirm RusselUs result that autoxidation involves the samemechanism under both initiation regimes. The value of g* did not vary with the mode ofinitiation and the gum analysis results were almost identical.The use of initiators increased the rate of autoxidation significantly and the kineticparameters are summarised in Table 4.7. Both the gum rate and rate constant kM were moresensitive to the concentration of benzoyl peroxide than of ABN, although the yield of gumwas generally larger with ABN (0.56-0.64) than with bP (0.41-0.51). Russell (1956c)reported similar trends with benzoyl peroxide initiation and attributed it to its tendency togenerate other oxidation products. Russell (1956b) reported that the rate of indeneoxidation was proportional to + [AJBN]), suggesting that R0 should vary with thesquare root of initiator concentration. This yields a mass transfer model where kR4 isproportional to initiator concentration; the data in Table 4.7 do not follow this trend. If R0is instead assumed to be linear in initiator concentration, Equations [4.33, 4.34 ] predictthat kR2 should be linear in initiator concentration. This gave(benzoyl peroxide) k12 = 6.888x105(1 + 461 [bPl) R2 = 0.990 [4.39](ABN) kR2 = 7.314x105(1 + 225 [ABN]) R2 = 0.988 [4.40]The initial gum rate did not fit a simple kinetic model, particularly at high initiatorconcentrations where the gum concentration overshot g* before approaching the saturationlimit. The relationship between gum rate and indene kinetics is complicated by the gumbeing a secondary reaction product and the kinetics are considered further in Section 4.6.These experiments showed that repeatable chemical reaction could be best achieved byusing lower initiator concentrations in the fouling runs.1254. Autoxidation of Model SolutionsTable 4.7 Chemically Initiated Autoxidation of Indene in ParaflexExperiment # Initiator g*(mmolJL)Yield RateConstantFitR2Initial Gum Indene RateRate ConstantkR(L.hr) (g/g) (lhJ(mol/L).hr)125 0.0 5.5 ±0.5 0.44 0.372 0.00890 0.992122 0.0 6.0 ±1.0 1.02 0.498 0.00804 0.960129 2.5 [bP] 6.5 ±0.5 0.94 0.411 0.01 148 0.993126 5.0 [bP] 7.0 ±1.0 1.38 0.408 0.01506 0.993128 7.5 [bP] 6.5 +1.0 2.6 0.509 0.01769 0.9921321- 2.5 [ABN] 7.0 ±1.0 1.18 0.562 0.01742 0.980131 5.0 [ABNJ 6.0 ±0.5 1.59 0.629 0.01213 0.989133 7.5 [ABN] 7.0 ±1.0 1.78 0.638 0.01421 0.9951- - problems in SCR air pressure: 0.41 M indene in Paraflex 79 kPa oxygen saturation : Tbulk 100°C1264. Autoxidation of Model Solutions4.4 Temperature Effects in AutoxidationThe effects of bulk temperature on reaction rates and gum properties were studiedusing thermally and chemically initiated model solutions of indene in kerosene and inParaflex.Figure 4.4 shows the variation of the polyperoxide gum solubility limit, g*, withliquid temperature in solutions of 0.41 mol/L indene in Paraflex. The figure shows that g*is independent of initiation mode and reactor configuration and increases monatonicallywith temperature. Equation [2.24] describes the solubility of a solvent free solute as afunction of temperature, solvation energy and activity coefficient Y2 The g* data did not ftthe expression for an ideal solution (Y2 = 1) very well, as expected for a polar solute in anon polar solvent. The data could be fitted to [2.11] by manipulating the magnitude of Y2The estimated value of Y2 necessary to fit the data was large compared to the solvationenergy term; the analysis was thus abandoned given the uncertainty in gum and solventproperties (in particular, dipole interaction parameters). The polyperoxide chain lengths arelikely to decrease with increasing temperature, which introduces further uncertainty in thecalculation of the gum mol fraction.Figure 4.5 shows the variation of g* with indene concentration in Paraflex at100°C. The gum solubility limit increases linearly with indene concentration and thus thearomaticity of the solvent. The values of g* in kerosene in Figure 4.5 are larger than inParaflex and confirm the hypothesis that g* is a physical property of the model solution.The effect of temperature on the autoxidation of indene in kerosene was studied byZhang et at. (1992) using thermally initiated solutions of 0.71 mol/L (10 wt%) indene inkerosene with = 329 kPa at temperatures ranging from 80-120°C. The PeroxideNumber behaviour shown in Figure 4.6 is similar to that observed by Hess and O1Hare1274. Autoxidation of Model SolutionsFigure 4.4 Temperature Effects on the Gum Solubility Limit, g*, in Paraflex30b.O 25*20& 1 C thermal initiation2.5mMABN0100 2.5mMbP5 0 PFRUA 1.OmMbP0 I I60 80 100 120 140 160 180Temperature(°C)Figure 4.5 Variation of the Gum Solubility Limit, g*, with Indene Concentration300.25 0 Paraflex* • kerosenei 200!2 0.8[indenej(molJL)1284. Autoxidation of Model SolutionsFigure 4.6 Temperature Effects in the Thermally Initiated Autoxidation of 10 wt%Indene in Kerosene : Peroxide Number Behaviour8070! ::403020100Time(hours)10 wt% indene in kerosene; Pair 329 kPa; performed by Zhang and Wilson120Table 4.8 Activation Energies in the Autoxidation of Indene in Model So’utionsSolvent Initiation Range Eact Indene IndeneInitial Overall ReactionGum Rate ER(°C) (kJ/mol) (kJ/mol) (k.J/mol)kerosene thermal 80-120 31.5 ±3 26.4±3 57.7Paraflex thermal 80-120 32.7 ±2 30.5±2 51.2Paraflex 2.5 mM bP 80-110 36.2 ±3 30.0±2 48.1Paraflex 1.0 mM bP 100-140 47.2 ±4 28.3±2 56.65 wt% indene in Paraflex: 10 wt% indene in kerosene: 79 kPa oxygen saturation0..-.:.0‘—. 00 ‘-...-.e.--0.- 81°C“.-. 91°CA-. 121°C--0-- 111°C‘-a’- 101°C0 20 40 60 80 100Peroxide GumYieldM,POx Range(kJ/mol)73.9 ±4 051-0.8235.0 ±3 0.50-0.660.41-0.710.44-0.711294. Autoxidation of Model Solutions(1950) in the autoxidation of linseed oil. The POx induction period varied with temperatureand corresponds to the period required to consume the inhibitor present in the kerosene.The induction period will thus be inversely proportional to the rate of generation of peroxyradicals and is unlikely to be subject to mass transfer effects (low reaction rate). Thechemical induction period r can then be written as oc R1. An Arrhenius plot of ‘r againstinverse temperature gave an activation energy of 79.3 (±5) kJ/mol, which is less than therange for thermal initiation of olefinic autoxidation in the literature (around 125 kJ/mol).Zhang fitted the initial increase in POx to Equation [4.36] and reported a similar activationenergy, of 73.9 ±4 kJ/mol for kM,po; this value suggests that the initial generation ofhydroperoxide in kerosene is not mass transfer limited. The formation of gum was alsoseen to lag the increase in POx in kerosene, which was not observed in the other solvents.This hypothesis was confirmed by the regression of the indene data to the mass transferkinetic models; the indene started to decrease significantly once gum formation started andfitted the model until the POx value declined. The regressed values of kR were used toestimate two activation energies; the overall activation energy, EM,0 and the reactionactivation energy ER which are related byEM,o = 0.5 Ediffusion + 0.5 ER ( Eprop + - Eterm)12) [4.41]ER/2 was calculated by plotting kRA/(CA,1 DA) in the Arrhenius plots. The value of ER wasfound to depend strongly on the diffusivity correlation used; the Hayduk and Minhascorrelation was used in the Paraflex calculations and the Wilke-Chang correlation inkerosene. A pseudo activation energy was also obtained from the initial gum formationrates. The estimated values are given with those obtained from the Paraflex experiments inTable 4.8. The yield of gum increased slightly with temperature across the range 0.50-0.72but there was considerable scatter in this trend.The activation energies of gum formation and indene consumption in Paraflex wereobtained using solutions of 0.41 mol/L indene in Paraflex with Pair = 379 kPa and threedifferent levels of initiation. The thermally initiated series was used to compare autoxidation1304. Autoxidation of Model Solutionsin Paraflex with the above results in kerosene. A series with 2.5 mM bP initiation was usedto compare thermal and chemical initiation while a series using 1 mM bP was used to studythe model solution selected for use in the TFU experiments. The activation energies aresummarised in Table 4.8.The thermally initiated experiments in Paraflex differed from those in kerosene.The chemical induction period was effectively zero and the Peroxide Number did notbehave as in Figure 4.6 but increased rapidly to a plateau value; any increase after this wasrelatively slow. Gum formation did not lag the increase in POx and regression of the indenedata showed that mass transfer effects were evident once gum formation had started. Theactivation energy for kM,po in Paraflex is almost half that found in the kerosene, whichconfirms that the initial increase in hydroperoxide in Paraflex is mass transfer limited. Thedecoupled activation energy was calculated to be 60 kJ/mol, which is similar to thekerosene value.The activation energies shown in Table 4.8 show good agreement within the limitsof experimental error. The initial gum rate and kR activation energies in the sametemperature range do not vary significantly with initiation mode and solvent nature andsuggest that a common autoxidation mechanism applies. The gum yield tended to increasewith temperature in the Paraflex experiments and this contributed to the higher gum rateactivation energy reported using 1mM bP initiation, where temperatures ranged up to140°C.The decoupled reaction activation energy ER 0. + Eprop - 0. SEterm wasfound to be approximately 52 kJ/mol. Howard and Ingold (1962) assumed that thebimolecular peroxy radical termination step activation energy Eterm 0, giving ER0. + Eprop. The activation energy for benzoyl peroxide decomposition (138.8 kJ/mol)is larger than that expected from thermal initiation (around 120 kJ/mol), but the values ofER do not reflect the differences in initiation method. Literature values for the propagationactivation energy range around 45 kJ/mol (Scott (1965)) giving an estimate of ER as 1001314. Autoxidation of Model SolutionskJ/mol. Howard and Ingold (1962) thus reported an ER value of 96.7 kJImol for styreneoxidation, which is significantly larger than the values in Table 4.8. The source of thediscrepancy in activation energies is likely to be the initiation step, which is neither welldefined nor understood under the conditions in these experiments.The activation energies were used to calculate mean values of kR for solutions of0.41M indene in Paraflex at 100°C and Pair = 379 kPa. The ratio of the thermally initiatedmean (0.01532 v’(molJL)/hr) and the 2.5mM bP initiated mean (0.0 1921 -./(molIL)/hr) was1.26, which is close to the value of 1.4.6 predicted by Equation [4.39]. The ratio of the 2.5mM bP initiated mean to the 1 mM bP initiated mean (0.0350 ./(molIL)/hr) was 0.549,although [4.39] gives a ratio of 1.21. This discrepancy is consistent with the mass transfermodel as the volume of liquid in the latter experiments was 0.59 that in the former, givingan estimated ratio of 0.7. This ratio was in reasonable agreement within the bounds ofexperimental error.The mean values of kR in the SCR were compared with the results from the foulingreactors. The 1 mM bP initiated SCR mean was close to the 1 mM bP initiated TFU mean(0.03254 v”(mollL)Ihr) while the estimated AM/V ratios were 12.6 rn1 and 5.98 m-1. Theestimation of TFU surface area was considered conservative as it excluded the jets ofrecirculated liquid inside the holding tank. The 2.5 mM bP initiated SCR mean was smallerthan the PERU value under similar conditions (0.0264 /(molIL)!hr)) while the thermallyinitiated values at 80°C were 0.00863 and 0.008 17 /(molIL)/hr respectively. The estimatedAM/V ratios (7.7 m1 (SCR) and 3.85 m (PFRU)) are not consistent with these resultsand suggest that the PERU system is more complex than expected.4.5 Autoxidation in the Presence of AntioxidantsThe effect of antioxidants such as the unknown additives in the keroseneexperiments was studied using thermally initiated model solutions doped with the1324. Autoxidation of Model Solutionssubstituted phenol, BMP. The aim was to establish the effect of BMP on the formation ofpolyperoxide gum, which had been identified as the fouling precursor in the initial foulingexperiments.Dopant concentrations of 0, 50, 100,200 and 400 ppm BMP were added to modelsolutions of 0.41 mol/L indene in Paraflex at 100°C and the results are summarised in Table4.9. The antioxidant extended the chemical induction period but the Peroxide Number, gumand indene concentrations all behaved as in the thermally initiated base case after theinduction period. The kinetic parameters in Table 4.9 confirm that the antioxidant did notaffect the autoxidation reaction once it was established. Howard and Ingold (1%2) studiedthe BMP-inhibited autoxidation of styrene initiated by AIBN and also found that theoxidation rate was constant after the induction period. The concentration of BMP wasmonitored using qualitative GC-PID analysis and confirmed that the end of the inductionperiod corresponded to the exhaustion of BMP.BMP acts as a peroxy radical scavenger, consuming two peroxy radicals permolecule in consecutive reactions. Assuming that radical generation in the induction periodproceeds via reaction [2.14], the steady state assumption for [R02] givesR1 = 2k [RH] [021 = 2 kAox [R02] [BMPJ [4.421The consumption of BMP is proportional to R1, which is effectively constant and zerothorder in BMP during the induction period. The induction period ris given by= [BMP]!(2k [RHI [021) [4.43]The data in Table 4.9 was fitted to Equation [4.43] and gave k1 as 1.22 x109m3/mol.swith R2 = 0.989.The initial gum formation rate, the gum analysis results and the gum solubility limit,g*, did not vary with BMP concentration. The gum formation process is thus unaffectedby the oxidation products of BMP. A similar result is expected for fouling; Asomaning andWatkinson(1992) similarly reported that the fouling behaviour of DCP in kerosene did notseem to be affected by the presence of t-but-catechol apart from the extended induction1334. Autoxidatjon of Model SolutionsTable 4.9 Autoxidation of Indene in Paraflex in the Presence of Antioxidants[BMP] Experiment Induction Initial Gum Gum Tbulk kR Gum# Period Rate Yield Analysis(ppm) (hr) (g/L.hr) (g/g) (°C) (mo1IL)/hr0 310 3.5 0.72 0.490 100 0.01264 C9H80021501’ 307 10 1.41 0.690 100 0.01446 -100 311 30 0.90 0.470 100 0.01108 C9H75020200 312 45 0.90 0.444 100 0.01123 C9H73022400 313 86 0.85 0.440 100 0.01480 C9H7,5023100 315 120 - - 80 - C9H7,5018100 318 85 0.30 0.511 90 0.01207 C9H81022100 316 11 1.01 0.391 110 0.03895 C9H75023100 317 4 1.99 0.458 120 0.01793 -0.41 M indene in Paraflex: Pair = 377 kPa; 1’ different indene batch1344. Autoxidation of Model Solutionsperiod observed in the inhibited solution. Autoxidation fouling in the presence ofantioxidants is discussed in Section 5.7.The effect of temperature on doped solutions was studied under similar conditionsin order to provide activation energies for comparison with uininhibited autoxidation. ThePeroxide Number and gum concentration behaviour did not differ from previousbehaviour, but the gum data contained considerable scatter and the pseudo activationenergy, 67±10 kJ/mol, is larger than the values in Table 4.8. The indene data was subjectto similar scatter and a least squares fit of kR gave an activation energy of 28.2 kJ/mol withR2 = 0.315. The results indicate that the autoxidation mechanism is not altered by thepresence of BMP after the induction period.The activation energy of the initiation rate constant was calculated from theinduction pçriod data to be 120±3 kJ/mol. This is larger than Zhang’s value for indeneinitiation in kerosene but is in the range (92-117 kJ/mol) reported by Lloyd andZimmerman (1965) for the inhibited autoxidation of cumene by several substituted phenols.The thermal initiation rate in pure indene under 750 mmHg oxygen at 50°C was thusestimated as 2.8x107mollm3s,which is 3x that calculated from Russell’s data. This wasreasonable as the analysis did not include other radical sinks or antioxidant efficiencies.Using the values of ER and in Equation [4.371 would make Eprop unreasonably closeto zero or negative. This confirms that the initiation step changes once the autoxidationreaction is under way.4.6 A Kinetic Model of Indene AutoxidationThe SCR studies confirmed that the autoxidation mechanism in the foulingexperiments differed from the literature due to the scale (mass transfer effects), extendedconversion (zeroth order kinetics) and solvent effects. The mass transfer effects limit thedirect application of the SCR results to the fouling reactors but the trends observed in the1354. Autoxidation of Model SolutionsSCR were all followed in the TFU and PFRU. The SCR results could be summarised intooverall kinetic model for indene autoxidation in Paraflex;CB(t)°5 = CB(0)°5- KM CA,1069 kR([bP) exp (-EM,JRT) t [4.44]where kR(bP)2 is given by Equations [4.39) and [4.40) and KM is a lumped kineticconstant. The value of KM at 1 mM bP initiation was calculated as 1.309 x106 (m3Imol)-069/hr; at 2.5 mM bP initiation, 5.74 x104(m3Imol)-°691hr;under thermal initiation, 3x104(m3/mol)-0691hr.The concentration of polyperoxide gum in solution cannot be predicted by a simpleequation such as [4.441 as the gum is a secondary oxidation product. The gumconcentration profiles lag the conversion of indene as the primary oxidation product, indenehydroperoxide, is soluble in solvents such as Paraflex. Norton and Drayer modelled this asA— B—b C — D, where B was the soluble hydroperoxide. Such models aresimplifications of the complex chemical reaction mechanisms involved and cannot berelated directly to the free radical kinetics.The models are formulated in order to explain theproduct distribution observed experimentally.The Norton and Drayer model would describe the reaction kinetics in the SCR bythe following series of differential equations, which can be solved analyticallydC1/dt = - k1C [4.45]dC2/dt = k1C- k2C [4.46)dC3/dt = k2C- k3C [4.47]dC4Idt = k3C [4.48]where C1 is indene, C2 indene hydroperoxide and C34 polyperoxides. This model did notexplain the exhaustion of indene observed in some SCR experiments and ignored the effectof decreasing indene concentration over an extended period.An alternative model is proposed which incorporates the autoxidation mechanism asunderstood from the SCR experiments. This model assumes a series of bimolecularreaction steps, which is a departure from the free radical kinetics involved. The probability1364. Autoxidation of Model Solutionsof forming a peroxide of m units is assumed to be proportional to the concentrations of rn-iperoxide and the monomer. This assumes that the concentration of free radicals of species iis in equilibrium with the concentration of species i. Ignoring oxygen effects, the semi-batch kinetics are described by a set of series-parallel reaction termsdC1/dt = - k1C - k2C1 - k3C1 [4.49]dC2/dt = e k1 C1 - k2CC1 [4.50]dC3/dt = ek2C1 - k3C1 [4.51]dC4/dt = ek3C1 [4.52]where e is a propagation efficiency, (1- e) representing the fraction of reacted peroxide thatdissociates to form inert products. Chain transfer does not affect the concentration of rn-erpolyperoxide so is not included here. Since C2.4 are all peroxides, it is assumed that k2 =k3. The equations do not have an analytical solution but the rates can be expressed in termsof the observed indene reaction rate.dC2 — C1(ek - k2C) — -(ek1 - k2C) (KC2-e)dC1— C1(-k - k2C -k3C) k1 + k2(C +C3) — 1 + K(C +C3) [453]dC3 — (ek2 - k3C) — -k2(eC C3) — K(C3-e2)d C1— C1 (-k1 - k2C -k3C) — k1 + k2(C + C3) — 1 + K(C2 + C3) [454]dC4— C1 ek3C — -ek3C — -eKC3dC1— C1(-k k2C -k3C)— k1 + k2(C +C3) — 1+ K(C2+) [4.55]where K = k2/k1.The reaction stoichiometry where e =1 (no dissociation) isa1C,0 = a1C + a2C + a3C + a4C [4.56]where a1= a2 = 1, a3 = 2, a4 = 3. In the presence of dissociation the stoichiometry willinclude dissociation products CD2, CD3, C1 whose concentrations will be governed byequations of the formd CD3 — (1 - e ) k2 C1 C2 — (e - 1) k2C — (e - 1)KC2dC1— C1 (-k1 - k2C -k3C) — k1 + k2(C +C3)— 1 + K(C +C3) [457]The mass concentration of productj is M1.C1, where is the molecular weight of productj. The molecular weights M2, M3 and M4 were assumed to be 148.1, 296.1 and 444.3respectively.1374. Autoxidation of Model SolutionsThis kinetic scheme could be expanded to include further features but this wouldincrease the number of parameters to be fitted to the experimental data. As proposed itinvolves three parameters which are to be regressed from three sets of concentration dataper experiment (indene, gum and Peroxide Number). The rate of indene disappearance isdescribed by Equation [4.34], givingdC1/dt = - 2 kM,R C°5 [4.58]The unknowns are thus the rate constant ratio K and the propagation efficiency, e. Equation[4.53] shows that there is a maximum in C2 when C2 = elK; this feature was useful inselecting combinations of parameters as described below.The model assumes that the reaction scheme does not change during the experimentand was thus unsuitable for thermally initiated studies, where the initiation step changesuntil the zeroth-order kinetic scheme is established. The model was compared with theresults from chemically initiated autoxidations of indene in Paraflex where all three sets ofdata were available.A FORTRAN 77 program was used to calculate the concentration profiles using aRunge-Kutte-Fehlberg subroutine to solve the set of ordinary differential equations. Theroutine was tested using Norton and Drayer’s model and gave results in good agreementwith the analytical solutions published in the same paper. Concentration units of mourn3were used throughout. Mass concentrations were expressed as g/m3 and the PeroxideNumber gave a concentration of (POx)12 mourn3.The values of kR in Section 4 wereconverted from lb/(mol/L)!lir to kMR (/(mol/m)!min) by multiplying by 0.52705.The data were compared with the model by assuming that the Peroxide Numbergave the concentration of the monomeric hydroperoxide alone (i.e POx C2), as the totalhydroperoxide concentration (C2 + C3 + C4) exceeded POx after a short period. The gumresults were mass concentrations and were assumed to be equal toC3(mass) +C4(mass).Numerical difficulties arose if K1m3/mol.min as the asymptotic value of C2 was so small(- 1 mourn3 cf. [indene] 370 mollm3).1384. Autoxidation of Model SolutionsFigures 4.7 and 4.8 show the variation in mass and peroxide concentrations as Kwas varied from 0.001-0.1 m3/mol.min at e = 1. The figures show the data from SCRruns 140 and 141 (Tbulk = 100°C, Pair = 79 kPa, 5 wt% indene in Paraflex, 1mM bPinitiation) for comparison. The indene reaction rate constant, 0.038086 q(moi/rn3)/rnin,was the mean value from the runs. The Figures show the expected trends in K; indene israpidly converted through the hydroperoxide intermediate to gum at high values of K,whereas at low K the indene forms a pooi of hydroperoxide which is slowly reduced.Figure 4.8 also shows that the asymptotic value of C2 at high K is lower and reached veryquickly, whereas at high K the asymptote is never reached. The change in C2 curve shapewith K is significant, as the experimental curves resemble those seen at low K; the lowvalues of POx involved, however, belong to the high K region. The gum mass resultsshow that the gum concentration increases in a pseudo-linear fashion, as observed in theexperimental data. By inspection, K 0.005m3/mol.min for these data.Figures 4.9 and 4.10 show the effect of the efficiency, e, on the product spectrum.A value of K of 0.025m3/mol.min was used which gave values of POx and gum close tothe data values. Figure 4.10 shows that the model POx curve does not change shape with eand is different from that observed in the data. The initiator concentration (1 molIm3)is nota significant factor in the curve shape. The model POx values are high, which suggests thata larger value of K be used; this would increase the gum formation rate, which Figure 4.9shows is already slightly high. Figure 4.8 also shows that the gum data rises after an initiallag, which was common to all the initiated experiments at moderate temperatures. The lagin the model gum curves is shorter and thus the model falls to account for this experimentalobservation. These deviations could be accounted for by incorporating more parameters,but this would require more data for verification.The lag in the rise of the gum concentration would also occur if the polyperoxidedimer was soluble in hexane and escaped detection in the gum assay, so that C(gum) = C4alone. The curves in Figure 4.9 were revised and are plotted in Figure 4.11. The figure1394. Autoxidation of Model SolutionsFigure 4.7 Kinetic Model of Indene Autoxidation: Effect of K on Mass Concentrationof Gum and Comparison with Data From Runs 140, 141. (e = 1)5000040000 K30000•‘1.-:-----‘-Io 20000.--.41,• ‘A10000 --,------ _-....1 ° . ..0- indene . K = 0.0010—gum 0 .1 •0 100 200 300 400 500Time(minutes)Figure 4.8 Kinetic Model of Indene Autoxidation: Effect of K on Peroxide Numberand Comparison with Data from Runs 140, 141. (e = 1)::50- K=0.0500 100 200 300 400 500Time(minutes)1404. Autoxidation of Model SolutionsFigure 4.9 Kinetic Model of Indene Autoxidation : Effect of e on Mass Concentrations= 0.025m3/mol.min: Data from SCR runs 140,14150000400003000020000100000Figure 4.1040II0 100 200 300 400 500Time(minutes)Kinetic Model of Indene Autoxidation: Effect of e on Peroxide NumberK = 0.025m3/mol.min: Data from SCR runs 140,14130201000I.- -.-.,.-.-.-.-..-.--0.7I. I I0I00100 200 300 400Time(minutes)5001414. Autoxidation of Model SolutionsFigure 4.11 Kinetic Model of Indene Autoxidation: Revised Mass Concentration Curvesat K = 0.025m3/mol.min with Data from SCR Runs 140, 14150000400003000020000100000II0 100 200 300 400 5000 100 200 300 400Time(minutes)500Figure 4.12 Kinetic Model of Indene Autoxidation: Revised Peroxide Concentrations atK = 0.025m3/mol.min with Data from SCR Runs 140, 14140e= .030 -e=Q.96=0820-- .,... -.-10-.s -. -.0 I I • ITime(minutes)1424. Autoxidation of Model Solutionsshows that the data fits the curve for e = 0.9 within experimental error up to the solubilitylimit. The solubility limit has been shown to be a physical parameter and would not bepredicted by a kinetic model without suitable modification.The POx curve was treated similarly and Figure 4.12 shows the C3 curves from thesame set of results plotted as POx. The curve at e = 0.9 shows good agreement with theexperimental data and thus suggests that the Peroxide Number method is measuring thehydroperoxide dimer rather than the monomeric hydroperoxide. This result was found toapply to other sets of experimental data. Figures 4.13 and 4.14 show the revised mass andperoxide concentrations predicted by the model for SCR run 142 (0.41 mol/L indene inParaflex, Tb1k = 120°C, 79 kPa oxygen saturation), where kM,R was calculated as 0.0734J(mol/m)/min. The revised concentrations predicted by the model at the mean conditionsin the TFU fouling runs (0.41 mol/L indene in Paraflex, Tblk = 100°C, 72 kPa oxygensaturation, kAl,R = 0.073403 /(mol/m3)/min) are plotted in Figures 4.15 and 4.16 alongwith the data from runs 501, 503 and 504. Both pairs of figures show that the data exhibitsthe revised trends surprisingly well; the same value of K was used in all the figures, at0.025m3/mol.min.The mass concentration curves for e = 0.9 in Figure 4.11 were used to calculate the‘yield’ of gum as defined by Equation [4.11 between the appearance of ‘gum’ at 60minutes and t, taken as 280 minutes. The value obtained, 0.66, was larger than theexperimental values (0.4,0.53) but was significantly smaller than 0.9, the value of e whichgenerated the curves. The model thus suggests that the yield is not a true measure of theadditionlabstraction ratio.The model gives good agreement with the experimental results when the original setof assumptions have been revised to include a ‘missing’ mono-hydroperoxide and a solublepolyperoxide dimer. The model mechanism may, however, be incorrectly formulated andthe observed agreement could be a fortunate coincidence of the numerical solutions and the1434. Autoxidation of Model SolutionsFigure 4.13 Kinetic Model of Indene Autoxidation: Comparison of Revised MassConcentration Curves and Data from SCR Run 142 at Tb1k = 120°C5000040000• 30000C0rIj2000010000Time(minutes)Figure 4.14 Kinetic Model of Indene Autoxidation: Comparison of Revised MassConcentration Curves and Data from SCR Run 142 at Tb1k = 120°CK = 0.025 m3/mol/minSCR run # 142Tbulk = 120°C: 1mM bP0.41 M indene inParaflex0 50 100 150 200 250 300 350 400Time(minutes)00 50 100 150 200 250 300 350 400403020100IC3e = 0.81444. Autoxidation of Model Solutions40000. 30000a0K = 0.025 m3/moLminTFU Data from runs 501, 503, 5040.41 M indene in Paraflex1.0 mM bP; Tb = 100°CFigure 4.15 Kinetic Model of Indene Autoxidation: Comparison of Revised MassConcentration Curves with Data from TFU Runs 501,503, 5045000020000100000Time(minutes)Figure 4.16 Kinetic Model of Indene Autoxidation: Comparison of Revised PeroxideConcentrations with Data from TFU Runs 501,503,504500II4030201000 100 200 300Time(minutes)400 5001454. Autoxidation of Model Solutionsdata. The ‘missing’ hydroperoxide hypothesis suggests that the Peroxide Number test wasnot sensitive to the initial oxidation product and requires experimental verification.4.7 Ageing of Polyperoxide GumsPeroxides and polyperoxides are known to undergo degradation at enhancedtemperatures such as those found on the heat exchanger surface in the fouling experiments.The initial fouling experiments in Section 5 indicated that polyperoxide gums were theprimary foulant precursor in autoxidative systems and that the deposits were composed ofpolyperoxide degradation products. The polyperoxide gums thus appeared to undergodeposit ageing as described by Nelson (1934). The rates and mechansim of the ageingprocess were not described in the literature, so the thermal degradation of polyperoxidegums was studied using TGA and the ageing oven apparatus. Gum samples were heatedfor prescribed periods in a nitrogen atmosphere to avoid combustion and because thedeposit phase was likely to be deficient in oxygen. A reliable method of ageing the gums insolution proved elusive so dry samples were used in these studies.The indene polyperoxide gum was obtained from an SCR experiment using a modelsolution in Paraflex as solvent. All Paraflex was washed out of the gum using hexane,which was in turn removed by drying in a vacuum oven. The gum was amber coloured andcontained both peroxide and carbonyl groups. On heating, the gum rapidly melted to form asticky, dark red coloured gum similar to that observed in veneers on the PFRU probe.This red gum often cracked on cooling, as observed in the veneers formed on coolersections of the PFRU probe. On prolonged heating, the red gum hardened to a dark browntar which cooled to form a brittle solid which was insoluble in acetone. The colour changesthus correspond to those seen on the PFRU probe during fouling runs.Figure 4.17 is a series of aged gum FTIR spectra from the oven ageing experimentat 200°C. The scale is a software construction and the absorbence values are written against146“-iC) C) -t 0—.C) C)—C) ciC) C)0 C) C) tTl C) C)— 0•rj -t C)0—200—400T=200CWavenumbers(cm—i)Res=1cm—i08/26/9215:40GainOOl4. Autoxidation of Model Solutionsthe major peaks. The spectra were generated by R. Lai (1992). The hydroxyl peak (notshown, 3400 cm-1) in the base gum disappeared rapidly on heating and the carbonylregion becomes more crowded as heating continues. Quantitative analysis of the carbonylregion (1700-1800 cm-1) gave inconsistent results. Thermal degradation here seems tofollow the mechanism reported at lower temperatures, where peroxides react to formcarbonyl adducts. The red colour changes are also consistent with such an ageing step.The mass of gum decreased during an experiment so this was used as a monitor ofthe rate of ageing. The loss of material is due to volatile products of condensation andscission which are removed by the nitrogen purge. Figures 4.18 and 4.19 show thereduction in gum mass observed in the ageing oven and TGA at temperatures ranging from160-240°C. The initially rapid loss of material is followed by a slower decay whichcontinues until the end of the experiment. If ageing corresponded to the exclusion of simplemolecular species such as water or C02,the mass would approach well defined asymptotes(i.e. C9H802- H20 = 87.8%; -C02 = 70.2%), but the data does not follow such patterns.Elemental analyses of the aged samples are given in Table 4.10 and did not show anytrends in C:H:O composition during ageing, confirming this observation. The chemicalmechanism of ageing is likely to be quite complex and further characterisation was notattempted.The data from the TGA include an induction period caused by the initial ramping tothe analysis temperature. The rates from the TGA are consistently lower than in the ageingoven, which was thought to be due to the differences in mass transfer between theapparatus; the ageing oven sample atmosphere is more vigorously agitated. The initial rateof devolatilisation in the ageing oven was plotted against inverse absolute temperature togiveInitialDegra&ztionRate dm]dt (%Imin) = 166 000 exp (-39750 (±1300)IRT(K)) [4.54]The TGA data gave a more reliable value of the final decay rate and this fitted to a similarexpression1484. Auloxidation of Model SolutionsFigure 4.18 Reduction in Gum Mass During Ageing Oven Experiments1009590858O7570IL555045400 10 20 30 40 50 60Time(minutes)50 60Figure 4.19 Reduction in Gum Mass During TGA Ageing Experiments100959085807570656055504540Time(minutes)160°C• 180°C0 • 200°Co •• 00 0 220°C00 •0 0 240°C.00000o o60.1 I I I0 10 20 30 401494. Autoxidation of Model SolutionsTable 4.10 Elemental Analyses of Aged Polyperoxide GumsExperiment # Temperature Time Carbon Hydrogen Oxygen(°C) (minutes) (wt%) (wt%) (wt% t)117 100 0 74.92 5.55 17.5374.63 5.55 19.73401 200 1 75.30 5.58 19.12401 200 10 76.18 5.18 18.64401 200 20 76.54 5.20 18.26401 200 90 77.90 4.72 17.38402 160 60 76.68 5.13 18.19403 240 4.5 77.84 4.64 17.52404 180 20 77.62 5.11 17.27405 220 18 75.43 4.70 19.87t - oxygen wt% calculated by difference: * - Source of gum for ageing experimentsAgeing experiments performed in conjunction with R. Lal1504. Autoxidation of Model SolutionsFinalDegradationRate dmldt (%/min) = 0.1182 exp (-16300 (± 1600)/RT(K)) [4.55]The activation energies in both cases are considerably lower than the peroxide 0-0 bondenergy, 146.5 kJ/mol. This result indicates that other reactions are involved or that masstransfer is involved in the ageing experiments. The activation energy for the initialdegradation rate is also considerably lower than the activation energy for the initial foulingrate (84 kJ/mol). The experiments did show that polyperoxide deposits will age undertemperatures typical of the fouling runs to produce the deposits reported in Section 5, thusconfirming the ageing hypothesis.4.8 Summary of Autoxidation StudiesThe batch kinetic studies confirmed that the oxidation of model solutions of indeneis an autocatalytic process and identified several characteristics which influence theperformance of batch fouling experiments performed under these conditions.Indene reacts with oxygen under these conditions to form yellow-amberpolyperoxide gums [(C9H802)H] as a major product. The solvent can interfere with theindene cooxidation reaction and prevent the formation of polyperoxides, or reduce the yieldof these products via chain transfer. The solubility of the polyperoxide products in aliphaticsolvents was found to be limited by the solvent nature; the solubility limit, g*, increasedwith solution aromaticity and temperature and is thought to be a physical parameter of thesystem. The soluble gum consisted of 2-4 peroxide units and FTIR indicated hydroxyl andcarbonyl activity. Polyperoxide was precipitated as globules of darker orange gum after thesolubility limit was reached. The thermal degradation of these gums was studied in an inertatmosphere at 160-240°C to simulate the ageing process on a heat transfer surface. Thegums melted to form dark redlbrown solids with increased carbonyl activity. The rate of1514. Autoxidation of Model Solutionsmass loss increased with temperature; activation energies of 40 kJ/mol and 16 kJ/mol wereobtained for the initial rate and final rate respectively.The disappearance of indene featured a chemical induction period whichcorresponded to the accumulation of hydroperoxides in solution. This induction period waseliminated by the use of chemical initiators and extended by the use of an antioxidant,BMP. The antioxidant was consumed during the induction period and its products did notinfluence the subsequent rate of reaction significantly. The activation energy of indeneinitiation was determined as 120 kJ/mol, which is in good agreement with the literature.The rate of indene autoxidation after the induction period was limited by oxygenmass transfer to solution. The process was modelled as mass transfer followed by a fastchemical reaction which was zeroth order in oxygen; the experimental data fitted thisscheme reasonably well. Temperature, initiation method, indene and oxygen concentrationswere all found to have significant effects on the rate of indene reaction and gum formation.The autoxidation of chemically initiated solutions of 5 wt% indene in Paraflex wasstudied in further detail. This model solution gave reasonably reproducible results but thekinetics were found to depend on the apparatus involved.A model of indene autoxidation was proposed which predicted the trends observedin the experimental data. Further experimental work is required to verify that the model is atrue representation of the mechanism of indene autoxidation in these model solutions.1525. Initial Fouling Experiments5. Initial Fouling ExperimentsAsomaning’s study of autoxidation fouling (1990) used model solutions of alkenesin kerosene and demonstrated the usefulness of using chemically simpler systems toinvestigate the role of different chemical species. Asomaning did not investigate the effectof temperature or flow rate, and adopted the solution recipe (10 wt% alkene in kerosene)used by Taylor (1969). This model solution presented various problems in analysis andoperation so a range of model solutions was studied in order to select a candidate for use inthe TFU experiments, where flow velocity and surface temperature would be theparameters varied. These initial experiments were performed in the PFRU and theinformation collected proved to be invaluable in the safe and reliable operation of the TFU.The scope of these initial experiments was subsequently extended to include the effects ofchemical initiators, antioxidants, flow velocity and surface temperature. The PFRU studiesthus constitute the major part of the current work.5.1 Model Solution SelectionAsomaning’s experiments used 10 wt% solutions of alkene in kerosene andidentified three alkenes which produced significant fouling under the experimentalconditions (Tblk 70-85°C, Tsurf = 180-205°C, Re = 9830); indene, dicyclopentadieneand hexadec-1-ene. Indene and hexadec-1-ene presented fewer problems in operation thanDCP, so DCP was not considered further at this stage. Hexadec- 1-ene was a strongcandidate for model solution studies as its autoxidation had been modelled by Norton andDrayer (1968). The four solvents selected were kerosene and three alternatives: an aliphatic(Paraflex), an aromatic (tetralin) and a polar aromatic (trichlorobenzene). The interaction ofthe different solvents and alkenes provided significant insights into the chemical processesinvolved in fouling in autoxidative systems.1535. Initial Fouling ExperimentsThe initial model solution experiments are summarised in Table 5.1. Theexperiments were usually started under the thermal conditions used by Asomaning; solutionsamples were taken and frozen for chemical analysis after the experiment. The soluble gumassay was not developed at this stage and GC analysis was developed during this period.The solvent densities varied so a common concentration of 12.5 v/v% (10 wt% inkerosene) was used in the first experiments. The Reynolds number also varied so the flowrate used by Asomaning was used where possible. The values of Re in Table 5.1 arecalculated using the expressions described in Section 3.1 and feature a smaller kerosenedynamic viscosity than that used by Asomaning.Blank runs of solvent were run for 48 hours in order to establish the thermalstability of kerosene, Paraflex and tetralin. No fouling was observed in Paraflex orkerosene and the POx values showed minimal increases after 48 hours. No fouling wasobserved in tetralin but chemical analysis indicated that significant autoxidation hadoccurred. Trichlorobenzene was not run as a blank and this solvent caused contaminationproblems in the PFRU which were repeated in the SCR. Trichlorobenzene developed agrey colour after equilibrating at 80°C overnight and the fouling runs generated depositswhich were very different from indene experiments in the other solvents. GC analysis fromrun 030 indicated that little indene had been consumed although heavy fouling hadoccurred. The deposit contained some chlorine, even though trichlorobenzene was thoughtto be inert to autoxidation under these conditions. Extensive cleaning was needed after thetrichlorobenzene experiments and thus trichlorobenzene was abandoned as a solvent. Thenature of the contamination reaction was not considered further.5.1.1 Tetralin as SolventNo fouling was observed in solutions of indene or hexadecene in tetralin, even afterthe surface temperature was raised to 234°C in run 007. A very thin gum was noticeable on154Table5.1SummaryofInitialFoulingExperimentsUiUiExperimentSolventAlkeneThulk(mol/L)(°C)T5fUoqReDurationFouling(°C)(W/m2K)(kW/m2’)(hours)CommentsI N001Paraflex-84.8202913108.7305048no003tetralin-79.92001375172.01072148noColourchange011kerosene-82.01811368145.61069093no004tetralinhexadecene83.22041398173.01072148nothingumlayer(0.388)•hexadecene83.02001404170.01072124nothinmattefilm007tetralin(0.205)78.22341404212.81072124no008tetralinindene82.82051470185.01072148no(0.507)016tnchloroindene82.31801250120.61360027yesContaminationbenzene(0.951)030trichloroindene80.01801250121136007.5yesContaminationbenzene(0.41)012kerosenehexadecene83.21801368137.21069072no(0.388)101.71801521112.414080120no102.72401521221.514080178yesinitiatoradded014kerosenehexadecene106.12451640227.11408041yesinitiatoraddded(0.388)013keroseneindene82.71831361138.01069033yes(0.878)002Paraflexhexadecene85.8204937113.6305048noDatalogger(0.466)78.2281937197.9305072yesfailure005Parallexhexadecene83.02501398161.0431824no(0.388)83.0280943197.0431845no009Paraflexhexadecene100.22501076157.33050122yesFlowRateloss(0.388)duetosampling006Paraflexindene81.4211910119.630508.3yes(0.840)Allexperimentsperformedat377kPaairoverpressure5. Initial Fouling Experimentsthe PFRU heated section after these experiments but it did not represent a significantfouling resistance. The GC analyses of the samples from these runs indicated that bOthsolvent and alkene were undergoing reaction. This was evident from the solution colourwhich changed from being colourless to a dark orange/red colour. The chromatographsshowed that the reduction in area of the tetralin peak (tres = 3.1 mm) over time wasaccompanied by the growth of a new product peak = 5.6 miii), whilst FTIR analysis ofsolution samples showed a corresponding increase in carbonyl activity. These results areconsistent with the autoxidation of tetralin reported in the literature, where tetralin is firstoxidised to tetralin hydroperoxide and thereafter to the ketone (tetralone) or other sideproducts. Figure 5.1 shows the peroxide analysis data. The Peroxide Number increasedrapidly to an equilibrium level; this behaviour is consistent with the autoxidation of tetralinreported in the literature. The equilibrium POx value and the rate of tetralin disappearanceboth decreased when the concentration of alkene was increased. The plateau hydroperoxideconcentrations do not correlate with the initial concentrations of tetralin and suggest that thealkenes inhibited the autoxidation of tetralin. This is probably due to chain transfer betweenthe more reactive tetralin peroxy radical and the alkene; Russell (1955) investigated thecopolymerisation of tetralin and cumene and reported similar results. Tetralin, as a morereadily oxidised species, thus acts as a fouling inhibitor as it interrupts the formation of thealkene polyperoxide chains thought to cause fouling.The tetralin results illustrate the complexity of real systems where many compoundscan interact depending on their relative concentrations and activities.5.1.2 Hexadec-1-ene as DopantAsomaning (1992) reported significant fouling from a solution of hexadecene inkerosene but this result could not be reproduced. Asomaning’s method differed from thecurrent work in that the alkene was added to the solvent before the system was warmed up1565. Initial Fouling ExperimentsFigure 5.1 Peroxide Number Analyses from Initial Fouling Experiments using ModelSolutions with Tetralin as Solvent2000..‘C‘C15OOx1000 x‘Cn500•x00 000 0 00U— I0 10 20 30 40 50 60Time(hours)•- pure tetralin; o - 0.507 M indene; x - 0.388 M hexadecene;- 0.205 M hexadecene1575. Initial Fouling Experimentsand pressurised, whereas in the current work the alkene was added to the solution after thesystem had equilibrated at the operating conditions. Fouling was observed in kerosene andParaflex when the surface temperature was increased, following a long induction period.The deposit formed was brittle and easily disturbed, apart from the material on the (hot)probe surface. This material was almost black in colour and vigorous cleaning was neededto remove it. The end of the fouling induction period (when thermal fouling becamesignificant) corresponded to a maximum in the Peroxide number, as shown in Figure 5.2for hexadecene in kerosene (run 014). The figure shows that the fouling induction periodwas controlled by the bulk chemical reaction. Norton and Draye?s hexadecene autoxidationmodel links the maximum in Peroxide Number to the generation of significantconcentrations of polyperoxide. The chemical analysis and fouling results thus supportAsomaning and Watkinson’s hypothesis that fouling in autoxidation systems is caused bypolyperoxides.Further hexadecene experiments were performed using an enhanced bulktemperature, (100°C), and a chemical initiator, 0.OIM benzoyl peroxide, to shorten theinduction period. Chemical reaction control of the induction period was demonstrated afterrun 009, where the fouled PFRU probe was quickly cleaned, replaced in the system andthe experiment restarted. A much shorter induction period (10 hours) was observed and theprobe fouled completely in 20 hours. Extended runs such as run 009 often featuredreductions in flow rate as liquid sampling eventually depleted the liquid level in the holdingtank; this factor and the physical strain of extended experiments highlighted the need forshorter experiments.Quantitative analysis of hexadecene concentration was not available at this stage.1585. Initial Fouling ExperimentsFigure 5.2 Fouling Resistance and Hydroperoxide Concentration Profiles from Foulingof Hexadecene in Kerosene in Run 014. c0.0002p0.00010.00040.000310985-i 76zd5i2 320 0.00002500Time(minutes)Run 014: 0.388 M hexadec-1-ene in kerosene: Tbulk 106.1°C, Tsurf 245°C: Pair= 373 kPa0 500 1000 1500 20001595. Initial Fouling Experiments5.1.3 Indene as DopantFouling was observed in the model solutions of indene in kerosene and in Paraflex.The solutions developed a noticeable orange colour and a very strong, ketone-type smellduring the course of the run. The red/orange coloured gum reported by Asomaning wasobserved on glass surfaces in contact with the liquid when fouling was occurring on thePFRU heated section. A fouling induction period was observed in both solvents, whichcorresponded to a maximum in the Peroxide Number. The fouling resistance-time plotsthen followed an accelerating profile. Figure 5.3 shows the Peroxide Number and foulingresistance results from run 013 in kerosene. The induction period for indene in kerosenewas significantly longer than that observed in Paraflex, but the Peroxide Number andfouling resistance profiles were similar. Measurement of indene concentration in kerosenewas not qualitative at this stage; quantitative indene analysis in Paraflex was underdevelopment and showed a decrease in indene concentration during the run.ETIR analysis of liquid samples from indene and hexadecene solutions in keroseneand Paraflex showed many common features. Broad hydroxyl peaks of low to mediumstrength were visible in the 3200-3600 cm1 range once the Peroxide Number started toincrease. These could be caused by hydroperoxides, acids (i.e. hydroperoxide dissociationproducts) or water generated by condensation steps. The most noticeable absorbances werefound in the 1600-1800 cm1 range, corresponding to carbonyl groups formed from thedecomposition of hydroperoxides. These absorbances were complex and difficult toassign, but confirmed that the product slate was consistent with autoxidation occurring insolution.Indene was selected as the dopant for further study as lower surface temperatureswere needed to generate significant fouling resistance measurements. The autoxidation of160) c:’45. Initial Fouling Experiments0.0012Figure 5.3 Fouling Resistance and Hydroperoxide Concentration Profiles from Foulingof Indene in Kerosene in Run 01330252015r0.001010I,..— ‘-ci50.0006-------0.0004000.0002200 400 600 800 1000 1200 1400 1600 1800Time(minutes)Run 013: 0.878 M indene in kerosene: TbuIk 82.7°C, Tsurf= 183°C: pair 373 kPa0.00001615. Initial Fouling Experimentsindene was also more extensively reported than that of hexadec-1-ene. Hexadec-1-enepresented greater problems for quantitative chemical analysis than indene and indenemethods were being further developed for a parallel study of indene fouling in keroseneperformed by G. Zhang.5.1.4 Deposit CharacterisationThe deposits formed on the PFRU heated section were photographed then removedfor further analysis. Figure 5.4 is a photograph of the fouled PFRU probe after run 014with hexadecene in kerosene which shows the general pattern of deposit found in thePFRU experiments. The unheated sections of the probe were usually free of any deposit,except in cases where significant amounts of orange gum were observed in solution or onglass surfaces. Under these circumstances the unheated sections featured a streaky coatingof a yellow/amber veneer which resembled the gum found in the solution samples.Deposition ended abruptly after the heated section. Figure 5.5 is the heated section depositthickness profile from run 012 (0.388 M hexadecene in kerosene) measured using verniercalipers. The deposit thickness increases as the local surface temperature increases and thethermal boundary layer becomes fully developed. The deposit profile was compared withthermal entry length data at similar Re; the experimental thermal entry length seems toextend further along the probe than expected. The PFRU probe is limited in this case as allthe thermocouples are located at the same axial position. The nature of the deposit alsochanges with axial position; as the distance from the thermal entry point increases and thesurface temperature increases, the deposit changes from a veneer to a cracked, powderysolid. Indene deposition profiles were harder to measure as the deposit was generallysofter and was compressed or removed by the vernier caliper’s action. Indene depositswere usually formed at lower surface temperatures and were not as carbonaceous as thehexadecene deposits. A thickness measurement probe (Elektro-Physic Köln Minitest162Figure 5.5IFigure 5.4 Photograph of Fouled PFRU Probe Following Fouling Run 014Thermally Initiated Autoxidation of 0.388 M Hexadecene in Kerosene5. Inilia! Fontin Experiments__AIlDWJIJJIII1IJ1}II1llujIITWVEF-itNV 1tOl ‘AtEca4 INDeposit Thickness Profile of PFRU Probe Fouled by the Autoxidation of0.388M Hexadecene in Kerosene., IHealed. SectionInletFLOW—>400350300250200150100501End ofHeatedSection0 I • I • 1 • I • • I • I • I0 10 20 30 40 50 60 70 80 90Axial Distance(mm)Run 012: 0.388 M hexadec4-ene in kerosene: Tbujk = 102.7°C, Tsurf = 240°C: Pair = 373 kPi1631005. Initial Fouling Experiments2000) based on the eddy current principle was tested but a suitable calibration wasunavailable.The composition of some of the initial deposits is given in Table 5.2 along with anestimate of their mean thermal conductivity based on the approximation Rj= whereOf is the deposit thickness. The indene deposit analyses in both solvents were inreasonable agreement with the values reported by Asomaning and Watkinson (1992) andthe calculated values for indene polyperoxide. The analysis results for the hexadecenedeposits show considerable scatter and do not compare well with the calculated values forthe polyperoxide. The calculated values of the deposit thermal conductivity ( 0.2 W/m.K)are an order of magnitude smaller than those given by Asomaning and Watkinson, but arein the range reported for amorphous organic deposits by Watkinson (1988). Values of Ajgreater than 1 W/m.K are associated with more carbonaceous materials.The FTIR spectra in Figure 5.6 show the chemical activity of the PFRU depositfrom run 013 and a sample of the yellow/amber gum precipitated from solution in the sameexperiment. The broad band at 3100-3600 cm1 in the gum confirms the presence of theOH group which could be caused by -OOH. Both spectra show significant carbonylactivity. The deposit carbonyl distribution is complex and indicates various C=O sources.The other peaks are consistent with the breakdown of indene polyperoxide. Thecomposition and FIIR results indicate that the foulant is an aged product of polyperoxide,as suggested by Asomaning and Watkinson.5.1.5 Initial Mechanistic InsightsThe study of candidate model solution combinations highlighted the importance ofthe chemical reaction in chemical reaction fouling. The initial fouling experimentsdemonstrated that a compound likely to undergo autoxidation to form fouling precursors inone solvent could be inhibited by another solvent or competing species.1645. Initial Fouling ExperimentsTable 5.2 Chemical Analysis of Deposits From Initial Fouling RunsRun Solvent Alkene C H 0 Rf final(wt%) (wt%) (wt%) (m2.K/W) (W/m.K)002 Paraflex hexadecene 71.95 5.02 23.03 - -009 Paraflex hexadecene 79.88 10.31 9.81 0.0020 -012 kerosene hexadecene 66.33 2.85 30.82 0.0044 0.19012 kerosene hexadecene 65.16 3.20 31.64 0.0044 0.19(gum)014 kerosene hexadecene - -- 0.0051 0.18015 kerosene hexadecene - -- 0.0028 0.23hexadecene 75.0 12.5 12.5 - -polyperoxide006 Paraflex indene 76.09 5.32 18.59 0.0010 0.19013 kerosene indene 82.19 6.33 11.48 0.0011 0.19indene 73.0 5.4 21.6polyperoxide1655. Initial Fouling ExperimentsFigure 5.6 FT’IR Spectra of PFRU Deposit and Soluble Gum from AutoxidationFouling of Indene in Kerosene in Run 013.ilt if•’ iiIII ri1 1P r1-fl- I PeroxideGum i’lli1‘1‘ I -ft m]f11}4____I JI— ——: ,J ff J14 -4- 4- 1r32G 2400 2800 1688Frequency (cm-i)-1 1 U] Deposit 11i [I -‘j4irJ I;in 1f 194 jl—:9 [[f I ‘- -mF1TrII WflC)U —S:r- 1T- --1 - Frequency(cm1) -3288 2400 2000 1688 1208 800 401665. Initial Fouling ExperimentsThe use of chemical analysis in the PFRU fouling experiments provided qualitativeevidence that autoxidation occurs in solution during the fouling runs. The analysesconfirmed that the fouling induction periods observed in Asomaning’s experiments werecaused by the bulk chemical reaction. The onset of fouling is linked to a stage in the bulkreaction associated with a maximum in the Peroxide Number. Norton and Drayer’s modelof hexaclecene autoxidation (1968) suggests that the maximum in hydroperoxideconcentration corresponds to the appearance of polyperoxide as the major product. Thedelay between the end of a chemical initiation period, marked by the increase in thePeroxide Number, and the increase in fouling resistance confirms that the fouling precursoris not a primary product of indene or hexadecene autoxidation. These observations supportAsomaning and Watkinson’s hypothesis that fouling was caused by polyperoxides ratherthan hydroperoxides. The deposit analysis results were consistent with a polyperoxidefoulant precursor model.The use of a chemical initiator introduced a possible method to eliminate lengthyinduction periods if the reaction mechanism can be preserved; Russell (1955b) reported thesame reaction pathway for thermally and AIBN initiated indene autoxidation. The use ofinitiators is discussed further in Sections 4.3 and 5.5.The lack of an aromatic solvent in which to pursue further fouling studies isunfortunate as this would provide a solvent in which indene (the selected dopant)autoxidation products would be most soluble. The role of insoluble species in chemicalreaction fouling was reviewed in Section 2.3.2 and an aromatic solvent would haveprovided valuable information in this area.5.2 Effects of Dopant ConcentrationA series of experiments was performed in order to study the effect of indeneconcentration on the fouling process. These runs were performed under identical thermal1675. Initial Fouling Experimentsand chemical conditions using both kerosene and Paraflex as solvent; this facilitated furthercomparison with Asomaning’s experiments and later work at UBC and ANL using theindene/kerosene system.Four concentrations of indene in kerosene (0.84, 0.74, 0.40, 0.15 molfL) were runat a bulk temperatures of 80°C, a surface temperature of 180°C and Re=10700 under 79.2kPa oxygen saturation (as air). All concentrations produced significant fouling deposits andshowed similar behaviour to Figure 5.3. The results from both solvents are summarised inTable 5.3. The chemical induction period was defined as the time at which the PeroxideNumber started to increase. The Peroxide Number showed sequential features in commonwith prior autoxidation research, i.e. induction period, linear increase, acceleration,maximum concentration, decrease in concentration. The maximum value varied between25-35 meqfL (12-18 mmol/L hydroperoxide) but did not show a strong dependence oninitial indene concentration. The maxima were accompanied by the appearance of an orangegum in solution which adhered to glass surfaces and the rotameter float. The chemicalinduction period preceded the fouling induction period and decreased as indeneconcentration increased. The fouling induction period coincided with the maximum inPeroxide Number; the fouling resistance curves followed an accelerating profile after themaximum and did not show any evident variation in form with initial indene concentration.Four concentrations of indene in Paraflex were studied (0.15,0.41,0.68,0.71mol/L) under the same thermal conditions, at Re 3050. The chemical induction periodswere shorter than in kerosene and again depended on initial indene concentration. ThePeroxide Number and fouling resistance data from run 025 (0.41 molJL indene) are plottedin Figure 5.7 and show similar features to the kerosene runs except that the PeroxideNumber shows a gradual rise to a plateau level rather than a maximum. The plateauPeroxide numbers (25-40 meqlL) did not vary significantly with initial indene concentrationand were similar to those reported in kerosene. The orange gum seen in the kerosene runswas also evident in the Paraflex experiments. The fouling resistances measured at higher1685. Initial Foulinf ExverimentsTable 5.3 Effects of Indene Concentration in Fouling ExperimentsRun Solvent jndenej Chemical d(VPOx)/dt [indenel Rate constantInduction rate constant fitPeriod kR(mol/L) (lv) /(meq/L)/hr ‘(mo1/L)/hr (R2)013 kerosene 0.84 21 ±1 0.450 (0.56,0.36) - -020 kerosene 0.71 22 ±1 0.398 (0.34,0.42) - —021 kerosene 0.35 83 ±1 0.215 (0.18,0.24) - —023 kerosene 0.14 91 +3 0.138 (0.12,0.15) - —027 Paraflex 0.74 3 ±1 0.335 (0.32,0.34) 0.01332 0.986024 Paraflex 0.68 10 ±1 0.325 (0.30,0.34) 0.01 152 0.922025 Paraflex 0.41 26 ±2 0.220 (0.20,0.23) 0.00793 0.995026 Paraflex 0.15 20 ±1 0.155 (0.14,0.16) 0.00546 0.985All runs at Tblk 80°C, Tsuff 180°C; Re (Paraflex) = 3050; Re (kerosene) = 11 000; P02 = 79.2 kPaFigure 5.7 Fouling Resistance and Peroxide Number Profiles from Indene in ParaflexF — —. Peroxide NumberI —-•---- Fou’ing Resistance1000 2000 3000[minutes]Run 025: 0.41 M indene in Paraflex; Tbulk = 80°C; Re = 3050; oxygen4035300.001825200.0014150.001210I0.0010a0.000650.0008000.0004Time_:.0.000240000.0000= 79.2 kPa; Tsurf = 180°C1695. Initial Fouling Experimentsindene concentrations did not increase uniformly but fluctuated after an initial uniformincrease; later inspection showed that considerable deposition had occurred. This isdiscussed further in Section 6.5. The similarity in results suggested that the samemechanism controls fouling in these non-polar, aliphatic solvents.The accelerating fouling resistance profiles under constant flow and surfaceconditions must be due to an increase in the concentration of foulant precursor. This wasnot reflected in the Peroxide Number, however, and prompted the development of the gumassay.The hydroperoxide induction period data was compared with a simple kineticmodel. Indene cooxidation occurs in the absence of significant hydroperoxideconcentration, giving the rate expression;-d[02]Idt = k1[RHj1.5 [O2]° [5.11where k1 is an overall initiation rate constant. If the end of the chemical induction period isdue to the buildup of hydroperoxide to a critical concentration, [ROO1I]*. whereunimolecular decomposition becomes dominant;kD[ROOH]* = k1[RH][02] [5.2]where kD is an initiationldecomposition rate constant. Assuming that all oxygen reactedappears as hydroperoxide (i.e. zero polyperoxide or epoxide formation and lowconversion), Equation [5.1] gives[ROOH]*= k1 [RHI1.5 [02]Ot [5.3]Equations [5.2] and [5.3] suggest that the chemical induction period, r, should beproportional to ./([O] [RH]), which is counter intuitive in its oxygen dependence.Assuming a constant critical value of [ROOHJM gives x inversely proportional to [RH]1.5and a more realistic oxygen dependence. A plot of the chemical induction period data inTable 5.3 did not fit either of these simple models; the induction period decreased linearly1705. Initial Fouling Experimentswith indene concentration. The departure of the fouling system kinetics from the literaturemodels prompted the study of indene autoxidation described in Section 4.The initial increase in Peroxide Number was found to obey Equation [4.36] and theplots of s/[ROOH] against time in Figure 5.8 show the expected linear dependence on/[RH]. The results from the thermally initiated autoxidation of indene in Paraflex in theSCR at 100°C showed the same trend as observed in the PFRU. The Peroxide Numberdiverged from Equation [4.36] as it approached the maximum, as reported in the SCRstudies in Section 4.1. Figure 5.8 indicates that the rate of indene autoxidation in Paraflexand kerosene is effectively equal following the induction period.The indene concentration data from the Paraflex experiments showed that indeneconsumption increased significantly once the Peroxide Number had started to increase,confirming that thermal initiation gives way to hydroperoxide-initiated oxidation. The datafitted the zeroth order kinetic model reasonably well but the values of kR in Table 5.3increase slightly with indene concentration. This variation was not seen in the rate constantsfor first order indene consumption, kND, (around 0.035 hr’).The discrepancies between the results from the indene concentration runs and theliterature prompted a rigorous study of autoxidation in these model solutions. Theautoxidation of indene proved to be similar in kerosene and Paraflex, with differences inthe chemical induction period. The extended induction periods observed in kerosene werecaused by inhibiting additives or components in the kerosene. Paraflex was thus chosen forfurther fouling studies on the basis of shorter induction periods, a wider range of operatingtemperatures and ease of chemical analysis. The larger viscosity resulted in smallerReynolds numbers for given flow rates but this is less important for ‘heavier’ petroleumfluids. The variation in fouling behaviour with indene concentration prompted the use ofone concentration, namely 5 wt% indene (0.41 mob’L).1715. Initial Foulin2 ExperimentsoI-Figure 5.8 Kinetic Fit of Peroxide Data from Initial Indene Fouling Runs to Eqn [4.36]0.500.450.400.350.300. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0‘Ilnitial indene concentration, /[RHJV’(mo1/L)1725. Initial Fouling Experiments5.3 Fouling in Model Solutions with Two DopantsThe tetralin solvent studies indicated that an active solvent could alter the foulingprocess by inhibiting the generation of fouling precursor. There was little published workdescribing the opposite case, namely the interaction of two active species which wereknown to cause significant deposition individually. It is very difficult to study thesynergism of different reaction mechanisms as autoxidation requires chemical conditionswhich preclude vinyl polymerisation (i.e. high oxygen concentrations). Indene anddicyclopentadiene (DCP) were thus used to study the interaction of two active speciesunder autoxidative conditions.Table 5.4 summarises the results from the three experiments performed underidentical flow conditions and thermal initiation. The gum analysis method was developedafter run 028 but the gum concentration in Paraflex could be deduced from the PeroxideNumber results; other indene runs in Paraflex at 80°C indicated that the increase in gumconcentration lagged the Peroxide Number by about an hour. The DCP experimentsproved difficult to perform owing to its very strong odour, and caused several difficultiesin chemical analysis. The gum produced in runs 03 3,4 was not readily soluble indichloromethane, which compromised the accuracy of the GC analysis method; DCP alsocaused a side reaction in the Peroxide Number method.Figures 5.9 and 5.10 show the fouling resistance and Peroxide analysis resultsfrom the runs described in Table 5.4. The chemical induction period in all three runs issimilar but the fouling characteristics differ greatly. The DCP-only run fouling resistance isalmost linear against time and started to increase once gum was generated in the bulksolution. The final increase in fouling rate corresponds to the Peroxide Number reaching aplateau level at 38 meq/L. The gum concentration had reached 4.3 gIL at this point, whichwas less than g* observed for indene at the same temperature. GC analysis indicated that17% of the DCP had reacted by the end of the experiment, which is considerably less than1735. Initial Fouling ExperimentsTable 5.4 Fouling from Solutions of Indene and Dicyclopentadiene in ParaflexAlkene Concentration Run First order rate Chemical Fouling Depositconstant induction period induction period AnalysiskND(mol/L) (1/hr) (hi-) (hi-)indene 0.410 028 0.0354 5-8 17 ±0.5 C9H54074DCP 0.321 033 0.0094 6.5 ±1 9 ±0.5 12867(polyperoxide) (C9H120)indene 0.415 034 0.0340 5.5 ±1 19 ±0.5 C128054DCP 0.139(soluble gum) (C9H10387)PFRU ; Tsurf 80°C; Tsurf 210°C; Re 3050; thermal initiation; 79.2 kPa oxygen overpressure1745. Initial FoullnxpeimentsFigure 5.9IfTJ4IFouling Resistance Profiles in Indene/DCP studiesTime(minutes)0 200 400 600 800 1000 1200 1400 1600 1800 2000Figure 5.10 Peroxide Number versus Time in Indene/DCP studies50403020100o indene• DCP•indene/DCPA••A8°•• A8A6AA.A r - - - I i - - I - - i - - - - - -0 200 400 600 800 1000 1200 1400 1600 1800 2000Time(minutes)1755. Initial Fouling Experimentsthe consumption during the indene run (50%). Asomaning (1990) also observed linearfouling curves with DCP in kerosene. The indene fouling profile shows a transitionbetween an initial, linear fouling rate and the accelerating fouling profile observed inSection 5.2 at lower surface temperatures. The start of the linear fouling profile coincideswith the Peroxide Number increasing less rapidly and ends when the Peroxide Number hasreached its plateau value of 24 meq/L. The DCP linear fouling rate is significantly largerthan the indene rate, despite a smaller initial concentration, in agreement with Asomaning’sresults.The differences in the fouling characteristic profiles of indene and DCP indicatesthat the fouling mechanism is sensitive to solute/solvent/polyperoxide factors. Zhang et al.(1993) and Asomaning (1993) subsequently performed fouling experiments using 10 wt%indene in kerosene with gum and indene analyses. The fouling resistance profiles weresimilar to the DCP result observed in Paraflex, where deposition was detected when thesoluble gum concentration in the bulk started to increase. These results suggest that thefouling mechanism for indene in kerosene is linked to bulk deposition rather than a surfacereaction process. Solvent effects on fouling behaviour require further study.Figure 5.9 shows that the deposition from the alkene mixture was different to thesingle alkene solutions. 0.139 M DCP was added to 0.41M indene to observe the effectson the indene kinetics rather than trying to maintain a constant wt% or mol fraction ofalkene. The hydroperoxide and gum concentration profiles and the first order rate constantfor indene disappearance were similar to those observed in the indene run. The DCPconcentration did not change significantly during the experiment. The fouling resistance didnot increase until the Peroxide and gum concentrations had reached their plateau values,after which deposition was very fast. No fouling was observed during the initial increasein gum concentration, contrary to both single alkene run results.Micrographs of the fouling deposits showed the foulant to be composed of twotypes of material; a closely packed matrix of small orange/yellow particulates (6-20 jim)1765. Initial Fouling Experimentsnearer the heated metal surface and a surface layer of randomly scattered, orange/red,larger (50-70 jim) particulates, described as ‘blobs’. Some yellow blobs were also seen onthe cooler sections; when removed and heated in an oven at 200°C, these blobs resembledthe red/orange analogues observed on the hotter section. The material nearest the heatedsurface was harder, most resistant to removal and insoluble in the acetone which dissolvedalmost all the other deposit. FuR analysis of the deposits all featured the chemical activityassociated with polyperoxide decomposition discussed in Section 5.1.DCP and indene both caused significant fouling in Paraflex. The foulant nature andthe sequencing of the fouling resistance and chemical reaction indicate that the foulingprocess is dominated by the generation of insoluble polyperoxide gums in the bulksolution. The fouling mechanism differed between the two alkenes and their combinedfouling behaviour was not the sum of individual contributions. The two species have tocompete for the oxygen available in solution, which would affect the chemical reactiongenerating the fouling precursor. Further synergism studies were postponed until a reliablefouling model was defined in order to interpret the results of such experiments.5.4 Temperature Effects in Thermally Initiated FoulingThe effects of surface temperature were studied in the PFRU once a suitable modelsolution had been selected. Fouling experiments were performed under identical chemicaland flow conditions (0.4] M indene in Paraflex; Tb1k 80°C; 79.2 kPa oxygenoverpressure; thermal initiation; Re 3050) and the initial surface temperature was variedbetween 180-240°C. The results are summarised in Table 5.5.The chemical analysis methods provided a monitor of the bulk reaction and Table5.5 shows that there is some variation in the reaction parameters with surface temperature.The gum analysis method was developed during this series of runs and the gum resultswere similar in the two runs involved. The solubility limit, g*, is considered to be a1775. Initial Fouling ExperimnetsTable 5.5 Temperature Effects in Thermally Initiated Fouling ExperimentsInitial Surface Run Indene Rate Chemical Initial Fouling Final GumTemperature Constant Induction Rate Fouling Rate SolubiltykR Period Limit, g*[°CJ [‘J(molfL)/hr] [hours] [m2.KJW.min] [m2.K!W.min] [gIL]180 025 0.00793 41.0 ±2 i.0xi0 10.5x -200 031 0.00861 24.0 ±2 1.9x107 7.3x105 4.8 ±0.7210 028 0.00849 8.5 ±1.5 5.0x107 6.9x105 -225 032 0.00766 6.5 ±1 5.7x10 8.7x105 5.0 ± 0.7240 029 0.00723 6.0 ± 0.5 14.4x 10- -PFRU Runs with Re 3050; Tbulk 80°C; 0.41M indene in Paraflex; 79.2 kPa oxygen overpressure1785. Initial Fouling Experimentssolvent-related property and did not vary with surface temperature. The other reactionparameters show that the bulk autoxidation reaction is coupled to the heat exchangeroperating conditions. This would be expected under thermal initiation, where the hotterPFRU surface is a significant source of free radicals. The chemical induction period, -r,shows a much stronger dependence on surface temperature than the indene rate constant,kR. Figure 5.11 is an Arrhenius-type plot of the data in Table 5.5 where rates are plotted inthe absence of clearly defined rate constants; the rate of radical formation is represented byil-u and gives a pseudo-activation energy of 53 kJ/mol. Zhang et at. (1993) reported avalue of 16 kJ/mol for induction periods in PFRU fouling runs using solutions of 10 wt%indene in kerosene under similar conditions. Both values are lower than the activationenergies for the respective induction period obtained from the SCR experiments in Section4.5. The variation of induction period with PFRU surface temperature is due to thecoupling of reaction and heat transfer in the PFRU apparatus and is considered further inSection 7.1.An alternative to thermal initiation for further studies was needed which avoided thecoupling of the bulk reaction with PFRU operating conditions and the extended inductionperiods at lower surface temperatures. This prompted the investigation of free radicalinitiators described in Section 5.5.The fouling resistance profiles are plotted in Figure 5.12 and indicate the twofouling regimes discussed in Section 5.3. The abscissa, adjusted time, refers to the periodafter evident chemical reaction has started (given by the chemical induction period, -u). Thetransition between the two regimes is dramatic; the fouling resistance increases initially in alinear fashion then accelerates to a rapid rate, when the run was terminated to save theprobe. The initial fouling rates in Table 5.5 were obtained by drawing a line through thedata points following the time when the fouling resistance had become significant(‘-j4 x1795. Initial Fouling ExperimentsFigure 5.11 Temperature Effects in Thermally Initiated Fouling Experiments in Paraflexo0 Initial Fouling Rate-6 A Final Fouling Rate-8 • 1/Induction Period-10I-12 A A A- 4-180.0019 0.0020 0.0021 0.0022 0.0023llTsurface (K)Figure 5.12 Effects of Surface Temperature on Fouling Resistance Profiles in ThermallyInitiated Solutions of Indene in Paraflex - Runs 025, 028, 029, 031, 032c).0020Surface0.0018 Temperature0.00 16 225°C0.0014 a 200°C 1.1)0.0012 + 180C ++210°C0.0010 : +a +240°C •+B * +0.0008 B +• a01 0. 00060.00040.0002 +,0.0000 I I +0—- 250 500 750 1000 1250 1500 1750 2000Adjusted Time(minutes)1805. Initial Fouling Experimentsi0-m2.KJW in Figure 5.12). The data in this region was reasonably linear and estimatesof fouling rate could vary by ± 20%, or more at lower temperatures where the fouling ratewas small. The final fouling rates are calculated from the last data points in the acceleratingstage. Figure 5.11 shows that the final fouling rates do not show a strong dependence onsurface temperature while the initial fouling rates fitted a modified Arrhenius equation;dRf/dt = 548 exp (-84 800IRT) (R2 = 0.953) [5.4]The activation energy for the fouling process, 84.8 ±13.2 kJ/mol, is at the higher end ofthe range reported for chemical reaction fouling in the literature. The value indicates thatchemical reaction steps are involved in the formation of deposit. Zhang et al. (1993)reported an activation energy of 39 kJ/mol for the initial fouling rate from solutions of 10wt% indene in kerosene under similar conditions, while Crittenden and Khater (1984)reported an activation energy of 70 kJ/mol for evaporation fouling caused by keroseneautoxidation. The magnitude of the final fouling rate and its weak dependence on surfacetemperature indicate that it is controlled by conditions in the bulk liquid rather than thesurface reaction zone.The differences between the initial fouling regime and the accelerating regimeindicate that different processes are involved in the generation of deposit. These processesare likely to be linked to changes in the chemical reaction as autoxidation continues in thebulk liquid. The importance of the bulk chemical reaction prompted the investigation ofmodel solution autoxidation described in Section 4. Figure 5.13 shows how the change indeposition mechanisms is related to the bulk solution chemistry. No fouling is observedprior to generation of gum in the bulk liquid, indicating that the surface reaction zone is notthe only region involved in the formation of deposit. The linear, initial fouling regimecorresponds to the increase of gum concentration to the solubility limit, g*, and beyond;some time after the solution has reached g*, the accelerating regime begins and heavyfouling follows. The heavy fouling regime alone is observed at lower surface temperatures,presumably because the reaction zone is not hot enough to generate significant deposition.1815. Initial Fouling EperimenIsRun 028: T8 225°C; Tbulk 80°C; Re 3050; 0.41M indene in ParaflexP0.0008Figure 5.13 Gum Concentration and Fouling Resistance in Thermally Initiated Fouling6 0.0010543T’l t200.00060.00040.0002Adjusted Time(minutes)0.000012001825. Initial Fouling ExperimentsThe variation in the initial fouling rate with inclene concentration could not be reliablydeduced from the runs in Section 5.3 as these were performed at relatively lowtemperatures (180°C).The transition in fouling mechanisms was confirmed by examining the depositformed at the different surface temperatures. All deposits showed the changes indeposition pattern with surface temperature described in Section 5.1. The foulantmorphology depended on the surface temperature and the length of a run. Lower surfacetemperatures, where fouling followed an accelerating profile alone, produced very chunky,soft deposits. Optical microscopy indicated that the surface was covered at random by largeblobs of a dark red material in a disordered structure, with greater surface coverage athigher surface temperature. This material lay on top of a hard dark brown deposit whichwas densely packed and much harder to remove. The chunky material was soluble inacetone but the hard material had to be removed mechanically. The blobs ranged indiameter from 50 to 100 jim and looked as if they had melted in to the deposit. The largerblob dimensions are comparable to the depth of the thermal boundary layer in theseexperiments and suggest that these larger blobs are agglomerates of insoluble gum formedin the bulk solution. Deposit from the enhanced surface temperature runs was significantlydifferent; the foulant at the hotter end consisted of an ordered array of small gumparticulates, of mean size 8 jim, densely packed together in a yellow/orange red matrix.Higher resolution microscopy showed this material to be quite porous and particulate innature down to the dark amorphous solid formed at the PFRU surface. Agglomerated gumparticles could be seen to lie on top of this denser matrix in runs where accelerating foulingalso occurred. The veneer formed on the less hot PFRU surface suggested that blobs hadadhered to the surface and melted in to the veneer. The foulant morphology thus indicated adeposition process involving precursor solubility, as hypothesised by Crittenden et al.(1987b) in their polymerisation fouling studies. Estimates of the deposit thermal1835. Initial Fouling Experimentsconductivity using Af = b/Rf ranged from 0.1-0.2 W!m.K, in agreement with earlierresults.Table 5.6 summarises the data from elemental analysis of the deposits. The C:H:Oresults are compared with the result for indene polyperoxide, C9H802the foulantprecursor. The deposits from kerosene and Paraflex were similar in composition as well asmorphology. The deposit shows reduced oxygen content, particularly in the materialexposed to the higher temperatures near the probe wall. FTIR analysis of the depositsindicated no hydroperoxide functionality but a complex mixture of aromatic carbonylgroups generated by the degradation of peroxide linkages. Deposit recovered from theprobe wall contained less oxygen and gave very complex FTIR spectra, indicating that thefoulant undergoes an ageing process once deposited. The FTIR spectra were very similar tothose reported by Lambourn and Durrieu (1983) in their studies of crude oil fouling in thepresence of oxygen. The deposits from the chemically initiated runs are discussed in detailin Section 5.5.The fouling resistance data and the deposit morphologies indicate that fouling inthese batch autoxidation experiments involves a transition between two depositionmechanisms;1. Linear Fouling Regime Polyperoxide formation and attachment involvingbulk and reaction zone phenomena;2. Accelerating Fouling Bulk precipitation of polyperoxide, dominated bybulk kinetics and mass transferTransitions in fouling mechanisms due to solubility effects has been reported by Fryer et al.(1990) in milk fouling, where the generation of whey protein foulant in the reaction zone issurpassed by the generation of agglomerates in the bulk when the bulk temperature reachesa critical value for protein denaturation. Fouling from asphaltenes in crude oil has beendescribed as a solubility phenomenon by Eaton and Lux (1984) and by Scarborough et al.1845. Initial Fouling ExperimentsTable 5.6 Elemental Analysis of Fouling DepositsIndene Initiation Solvent Surface Elemental Origin of DepositConcentration Temperature Analysis[mol lU [°Cj0.40 thermal kerosene 180 C9 H727 0130 Deposit surface0.41 thermal Parallex 180 C9 H967 °0.97 Less hot surface180 C9H101 01.16 Deposit surface180 C9 133 0056 PFRU Wall0.41 thermal Paraflex 240 C9 H98 01.77 Gum from Solution240 C9H898 00.87 Deposit surface240 C9 Hg97 00.75 PFRU Wall0.41 chemical Paraflex 240 C9H104 Oi 300 minutes0.41 chemical Parallex 240 C9H924 01.03 400 minutes0.41 chemical Parallex 240 C9H913 °094 480 minutes- -.. (C9 H8 02)nH Indene polyperoxideTable 5.7 Effects of Bulk Chemical Parameters on Fouling and Reaction BehaviourRun Tbulk Re Initiation Induction kR Gum Yield g* Initial Fouling DepositPeriod ((mol Rate (x107) Analysis(°C) (hr) 7L)ihr) (g/g) (giL) [m2.KlW.minj028 80.0° 3050 thermal 14.5 0.00849 - - 5.01 C9H109080031t 80.0° 3050 thermal 24.0 0.00861 0.59 4.8 1.92 C9H10308035 100.0° 3295 thermal 1.6 0.01574 0.50 9.3 5.6 C9H108087038 80.7° 3050 2.5mMABN 1.0 0.01041 0.45 5.5 3.01 C9H106O20039 80.8° 3050 2.5mM bP 1.0 0.01439 0.41 5.0 2.59 C9H104O26042 90.0 3011 2.5mM bP 0.0 0.02188 0.63 9.0 7.27 -049 90.0 3011 2.5mM bP 0.0 0.02342 0.59 10.0 6.25 -040 100.0 3295 2.5mM bP 0.0 0.02869 0.61 11.5 10.8 C9H10125t Tgf 200°C, included for diagnostic comparisons1855. Initial Fouling Experiments(1979), who reported deposits generated from small organic particulates. Further work wasconcentrated on mechanism (1) as this corresponds more closely to conditions incommercial heat exchangers. There are considerable experimental difficulties involved infurther study of mechanism (2) owing to the insoluble nature of the precursors.5.5 Velocity and Surface Temperature Effects inChemically Initiated FoulingChemical free radical initiators were investigated as a means of eliminating thechemical induction periods and scatter in reaction parameters observed in the thermallyinitiated experiments. This could also be achieved by increasing the bulk liquidtemperature, which would in turn increase the operating Reynolds number away from thetransition regime threshold (Re 2300). Table 5.7 is a summary of the results from aseries of runs performed to seek an optimum set of operating conditions for further foulingstudies. The optimum criteria were chiefly those of experimental feasibility; negligibleinduction period, reasonable run lengths and fouling behaviour similar to that observed inSection 5.4. The runs were performed using solutions of 0.41 mol!L indene in Paraflexunder 79.2 kPa oxygen overpressure in the PFRU operating at a surface temperature of210°C except where noted and at Re 3000. The SCR experiments had showed that 2.5mmol/L was a suitable concentration of initiator and this was used for the rest of the initialfouling experiments. Reaction diagnostics are reported as described in Section 4.Both chemical initiators (benzoyl peroxide, ABN) eliminated the chemical inductionperiod and increased the rate of reaction, as observed in the SCR. The initiated runs at 80°Cgave similar initial fouling rates to the thermally initiated run but the experiments weredeemed to be long, lasting over 20 hours. The initiated runs at 90°C were repeated becausethe PFRU gave unusual fouling resistance profiles; Rf increased linearly while the gum1865. Initial Fouling Experimentsconcentration increased and levelled off, as observed previously, but then decreased orfluctuated rather than accelerating. This was thought to be due to spalling but inspection ofthe probe showed that a heavy deposit layer had collected; this deposit contained largeragglomerates of insoluble gum than observed previously and presumably yielded a negativefouling resistance via surface roughness effects as described for particulate fouling byCrittenden and Alderman (1988). The accelerating Rf profile was not observed at 90°Cdespite several attempts. The runs at 100°C showed the same trends observed at 80°C; thechemically initiated runs involved increased reaction rates and slightly higher fouling rates,while the fouling resistance profile remained similar. The initial linear fouling rate operateduntil after the gum solubility limit, g*, was reached and was followed by the acceleratingprofile. The increased reaction rate at higher temperatures meant that the initial rate periodwas shorter than that observed under thermal initiation. These results confirmed thatoperating at 100°C with chemical initiation did not change the initial fouling rate mechanismwhilst yielding shorter runs with advantages in control and Re range.The reaction parameters listed in Table 5.7 and the raw data showed that theautoxidation reaction in the PFRU followed the same trends observed in the SCR inSection 4. Chemically initiated solutions featured enhanced indene reaction rates andslightly lower gum yields compared to thermally initiated runs. Activation energies forthermal initiation were estimated for kR (39.5±4 kJ/mol) and the initial gum formation rate(56 ±11 kJ/mol); these activation energies are larger that those reported in the SCR. Thistrend was also seen in the activation energies under benzoyl peroxide initiation; 52±6kJ/mol for kR and 59±9 kJ/mol for the initial gum formation rate. The SCR studies alsoreported increased activation energies for initiated solutions. The error estimates are quitelarge and reflect the difficulty in achieving experimental reproducability in a larger reactorwith different indene batches. The presence of the hot fouling probe could also beresponsible for the larger activation energies observed in the PFRU Fouling runs. The gumsolubility limit, g*, shows reasonable reproducability at each bulk temperature irrespective1875. Initial Fouling Experimentsof the mode of initiation. One notable feature was the tendency for the gum level toovershoot g* in later fouling experiments; this irregularity was not seen under thermalinitiation and suggested that a lower initiator concentration would be preferable in thesubsequent TFU runs.The foulant composition and morphology were similar across the range ofconditions used and confirm that the same fouling mechanism is involved. The sequence offouling events was also similar to that observed in Section 5.4, though the timescale wasshortened as reaction rates increased. At higher bulk temperatures, insoluble yellow gumtended to collect on the unheated surface of the PFRU probe; this was soluble in acetoneand its FEIR spectrum indicated that it contained indene poiy- and hyciroperoxides.The initial fouling rates in Table 5.7 increase with bulk temperature, which excludesa simple surface reaction fouling model. No further conclusions were drawn as the surfaceshear stress and residence times also varied in these runs. Panchal and Watkinson (1993)reported similar results for indene in kerosene, although their experiments were performedat constant Urn rather than Re.Bulk temperatures of 100°C and 2.5 mM benzoyl peroxide initiation were thus usedin subsequent fouling studies. Each set of runs used the same batches of indene andParaflex to reduce variations in the bulk reaction caused by different indene batchesobserved in the SCR. Table 5.8 sunimarises the reaction diagnostics; the mean initial gumrate was 2.5 ±0.2 g/L.hr, mean kR was 0.0237 ‘./(molIL)!hr and the gum yield varied from0.52-0.77.Figure 5.14 is an Arrhenius type plot of the initial fouling rates obtained from runsat Re = 3295 where Tsurf ranged from 210°-255°C. Figure A.2.1 shows the foulingresistance profiles from the individual runs. The initial fouling rates were calculated asbefore and increase with surface temperature, fitting an expression of the formdRf!dt = 977 exp [- (81 900 ±16 400)/RT(K)] [5.5]1885. Inital Foulin,ç’ ExperimentsModel solution: 5wt% indene in Parallex at 100°C; 79.2 kPa oxygen blanket; 2.5 mM bP initiator:z!. -,lIT surface (K)Thermal initiation- Tbulk 80°C, 100°C: Chemical initiation- Tbulk 100°C, 2.5mM benzoyl peroxideTable 5.8 Reaction Diagnostics from Chemically Initiated PFRU Fouling ExperimentsRun # Tsface Re [indene]0 k4 Gum Yield R2 Tbulk(molIL) ((mol/L)(CC) /hr) (g/g) (kM) (°C)040 207.1 3295 0.331 0.02869 0.613 0.998 100.4043 224.6 3295 0.340 0.03325 0.772 0.987 100.4044 239.5 3295 0.361 0.03297 0.524 0.996 100.2050 247.4 3295 0.301 0.02205 0.535 0.988 100.3055 252.4 3295 0.295 0.02226 0.736 0.990 100.4049 254.6 3295 0.365 0.02342 0.682 0.992 100.6045 222.1 4920 0.372 0.02526 0.527 0.994 100.6047 218.0 6513 0.368 0.02741 0.603 0.998 99.9056 246.7 1020 0.365 0.02485 0.700 0.984 100.0054 247.0 1920 0.302 0.02271 0.593 0.977 100.2053 246.7 6513 0.3 12 0.02496 0.630 0.979 100.2Figure 5.14 Surface Temperature Effects on Initial Fouling Rate in Thermally andChemically Initiated Fouling Runs-10• Chemical Initiation (iO0C)1] 0 Thermal initiation (80°C)Zhang et al. (kerosene)Thermal Initiation (1 00°C)- s ••-14- N.&N0-15- ‘N0-’-16--17-—18 —‘——‘ I .. . . ‘_ ..0.0018 0.000 0.002 0.0024 0.00261895. Initial Fouling ExperimentsThe activation energy is similar to the value reported for thermally initiated fouling (84.8kJ/mol) but is subject to a larger estimate of error due to the scatter in the data. The initialfouling rates from thermally initiated runs at TbUlk 100°C were higher than those reportedat 80°C at the same surface temperature but less than those reported in the chemicallyinitiated runs. The increase in initial fouling rates could be due to the increased rate ofreaction in the bulk liquid, the reduced Urn used at 100°C or the change in solutionproperties with bulk temperature. The flow velocity at 100°C (0.506 m/s) was smaller thanthat at 80°C (0.687 mIs) but the ratio of chemical/thermal fouling rates is 3.75, whichsuggests that other factors are involved. The experimental data indicates that the initialfouling rate is linked to the bulk reaction rate; Panchal and Watkinson (constant Urn) and theruns described above (constant Re) both reported increased initial fouling rates withincreased bulk temperature and hence bulk reaction rate. This result would be consistentwith a reaction zone fouling model such as those described by Panchal and Watkinson, andPaterson and Fryer. Figure 14 also shows the data reported by Zhang et al. (1993) forfouling from solutions of lOwt% indene in kerosene in the PFRU under similar chemicalconditions (TbUlk 86°C, thermal initiation, 86 kPa oxygen overpressUre) but at Re11000. The activation energy in the kerosene solution was reported as 39 kJ/mol, which issignificantly smaller than that observed in Paraflex.The deposits obtained from the chemically initiated runs showed the samemorphologies and composition trends as observed in the thermally initiated runs. The runsat highest surface temperature involved greater surface coverage, consistent with thedevelopment of a thermal boundary layer and a local surface temperature effect indeposition. Photo-micrographs of the deposits confirmed the link between the onset ofaccelerated fouling and the appearance of large dark red particulates of insoluble gum.The effect of flow velocity was studied at two different surface temperatures (—222and 247°C) using different batches of indene in each series. The initial and final fouling1905. Initial Fouling Experimentsrates defined in Section 5.4 are summarised in Table 5.9 and plotted on log axes in Figure5.15. The final fouling rates are an order of magnitude greater than the initial fouling rates,as observed under thermal initiation. Figures A.2.2.a-d show the fouling resistance andgum concentration profiles for each series and confirm that the onset of accelerated foulingcoincides with the gum level reaching the solubility limit, g*, of 11 gIL. The initial foulingrates decrease with increasing flow velocity while the final fouling rates do not, confirmingthe change in deposition mechanism. There is significant scatter in the initial fouling rates,which reflects the scatter in the bulk reaction diagnostics seen in Table 5.8. The figureshows that the rates could be fitted to Red, where -2< n <-1. Zhang et al. (1993) observedsimilar scatter in their data from indene in kerosene and found that n -1. The initialfouling rate was thus proportional to exp (-E/RTsurfe) u ; Paterson and Fryer’s modelpredicted n = -1, whereas Epstein’s model (1993) predicts n -1.75 when chemicalreaction controls the fouling process. Fouling models are discussed further in Section 7.2.The final fouling rates at different surface temperatures should not be compareddirectly as they are limited by the maximum surface temperature on the PFRU probe. It isevident, however, that the final fouling rate increases as Re exceeds 3000 and the flowenters the turbulent transition regime. Bulk deposition processes are expected to be fasterunder such conditions due to the contribution of turbulence to mass transfer rates.The deposit morphology did not change noticeably across the range of flow ratesstudied and deposit analyses confirmed that the foulant chemistry was consistent with theageing of polyperoxide gums. The thermal entry length on the PFRU probe was shorter atlarger Re, increasing the proportion of the probe surface covered by deposit.5.6 Stages in Autoxidation FoulingA series of runs was performed under identical chemical and flow conditions inorder to study the sequence of events in the fouling experiments. These were performed1915. Inital Fouling ExperimentsFigure 5.15 Log-log plot of Velocity Effects in Chemically Initiated Fouling1000Final, 222°C0 Final 240°C 0 00100A__ __A A___- A.o A Initial, 222 C• Initial, 240°C1000 10000Reynolds NumberTable 5.9 Effect of Flow Rate on Initial Fouling Rate in Initiated PFRU ExperimentsRun# Re urns Tsurf Uref kR Initial Fouling Final FoulingRate Rate(x107 (x107)(mis) (°C) (W/m2.K) (rnol/L)/hr (m2.K!W/min) (m2.K/W/rnin043 3295 0.506 224.6 790 0.03325 33.3 154±20220 27.2t045 4920 0.756 222.1 1018 0.02526 9.3 200±20047 6513 1.000 218.0 1245 0.02741 5.6 200±20050 3295 0.506 247.4 814 0.02205 15.2 90±10053 6513 1.000 246.7 1262 0.02496 7.4 150±20054 4920 0.756 247 1053 0.02271 10.3 158±20055 3295 0.506 252.4 800 0.02226 26.7 87± 10247t 21.9t056 4020 0.617 246.7 922 0.02485 12.0 169±15t - estimated using activation energy 83 kJ/mol1925. Initial Fouling Experimentsunder severe fouling conditions (Tblk 100°C, Tsurfo.ce = 240°C, Re 3295) usingthermally initiated solutions of 0.41 mol/L indene in Paraflex. The reaction diagnostics aresummarised in Table 5.10 and show similar scatter in reaction diagnostics as those inSection 5.5. This scatter is reflected in the initial fouling rates in Table 5.10The runs were stopped at the start of the initial fouling rate period; at the end of theinitial fouling period, when the gum concentration had reached g*, and during theaccelerated fouling period. No fouling was observed until gum was observed as cloudinessin the bulk solution. Figures 5.16.a-c are photomicrographs of the surface deposit in theheavily fouled section and show distinct changes in deposit morphology during the foulingprocess. The deposit in Figure 5. 16.a appeared as a smooth, even veneer to the naked eyebut is clearly composed of striations of gum in the direction of flow. These striations arecomposed of globules of orange gum (diameter 14-22 jim) with occasional larger globules(48 jim) or agglomerates of smaller globules. The striation separation ranges from. 18 to 22jim. The deposit in Figure 5. 16.b appeared as a smooth red/brown material and can beseen to consist of many gum globules (15-20 jim) which have adhered to the surface andmelted into the deposit. More large globules of insoluble gum (30 jim) are visible. Thedeposit in Figure 5. 16.c was a relatively uniform, rough, dark brown solid; the figureshows that this consists of larger globules of red insoluble gum and agglomerates (size 44-65 jim) on top of a fine particulate matrix (size 10 jim) which is presumably the agedpolyperoxide gum deposited earlier. Scraping off the earlier veneers often uncovered thinlayers of this orange matrix material. Figure 5. 16.d shows the smooth orange veneer foundin the thermal entry length of the 500 minute run; melted gum globules are evident in theveneer. The surface beneath shows the darkened (aged) early striations and the scratchedsurface of the PFRU probe. The scratches in the veneer surface occured during cooling andwere observed in the ageing studies in Section 4.7These interrupted runs confirmed that the change in fouling mechanism is due to theonset of bulk precipitation of insoluble gum agglomerates. The deposit formed in the initial1935. Initial Fouling ExperimentsTable 5.10 Reaction Diagnostics and Fouling Summary of Interrupted Fouling RunsDuration Induction kR Gum Fouling Final Initial Fouling Deposit [indene]Period Yield Induction Rf Rate Analysisv’(mol/ Period (x103) (x107)(minutes) (minutes) (°C) L)lhr) (minutes) (m2JJr (m2K/Wmin) (moIIL)300 120 ±50 237.7 0.0302 0.464 240 ±18 0.06 7.69 C9H1041 0.400400 120 ±30 237.4 0.0213 0.439 240 ±18 0.10 5.67 C9H240103 0.360500 90±30 236.6 0.0337 0.446 180 ±15 0.75 6.66 C9H9130J94 0.401Tbulk 100°C; Re 3295; P 79.2 kPa oxygen overpressure; thermal initiation1945. In i/ia! Fouling ExpertmenIsFigure 5.1 6a,b Photographs of PFRU Deposit from Interrupted Fouling Runs usingThermally Initiated Solutions of Indene in ParaflexFig. 5. 16 a (t = 300 minutes, 272x)Fig 5. 16 b (t = 400 minutes, 272x)‘J:w__’• ----..-,.. Li..:-—c.. —._s ...‘— -. .. ‘ ‘- .— .-* -..__-J3• _4. --‘.——.--—.——-‘1955. Initial Fouhin2 ExperimentsFigure 5.16c,d Photographs of PFRU Deposit from Interrupted Fouling Runs usingThermally Initiated Solutions of Indene in ParaflexFig 5.16 c (t 500 minutes, 272x)Fig 5.16 d (t = 500 minutes, 272x)5. Initial Fouling Experimentsstages of fouling consisted of smaller gum globules or veneers which were derived fromgum melting into the deposit as part of the ageing process. The initial fouling mechanismthus involves gum insolubility and the adhesion of gum globules to the surface. Onceadhesion occurs, the polyperoxide gum ages to the darker coloured carbonyl materialsdiscussed earlier. The analysis results in Table 5.10 show that the deposit is low in oxygencompared to its parent polyperoxide gum at even the earliest stage of deposition. This isconsistent with an adhesion step involving a chemical reaction, which explains the largeactivation energy reported for the fouling rate. The estimated values of A1 ranged around0.18 W/m.K, which agreed with the values reported earlier and suggested that the ageingprocess did not alter the deposit thermal conductivity.The globule sizes in the initial fouling phase (<20 jim) can be compared with theestimated momentum and heat transfer boundary layer thicknesses. The former can beestimated fromy = 5, giving Omom = 153 jim; the latter from O1her,1 = A/U0, where A isthe thermal conductivity of the liquid, giving ôtIzernwl = 148 jim.; the latter figure isexpected to be smaller as Pr> 1. The gum globules could thus be formed in the viscoussublayer or the bulk solution. Several attempts were made to measure the size of gumglobules in solution but these were complicated by the tendency of the gum to agglomerateon cooling. Insoluble gum globule sizes were found to lie in three ranges; 20-24 jim, 45-70 jim and 110+ jim, the latter including very large gum agglomerates up to 1 mm. Thesoluble gum size estimates seemed to fit a similar distribution.5.7 Fouling in the Presence of AntioxidantsThe ability of antioxidants to suppress autoxidation in model solutions of indene inParaflex at 100°C was demonstrated in Section 4.5. The effectiveness of BMP in theharsher operating conditions of a heat exchanger was unknown, although the literature doesrefer to antioxidant ceiling temperatures above which the antioxidants become relatively1975. Initial Fouling Experimentsineffective. Lloyd and Zimmerman (1965) studied the effect of 2,6-di-butyl-4-cresol on theautoxidation of cumene and found that above the effective ceiling temperature, the inductionperiod varied as the log of the initial cresol concentration. Below the ceiling temperature (at126.5°C), the induction period varied with the square root of the antioxidant concentration.Morris et at. (1988) studied the antioxidation performance of an analogue of BMP andfound that it was less effective than metal ion inhibitors at temperatures above 100°C. Theeffect of antioxidants on fouling rates after the induction period was also unknown as theliterature usually measures antioxidation effectiveness in terms of the induction periodalone. This was of particular interest to the fouling runs performed in kerosene as thesefeatured extended induction periods caused by commercial additives.A series of fouling runs was performed under identical fouling conditions in thePFRU (Re = 3300, T5,.f = 240°C) using a model solution of 0.41 mol/L indene in Paraflexat Tblk = 100°C and 79.2 kPa oxygen overpressure in air. Thermal initiation was used inthe presence of 0, 50, 100 and 200 ppm BMP. The results are summarised in Table 5.11and the gum analysis data are plotted in Figure 5.17. This figure and the data in the tableshow that there is little effect of BMP addition to the reaction diagnostics after the chemicalinduction period. The scatter in the reaction diagnostics is consistent with that observed inother thermally initiated experiments. GC/PID analysis showed that the chemical inductionperiod ended when the BMP was exhausted. The chemical reaction in the PFRU systemthus behaved similarly to the antioxidation experiments in the SCR.Figure 5.18 shows how the fouling resistance profiles behaved differently at higheradditive concentrations. The initial fouling rates at lower BMP concentrations did not showany significant change in fouling behaviour. At 200 ppm BMP addition, fouling startedbefore autoxidation in the bulk solution and featured the formation of a thin grey deposit onthe hottest part of the heat exchanger. The fouling rates reported in this region are smallerthan the initial fouling rate observed in the period when the gum concentration increasestowards g*. This grey material was covered by the orange and red/brown indene1985. Initial Fouling ExperimentsTable 5.11 Fouling in the Presence of an Antioxidant (2,6,di-t-butyl 4-methyiphenol)[BMPJ Reaction Fouling Fouling Initial kR Gum DepositInduction Induction Regime Fouling Rate [R2] Yield AnalysisPeriod Period {x107j(ppm) (hours) (hours) (rn2.K!W.min) (mol/L)/hr0 2 ±0.5 3.5 ±0.5 increasing 13.0 0.03368 0.464 C9H03097gum 0.98650 10 11.5 increasing 12.3 0.03217 0.560 C9H350108gum 0.993100 17t 18 increasing 13.1 >0.0171 0.503 C9H88305gum200 25.5 12 pre-gum 1.58 0.03114 0.509 C9H668011522.5 increasing 13.1 0.992gum200 21 ±1 13.3 pre-gum 4.67 0.02549 0.677 C9H0009521.8 increasing- 0.989gumt - estimated value1995. Initial Fouling ExperimentsFigure 5.17 Autoxidation Fouling in the Presence of Antioxidant; Gum ConcentrationsFigure 5.18 Fouling Resistance Profiles in the Presence of an Antioxidant0.00100.00090.00080.00070.00060.00050.0004X 0.00030.0002ri-4 0.00010.000018000 0 ppm 50 ppm O 100 ppm 200 ppm 200 ppm1200 A o10 A080 El67 bI El‘_-. .—.4 0AEl2Time(hours)20 25 300 300 600 900 1200 1500Time[minutes]2005. Initial Foiling Experimentspolyperoxide deposit after the bulk reaction started. Deposit analysis did not show anysignificant differences, which is expected as the BMP oxidation products would be difficultto determine from the polyperoxide products unless isotopic labelling of the BMP wereused. An interrupted run would have yielded further information in a complete study of thephenomenon. Contamination was dismissed following thorough cleaning and a repeatedrun. This early, linear fouling regime was not observed at lower BMP concentrations and isthought to be caused by reaction of the BMP or its oxidation products at the heat exchangersurface.Figure 5.19 is a comparison of induction periods in the SCR and the fouling runsand shows that the heat exchanger affects antioxidant performance. Chemical inductionperiods are shorter in the fouling system and are no longer proportional to the initial BMPconcentration. The chemical induction periods are expected to be reduced by the enhancedtemperatures in the heat exchanger but the change in concentration dependence indicatesthat effects similar to those reported by Lloyd and Zimmerman are also involved. Thefigure shows a logarithmic curve fitted to the PFRU induction period data and the fit isreasonable. These results are consistent with the effects of the heat exchanger operatingconditions on the chemical induction period observed in Section 5.4. Further experimentswould be required to investigate antioxidant ceiling temperatures but this was considered tobe beyond the scope of the current study.The antioxidant experiments confirmed that at low concentrations, phenolicantioxidants such as BMP extend the chemical reaction induction period but do not affectthe subsequent autoxidation and fouling rates. The kerosene used in the fouling runs likelycontains low levels of phenolic inhibitors to resist atmospheric degradation; the inhibitorsare thus unlikely to affect the fouling rates and autoxidation kinetics. The foulingexperiments also demonstrated that antioxidants have to be selected for the duty required.BMP is unsuitable for use under the fouling run conditions and an operator would actually2015. Initial Fouling ExperimentsFigure 5.19 Effect of Antioxidant (BMP) on Reaction and Fouling Induction Periods5045 - 0 PFRU Reaction• SCR Reaction -,40PFRU Fouling35.0 50 100 150 200 250[BMPJo(ppm)0.41 M indene in Paraflex: TbuIk 100°C : 377 kPa air overpressure: Tsurf (PFRU) = 240°C2025. Initial Fouling Experimentsmagnify the fouling problem if the BMP dosage were increased in order to suppressautoxidation. The selection of suitable antioxidants has already been described as acommercially sensitive area. The combination of SCR and PFRU experiments doesdemonstrate, however, the pitfalls in extending simple batch tests to antioxidantperformance in a plant. Conversely, the understanding of the mechanisms involved inautoxidation fouling gained in these experiments suggests that the existing fuel stabilityinformation can be used to identify streams likely to form gums and foul heat exchangers.5.8 Summary of Initial Fouling ExperimentsCombinations of two dopants and four solvents were tested in order to identify asuitable model solution for use in fouling studies. These model solution experiments werealso used to develop suitable experimental procedures and chemical analyses for furtherfouling experiments. A solution of 5 wt% (0.405 mol/L) indene in Paraflex was identifiedas the optimum model solution for fouling studies.The chemical analyses confirmed that autoxidation was occurring in the bulk liquidand controlled the fouling behaviour in the batch fouling experiments. The foulantprecursors were identified as soluble and insoluble polyperoxides generated by thecooxidation of the alkene, confirming the hypothesis of Asomaning and Watkinson (1992).Parameters which influenced the formation of polyperoxides in the bulk liquid had similareffects on the observed fouling behaviour; this result could be used to rationalise previousfouling studies and help to relate fuel stability studies’ findings to heat exchanger fouling.Solutions of 5 wt% indene in Paraflex exhibited two distinct regions of foulingresistance behaviour; an initial, pseudo-linear fouling rate region followed by a later,accelerating region where the fouling resistance increased at a markedly larger rate. Thetransition between the regions coincided with the polyperoxide gum concentration reachingthe solubility limit, g*, after which time globules of insoluble gum were found in solution.2035. Initial Fouling ExperimentsThe initial fouling rate increased with surface temperature and decreased with increasingflow velocity, whereas the later fouling rates appeared to increase with flow rate and didnot show any systematic variation with surface temperature.The transition in fouling behaviour was shown to correspond to a change in foulingmechanisms, from a surface reaction zone process to the deposition of insoluble gumglobules. This transition was confirmed by a series of interrupted fouling experiments.These runs found that the deposit was composed of small particulates of insoluble gum,ranging in size from 7-22 jtm, at even the earliest stages of fouling. These particulatesmelted to form veneers on cooler regions of the heat transfer surface. Chemical analysisindicated that the gum aged soon after adhesion occurred. The deposits were depleted inoxygen and contained strong carbonyl group activity consistent with a thermal degradationageing step.The chemical induction periods observed in thermally initiated solutions of indenein Paraflex and kerosene were affected by the conditions in the fouling probe. Chemicalinitiators were used to eliminate the chemical induction period and to ensure reactionreproducibility. The chemical initiators increased the reaction rate and also the initial foulingrate.The fouling behaviour of mixtures of two active alkenes was studied using indeneand dicyclopentadiene. The chemical analysis and fouling results showed non-additiveeffects which were attributed to the competition for available oxygen in solution.The effect of an antioxidant on fouling behaviour was studied under severe foulingconditions using a commonly used gasoline antioxidant, BMP. The surface temperatureconditions were above the ceiling temperature for BMP reported by Lloyd and Zimmerman(1965) and the antioxidant efficiency was reduced markedly. The end of the extendedchemical induction period coincided with the exhaustion of the antioxidant. The2045. Initial Fouling Experimentsautoxidation kinetics and fouling behaviour following the end of the chemical inductionperiod showed little difference from the thermally initiated runs at low BMP concentrations.Unusual behaviour was observed at larger BMP concentrations.2056. TFU Fouling Studies6. Fouling Experiments in the Tube Fouling Unit6.1 Autoxidation of indene in the TFUThe effects of surface temperature and flow velocity on the initial fouling rate anddeposit morphology in the TFU were studied using a model solution of 0.405M (5 wt%)indene in Paraflex initiated by 1 mM benzoyl peroxide at Tblk 100°C and Poxygen 72kPa. Lower Pair and initiation levels than in Section 5.5 were used in order to extend theinitial fouling rate duration and ensure repeatable chemical reaction in the fouling apparatus.The same batch of indene was used in the fouling runs and associated SCR experiments.Figure 6.la-c show the Peroxide Number, gum and indene concentration resultsfrom batch reactions performed in the SCR, PFRU and TFU. The air flow rate wasmaintained at excess oxygen (300 mL/Lsoln/min) in all cases. The figures show similarbehaviour in all three reactor configurations.The kinetic parameters are given in Table 6.1 with the results from the fouling runsin the TFU and PFRU under the same chemical conditions (except P02). The high indeneconcentration in run 065 was caused by an experimental error. The autoxidation of indeneproceeds faster in the TFU and PFRU than in the SCR, despite the discrepancy in oxygenpartial pressure conditions. The discrepancy in autoxidation rates is thought to be caused bydifferences in oxygen mass transfer between the units. Runs 501 and 502-12 did not showany significant effect of the heated section on the reaction kinetics; individual fouling runkinetics were subject to some variation, though this was than less than observed in thePFRU runs using 2.5 mM bP initiation. The chemical analysis results confirmed thatreasonably reproducible reaction was achieved in the TFU Fouling runs.Figures A.2.3a-c and A.2.4a-c show the Peroxide Number, gum and indeneconcentration results from the TFU fouling runs and the kinetic parameters from these runsare summarised in Table A.2.1. The gum profiles show considerable scatter after t,2066. TFU Fouling StudiesFigure 6. la-c Comparison of Indene Autoxidation in SCR, TFU and PFRU0 SCR 14550 • SCR 147TFU5O1 040-PFRIJO65 0830A20 e100I.,,,a.POx 0 1 2 3 4 5 6 7 8 9 102018 145 SCR16 147 SCR14 501 TFU12 065 PFRU 2 A AAA A10 A 08 A•O64 A.00°2 AA0I•G• I...Ib.[gurnj 0 1 2 3 4 5 6 7 8 9 100.45k- 0 147SCR040- A A • 145 SCR0.35 A A 501 TFU0.30 -.A 065 PFRUfl2c.0.20-A00.15- 00.10-0.05 -0.000 1 2 3 4 5 6 7 8 9 10c. [indene] Time(hours)2076. TFU Fouling StudiesTable 6.1 Comparison of Indene Autoxidation Kinetics under TFU ConditionsExperiment Reactor Tbulk P02 kR Initial Gum Yield g*Rate____________ ____________(°C) (kPa) i/(mol/L)/hr (g/L.hr) (g/g) (gIL)145 SCR 100 79 0.02692 2.0 ±0.2 0.65 10.5 ±1147 SCR 100 79 0.02723 2.1 ±0.1 0.52 11.5±1501 TFU 100 72 0.03253 3.1 ±0.1 0.57 11 ±0.5502-12t TFUt 101.7 72 0.03254 2.7 +0.1 0.53 11.5 ±1(avemged)065 PFRUt 100.2 79 0.03996 3.6 ±0.1 0.49 12 +10.405 M indene in Parallex; air flow rate 300 mLIL.min (ntp); - fouling experiments;- 0.43 M2086. TFU Fouling Studies(around 300 minutes), which is caused by the entrainment of insoluble gum in solution.This scatter in g* was not observed in the SCR tests.The TFU batch reactor run (501) was also used to assess operating procedures andtube processing methods. An electric grinder was used to cut the 501 test section intodifferent lengths but the metal dust caused errors in the deposit mass assay. The tubesurface was marked by striations of yellow/orange gum which dissolved readily in acetone.The striations ran parallel to the fluid flow and were approximately 0.5mm wide andregularly spaced (separation 1-2 mm). The deposit mass coverage increased rapidly fromzero over the first 15 cm then increased gradually with axial distance. The striations arecaused by precipitation of polyperoxide gum at the unheated tube surface, where T5,f<Tb1k due to the small insulation losses. The gum solubility limit, g*, is a strong function oftemperature and polyperoxide gum is thus precipitated at the surface. This form ofcrystallisation fouling was responsible for the loss of cooling capacity observed in thefouling runs, where the cooling utility was mains water at 12-15°C. The striationphenomena are discussed further in Section Initial Fouling StudiesExperiments 503 and 504 were performed under identical chemical and foulingconditions 200°C, W 0.105 kg/s) to study the reproducibility of TFU foulingdata. Run 502 also involved these conditions but was terminated by a power failure after300 minutes.Good agreement was found in the chemical analysis results and in heat transferperformance. The overall fouling resistance profiles from both runs are plotted in Figure6.2 and the agreement is well within the estimated error in calculating Rf. The overallfouling resistance, Rf0, was calculated using data from those thermocouples which were2096. TFU Fou1ing StudiesFigure 6.2 Fouling Resistance Profiles Showing TFU ReproducibilityIo 503• 5040.000090.000080.00007,_. 0.000060.00005Q0.000040.00003z00.000020.000010.00000-0.00001•0000o .00OOo•0• o’9o O•o 1’•0 O%%0o 00oo o• o00•...••0 000•00.00 100 200 300Time(minutes)400 500 6002106. TFU Fouling Studiesconsidered to be downstream of the thermal entry length. The data exhibited more scatterthan that observed in the PFRU runs due to magnetic field interference and the effect ofaveraging over the whole heated section. The runs confirmed that the TFU gavereproducible results.The fouling resistance profiles in Figure 6.2 show a transition from an initially lowfouling rate to a significantly larger fouling rate at approximately 300 minutes. The initialrise in fouling resistance was caused by the system adjusting to the change in heat transfercoefficients after indene had been added. The system data were averaged over the first 25minutes in order to give the initial values. Figure 6.3 shows the fouling resistance and gumconcentration profiles for run 503 and confirms that the transition in fouling rates (andpresumably mechanisms) coincides with the gum concentration approaching g*. All theTFU fouling runs showed the same relationship between gum concentration and foulingmechanism as reported in the PFRU fouling runs in Section Tube Pressure DropThe pressure drop across the tube changed during a fouling experiment as materialdeposited on the tube wall narrowed the flow channel and affected the roughness of thetube surface. The variation of the pressure drop with time confirmed the transition infouling mechanisms from a surface reaction zone process to the deposition of insolublegum. Figure 6.4 shows the tube pressure drop and soluble gum concentration data fromrun 505. The tube pressure drop decreased slightly during the initial fouling rate period (t <t*) but increased significantly after t. Similar behaviour was observed in all fouling runs.Deposition increased the tube pressure drop by reducing the channel size (increasing u1)and by increasing the friction factor via roughness effects. Roughness effects are discussedfurther in Section TFU Fouling StudiesFigure 6.3 Overall Fouling Resistance and Soluble Gum Concentration in Run 5030.00009 • 18.00 R °0.00008 - f,o—16.0•°• Gum•°0.00007 - — 14.0—— 0• oQ000 •0.00006— 0 0 12.0• — 008.0.00005—<‘ 10.0•00— 0 0 00.00004 — 0 0 8.0— ,@000.00003 • 6’ °, 6,0—000.00002 0 & 4.00 ,p Oo‘. °o8° 00.00001 C’02.0I I I 0.00 100 200 300 400 500 600Time(minutes)2126. TFU Fouling StudiesFigure 6.4 Tube Pressure Drop and Soluble Gum Concentration in Run 50514.5 20.0o APtube18.0• Gum16.014.0 — °— -— —14.0.00o.—0 00 12.00 0000 0013.5-• 0 - 10.0——Uci)•—- 8.0 ç•0 0 —H 000- 6.0—• 00 0-o13.0 o0 020 0 00 0000 00 0000 000 oo 0000 0 - 4.0OOQJ%—00o 9 OOo%0 0 0 00 000 0 2.0O0 0000 00—0 0-— •— 012.5 I 0.00 100 200 300 400Time(minutes)2136. TFU Fouling StudiesTable 6.2 is a summary of the initial conditions in the fouling runs and the observedtube pressure drops. AP increases with u,, as expected from Equation [3.33]. The lowvalue of initial AP in run 512 was probably caused by trapped air in the lines. Runs 510and 512 involved very large changes in A.P1, caused by heavy fouling. Run 508 wasoperated at close to maximum heat flux but cooler fouling caused the bulk temperature torise steadily after 200 minutes. The deposit morphology was noticeably different from thatof runs 503-4 and appeared to be caused by bulk precipitation of insoluble polyperoxides.6.2.2 Local Fouling BehaviourThe local heat transfer coefficient was not uniform along the heated length owing tothe development of the thermal boundary layer in the tube. This produced a non-uniformtemperature profile along the heated length, which would be expected to affect the foulingrate as this is a strong function of temperature. The local fouling behaviour variedsignificantly with axial location x, but this variation could not be explained simply in termsof the surface temperature or the development of the thermal boundary layer. Table 6.3summarises the local heat transfer and fouling resistance variation observed in run 504. Theoverall values were calculated using the thermocouples located between x = 219 mm and x= 708 mm. There is a local maximum in the initial fouling rate and surface temperaturearound 300 mm, which may be caused by variations in tube thickness. The depositcoverage in this region was also larger than average, which indicated that thethermocouples were operating correctly. Most runs showed some deviation from theexpected temperature profile along the heated section so a proposed simulation of the TFUwas not performed.Table 6.3 shows that negative fouling resistances were observed in the first 150mmof the heated section after the transition to the bulk precipitation fouling mechanism.Figures 6.5a-c show the local Rf profiles from the 44.4, 104.8 and 219.1 mm positions.2146. TFU Fouling StudiesTable 6.2 Initial Conditions and Tube Pressure Drops in TFU Fouling RunsRun T5f,0 Tbulk W Uo q Urn Initial Initial Final(qt) ± 0.01 LPtube I AP1j(°C) (°C) (kg/s) (W/m2.K) (kWim2) (mis) (kPa) (kPa)502- 101 - --- --503 199.7 102.8 0.11 1901 171.6 2.18 13.6 0.00767 14.3(170.5)504 198.3 101.3 0.11 1873 166.9 2.18 13.2 0.00739 13.5(171.8)505 222.6 101.4 0.102 1886 207.7 2.03 13.1 0.00863 13.8(200.8)506 187.4 100.2 0.104 1843 146.6 2.07 13.3 0.00845 13.7(145.2)507 211.5 100.8 0.104 1883 189.4 2.08 13.1 0.00820 13.9(185.4)508 199.4 102.3 0.174 2883 256.0 3.47 32.2 0.00716 33.2(256.0)509 212.4 103.6 0.137 2476 248.3 2.73 20.2 0.00726 27.8(242.7)510 211.3 101.9 0.056 1153 113.3 1.12 4.8 0.01016 6.4(112.8)511 209.1 101.5 0.087 1758 176.1 1.73 9.2 0.00837 9.3(171.9)512 211.3 101.5 0.056 1188 118.2 1.12 3.3 0.00704 6.8(117.5)0.405 M indene in Paraflex, 1mM bP initiation; t - average heat flux over Run; f - Fanning friction factor2156. TFU Fouling StudiesTable 6.3 Local Heat Transfer and Fouling Behaviour in Run 504Distance along Tsurf,o h(x) Initial Fouling Final Fouling Final Fouling Depositheated section, Rate Rate Resistance CoverageX (x108) (x108) (x10)(mm) (°C) (W/m2.K) (m2.K!W.mjn) (m2.K/W.min) (m2.K!W) (mg/cm2)44.4 185.91 1950 4.4 32t 0104.8 195.24 1783 5.7 -136,-33 -8.0-158.8 197.5 1760 10.3 0 2.0 0.48219.1 195.6 1811 10.2 34 5.5 0.54323.9 201.5 1747 8.6 42 7.1 0.50435.0 197.4 1869 8.6 40 7.3 0.65546.1 194.8 1976 7.4 40 7.0-603.3 199.8 1888 6.6 36 6.5 0.65657.2 196.1 1995 5.9 35 5.5 0.47708.0 210.7 1891 6.2 36 6.5 0.50overall 198.3 1873 7.7 37 7.2 0.60- initial rate which tends to zero after 30 minutes: - initially rapid rate, which decreased in magnitude2166. TFU I’ouliizg StudiesFigure 6.5 Local Fouling Resistances in Run 5042 0.000090.00008 -0,00007 . a. x = 44.4mm0.000060.000050.000040.00003 -.1.0.000024.’.—0.00001 • •• . .0.00000 Iv.- -0.00001-0.000020 50 100 150 200 250 300 350 400 450 5000.00004N.# .$•• ••.i- dPh1M-0.00000-0.00002-0.00004.-0.00006b.x=104.8mm-0.00008-C-0.000100 50 100 150 200 250 300 350 400 450 5000.000060.00005 c. x = 219.1 mm• .&• -N••.•••0.000030.00002- S.% •.1• •.5 •0.00001 •.-0.00000C-0.00001-0 50 100 150 200 250 300 350 400 450 500Time(minutes)2176. TFU Fouling StudiesNegative thermal fouling resistances were not reported in previous chemical reactionfouling studies but have been reported in particulate fouling (Crittenden and Alderman,1988). The positive mass deposit coverage values indicated that this was not a spallingeffect but was caused by enhancement of the local heat transfer coefficient. A negativevalue of R/x) indicates that h(x) >h(x)0,(and at constant heat flux, Tsurj(X) < T(X)surjo),which occurs when the polyperoxide gum particulates are large enough to disturb thedeveloping thermal boundary layer. The effect on the friction factor is expected to be lesssignificant as the momentum boundary layer thickness, ömom, is larger than the thermalboundary layer thickness, ôth, for fluids with Pr > 1.A negative fouling resistance at constant heat flux invalidates the constantdepositlliquid interface temperature assumption which is desired in chemical reactionfouling studies. Equation [5.4] shows that the fouling rate is very sensitive to surfacetemperature so that any variation in temperature will be reflected in the experimental data.The absence of negative fouling resistances after the thermal entry length (shown in Figure3.12 to be approximately 180 mm) indicated that this was a localised phenomenon in thethermal entry region and that data from downstream were not affected. This was confirmedby the inspection of the deposit morphology described in Section Deposit Distribution and MorphologyThe foulant was inspected in situ using optical microscopy, SEM and a simplegravimetric analysis. More information about deposit structure and properties could beobtained using the techniques recommended in Section 9.The deposit morphology varied with position in the fouled heated sections. Figure6.6 is a photograph of the fouled test section from run 507 after being rinsed of solvent,dried and cut into sections. The numbers in the figure refer to the tube section labels. Thephotograph shows that deposition occurred in four different regions which are shown in218Flow Sec/ion36 NumberI IFigure 6.8 Photograph of Thermal Entry Length Deposition in Run 50726Figure 6.6 Photograph of Fouled TFU Tube from Run 507 after Division into Sections‘V ,(.,-6. JTU Foniin Sindies./- ‘t?IFigure 6.7 TFU Test Section Depositon Regions• 501 V 5fr 131] 13-Unheated Entry Themial Developed ExitLength Entry Length Region LengthIll lIi!IIIfIIIIiIlJTTlJTTTjIllIIIijIlilI!Ij1I ;I!IIIIIllIl1l1lI1l1III1I1(I11T11111[i]J1i1IIIIIIi1I[I9iII1iTf_________ii- MAO IN ENGL LOil O)I O6O O9O CV rL-.6. TFU Fouling Studiesschematic form in Figure 6.7. Figure 6.8 shows the transition between the unheated entrylength (UEL) and the thermal entry length (TEL) in greater detail. The striations observedin the unheated run (501) were found in the cool entry length but not in the cool exit length;the surface in the latter region was relatively clean in all the fouling runs. This behaviour isconsistent with the striations being generated by the deposition of insoluble gum discussedin Section 6.1. The liquid leaves the heated section with Tb,o > Tj so that the gumconcentration is no longer at the solubility limit and the driving force for precipitation issignificantly reduced. The striation widths and separation were much larger than the visiblegrain structure in the drawn tubing, which suggested that the striations were not caused bychannels in the metal surface. The striation patterns were found to vary with liquid velocity,being wider and further apart at lower Re; the deposit mass coverage was also larger atlower Re. At low Re, smaller striations coexisted with larger ones and often merged withother striations. The striation deposition patterns are thought to be caused by localchannelling in the surface turbulence structure, in conjunction with irregularities in the tubesurface. Yung et al. (1989) described the locally organised motion in turbulent boundarylayers and reported dimensions of the variable speed fluid streaks. The streaks averagewidth was 15-20 v/u* and they were spaced randomly apart, at a mean separation of 80-100 v/u. For the TFU, these dimensions were calculated as widths ranging from 251-77jim and mean separations ranging over 1100-420 jim. The streak widths and separationsdecrease with u, as observed in the TFU, and the streak separations were comparable tothe striation separations observed in the experiments. The variation in deposit coveragewith Re may, however, be related to removal and deposition mechanisms. Further study ofthe striation patterns was recommended.Figure 6.8 shows the fouled thermal entry length from run 507 and the transitionbetween the unheated entry length and the developed fouling regions. The length of thefouling transition region varied with Re in the same manner as the heat transfer entrylength. The dimensions of the striations in this region were similar to those in the cool entry2206. TFU Fouling Studieslength but the gum itself was red or darker, indicating that the material had been aged by theenhanced surface temperature. The striations then fanned out into a chaotic region wherethe deposit consisted of randomly sized and distributed globules of aged gum. The depositpatterns bear a striking resemblance to photographs of the onset of turbulence in smoketracer/wind tunnel studies, where turbulent spots occur and their wakes expand into thechaotic region of developed turbulence.The gum globule sizes in the heated entry length region ranged up to 1 mm, whichwas significantly larger than those measured downstream in the developed region. Thelarger globules were thought to originate as entrained agglomerates of insoluble gum whichwere deposited on contact with the first section of heated surface. The thermal entry lengthwould then act as a trap for such large agglomerates, removing them from solution beforethey reached the developed region.The thermal entry length observed in the deposit from the PFRU runs wasnoticeably shorter than in the TFU runs. This suggests that entrained globules of insolublegum could be deposited in the developed region of the PFRU which would not bedeposited in the TFU. This mechanism may explain the differences in fouling resistanceprofiles reported by Panchal and Watkinson (1993) between a longer, tubular probe and thePFRU unit operating under identical chemical and thermal conditions (equal u, cf. Re).The increase in Rf was more gradual in the tubular unit and resembled the TFU foulingresistance profiles.The foulant observed in the developed region consisted of randomly depositedglobules of gum which had aged to form a red/brown solid. Figure 6.6 shows a grainstructure in the developed region which was not always apparent on the microscopic scale;the striation orientation parallel to the flow direction was not as evident in this region.Figure 6.6 also shows that the foulant was not uniformly deposited around the tubesurface; some segments experienced heavier fouling, which was thought to be due tovariations in the tube thickness.2216. TFU Fouling StudiesThe transition from the developed region to the exit length was always very sharplydefined, confirming that conduction losses through the terminals were greater than thosealong the unheated tube.Figure 6.9 shows optical and electron micrographs of the deposit from run 507,section 17 (Tsurf 210°C). The figure shows some of the deposition patterns observed inthe PFRU experiments; a more even initial deposit formed from smaller gum globulesfollowed by the random deposition of larger globules and agglomerates of insoluble gum.Run 502 was terminated during the transition to the accelerating fouling regime and showedsignificantly fewer globules of dark insoluble gum and almost no agglomerates of thismaterial. The deposit coverage observed in run 502 was also smaller due to the lowersurface temperature involved (200°C).The deposit coverage profiles from runs 502-4 are plotted in Figure 6.10 and showthe deposition patterns observed in Figure 6.6 in a quantitative form. The missing datapoints refer to sections retained for further analysis. The coverage in the cool entry lengthincreases in the first 300 mm to a mean value of 1 mg/cm2irrespective of the run duration(502 - 5.5 hrs; 503 - 8.5 hrs; 504 - 8 hrs). The reduced value at the inlet was probably dueto increased turbulence in this region following the tube fittings. The developed regioncoverage profiles show the local hot spot reported in 503 and the effect of run duration onthe final deposit mass. Experiments 503 and 504 differed in duration by 30 minutes but thedifference in coverage is greater than that between 504 and 502; this result indicated that thecoverage data should not be used to compare fouling rates between runs featuring differentdurations. The run 502 data confirmed that the initial fouling rate period in Paraflexinvolved the deposition of relatively small amounts of material under the conditions used.Globule diameters, d, were estimated from the micrographs and found to belong tofour size ranges. The smallest, ranging in from 7-18 Jtm, were the amber/orangeparticulates involved in the formation of the yellow matrix and the initial, even deposit. The2226. 1 FU Foiling StudiesFigure 6.9 Optical and Scanning Electron Micrographs of Deposit fom Run 507a. Optical (136x)TFU Run 507 Section 16b. SEM (bOx)2236. TFU Fouling StudiesFigure 6.10 Variation in Deposit Coverage with Axial Position in TFU Runs 502-5042.cL8 - Heated Zone 0 502I.5031.61.4A5041.2-•0 • *.1)•0.t.. .‘ L0 •AA A. .0.8*0A0.60I4 A A 00 ec 0.4 Flow 0 .0 00 00 •0.2- o0o0 0.•0000.0 I I .I0 200 400 600 800 1000 1200 1400 1600 1800x (mm)Table 6.4 Comparison of Sand Roughness Criteria with Gum Globule Dimensions inTFU Fouling Run 503Globule Size Range (d/d1lRe RoughnessRegime(Fm)(Tbuak= 108°C) (Tfjlm = 154°C)___________ ___________7-18 0.5-1.3 0.9-2.4 8-19 hydraulicallysmooth30-40 2.1-2.8 4.0-5.3 32-43 transitionalroughness66-75 4.6-5.3 8.8-9.9 74-84 transitionalroughness250+ 17.6+ 33.2 71+ fully rough2246. TFU Fouling Studiessmaller dark red/black globules observed in the later stages of the initial fouling rate rangedaround 30-40 jim in diameter. The larger black/red globules observed in the bulkprecipitation phase ranged from 66-75 p.m and then — 250 p.m+ in diameter and wereevidently agglomerates of smaller units; these approached millimetre dimensions in thethermal entry length.These dimensions can be compared with estimates of the ‘boundary layer’thicknesses in the flow through the tube. The viscous sublayer thickness is estimated as6mom = 5v/u*; the data in Table 6.2 gave u for run 503 0.1354 mIs. The kinematicviscosity of 1.925x 106 m2/s at Tb,m 108°C gave a sublayer thickness of 71 jim; usingTfilm 154°C to calculate V (1.02 xlO-6m2/s) gave 3mom 38 p.m. These estimatesconfirm the observations in Section 6.2 where the friction factor was not significantlyaffected by the formation of deposit until the generation of the larger insoluble gumglobules. The larger globules present in the bulk precipitation process were expected tocause measurable roughness effects. White (1991) described the following sand roughnesseffect criteria based on d:< 4 hydraulically smooth wall (dId) Re < 104 < d < 60 transitional roughness regime> 60 fully rough (no viscosity effect) (dLjdt) Re > 1000The globule dimensions are compared with these criteria in Table 6.4. The largest globulesize range belonged to the transitional roughness regime in both cases, whereas the largeagglomerates (250 p.m+) are expected to give fully rough behaviour.:The effective heat transfer sublayer thickness can be estimated as öth = A/U, whichgave 6thm in run 503 as 61 p.m (based on the film temperature). The heat transfer sublayerthickness is thus comparable with the viscous sublayer thickness (6monz = 38-7 1 jim),which is unexpected in turbulent flow where Pr>]. This result indicates that the d sizes areof the same order of magnitude as 5th• Paterson and Fryer’s fouling model describes thegeneration of foulant as occurring within the thermal ‘boundary layer’; the relative sizes2256. TFU Fouling Studiesconfirm that the larger globules must have been formed in the bulk fluid. Fouling modelsare discussed further in Section 7.3.The concentration ‘boundary layer’, ômass, can be estimated using the ChiltonColburn mass transfer analogySt Pr 2/3 = St’ Sc 2/3 = f12 [6.1]Assuming St’ dr/Re Sc ömass givesómass = ôth (Pr/Sc)”3 [6.2]At the film temperature in run 503, & = 17.9; Sc(indene) = 475 and Sc(oxygen) = 147,giving ômass as 20.4 pm (indene) and 30.2 pm (oxygen). The concentration boundarylayers thus lie completely within the viscous sublayer. Both ranges of insoluble gumglobule sizes are larger than ömass and thus indicate that this deposit precursor wasgenerated in the bulk fluid.6.3 Surface Temperature Effects in TFU FoulingThe effect of surface temperature was studied using the model solution described inSection 6.1 at a mass flow rate of 0.104 kg/s and temperatures ranging from 187-225°C.No cooling capacity difficulties were noted during these runs. The initial fouling rate,deposit coverage and final fouling resistance increased with initial surface temperature, asreported in Sections 5.4 and 5.5. The fouling resistance profile at the highest temperature(run 505; 225°C) exhibited an earlier acceleration to the precipitation fouling regime than theother experiments. This run was terminated early in order to avoid thermocouple damage.The deposit in this case contained larger amounts of insoluble gum globules than observedpreviously.The fouling data for different surface temperatures and flow velocities aresummarised in Table 6.5. The final Rfvalue and the mean deposit coverage both reflect thetrends observed in the initial fouling rate. Figure 6.11 is an Arrhenius plot of the averaged2266. TFU Fouhin2 StudiesTable 6.5 Surface Temperature and Flow Velocity Effects on Initial Fouling Ratet - final rate becomes negativeo - local initial fouling rate; • - local final fouling rate; • average initial fouling rateRun Duration(hr)Tsurf,o(°C)w(kg/s)Urn(mis)Initial FoulingRate(x 108)(m2.K1W.min)Final FoulingResistance(xlO5)(m2.K/W)Mean DepositCoverage(Developed Region)(mg/cm2)5035045055065075085095105115128. 2.05 8.1 8.0 1.150.105 2.04 7.8 7.2 0.610.102 2.04 23.4 24 4.060.104 2.07 5.8 3.6 0.130.104 2.09 13.0 7.7 1.480.174 3.48 2 -O 0.040.137 2.75 6.6t -8.6 1.850.056 1.12 16.Ot -15.5 6.030.087 1.73 22.0 19.5 3.540.056 1.12 24.Ot -14.0 8.37Figure 6.11 Arrhenius Plot of Initial and Final Fouling Rates in TFU Experiments-12-13-14-15-16-17-18-190.00200 0.00205 0.00210 0.00215 0.002201/Ts,o(K)0.002252276. TFU Fouling Studiesinitial fouling rates and also shows the local initial and final fouling rates. The initial foulingrates were obtained by drawing a line through the fouling resistance data where t<t*; thescatter in the data suggested little benefit from the use of a regression program. Theaveraged initial fouling rates were fitted to a modified Arrhenius equationdRf/dt = 23.9 exp (-7M00/RT) (R2 = 0.970) [6.3jThe activation energy, 76.4 ±8.0 kJ/mol, is reasonably close to the values of 81.9 ±16.4and 84.8 ±13.2 kJ/mol reported in the PFRU studies and indicates that the same foulingmechanism is operating in the TFU runs. The prefactor, 23.9m2.KIW.min, is an order ofmagnitude smaller than those reported in the PFRU; this is likely linked to the changes inprobe geometry and the higher flow rate in the TFU experiments. The final fouling rates arealso significantly smaller than those reported in Section 5.4. Comparisons are less useful inthis case as the runs were terminated after 8 hours where possible, rather than allowing theprecipitation fouling process to accelerate as in the PFRU runs. The 8 hour limit wasadopted to avoid excessive precipitation of insoluble gum in the TFU.The similarity in mechanisms was confirmed by the deposit analyses summarised inTable 6.6. The gum recovered from the unheated entry length was similar in composition tothe indene polyperoxide gums described previously. Gums recovered from the heatedsections showed a marked loss in oxygen content, which was more pronounced at higherinitial surface temperatures. The gum from the thermal entry length, where the surfacetemperatures were not as severe, showed an oxygen content between those of the two otherregions. The results are consistent with the gum ageing process described in Section 5. Thedeposits from runs 509, 510 and 512 show significantly higher oxygen content thanreported in the other TFU runs. This is consistent with the negative final fouling resistancesreported in these cases; the lower final surface temperature would reduce the rate of ageiñgand thus the loss of oxygen in the deposit.2286. TFU Fouling StudiesThe effective thermal conductivity, of the deposit can be estimated from thedeposit coverage and the final fouling resistance when the latter is positive. Assumingconstant deposit density p, the final fouling resistance of a mass of deposit (mass = areaöf.Pf) isRf = öJfI?f = mass/pf.f.area = (coverage)/pf.f [6.4]Figure 6.12 is a plot of final fouling resistance against surface coverage for all local foulingresistances where the local surface coverage was also known. The hand drawn line gives avalue of pfAf as 184.6 W.kg/m.K; the density of indene is approximately 990 kgm3,giving an estimated value of as 0.19 W/m.K, which is in good agreement with thevalues estimated from deposit thicknesses in Section 5. The deviation from the line islargest at low surface coverage (and low surface temperature). These coveragemeasurements involved the largest percentage errors and the deposit contained moreoxygen, which would decrease 2 and thus raise the points above the line. The TFU thusgives a direct measurement of pf? a quantity which is usually estimated in fouling modelcalculations.6.4 Velocity Effects in TFU FoulingThe effect of flow velocity was studied using the same model solution of indene inParaflex and a surface temperature of 210°C. A 770 mm long heated section was used andthe bulk fluid velocity varied from 1.12 - 2.75 mIs. The range of velocities andtemperatures involved was limited by the heating and cooling capacity of the system underfouling conditions. Runs 508 and 509 were performed at high heat fluxes and experiencedcooling capacity problems due to cooler fouling. The reported initial fouling rates weresmall and subject to error estimates of ±15%; the rates could have been increased by usinghigher surface temperatures but this would have required a new transformer or shorter2296 TFU Fouling StudiesTable 6.6 Elemental Analysis of Deposits from TFU RunsIRun, Section# Position Tsurf,o(‘C)wt%C H 0As peroxide504.13 TEL 195 74.26 4.74 21 C9H690191505. 8 UEL 100 70.0 5.8 24.2 C9H00232505.15 DR 225 84.11 7.32 8.57 94069505.20 DR 225 81.04 6.57 12.39 CH88103505.23 DR 225 80.2 5.51 14.29 974120507. 5 UEL 100 71.71 5.33 22.96 C9H800216507.15 DR 210 78.38 4.96 16.66 68143507.21 DR 210 77.07 5.25 17.68 C9H740155509.16 DRt 210 72.85 4.70 22.45 C9H700208510.16 DRt 210 74.04 4.68 21.28 68194511.20 DR 210 80.25 5.35 14.4 C9H720121512.20 DRt 210 78.21 4.91 16.88 68146t - negative final Rf: UEL - unheated entry length; TEL - thermal entry length; DR - developed regionFigure 6.12 Plot of Final Local Fouling Resistance against TFU Deposit Coverage forEstimation of Deposit Thermal Conductivity0.000400.000350.000300.000250.000200.000150.000100.000050.000000.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0Deposit Coverage (mg!cm2)2306. TFU Fouling Studiesheated lengths. The study was interrupted by pump failure; no further runs wereundertaken.The results are summarised in Table 6.5 and show that the initial fouling ratedecreased with flow velocity, as reported in Section 5.5. The thermal entry lengths wereshorter at larger flow velocities, as reported in the literature. Run 511 showed someanomalies in its fouling behaviour; the initial fouling rate was larger than expected and theexperiment had to be terminated early as the surface temperature alarm reading was highand larger than the surface temperatures recorded by the PC. The deposit obtained from run509, after cooling capacity failure, was quite different to that observed in other TFUfouling experiments; the foulant was a shiny, dark mass of randomly deposited, large gumaggregates of size > 200 pm. This caused negative fouling resistances in the later part ofrun 509.Negative fouling resistances were also reported in run 510 and its repeat, run 512,which were performed at the lowest TFU flow rate. Cooling failure was not responsibleand the onset of decreasing Rf coincided with t and the transition to the bulk agglomeratefouling regime. The deposit consisted of an orange/red veneer composed of dark,amorphous gum agglomerates ranging in size from 150 pm upwards. The agglomeratesthus increased the local heat transfer coefficient and caused a cooling effect at the heatedsurface. The effect of these large agglomerates on the viscous sublayer was confirmed bytube pressure drop, which increased rapidly after t as Rf decreased. This can be clearlyseen in Figure 6.13, which shows the fouling resistance and pressure drop profiles fromrun 510. The increase in pressure drop is caused by both the change in friction factor andthe narrowing of the duct by the accumulation of deposition. The increase in friction factoralso enhances heat transfer and a rough comparison of these effects can be made using theheat and mass transfer analogy in Equation [6.1]. Substituting for the friction factor in anapproximate tube pressure drop expression based on Equation [3.33] gives2316. TFU Fouhin2 StudiesFigure 6.13 Tube Pressure Drop and Overall Fouling Resistance in TFU Run 5100.00010ci)zC10I -- I- Io Rf• LP8&&o0 .000 080 0co 000 0 a, o 0 oO .•ooo%)00 o0 8 0 •aoo 0 •00 00•0•... •0.000‘DO0.000050.00000-0.00005-0.00010-0.00015-0.000206.56.005.5-CH5.04.5.1•.•••?* .• •..•._.••..1..•...• IIp•0•i..•100 200.300 400 500Time(minutes)2326. TFU Fouling StudiesAP = 4 f (L/d (Pum2)/2 = 4 (ff2) Re (L/dt2)um i =4 Nu Pr“3(L/dt2)Um 11 [6.51Assuming constant fluid properties and flow rate, the ratio of pressure drops and Nusseltnumbers is given byAP,0/zP(t) = [Nu0fNu(t)] [d(t)4I ,0 = [U0IU(t)] [ d(t)3/d,0l [6.6]Equation 6.6 shows that the pressure drop is more sensitive than the Nusslet number tochanges in the duct geometry. Data from run 510 were: U increased from 1153 to 1400W/m2K; zIP increased from 4.8 to 6.45 kPa. Assuming that the pressure drop outside theheated section remains constant, at 4.8x1.04/1.81 2.76 kPa, Equation 6.6 gives d(t)7.89 mm and an approximate foulant thickness of 0.5 mm. This is in reasonable agreementwith a visual estimate of the deposit thickness; the material was too brittle to measuremechanically. The negative fouling resistances observed were thus consistent with thedeposition of large gum agglomerates which interrupted the viscous sublayer. Theseaggregates are presumably present at higher flow rates also but do not cause similar effectsthere; this indicated that the adhesion of the aggregates is controlled by the residence time atthe surface (which decreases with u *) or by the shear forces given by ‘r acting at thesurface (which are larger at higher flow rates). The fouling mechanism is discussed furtherin Section 7.The initial TFU fouling rates decreased with flow velocity, varying as Um where -1<n <-2. The initial rate reported in run 511 was larger than those observed at a lower flowrate (510,512) and suggested that there was a local maximum in the fouling rate - velocitycurve. Such maxima are a feature of Epstein’s model (Section 2.4). The unusual foulingbehaviour in this run rendered the rate value suspect and a repeat experiment was requiredto verify the result; unfortunately a repeat run could not be performed.2336. TFU Fouling Studies6.5 Comparison of TFU and PFRU Fouling ProbesNegative fouling resistances were reported in the initial PFRU fouling experimentsat higher concentrations of indene in Paraflex in Section 5.2, but not at the 5 wt% level.The TFU studies also operated at larger flow velocities than the PFRU so a comparison runwas performed using the same model solution in the PFRU and T,j= 210°C. The PFRUhad been modified since run 064 and was operated at its maximum flow rate, which gaveUrn 0.774 mIs and Re 5040. The flow velocity was less than in runs 510 and 512 (1.1mis) but the Reynolds number was similar (5360). The heat flux calibration had also beenmodified and the calculated value, 145.8 kW/m2,yielded a higher value of U0 thanreported in Section 5. The PFRU was operated under similar chemical conditions as theTFU and Section 6.1 describes how the reaction rate was slightly faster in the PFRU.The comparison experiment, run 065, was run for eight hours and showed identicalfouling behaviour to that observed in the TFU. The negative fouling resistances thereforearose because of conditions leading to bulk precipitation rather than differences in probegeometry. Figure 6.14 shows the Rfprofiles from runs 512 and 065; the fouling resistancedecreased after t and the onset of the bulk precipitation fouling mechanism. The drop in Rfin run 512 around 250 minutes was caused by a transient in heater power and heat flux.The figure confirmed that the same fouling mechanism was involved in both foulingprobes. The initial fouling rate in the PFRU (3.36x107m2.KIW.min) was larger than inthe TFU (1.6 and 2.4 xlO-7m2.KIW.min), which was attributed to the lower flowvelocity. This result suggested that the dimensionless velocity (given by Re) was lessimportant than Urn in the fouling mechanism. The fouling resistance data from the PFRU inFigure 6.14 shows considerably less scatter than the data from the TFU. This is becausethe PFRU measures a local fouling effect whereas the TFU monitors the overall unit2346. TFU Fouling StudiesFigure 6.14 Comparison of Fouling Resistance Profiles in TFU and PFRU0.00020o 065(PFRIJ) .• •• ,..• 512(TFU) •0.000150• •i ’0.O0010• •% • 0Q •0 •• ••::004-e• ••:. 00.00005-•.••00•00-000000 C9- .•0 00 0• 0•()0 •o000 s-0.00005- 00 00 00-0000100 50 100 150 200 250 300 350 400 450 500Time(minutes)2356. TFU Fouling Studiesperformance. The differences in the TFU and PFRU Rj profiles before t>’ were similar tothose reported by Panchal and Watkinson (1993).The PFRU deposit was remarkably similar to that observed in the correspondingTFU runs. The unheated entry length was coated with yellow gum striations of varyingwidths, developing swirled patterns at positions further downstream from the probeentrance. The swirl was probably caused by the rotational motion of the fluid imparted bythe sharp turn at the probe inlet. The thermal entry length was 25 mm long and containedclearly defined red/orange striations running parallel to the flow. This entry length wassignificantly shorter than that observed in the TFU runs (100-150 mm) and did not involvethe ‘turbulent spot’ zone reported in Section 6.2.3. The heavily fouled region featured verydisordered deposition by thick, shiny red/brown agglomerates of insoluble gum whichranged in size up to 1 mm. Occasional light yellow strands of gum were evident in the exitlength.The TFU studies showed the same initial fouling rate behaviour as observed in thethermally and chemically initiated PFRU studies described in Section 5.2. The depositmorphology and composition also confirmed that the same fouling mechanisms occur inboth probes.The negative Rf behaviour reported at low velocities in runs 510, 512 and 065 isunusual. A heat exchanger subject to this type of fouling behaviour would suffer increasedpressure drops before there was any thermal evidence of deposition. It is likely that thedeposit would eventually lose the enhancement in heat transfer coefficient caused byroughness and the fouling resistance would then increase rapidly. Further investigation ofthe negative fouling resistance phenomenon requires a study of the reaction andagglomeration processes which generate the large globules involved.Similarly unusual Rf behaviour was observed in PFRU experiments 024 and 027,where thermally initiated solutions of 9 and 10 wt% indene were studied at T5U,j= 180°C,2366. TFU Fouling StudiesTbik = 80°C, Re = 3050 and 79 kPa oxygen saturation. The low surface temperature gavenegligible initial fouling rates so that the fouling behaviour was dictated by bulkprecipitation. Figure 6.15 shows how the fouling resistance profiles for these runs deviatedfrom that observed in run 025 (5 wt% indene). Similar deposit morphologies to run 065were also reported. Experiments 024, 027 and 065 featured larger indene reaction ratesthan run 025, which may be an important factor in the formation of agglomerates.6.6 Summary of TFU Fouling StudiesThe Tube Fouling Unit (TFU) was shown to give reproducible heat transfer andthermal fouling resistance results, although differences between individual tubes did causevariations in the fouling data. The range of operating conditions was limited by the currentdesign and by the fouling of the cooling system by the polyperoxide gums which alsofouled the heated section. This cooling fouling phenomenon is linked to the solubility of thepolyperoxide gums and was not noticed in the PFRU.Ten fouling experiments were performed in the TFU to investigate the effects ofsurface temperature and flow velocity on the initial fouling rate. Chemical analysesconfirmed that conditions of reproducible reaction were achieved in the bulk liquid. Theinitial fouling rate showed the same trends in surface temperature and velocity reported inthe PFRU experiments. Anomalous fouling resistance behaviour (negative Rf) wasobserved at the lowest velocity ( lm/s), where the deposition of large agglomerates ofgum enhanced the heat transfer coefficients via roughness effects. This behaviour was alsoobserved in a comparative experiment performed in the PFRU.The experimental evidence confirmed that the same deposition mechanisms wereinvolved in the annular PFRU and tubular TFU fouling probes. The tube pressure dropmeasurements confirmed the transition in fouling mechanisms observed in the thermalfouling data. The TFU permitted the examination and analysis of deposit in situ and2376. TFU Fouhin2 StudiesFigure 6.15 Fouling Resistance Profiles from Initial PFRU Runs at High IndeneConcentrations showing Unusual Fouling Behaviour0.00060.00050.00040.00030.00020.00010.0000Concentrations of indenein Paraflex in wt%Time(minutes)3500 4000 4500 50002386. TFU Fouling Studiesrecorded local as well as overall fouling behaviour. Noticeable variations in local foulingbehaviour were caused by the development of the thermal boundary layer and precipitationof entrained gum. SEM and optical micrographs confirmed that the foulant consists ofparticulates of insoluble gum. The profiles of deposit coverage provided a usefulcomparison for the thermal fouling results and permitted the direct calculation of the factorPfAf.The unheated regions of the test section were coated with striations of insolublegum which require further study.2397. Models of Aspects of Fouling7. Models of Aspects of Autoxidation FoulingThe experimental studies showed that chemical reaction fouling underautoxidative conditions is a complex phenomenon which cannot be explained by a simple,global fouling model. A qualitative model of the fouling process was developed, featuringthe following mechanisms.a. FOULANT GENERATIONThe foulant precursor was identified as polymeric peroxides, formed by the reactionof alkenes with dissolved oxygen. The tendency to form polyperoxides is determined bythe alkene structure and by the chemical conditions in solution. Chain transfer to the solventor other components, free radical scavengers and low oxygen concentrations were all foundto interfere with the formation of polyperoxides. Fouling was not detected in the absence ofpolyperoxide gums, except at high antioxidant levels, so that the fouling process can bedescribed as subject to bulk chemical reaction control.The induction periods observed in thermally initiated fouling experiments were dueto the accumulation of hydroperoxide to a concentration large enough to cause theautocatalytic reaction with oxygen which generated polyperoxides. The reaction kinetics inthe fouling apparatus were found to be limited by mass transfer of oxygen into solution.The batch autoxidation work showed that autoxidation in the the model solutions was notadequately described by the kinetic schemes in the literature.Significant solvent effects were observed in the behaviour of the polyperoxidereaction products. The aromatic polyperoxide has a limited solubility in aliphatic solvents;polyperoxide was precipitated as agglomerates of insoluble gum after the polyperoxideconcentration exceeded the solubility limit.2407. Models of Aspects of Foulingb. DEPOSITIONFouling studies were only performed in aliphatic solvents as a suitable aromaticsolvent was not identified. Two deposition regimes were observed, involving differentdeposition mechanisms; an initial, pseudo linear fouling regime followed by an‘accelerated fouling’ regime. Fouling rates were much higher in the latter regime andswamped any contribution from the mechanism generating deposit in the initial regime. Themechanism in the accelerating regime involved the deposition of relatively large (30 jim+)globules and agglomerates of insoluble gum present in solution after the solubility limit wasreached. This deposit was uneven and rough enough to disturb the fluid momentumboundary layer.The initial fouling regime was observed when the dissolved polyperoxide gumconcentration was lower than the solubility limit. The deposit consisted of relatively small(6-20 jim) globules of soluble gum and its ageing products. Veneer-like deposits would beexpected if the precursor was formed on the surface alone but these were only observed inregions of low surface temperature where the gum could melt into the surface. Depositsformed at high temperature were found to be composed of an even, ordered ‘coral-like’material formed from the small gum globules and their ageing products. The initial foulingrate increased with surface temperature and decreased with flow velocity. It is this regimewhich is of prime interest for modeling because of its industrial importance. Cases ofaccelerating fouling could arise in closed heat transfer systems using hydrocarbon oils asthe heat transfer medium.c. DEPOSIT AGEINGThe material recovered from the fouled surface was chemically different from thepolyperoxide foulant precursor. The deposit aged to form polyperoxide degradationproducts on exposure to the enhanced temperatures on the heat exchanger surface. The2417. Models of Aspects of Foulingageing of deposit followed soon after adhesion to the surface and may be involved in theadhesion process itself. Deposit removal was not observed in the fouling studies.The current work did not provide a complete understanding of all aspects of thefouling process. A complete mathematical model of the phenomenon requires furtherresearch and experimentation and was beyond the scope of this work. Certain aspects ofthe fouling process were investigated and are described in the following sections.7.1 Autoxidation Kinetics in the Fouling ApparatusThe chemical induction periods in the thermally initiated fouling runs reported inSection 5.4 and by Zhang et at. (1993) were coupled to the operating conditions in thePFRU. No systematic variation with PFRU conditions was observed in the gum formationrate or indene kinetic constants in these runs; this is consistent with the oxygen masstransfer limitation, but could also be due to the lower activation energy involved in theseprocesses. A simple model of the PFRU system was developed to study the coupling ofbulk chemical reaction and fouling conditions.7.1.1 Model DevelopmentThe PFRU system was modelled as a well mixed batch reactor with a pumparoundthrough the fouling probe as shown in Figure 7.1. The system is assumed to be adiabaticapart from the cooling coils which maintain the holding tank at the set temperature, T1.Volume 2 represents the pump and the piping to the PFRU, where T2 was slightly hotterthan T1 due to the dissipation of heat in the pump. The difference was estimated as 0.5-16C. Volume 3 represented the volume of liquid exposed to the high temperatures at thePFRU surface (T1f= T3), assumed to be the volume of the thermal boundary layer. It is2427. Models of Aspects of FoulingFigure 7.1 Compartmental Kinetic Model of PFRU SystemC3PFRUVolume 3C2Table 7.1 PFRU Reaction Coupling Model ParametersParameter Paraflex Simulation kerosene SimulationTotal Liquid VolumeVolume 2Volume 4ThulkFlow rateEactT5fU010 L0.84 L0.42 L80°C17 L/min40- 120 kJ/mol180-240°C880-950 W/m2K10 L0.84 L0.42 L80-84°C7-15 L/min40-74 kJ/mol165-216°C880-1750 W/m2K2437. Models of Aspects of Foulingthis part of the fouling ioop which operates at higher temperature and thereby gives rise todifferences between kinetics in the isothermal and in the fouling experiments. The areaavailable for heat transfer was known and the thermal boundary layer thickness, öth, wasassumed to be given by öh = ?JU0, so that V3 = öth.(.rr da,i L), where L is the PFRUheated length. Thermal boundary layer development was ignored in this analysis. Thetemperature in the piping back to the reactor (Volume 4) was calculated fromT4 = T2 + U0(3t da,i L) (T3 -T2)!(W.Cp) [7.1]Volume 1 was calculated knowing the total volume of liquid and the piping volumesV1 = Viiq - V2 - V3 - V4 [7.21The effect of the enhanced temperature zone is modelled by considering the rate of reactionof a general species, concentration C. Assuming constant physical properties and flow rate,a mass balance onV1 gives=- C1(t)) + R1dt [7.3]where x1 is the space time in Volume 1, given by V1 p1W. R1 is the volumetric reaction ratein V1 and has the form R1 = - k1 C1“. Equation [7.3] reduces to the isothermal batch reactorresult when C1 = C4. C4 was calculated by assuming plug flow in the pumparoundvolumes; for a first order reaction, where R1= - k1C, C4 is given byC4(t) = C1(t --- t) exp -[k2t+k3t + k4t] [74]where is the space time in Volumej. This givesdC1(t)=-ç1(t) (1 + kjt1) + C1(t - -r- t3 - t4) exp-[k2t+3r+ k4t1 [7.5]In the case of zeroth order kinetics, where R1 = - k1, [7.3] becomesdC1(t) -C1(t) 1= +—(C(t-t2--u3t4) kdt Ti [7.6]Equation [7.61 is a linear o.d.e. with a delay term (t-t2 + t3 + t4) which does not have ananalytical solution. The equations show that the isothermal rate constant is augmented by a2447. Models of Aspects of Foulingcontribution from the pumparound which depends on the pumparound volumes and therelative sizes of the kinetic constants.The effective rate constant for the thermal boundary layer is a function of thetemperature gradient across the layer. Paterson and Fryer (1988) obtained the followingresult for the overall rate constant in a boundary layer with a linear temperature gradientacross its thickness, kr, where the rate constant obeyed the Arrhenius equation;kr = _B_ k(Tsurf) ( Tf i - exp (Tb - Tsuris)EactFact (Tsurf - Tb)1\ R Tf [7.7]T5,fand Tb are the surface and bulk liquid temperatures respectively. This approximationwas used in calculating k3.Equation [7.5] was solved numerically using a FORTRAN 77 program based on asimple Euler method to estimate the derivative. The initial conditions in the pumparoundwere represented byC4(t) = C1(0) t < t2 + t3 + t4 [7.81and short time increments were used (At = O.2-O.3(r + r3 + t)). The isothermal case wascompared with the analytical solution and showed negligible solution inaccuracy.The effect of the unheated flow loop and the heated section was quantified bycomparing the value of C1(t) with the isothermal result after a reasonable time period. Afirst order indene rate constant of 8.5 x104miw1 was regressed from the data for thethermally initiated autoxidation of indene in Paraflex in SCR run 122 shown in Figure 7.2.A time period of 200 minutes was selected and the coupling effect expressed as a ratio ofeffective rate constants. For first order kinetics,kcoupledfkiso = in (Ciso(t)/Ccoupled(t)) [7.91For zeroth order kinetics,kcoupledlkiso = [C(t = 0) Ccoupled(t)lI[C(t = 0) -C0(t)] [7.101The model parameters are summarised in Table 7.1.2457. Models of Aspects of Fouling7.1.2 Model PerformanceA sensitivity analysis of the model showed that the rate constant activation energywas the most important parameter within the range of experimental conditions. The effectson the effective rate constants was similar under zeroth and first order kinetics. Figure 7.3shows the effect of total volume on the rate constant ratio at activation energies of 120 and50 kJ/mol at T1 = 80°C, T3 = 240°C, U0 = 950 W/m2.K and W = 0.23 kg/s. The activationenergies are those obtained in Section 4 for the initiation and propagation stages in theautoxidation of indene. The effect of changes in liquid volume is larger at the higheractivation energy, which indicates that liquid sampling would increase the effective rateconstant and hence reduce the observed induction period. The model showed little effect ofsystem parameters such as surface temperature at the lowest activation energy, whichagrees with the experimental data in Table 7.2. This small effect of temperature could bedue to an oxygen mass transfer effect, however, where the bulk oxygen concentration wasnegligible and there was no reaction in the bulk liquid. Such mass transfer effects areunlikely to apply to the induction period and reaction coupling would then be significant.The model was compared to the data from the thermally initiated fouling runsinvolving indene in Paraflex in Section 5. The chemical induction period was assumed tobe inversely proportional to the initiation rate and the rate enhancement predicted by themodel was transformed into induction periods by using the data point atT5,f= 180°C as areference, i.e.tjnd(Tsurf) = tind (180°) keff (1 80°C)/keff(Tsuff) [7.11]The results are plotted in Figure 7.4 and show that the model did not predict the trend inthe experimental data. The reaction zone volume had to be increased to unrealistic values inorder to yield results close to the experimental data.Zhang et al. (1993) reported the induction periods from their thermally initiatedPFRU runs using indene in kerosene and their data showed the same trend as the Paraflex246I7. Models of Aspects of Fouling0.41 M indene in Paraflex: 79 kPa oxygen saturation: TbUlk = 80°C: thermal initiationFigure 7.3 PFRU Reaction Coupling Model Predictions of Volume Effects inKinetic Parameters in the PFRU System0.5 0 Eact = 1 20 kJ/mol4 5 6 7 8 9 10 11Volume (L)0.350.300.250.20Figure 7.2 Regression of First Order Indene Rate Constant from Data from SCRExperiment 1220.40 0-2‘S-4‘-5- 5 10 15 20 25 30 35 40Time(hours)2.5 • •02.0 00001.51.0 e .- .- -.• Eact = 50 kJ/mol0.02477. Models of Aspects of FoulingFigure 7.4 Comparison of Reaction Coupling Model Predictions and ExperimentalInduction Periods in Thermally Initiated Fouling Runs in Paraflex::2015-100 Experimental• Model0I I • I160 180 200 220 240 260T surface (°C)Table 7.2 Experimental Data from Thermally Initiated Fouling Experiments0.41M Indene in Paraflex; 79 kPa Oxygen SaturationSolvent Ner Tsurf Re Induction Period kp.(experimental) /_____________________________(CC)_______________(hours) v(mol/L)IhrParaflex 025 180 3050 26±2 0.00793Parallex 031 200 3050 20 0.00861Paraflex 028 210 3050 7-10 0.00849Paraflex 032 225 3050 6±1 0.00765Paraflex 029 240 3050 6±1 0.007232487. Models of’ Aspects of Foulingdata shown in Figure 7.4. They observed little variation in induction period with flow rate,whereas the model predicted an increase in induction period with flow rate due to thedecrease in t5rh.The model confirmed that thermally initiated reaction kinetics in the PFRU wouldbe coupled to the heat exchanger performance but gave very poor agreement with theexperimental data from the thermally initiated fouling runs. A more detailed analysis wasnot considered appropriate. The model did show that reproducible chemical reaction isunlikely in thermally initiated batch experiments if the PFRU surface temperature is varied.This conclusion supported the use of chemically initiated model solutions, which yieldedmore reproducible reaction behaviour.7.2 Fouling Mechanisms in the Initial Fouling RegimeThe mechanisms governing foulant generation and adhesion in the initial foulingregime were not completely understood. Existing fouling models were compared with theexperimental data and a fouling model using the Paterson and Fryer approach wasproposed.7.2.1 Particulate Fouling ModelsThe TFU runs confirmed that the deposit consisted of small globules of insolublepolyperoxide gum which could have been generated in the bulk liquid and deposited on thehot surface. The initial fouling mechanism would then be a case of particulate fouling,corresponding to Case lb in Panchal and Watkinson’s analysis. The gum particulate sizesestimated from the TFU micrographs were used to compare the observed fouling rates withestimates of particulate fouling rates. The unagglomerated globule dimensions fell within2497. Models of Aspects of Foulingthe range of O-5Ojm reported by Crittenden and Alderman (1988) in their discussion ofparticulate fouling.The rate of fouling by particulates transported from the bulk liquid to the wall in theabsence of boundary layer gum generation is given bydRf/dt = K0 C /p ?f [7.11]where is the mass concentration of particulates in the bulk liquid and K0 an overall masstransfer coefficient including adhesion (ka) and mass transport (km) coefficients similar tothose in Equation [2.33]1/K = 1/km + 1/ka [7.12]The adhesion term often includes a surface temperature dependency and a stickingprobability. Equation [7.12] assumes that the adhesion process is first order in particulateconcentration, following the result of Ruckenstein and Prieve (1973).Mass transport of particulates in turbulent flow can occur via diffusion, inertialcoasting of particulates through the viscous sublayer or by particle impaction (Melo andPinhero, 1988). The dominant mechanism is determined by the particle relaxation time, t;u2 d2t-- f___p 18v2 [7.131where 4 is the particle diameter and Pp the particle density. Melo and Pinhero gave thefollowing ranges for each mechanismt<O.1 diffusion k = u*/11.8ScO.67 [7.14]0. l<t<l0 inertial coasting km = (t u*/5.23) (PIPp) exp (0.48t) [7.151t> 10 inertial impaction [7.16]Turner (1993) discussed other forces (lift, drainage effects) involved in inertial coastingand suggested that the diffusionlcoasting bound (t = 0.]) should be lower in solid/liquidsystems than the bounds above, which were obtained from aerosol studies. The Schmidtnumber in Equation 7.14 was calculated using the Stokes-Einstein equation:= 1.38x1023T(K)/(3njd) [7.17]2507. Models of Aspects of FoulingThe particle sizes reported in Section 6 in Paraflex yield Sc> 106, which is outside theoriginal range of Metzner and Friend’s correlation (Equation 7.14); its use at high values ofSc was defended by Muller-Steinhagen et at. (1988). The values of for the four globulesize ranges were calculated at the conditions used in the fouling runs and are summarised inTable 7.3. The film temperature was used in calculating the fluid properties and the gumdensity was estimated as that of indene, 990 kg/rn3.The friction factor in the PFRU runswas estimated using (Rosenhow and Hartnett, 1973)f = 0.085 (Re-O.25) [7.18]Table 7.3 also shows the estimated mass transfer coefficients calculated using Equations[7.14, 7.15].Table 7.3 shows that the dominant mass transport mechanism for theunagglomerated globules is by inertial coasting; the larger agglomerates belong to theparticle impaction regime. The fouling rates in the absence of adhesion effects wereestimated using [7.11] and the value of p2 reported in Section 6.3. The maximum initialfouling rates for the group A particles were calculated and are given in Table 7.3 along withthe experimental results. Bulk particulate concentrations of 5 g/L and 11 g/L (g*) were usedat Tblk = 80 and 100°C respectively. The estimates for mass transfer controlled fouling areorders of magnitude too large and do not show the observed dependence on surfacetemperature. This suggests that adhesion effects may control the rate of particulate fouling,as described by Epstein’s model of mass transfer with adhesion discussed in Section 2.Chen (1993) found that the rates of submicron particle deposition from an isothermalturbulent flow regime were adhesion controlled and fitted the data to Epstein’s model withsome success. The particles in his study belonged to the diffusion transport regime.The initial fouling rate regime, however, involves certain features which cannot beexplained by a particulate fouling model. Firstly, the particles are relatively large and aretransported to the surface by inertial coasting, which varies as u3. The initial fouling ratedecreased with increasing u*; an adhesion term such as that in Epstein’s model would have251Table 7.3 Particle Relaxation Times and Mass Transfer Coefficients in ParticulateFouling Calculated at Conditions used in Fouling Runs, Assuming MassTransfer ControlConditions Tsuff Particle Size A B C D Fouling7-18j.tm 30-401.tm 66-75.tm 250j.tm+ Ratet(°C) Quantity (m2K/Wmin)Thermally 180 tp 0.005-0.03 0.08-0.53 0.41-0.53 5.9 4.06x10Initiated kD (mis) x108 6.3-4.6 2.4-1.99 1.42-1.3 0.58 [1.0x107]Tbulk 80°C kim (mls)x105 4.025 66-130 400540 7970um=O.687 miS(PFRU) 240 tp 0.009-0.064 0.18-0.32 0.86-1.11 12.4 8.3x104kD (mis) x108 11-6 4.2-3.4 2.47-2.3 0.96 [14.4x107kim (mis)x105 7.5-51 150290 1010-1480 37100Chemically 200 tp 0.004-0.029 0.08-0.53 0.41-0.53 5.9 5.72x104Initiated kim (mis)x105 2.5-16 47-86 300-360 4480Tbulk 100°Cum=O.SO6mis255 tp 0.008-0.053 0.146-0.26 0.71-0.914 10.15 10.7x104(PFRU) kim (mls)x 10 4.4-30 88-160 560-800 74400 [26.7x i0-]Uml.OO mis 200 tp 0.014-0.09 0.262- 1.268- 35.8x i0kim (flL/s)x104 1.5-10 0.466 1.63830-60 242-373255 t 0.026-0.173 0.48-0.856 2.33-3.01 67.9x104kim (mls)x104 2.7-19 62-130 725-1300 [7.4x10 ]Chemically 187 tp 0.042-0.028 0.784-1.39 3.79-4.91 0.0225Initiated kim (mis)x104 8.5-63 220-530 4590- [5.8x108]Tbulk 100°C 10090um=2.O4misf= 0.008222 0.065-0.429 1.193-2.12 5.77-7.46 0.0361(TFU) kim (mls)x103 1.3-10.1 41-113 1780-5160 [23.4x108]- experimental = 247°C: t - calculated for 181.tm particles using Equation [7.11] where lfka 0;kD - diffusive mass transfer coefficientkim - inertial coasting mass transfer coefficient[]denotes experimental values2527. Models of Aspects of Foulingto vary as 1/u*4,which is difficult to explain in physical terms. The particles in Chen’swork involved diffusion mass transport, which varies as u, and the adhesion term wasinterpreted in terms of the particle residence time at the surface. The fouling rates reportedin the bulk precipitation regime in the PERU increased with flow velocity, however, whichis consistent with the hypothesised change in fouling mechanism to deposition byagglomerate particulates.Secondly, a particulate fouling model does not account for the observed pseudo-linear Rfbehaviour in the initial fouling rate period. The fouling rate is proportional to C,which increases during this period and thus predicts an accelerating Rf profile. Thisdiscrepancy could not be explained by an increase in the mean particle size as the bulkreaction progressed, as the mass transfer rate (which increases as d increases in thediffusion and inertial coasting regimes) would thus increase over time. A complexadhesion/deposit removal term would have to be invoked in order to predict the psuedolinear Rfbehaviour.Thirdly, the particulate flux towards the surface moves along a temperature gradientand would be subject to thermophoresis effects in the thermal boundary layer. Thethermophoretic velocity acts against the temperature gradient and would thus preventparticles reaching the surface. The thermophoretic velocity, Utplj, can be estimated from(McNab and Meisen, 1973)— -0.26 ? v VT — 0.26 qUtph — —_______—(2 + T (2 + )T [7.19]where A and are the fluid and particle thermal conductivities respectively. Assuming thatA 0.12 W/m.K, the thermophoretic velocity at a film temperature of 130°C (vl.37x106m2/s)and a heat flux 90 kW/m2 is estimated as 2.2x10 m/s, which is of thesame order of magnitude as the inertial coasting velocity given in Table 7.3. MUllerSteinhagen et al. (1988) reported that Equation [7.19] overestimated Utph in their alumina2537. Models of Aspects of Foulingparticulate deposition studies and suggested a correction factor of 0.12; this still yields avalue of u1 comparable to km for the group A globules. Thermophoretic effects wouldthus be important in the initial fouling rate period. The values are suspect, however, asEquation [7.19] was based on experimental data from an aqueous, non-reacting system(latex particles in water). Equation [7.19j indicates that Utpli increases slightly with surfacetemperature (2.3x104mis at a film temperature of 160°C) and would thus not contribute tothe temperature dependency observed in the experimental data.Particulate fouling analyses are further complicated by the complex and poorlydefined particle size distribution in the fouling solutions. The assumption that negligiblereaction occurs in the boundary layer region is also highly suspect. Simple particulatefouling models were thus unable to explain the observed fouling behaviour.7.2.2 Chemical Reaction Fouling ModelsPanchal and Watkinson (1993) identified two chemical reaction fouling model caseswhich showed reasonable agreement with the experimental results from autoxidationfouling of indene in kerosene. These cases are described in Section 2.4.2 and Figure 2.5.Case la involved mass transfer of foulant precursor from the bulk to the wall toform foulant. The polyperoxide gum was assumed to be soluble so that convective masstransport applied. This case corresponds to Epstein’s model if adhesion effects are includedin the foulant generation step and the polyperoxide is assumed to form insoluble globules atthe wall. Adhesion effects are likely to control fouling as the convective mass transfercoefficients for the polyperoxide gum are large. Convective mass transfer coefficients kmfor indene and its monomeric, dimeric and trimeric peroxides under the fouling conditionsin run 503 are estimated in Appendix A.3.4 using both Metzner and Friend’s result and themodified form proposed by Epstein (1993b) in his analysis of the fouling data ofCrittenden et al. (1987a). The values range from 2.8 to 1.5 xl0 mis and 1.2 to 0.7 x1052547. Models of Aspects of Foulingmis, respectively. A maximum fouling rate can be estimated by assuming that all gum isconverted to foulant (i.e. simple stoichiornetry, sticking probability = 1) and that thesoluble gum consisted of peroxide dimers, ROOROOH. Using k1 = 0.85 x105mis and agum concentration of lOg/L, this yielded a mass flux of 0.05 1 kg/m2.s and a maximumfouling rate of 2.78 x105rn2.KIW.min.This rough estimate is two to three orders of magnitude too large and indicates thatadhesion effects or the surface reaction rate dominate deposition in this case. This modeldoes not explain the observed linear fouling rate behaviour, but does indicate that surfacereaction mechanisms are significant.The rate limiting step in the reaction at the surface could also be the mass transfer ofoxygen from the bulk fluid, where the concentration of oxygen is always less than that ofindene in these experiments. If the autoxidation reaction in the bulk liquid is controlled byoxygen mass transfer (see Section 4.2), the concentration of oxygen in the bulk will bevery small and will control any oxidation process occurring at the heated surface. Thisconcentration is likely to remain constant during an experiment and would thus give rise toa constant fouling rate. A similar maximum rate argument to that above can be applied tothe initial fouling rate in experiment 503 (8.1x10rn2.K/W.min) to estimate theconcentration of dissolved oxygen in the bulk fluid. Equation [7.11] is used with ilka 0and a stoichiometric assumption that 1 mol of oxygen produces 0.1481 kg foulant (i.e. allproducts attach; no kinetic resistance; no loss of material on ageing); this is unlikely butyields an order of magnitude estimate. Equation [7.111 givesCo2 (8.1x10!60) (184.6) I [(0.1481)(2.17x105)] 0.08 mol/m3 [7.20](mol/m3) (m2.KIW.s) (kg.W/m4.K) (rnol/kg) (s/rn)The dissolved oxygen concentration under saturation conditions in run 503 was estimatedat 4.9 mourn3,which is - 60x larger than this estimate of the bulk oxygen concentration.This value is consistent with the autoxidation kinetic scheme described in Section 4 wherereaction in the diffusive boundary layer gave rise to negligible oxygen concentrations in the2557. Models of Aspects of Foulingbulk liquid. The stoichiometric assumption and the mass transfer coefficient are the primarysources of error in this analysis. If all the oxygen transported to the surface does not formfoulant, the estimate of C02 will be increased and could approach the saturation value. Thefouling rate would still be controlled by the oxygen concentration and would still give riseto a linear Rtime curve.The case la model does not, however, predict the reduction in fouling rate withincreased velocity noted in the fouling experiments. Oxygen mass transfer would beincreased at higher flow rates, confirming that surface reaction and/or adhesion steps areinvolved in the fouling mechanism.The case 2 scenario proposed by Panchal and Watkinson was an attempt to modelthe complex reaction dynamics in the ‘reaction zone’ in the viscous sublayer next to the heattransfer surface. Convection in and out of the zone as well as enhanced reaction rates in thezone were estimated using a consecutive reaction scheme based on that of Norton andDrayer (1968). This scheme ignored the effect of indene concentration on the formation ofinsoluble polyperoxide and did not include adhesion effects, but did include oxygen masstransfer. The data from the tubular apparatus showed a quasi-linear initial fouling rate butthe Case 2 calculations did not show this behaviour. This discrepancy suggests that thereaction dynamics are more complex than those used and require further investigation. Thismodel assumed that the bulk reaction (and hence the surface reaction) was not oxygen-masstransfer limited. Panchal and Watkinson concluded, however, that the reaction zoneanalysis did yield the best agreement with the experimental results.There are several problems in modelling the reaction zone dynamics. The kineticmodel of indene autoxidation in Section 4.6 indicated that the reaction stoichiometry iscomplex. The initial fouling rate activation energies observed in the fouling experiments aresignificantly larger than the activation energies obtained for indene autoxidation in the SCRexperiments. This result suggests that the rate determining step in foulant generation differsfrom the bulk reaction kinetics, particularly as the dissolved oxygen concentration is2567. Models of Aspects of Foulingsuspected to be negligible throughout the liquid. The foulant generated, however, containsa significant amount of oxygen and the role of oxygen in these systems remains unclear.The influence of the gum solubility limit, g*, was also unknown. The complexity of thereaction zone dynamics in the current work prevented a quantitative comparison with thecase 2 scenario.7.3 Analysis of Fouling Rate BehaviourPaterson and Fryer’s ‘reaction engineering’ approach to milk fouling treated thecomplex reaction zone kinetics as a ‘black box’ and analysed the phenomenon in terms ofparameters likely to influence deposition. This approach will be used here, invoking thefollowing assumptions;(a) Bulk Chemical Reaction Control of the timing of fouling behaviour. The transitionfrom the induction period to the initial fouling rate period is determined by the bulk reactionwhere most free radical formation occurs. The transition from the initial fouling rate periodto the bulk precipitation phase is determined by the solubility of the polyperoxide gumproducts.(b) Boundary Layer Reaction Control of deposition during the initial rate period.Deposition begins soon after the formation of polyperoxides in the bulk as those formedinitially in the reaction zone would be convected out. The formation of deposit isdetermined by the rate of reaction in the reaction zone and the kinetics of the adhesionmechanism. The rate of reaction is assumed to be proportional to the bulk reaction rate viathe population of free radicals, as the fluid at the wall is continually being renewed by‘fresh’ elements of fluid. The reaction rate and the adhesion process are both affected bythe fluid residence time in the reaction zone, 0.2577. Models of Aspects of Fouling(c) Rapid Ageing. The adhesion process is assumed to involve part of the ageingmechanism so that the deposit thermal conductivity does not change significantly afterdeposition has occurred.(d) Negligible Removal. No evidence for a removal mechanism was observed.7.3.1 Formulation of a Lumped Parameter Fouling ModelA compartmental fouling model such as case 2 above requires a workingknowledge of the reaction kinetics in the reaction zone. A lumped parameter approach isused here and its predictions compared with the trends observed in the experimental data.Fouling rates will refer to initial fouling rates unless otherwise specified.The initial fouling rate is assumed to be controlled by the generation of foulingprecursor in the reaction zone, which lies in the viscous sublayer adjacent to the heattransfer surface. The fouling rate can be described in terms of a surface renewal model or afilm model, shown schematically in Figure 7.5. In the former, elements of fluid remain atthe surface for a given residence time during which insoluble gum is generated, which canthen react at the wall to form deposit. A surface renewal model involves mean residencetimes and surface reaction parameters which are not available for this case, so the filmmodel approach will be used.The film model is similar to that described by Epstein (1993b), where the foulingrate is given bydRf/dt = Rf = (l/pf) (rate of reaction at surface)(attachment factor) [7.21]2587. Models of Aspects of Foulingwhen mass transfer effects are negligible. Epstein assumed that the attachment factordepended on a chemical reaction step and the mean fluid residence time at the surface, 0,i.e. oc 0 exp (Eat/RTsurf).Figure 7.5 Schematic of Fouling Model MechanismBulk Fluid 0o 0 elements offluid in bulkrapid .massi i ‘ renewaltransfer. : / elements ofBoundary Layer :-fluid at surfaceI Heat Transfer Surface IFilm Model Surface Renewal Model0 was taken as proportional to v/u*2,which mirrors the periodicity of turbulent bursts inCleaver and Yates’ model for particle deposition. Assuming that the initial fouling rate,chemical reaction rate and attachment factor are described by Arrhenius-type kinetics, thetemperature dependence in equation [7.20] is given byEpj = EG + Eatt + E0 [7.22]where EG is the activation energy of the reaction generating polyperoxide gum precursorsand E0 is the temperature dependency in the residence time at the surface. The overallfouling rate is then given byRf. oc (l/fpf) (reaction rate) 9 exp (EattfRTsuri’) [7.23]Rf = 6 (l/?pf) RG(CJ) 0 exp (EattlRTsurf) [7.24]where RG is a gum formation rate which involves an overall activation energy EG and theconcentrations of indene and its oxidation products, {C}.J3 is a constant which has theunits <mis> when RG is given in <kglm3.s>.. Assuming that 0 cc v/u*2, Equation [7.24]becomes2597. Models of Aspects of FoulingRf = 13’ (lIX-pf) R0(C1vlu*2 exp (Ealt/RTsurf) [7.251The trends in the experimental data were compared with the behaviour predicted byEquation [7.25].7.3.2 Analysis of Experimental Fouling Datainspection of Equation [7.25] shows that the initial fouling rate varies with the rateof chemical reaction, the surface temperature and the flow velocity. Multiple parameterregression analysis was not used owing to the scatter in the data. The approach taken wasto determine the dependency on temperature and reaction rate before investigating velocityeffects. The calculation results are summarised in Appendix A.4.The effect of surface temperature on the initial fouling rate is shown in theArrhenius plots in Figure 7.6. Each series of runs was performed with the same indenebatch, flow rate and initiation method. The activation energies are given in Table 7.4 andcan be seen to lie within one standard deviation of each other. The difference in standarddeviations was not significant for this sample size. The correlation coefficient (R2) for eachactivation energy is not significant at the 1% level, indicating that the rates are related by theArrhenius equation (for these sample sizes). The initial fouling rate at 187°C in the TFUinvolves the largest experimental error; in its absence, the TFU activation energy figurewould be 88 ±6 kJ/mol, which is closer to the PFRU values.A comparison of flow velocity effects must be performed at a constant surfacetemperature. The TFU flow variation runs were performed at 210°C so this was used as thereference temperature. Equation [7.22] indicates that the fouling rate activation energyincludes a contribution from the mean residence time, mainly via the kinematic viscosity.The fouling rates were scaled to 210°C using(Rf7v)(2Iooc) = (Rf/V)(Tsur0 exp [-EactIR (1/483.15 - 1/T)1 [7.26]2607. Models of Aspects of Foulin2IIFigure 7.6 Arrhenius Plot of Initial Fouling Rates—11-12-13-14-15-16-170.0018 0.0019 0.0020 0.0021 0.00221/T(K)0.0023Table 7.4 Comparison of Initial Fouling Rate Activation EnergiesFoulingProbeInitiationModeEact(kJ/mol)Statistical Parameters0J3 R2 t-test ‘t ti”(kJ/mol) (data) (comoarison)PFRU thermal 84.8 13.2PFRU chemical 81.9 16.4(2.5 mM bP)TFU chemical 76.4 8(1.0 mM bP)t calculated using Ernean = 81 kJ/mol; tt’ = abs[Eact-811/°E0.953 2.32 0.290.976 4.99 0.050.968 2.32 0.582617. Models of Aspects of Fouling14‘S-80.0018Figure 7. 7 Arrhenius plot of (Rf7V) against Surface Temperature-2-3-4-5-6-70.0019 0.0020 0.0021 0.00221JT(K)0.0023Table 7.5FoulingProbeComparison of Activation Energies of Adjusted Fouling Rate (Rf/v)Statistical ParametersInitiation Eact t—test t ‘tt’Mode(kJ/mol) (kJ/mol) (data) (comparison)PFRU thermal 96.6 11 0.963 -8.79 0.23PFRU chemical 97.2 10 0.946 -9.36 0.32(2.5 mM bP)TPU chemical 88.2 8 0.977 11.4 -0.74(1.0mM bP)t calculated using Enean = 94 kJ/mol; ‘tt’ = abs[Eact -2627. Models of Aspects of FoulingThe activation energies for use in Equation [7.26] were calculated by plotting (Rf/v)against l/Tsurf as shown in Figure 7.7; the viscosity was calculated at the film temperature.The activation energies given in Table 7.5 agree within one standard deviation and rangefrom 88-97 kJ/mol, which is close to the sum of the energies of the initial gum degradationrate (40 kJ/mol) and the indene consumption rate (ER 48-57 kJ/mol). This resultsuggests that the oxygenated reaction occurs in the boundary layer region, which infers thatthe bulk oxygen concentration is not negligible. A reliable method of measuring thedissolved oxygen concentration is required to verify this hypothesis. The fouling rate datawas corrected to 210°C by calculating (Rf/v) and applying Equation [7.261 using theactivation energies in Table 7.5. Each series of data in the Arrhenius plots was representedby a single value (taken from the regressed line at Tsurj = 210°C) for subsequent checkingof velocity effects.The effect of the chemical reaction rate is demonstrated by the PFRU data in Figure7.7, where the chemically initiated fouling rate is markedly faster than the thermallyinitiated fouling rate. The flow velocity conditions were similar (0.51 mIs vs. 0.68 mIs) butthe bulk reaction rate constant was markedly larger in the chemically initiated case. This isconsistent with a surface renewal model, where the reaction at the wall will be dependenton the concentration of free radicals brought to the surface in a fluid packet. The surfacereaction rate in the film model would be expected to be proportionately higher than the bulkreaction rate. The chemical reaction rate in the bulk varied between series of experimentsdue to differences in indene batch, reactor configuration and bulk liquid temperature. Thereaction stoichiometry is also unknown. For the purposes of comparison, the chemicalreaction rate was assumed to be proportional to the maximum reaction rate of indene in thebulk liquid. The investigation of indene autoxidation in Section 4 showed that indene reactsto form other products than the gum fouling precursors. Equation [4.34] gives the bulkreaction rate of indene in kg/rn3.s as proportional to kR and the concentration of indene.R0 a kR Vtindenej (116.1)/3600 [7.27]2637. Models of Aspects of Foulingwhere kR is in ‘J(mo1JL)/hr. Equation [7.25] can be arranged to give[(Rf) (pf2f)]I(v RG) B” [1/u*2Jexp [-ERfIRTSUff] [7.28]The reaction rate activation energy has been included in the overall activation energy, ERJ.Writing (Rf)(pf2c)/(v RG) as a ratio of the fouling rate to the gum generation rate, e, gives= B” [l/u*2]exp [-ERfIRTSUff] [7.29]The PFRU data from Figure 7.7 were replotted as an Arrhenius plot of Equation [7.29] inFigure 7.8; the difference in flow velocities has been ignored in this figure. The value ofpf)t.f used was that obtained from Section 6, 184.6 kg.W1m.K. Figure 7.8 shows that thetransformed fouling rates seem to be in good agreement, so this basis was used incomparing the fouling data. This reaction rate assumption expressed in Equation [7.27]was highly suspect but offered some basis for comparison of the experimental data withouta working model of the boundary layer kinetics. The maximum difference betweentransformed values in Figure 7.8 occurred at 225°C and was 50%, which gives anindication of the size of errors expected in the model. When the fouling rate data iscorrected to a common surface temperature, 0 will then represent the velocity dependencein the attachment factor.The fouling model in Equation [7.25] indicates that the fouling rate is inverselyproportional to u2, which is given by Um2f/2. The friction factor in the PFRU had to beestimated as pressure drop data were not available. An estimation method based on the heatand mass transfer analogy represented in Equation [6.1] is described in detail in AppendixA.4. This analysis gave the following correlations for use in the fouling model calculations;f (PFRU) = 0.295 (±0.03) Re0344 ±0.02 [A.4.3]f (TFU) = 0.059 (±0.0 19) Re0203 [A.4.2]The fouling rate data was used to calculate (Rf/v) at T,-1’= 2 10°C using Equation[7.26]. This value was used to calculate O(2100c), which is plotted against the estimated2647. Models of Aspects of FoulingFigure 7.8 Arrhenius Plot of e v. Inverse of Surface Temperature for PFRU Runs12o thermal initiation• 2.5 mM bP initiation••10•09 0•8 007 I I0.0018 0.0019 0.0020 0.0021 0.0022 0.00231!T(K)Thermal initiation : 0.41 M indene in Paraflex, Tbujk = 80°C, Urn = 0.68 rn/sChemical initiation : 0.41 M indene in Paraflex, Thulk = 100°C, Urn = 0.51 mIs2657. Models of Aspects of Foulingfriction velocity in Figure 7.9. The Figure is a log/log plot and shows thatO decreases asfriction velocity increases. There is considerable scatter in the experimental data, which isexpected given the degree of estimation involved in generating the plot. The data point atthe largest friction velocity involved a very small fouling rate and was subject to largeexperimental errors (-.50%). The figure shows that the TFU data seem to follow the sametrend as the PFRU data, suggesting that the same fouling mechanism is involved in bothfouling probes. The TFU data are widely scattered and more information is required toconfirm this hypothesis. The figure shows the line of best fit obtained by least squaresregression of the data, which gavelog 0 = -1.98 (±0.3) log u* + 0.97 ±0.30 (R2 = 0.74) [7.30]The initial fouling rate can thus be expressed byRf = B” Vu198 RG exp (-11306/Tff) [7.31]where RG = 0.03225 kR /[indene] (kg/m3.s); kR is in (1(mol/L)/hr), [indene] in (molJL)and B” = 7.4 x108m498K/W.kg.s°98The activation energy in Equation [7.31] is themean value of the values given in Table 7.5.Equation [7.29] predicts a friction velocity dependence of -2, which is close to thatobserved. The average absolute deviation from the regressed result [7.30] was calculated as49%, which is larger than the experimental error in most runs. The systematic error fromthe reaction and friction factor estimates was not known. Further regression analysis wasnot performed given the degree of estimation involved in generating the data.Fortuin et al. (1992) developed an Extended Random Surface Renewal model forturbulent flow in liquids, which predicted that the mean residence time varied as 4v/u*2fThis result would also give a velocity dependence similar to that observed in Figure 7.9,but verification requires better quality data.2667. Models of Aspects of FoulingFigure 7.9 Fouling Model Analysis: Variation of 0(210°C) with Friction Velocity• • •D0\3.5 DQ \D C[3 ee2.52 I • I • I • I-1.6 -1.4 -1.2 -1 -0.8 -0.6log {u* (mis)]C PFRU, 2.5 mM bP PFRU, 2.5 mM bP, Tserieso •PFRU, 1 mM bP TFU, 1mM bP, Tseries0 PFRU, thermal, Tseries TFU, 1mM bP2677. Models of Aspects of FoulingPaterson and Fryer’s analysis (1988) pictured the reaction zone as a differentialreactor where the volume was proportional to the thermal boundary layer thickness, 3th•Incorporating this factor into the film model gives a fouling model of the formRf = B” (l/2f’pf) RG(Cj) th vlu*2 exp (Eatt/RTsurf) [7.32]wherej3” has units of s. This model can be developed using the same assumptions asabove and gives the resultOthth = B” [1/u*2]exp [-ERfIRTSUffJ [7.33]The thermal boundary layer thickness can be estimated from U1,/2i and was used to calculate0/öth• The temperature variation in )L. was ignored in correcting the data to 210°C. Figure7.10 is a log-log plot of the experimental data plotted as Equation [7.33] and was regressedusing least squares to givelog (e/h) = 5.9 - 1.12 logu* (R2=0.501) [7.34]This approach increases the scatter in the data and gave a larger deviation from the predictedvelocity dependence.Further fouling model analysis was not performed as the uncertainties in thereaction dynamics and friction velocity hinder numerical verification. The experimental datainvolved relatively small fouling rates and substantial variations in kinetic parameters,which introduced significant experimental errors into any modelling studies.The lumped parameter model described here provides a rationalisation of theobserved fouling behaviour. Further modelling studies requires better quality data andfurther investigation of the reaction at the heat transfer surface.2687. Models of Aspects of FoulingFigure 7.10 Paterson and Fryer Fouling Model Analysis:Variation of G(2100C)/&h with Friction Velocity8• • •D o7.5 -DDD6.5-:°6 I • I-1.4 -1.2 -1 -0.8 -0.6log [u* (mis)]D PFRU, 2.5 mM bP PFRU, 2.5 mM bP, Tseries0 PFRU, 1mM bP TFU, 2.5 mM bP, Tserieso PFRU, thermal, Tseries TFU, 1 mM bP2698. Conclusions8. ConclusionsChemical reaction fouling caused by the products of the autoxidation ofhydrocarbons was studied using model solutions of alkenes under oxygenated conditions.A model solution was selected to study the effects of surface temperature and flow rate onthe fouling rate in two different fouling probes operating under turbulent flow conditions.The current study represents the first systematic study of autoxidation fouling inheat exchangers operating under conditions of turbulent single phase heat transfer.Chemical monitoring of the process was used successfully to relate the observed foulingbehaviour to the known reaction chemistry and to formulate a qualitative model ofautoxidation fouling in these systems. This model included deposit ageing, which has beenreported in other chemical reaction fouling systems. Chemical initiation was used todecouple the effects of reaction rate and surface temperature in the analysis of fouling ratedata.A new fouling probe was commissioned and compared favourably with an existingdevice. The study of antioxidant effects in fouling (performed in conjunction with R. Lai)represents the first systematic study of such commonly used additives in these industriallyimportant systems.1. MODEL SOLUTIONSThe fouling behaviour of model solutions of indene and hexadec-l-ene in kerosene,tetralin, Paraflex (a light lubricating oil) and trichlorobenzene were studied in the annularPFRU fouling probe in order to select a model solution for further study. The experimentswere performed in the batch mode and chemical analyses were developed to follow theprogress of the autoxidation reaction. Fouling was caused by deposition of the insolublepolymeric peroxide products of the autoxidation reaction, confirming the hypothesis ofTaylor and Wallace (1967). Polyperoxide formation was determined by the reactivity of the2708. Conclusionsalkene peroxy radical and was controlled by chemical factors including oxygenconcentration, alkene structure, solvent nature and solvent structure. Fouling was thuscontrolled by the bulk chemical reaction, which explained the fouling behaviour observedby Asomaning and Watkinson (1992) and Zhang et al. (1993). A solution of 5wt% indenein Paraflex was selected for further fouling studies.The fouling and kinetic behaviour of model solutions of mixed alkenes (indene,dicyclopentadiene) could not be explained by the results in single alkene mixtures.2. FOULING BEHAVIOUR IN AUTOXIDATION FOULINGThe fouling behaviour was controlled by the bulk chemical reaction. A chemicalinduction period was observed under thermally initiated conditions; this was linked to theaccumulation of free radicals in solution, which was affected by the temperature in thefouling probe. A simple model of the annular probe recirculating system (PFRU) did notexplain the trends observed in the induction period data of Zhang et at. (1993) and in thecurrent work.The fouling resistance in Paraflex increased at a nearly constant rate until the solublepolyperoxide concentration reached a solubility limit, g*. This initial fouling rate is thoughtto be controlled by the complex reaction and adhesion processes in the hot boundary layerregion adjacent to the heat transfer surface. Once the gum concentration exceeded thesolubility limit, fouling was controlled by the deposition of globules and agglomerates ofinsoluble gum entrained in the bulk fluid. The fouling rate in this phase increased with timeand was orders of magnitude larger than the initial fouling rate. Similar fouling resistancebehaviour was observed in both the annular PFRU and tubular TFU fouling probes. Thepressure drop caused by fouling increased significantly in the later fouling period and thelarge globules and agglomerates caused roughness effects. Negative fouling resistanceswere reported at low velocities when large agglomerates deposited, disrupting the heattransfer boundary layer.2718. ConclusionsThe initial fouling rates were relatively small (106108m2.K/W.min) andincreased with surface temperature. Activation energies were estimated assuming anArrhenius relationshipdRf/dt = a exp (-Eact/RTsf) [8.1]as 84.8 ±13.2 kJ/mol (PFRU, thermal initiation), 81.9 ± 16.4 kJ/mol (PFRU, chemicalinitiation) and 76.4±8 U/mo! (TFU, chemical initiation). The initial fouling rate decreasedwith flow velocity, varying as u where -1<n<-2. The later fouling rate did not varysignificantly with surface temperature and appeared to increase with flow velocity, whichwas consistent with fouling controlled by the deposition of large particulates. The initialfouling rate variation with bulk temperature was complicated by bulk reaction and velocityeffects.3. AUTOXIDATION KINETICS AND BEHAVIOURThe autoxidation of indene in kerosene and Paraflex was studied in a semi-batchreactor under the conditions used in the fouling experiments. Solvent effects were observedin the indene reaction rate, autoxidation mechanism and the polyperoxide gum behaviour.The solubility of the polar, aromatic indene polyperoxides was in the range 10-12 g/L in themost common mode! solution used (5 wt% indene in Paraflex at Tblk = 100°C) anddepended on the physical nature of the solvent (temperature, aromaticity). Insolublepolyperoxide was precipitated as gum globules ranging in size from 30 jm+, once thesolubility limit was reached.The indene reaction kinetics in the kinetic and fouling experiments were subject tooxygen mass transfer effects after the chemical induction period. The data was successfullyfitted to a kinetic scheme based on mass transfer with zeroth order chemical reaction inoxygen, first proposed by Van de Vusse (1961). The activation energies reported for thedisappearance of indene (48-58 kJ/mol) did not agree with the literature values for welldefined initiation schemes.2728. ConclusionsA kinetic model was proposed based on a set of series-parallel reactions whichshowed the trends observed in the experimental data, but requires experimental verification.4. ANTIOXIDATIONA study of antioxidation was performed using 2,6,di-t-butyl,4-methylphenol(BMP) in model solutions of 5 wt% indene in Paraflex. The substituted phenol extendedthe chemical induction period but did not affect the autoxidation rate, as reported byHoward and Ingold (1962). GC-PID analysis showed that the end of the induction periodcorresponded to the exhaustion of BMP. The activation energy of indene initiation wasestimated as 120 kJ/mol.The antioxidant was less effective under heat transfer conditions, marked by areduction in the length of the extended induction period. The fouling and autoxidationmechanisms did not show any noticeable change after the induction period. Highconcentrations of phenol caused extra fouling in the heat exchanger and indicated that suchantioxidants should not be used above their ceiling temperatures.5. DEPOSIT MORPHOLOGY AND AGEINGThe deposit formed during the initial fouling period consisted of coloured gumveneers on cooler surfaces and small (6-17 pm) globules of insoluble material on hottersurfaces. The deposits formed during the later fouling regime consisted of larger, randomlydistributed globules of insoluble gum.The deposit chemical activity was consistent with ageing of the deposit by thedegradation of polyperoxides on the hot surface. This thermal degradation mechanism wasconfirmed by ageing simulation experiments. Ageing was thought to be involved in theadhesion of the polyperoxides to form deposit.The thermal conductivity of the deposit was estimated by a novel method in theTFU as 0.19 W/m.K, which is in good agreement with the literature values.2738. Conclusions7. TUBE FOULING UNIT (TFU)The TFU is a novel device for studying fouling which was designed andcommissioned during the current work. The device allows the inspection of deposits in situand was found to give reasonably reproducible fouling behaviour. Polyperoxides were alsofound to cause cooler fouling in the TFU.8. FOULING MODELSThe initial fouling rate behaviour could not be explained by particulate foulingmodels, or by chemical reaction fouling models which did not account for reaction andadhesion effects in the reaction zone next to the heat transfer surface. The fouling resistanceprofiles in the initial fouling rate period could not be quantitatively explained by the currentwork. A lumped parameter fouling model was shown to predict the trends observed in theexperimental data but involved two parameters which were difficult to quantify.Novel aspects of the study include the investigation of parameters involved inselecting a suitable model solution; the use of chemical initiators to decouple the effects ofreaction rate and surface temperature; the ageing simulation work and many features of theTFU, which represents the application of techniques used in two different cases ofchemical reaction fouling to provide extra insight into the fouling processes in thesehydrocarbon systems.2749. Recommendations9. Recommendations for Further StudyThe current work explained some of the features observed in autoxidation foulingbut also generated questions which require further further investigation. These are1. Oxygen in Autoxidation Fouling. A reliable method of determining dissolvedoxygen concentration was not available during the current work. The oxygen mass transferlimitation reported in Section 4 is based on inference and requires experimental verification.Measurements of dissolved oxygen concentration in the bulk liquid are also needed in thedevelopment of a reliable fouling model.2. Autoxidation Mechanisms. The product analysis methods could be improved toyield further information required by a kinetic model of indene autoxidation. A method todetermine insoluble gum agglomerate sizes in solution would be useful in studying the finalfouling rates. The chemical reactions involved in the reaction zone are not completelyunderstood and require further investigation in the development of a reliable fouling model.3. Solvent Effects. The current work did not identify a suitable aromatic or polarsolvent for model solution studies. Fouling behaviour was found to be linked topolyperoxide solubility in aliphatic solvents and it would be useful to investigate theinfluence of aromatic or polar solvents where the autoxidation products were more soluble.4. PFRU Apparatus. Several features of the PFRU could be improved on beforeperforming further autoxidation studies. The probe alignment, gas sparging system,cooling system and data collection methods need attention. This device was used to a muchgreater extent than was initially expected.2759. Recorntnenda lions5. TFU Apparatus. The TFU pump and connections to the heated section should bereplaced. The configuration of the device requires revision if it is to be used to studyautoxidation fouling further as the cooling system and transformer imposed limits on theoperating range in the current work.Further TFU runs would be useful in verifying the effect of flow velocity inautoxidation fouling. The TFU fouling rates in the current study are small and these runsshould be performed at higher surface temperatures if possible.6. Deposit Ageing. There is considerable scope for the use of surface science methods(such as FTIR and XPS) in examining samples of deposit in situ from the TFU. Thesemethods could provide further information about the deposit structure and thus its history.7. Comparison of Data. A large amount of fouling data has been generated at UBCand Argonne National Laboratory using the indene/kerosene system to study autoxidation.This material should be compared with the current work in order to derive a more generalmodel of fouling under autoxidative conditions. The aim of this model would be theextension of the batch fouling studies to continuous systems as found in the petrochemicalindustry.276AbbreviationsAbbreviationsABN 2-2’-azobis-2-methylpropionitrileAIBN AzodiisobutyronitrileBMP 2,6-di-t-butyl-4-methyl-phenolbP Benzoyl peroxideDCP Di-cyclopentadieneRD Flame lonisation DetectorFTIR Fourier Transform Infra Red (Spectroscopy)GCMS Gas Chromatography - Mass SpectrometryHWP Hot Wire ProbeMIBK methyl-isobutyl-ketonen.m .r. Nuclear Magnetic Resonance (Spectroscopy)n.t.p. Normal Temperature and Pressure (20°C, 101.3 kPa)PFRU Portable Fouling Research Unit fouling apparatusPID Photo-lonisation DetectorPOx Peroxide NumberSCR Stirred Cell ReactorTBA Tri-t-butylamineTCD Thermal Conductivity DetectorTFU Tube Fouling UnitTFU Tube Fouling UnitTGA Thermal Gravimetric AnalysisTHF TetrahydrofuranXPS X-ray Photoionisation Spectroscopy277NomenclatureNomenclatureA surface area m2a generic constanta1 stoichiometric constant, species i -AM surface area, mass transfer m2b time constant (Kern and Seaton model) 1/sBif fouling Biot number -C1 concentration, speciesj moIIm3Cor orifice discharge coefficient -C concentration of precursor or particulates mourn3,g/m3Cp heat capacity J/kg.KD diffusion coefficient m2/sdh hydraulic diameter md0, diameter; outer,inner me kinetic model, propagation efficiency -Eatt activation energy of attachment process J/molactivation energy of gum formation process J/molactivation energy of stepj J/molEj,0 overall activation energy J/molactivation energy/temperature dependence in residence time J/molEpj overall activation energy of fouling process J/molF general function -f friction factor (Fanning, unless specified) -G mass flux (Epstein, 1993b) kg/rn2.sg* soluble gum solubility limit g/LH enthalpy J/molh local heat transfer coefficient W!m2.KI current ampsSouder viscosity constant -J mass flux kg/m2.sK kinetic model - ratio of kinetic constants 3/mol.minK, K pressure drop coefficients; entry, exit -kciep mass transfer factorkD rate constant, Equation [5.2]278Nomenclaturek1 rate constant, initiationk1 overall rate constant, initiation molJL.minklmG mass transfer coefficient; liquidlgeneral, gas phase mis, mol/N.sKm lumped kinetic constant, equation [4.**]kM,po kinetic constant, mass transfer, peroxide ./(mol/L)/hrk11,34 kinetic rate constant, step n , i.e. Equation [341kj first order kinetic rate constant (Norton and Drayer model) 1/hrlcd, k’0 zeroth order reaction rate constants 1I(L/mol.hr)K0 overall mass transfer coefficient rn/skpD kinetic constant, peroxide decompositionkR kinetic constant inc. mass transfer, Equation [4.34] /(mo1/L)Ihrkr general reaction rate constantKVa,b,c constants in kinematic viscosity correlation, Equation [3.5]L dimension, length mOstwald coefficient, oxygen -M Molecular weight -m mass kgN1 total number of mols, species i molNr foulant generation rate, Paterson and Fryer model kg/sNu Nusselt number -P pressure Papair saturating air pressure kPaPOx Peroxide Number meq/LPr Prandtl number -q heat flux kW/m2Qe heat supplied as electrical power WQi heat losses to ambient WQ heat transferred to liquid WR universal gas constant J/mol.KRegression coefficient of variation -[RH] concentration of species RH molIm3Rf fouling resistance m2.K/WRf,0 overall fouling resistance, TFU m2.KJWRf* asymptotic fouling resistance m2.K/WR linearfouling rate .K/W.minRG volumetric reaction rate, foulant generation moIIm3.s279NomenclatureR free radical initiation rate mourn3.sR volumetric reaction rate, species j mourn3.sr1,0 radius; inner, outer mR0 volumetric disappearance rate, zeroth order mourn3.sRe Reynolds numberS sticking probabilitySc Schmidt numberSh Sherwood numberSt, St’ Stanton, modified Stanton number -T temperature °C, Kt time s, mm, hrt time taken to reach g* hrtD diffusional contact time sdimensionless particle relaxation time -tres residence time sTtr triple point KU overall heat transfer coefficient W/m2.KU0 clean overall heat transfer coefficient W/m2.Kfriction velocity rn/sbulk mean velocity m/sUtph thermophoretic velocity rn/sV volume m3v voltage VoltsV1 molar volume, speciesj rnlJmolW mass flow rate kg/sw weight fraction, species j -x dimension mXE dimension - oxygen exhaustion mXf deposit thickness mx PFRU thermocouple depth mSubscriptsa,att adhesion,attachmenta,i/o annulus, inner, outeract activation energy280Nomenclatureageing ageing processesamb ambientAOx antioxidantb,bulk bulk liquidback back diffusion (Crittenden et at. 1987b)cony convective heat transfercoupled coupled kineticsD, diff diffusiondep deposition, depositeff effectivef foulantf.pt. freezing pointg gas phasein inletmd inductionunit initiationins insulationmt gas/liquid interfaceiso isothermalI liquid phasem meanM,mass mass transfermet metalmix mixtureor orificeout outletp particle or precursorPOx peroxideprop propagationrem removalrxn reactionRM Russell’s kinetics, including mass transfer effectsRP Russell’s kinetics, excluding mass transfer effectss, surf surfacet, i, o tube, inner, outerterm termination281Nomenclaturew, i, o tube wall, inner, outerGreekB,B1 etc. general constantHildebrand solubility parameter (J/cm3)05AHf enthalpy change, solvation i/mo!AP pressure drop Patemperature difference KAL time increments so filmldeposit thickness mE Hayduk & Minhas factor -0 fouling flux m2.KJJmol fraction, species i-y activity coefficient -thermal conductivity W/m.Kdynamic viscosity Pa.sv kinematic viscosity m2/se ratio of fouling rate to reaction rate (s/rn)0 residence time sp density kg/rn3a standard deviationinduction period, space time hrwall shear stress PaWD deposit strength factor -282ReferencesReferencesAlexander, J.M. (1990). Gas Phase Ignition in Liquid Phase Air OxidationProcesses, Trans.I.Chem.E. , 68B, p17-23.Asomaning, S. (1990). The Role of Olefins in Fouling of Heat Exchangers, M.A.Scthesis, University of British Columbia.Asomaning, S. & Watkinson, A.P. (1992). Heat Exchanger Fouling by Olefin-KeroseneMixtures, Can. J. Chem. Eng. , 70, p444-451.Asomaning, S. (1993). Progress Reports in Aspects of Autoxidation Fouling: Contract#11982401, Argonne National Laboratory/University of British Columbia.Astarita, G. (1967) Mass Transfer with Chemical Reaction, publ. Elsevier, AmsterdamAtkins, G.T. (1962). What to Do About High Coking Rates, Petro/Chem Engineer, 34,p20-25.Belmar-Beiny, M.T. & Fryer, P.J. (1992). The Initial Stages of Fouling from WheyProtein Solutions. Proc. 3rd UK Natl. Heat Tfr. Conf., Birmingham, UK,I.Chem.E., Rugby, UKBohnet, M. (1987). Fouling of Heat Transfer Surfaces, Chem. Eng. Technology, 10,p113-125.Bott, T.R. & Walker, R.A. (1971). Fouling in Heat Transfer Equipment, The ChemicalEngineer , #255,p391-5Bott, T.R. (1988). General Fouling Problems. in Fouling Science and Technology.pubi. Kiuwer Academic Publishers, p 1-14.Bott, T.R. (1990). Fouling Notebook. pubi. I.Chem.E., Rugby, UKBowen, B.D. (1978). Fine Particle Deposition in Laminar Flow. Ph.D thesis, Universityof British Columbia.Bradley, S.E., Fryer, P.J., Griffin, T.R. & Wilson, D.I. (1989).The Mechanisms ofFood Fouling and their Implications for Fouling-Resistant Heat Exchanger Design,in Fouling and Cleaning in Food Processing. eds. Kessler, K.G. & Lund, D.B.,Munich, p12-24.Bradley, S.E. & Fryer, P.J. (1992). A Comparison of Two Fouling-Resistant HeatExchangers, Biofouling, 5,p295-314Braun, R. & Hausler, R.H. (1976). Contribution to Understanding Fouling Phenomena inthe Petroleum Industry, Paper 76-CSME/CSChE-23, 16th National Heat TransferConference, Aug. 8-11, St. Louis, MO,Brill, W.F. (1965). Selectivity in the Liquid Phase Autoxidation of Olefins. Adv. Chem.Ser. 51, p70-8.283ReferencesButler, R.C. & McCurdy, W.N.J. (1949). Fouling Rates and Cleaning Methods inRefinery Heat Exchangers, Trans. A.S.M.E. , 71, p843-849.Canapary, R.C. (1961). How to Control Refinery Fouling, Oil & Gas J. , 9, p1 14-118.Carisson, D.J. & Robb, J.C. (1966). Liquid Phase Oxidation of Hydrocarbons, Trans.Faraday Soc., 62,p3403-15Chantry, W.A. & Church, D.M. (1958). Design of High Velocity Forced CirculationReboilers for Fouling Service, Chem. Eng. Frog. , 54,p64-7.Charlesworth, D.H. (1963) Atomic Energy of Canada Ltd. Report, Report #AECL-1 761Chen, C.Y. (1993). Measurement of Fine Particle Deposition from Flowing Suspensions.M.A.Sc. thesis, U.B.C.Chenoweth, J.M. (1988). Liquid Fouling Monitoring Equipment. Fouling Science andTechnology. Kiuwer Academic Publishers. p49-66.Chenoweth, J.M. (1990). Final Report of the HTRI/TEMA Joint Committee to Review theFouling Section of the TEMA Standards, Heat Tfr. Eng., 11,(1), p73-107Chhabra, R.P. (1992). Prediction of Viscosity of Liquid Hydrocarbon Mixtures,A.I.Ch.E. Journal , 38,p1657-61.Cleaver, J.W. & Yates, B. (1975). A Sublayer Model for the Deposition of Particles fromTurbulent Flow, Chem. Eng. Sci. , 30,p983-92.Cleaver, J.W. & Yates, B. (1976). The Effect of Re-entrainment on Particle Deposition,Chem. Eng. Sci. , 31,p147-51.Coulson, J.M., Richardson, J.F. & Sinnott, R.K. (1983). Chemical Engineering Vol.6.Pergamon, OxfordCowan, J.C. & Weintritt, D.J. (1978). in Water Formed Scale Deposits. pubi. Gulf,Houston, TXCrawford, J.E. & Miller, J.M. (1963). Refinery Process Fouling - Effect of Trace Metalsand Remedial Techniques, Proc. A.P.I. Sect III (Refining), 43, p106-i 14.Crittenden, B.D. & Khater, E.M.H. (1984). Fouling in a Hydrocarbon Vaporiser. Proc.1st. UK Nati. Heat Transfer Conf., Univ. of Leeds, UK, I.Chem.E.Crittenden, B.D., Hout, S.A. & Alderman, N.J. (1987a). Model Experiments ofChemical Reaction Fouling, Chem. Eng. Res. Dev. , 65, p165-170.Crittenden, B.D., Kolaczkowski, S.T. & Hout, S.A. (1987b). Modelling HydrocarbonFouling, Chem. Eng. Res. Des. , 65,p171-9Crittenden, B.D. (1988a). Basic Science and Models of Reaction Fouling, in FoulingScience and Technology, pubi. Kiuwer Academic Publishers,p293-314.Crittenden, B.D. (1988b). Chemical Reaction Fouling of Heat Exchangers. in FoulingScience and Technology, publ. Kiuwer Academic Publishers. p315-332.284ReferencesCrittenden, B.D. & Alderman, N.J. (1988). Negative Fouling Resistances: The Effect ofSurface Roughness, Chem. Eng. Sci. , 43,p829-38.Crittenden, B.D., Kolaczkowski, S.T. & Downey, I.L. (1992). Fouling of Crude OilPreheat Exchangers, Chem. Eng. Res. Des. , 70,p547-5Crittenden, B.D., Kolaczkowski, S.T. & Takemoto, T. (1993). Use of In-Tube Inserts toReduce Fouling from Crude Oils. Proc. Heat Transfer, Atlanta, GA, A.LCh.E.Symp. Ser. 295,p300-7Danckwerts, P.V. (1970). Gas-Liquid Reactions. McGraw-Hill, LondonDaubert, T.E. & Danner, T.E. (1989). Physical and Thermodynamic Properties ofPureChemical Compounds. publ A.I.Ch.E., NYDavies, A.G. (1961). Organic Peroxides. publ. Butterworths, LondonDickakian, G. & Seay, S. (1988). Asphaltene Precipitation Primary Crude ExchangerFouling Mechanism.Oil & Gas J., 86(10), p 47-50.Dickakian, G. (1990). Fluid Catalytic Cracker Unit Fouling: Causes and Mitigation. Proc.A.I.Ch.E. Summer Meeting, San Diego, CA,Drews, A.W., Ed. (1989). ASTM Manual of Hydrocarbon Analysis, 4th edn. pubi.ASTM, Baltimore, MDDugan, C.F., Van Nostrand Jr., W.L. & Haluska, J.L. (1978). How AntifoulantsReduce the Energy Consumption of Refineries, Chem. Eng. Frog., 74,p53-7.Eaton, P. & Lux, R. (1984). Laboratory Fouling Test Apparatus For HydrocarbonFeedstocks, ASME HTD 35, p33-42.Epstein, N. (1981) Fouling: Technical Aspects, in Fouling ofHeat Transfer Equipment,eds. Somerscales, E.F.C. & Knudsen, J.G., publ. Hemisphere, p31-53Epstein, N. (1983). Thinking About Heat Transfer Fouling : A 5x5 Matrix, Heat TransferEngng., 4, p43-56.Epstein, N. (1988). General Thermal Fouling Models. in Fouling Science andTechnology., publ. Kiuwer Academic Publishers. p15-30.Epstein, N. (1993a). The Velocity Effect on Initial Fouling Rates, Israel ChemicalEngineering, 22,p32-7.Epstein, N. (1993b). A Model of the Initial Chemical Reaction Fouling Rate For FlowWithin a Heated Tube and its Verification, accepted for presentation at the 10thIntnl. Ht. Tfr. Conf., Brighton, UK (1994),Fernandez-Baujin, J.M. & Solomon, S.M. (1976). in Industrial and LaboratoryPyrolyses. eds. Aibright, L.F. & Crynes, B.L. Washington, A.C.S. Symp. Ser.32.p345-72285ReferencesFetissoff, P.E., Watkinson, A.P. & Epstein, N. (1982). Comparison of Two HeatTransfer Fouling Probes. Proc. 7th Int’l Heat Transfer Conference, MunichFetissoff, P.E. (1982). Comparison of Two Fouling Probes. M.A.Sc. thesis, Universityof British Columbia.Fisher (1959). Fisher Density/Temperature Tables. Fisher/TAG Manualfor Inspectors ofPetroleum. 29th edn., publ. Fisher ScientificFortuin, J.M.H., Musschenga, E.E. & Hamersma, P.J. (1992). Transfer Processes inTurbulent Pipe Flow Described by the ERSR Model, A.I.Ch.E.J. , 38,p343-60.Frankenfeld, J.W., Taylor, W.F. & Brinkman, D.W. (1983). Storage Stability of SynFuels from Oil Shale - Parts 1,2,3, md. Eng. Chem. Prod. Res. Dev. , 22, p608-627.Froment, G.F. (1981). Thermal Cracking for Olefins Production. Fundamentals and theirApplication to Industrial Problems, Chem. Eng. Sci. , 36,p1271-82Fryer, P.J. & Slater, N.K.H. (1985). A Direct Simulation Procedure for ChemicalReaction Fouling in Heat Exchangers, Chem. Eng. J. , 31,p97-107.Fryer, P.J. (1987). The Fouling of Heat Exchanger Networks. in Fouling Science andTechnology. publ. Kiuwer Academic Publishers, p51 1-528.Fryer, P.J., Bradley, S.E., Griffin, T.A. & Wilson, D.I. (1990). The Design of RobustAntifouling Heat Exchangers. presented at A.I.Chem.E. Summer Meeting, SanDiego, CAGarret-Price, B.A., Smith, B.A., Watts, R.L., Knudsen, J.G., K., Marner, W.J. &Suitor, J.W. (1985). Fouling ofHeat Exchangers; Characteristics, Costs,Prevention, Control and Removal. pubi. Noyes Publications, New JerseyGillies, W.V. (1979). Fouling and Its Control by Chemical Additives in HydrocarbonStreams. Proc. Inst. Corrosion Science and Technology, and Inst. of Chem. Eng.Fouling Conf., Univ. of Surrey, Guildford, UK,Gnielinski, V. (1986). in Warmeubertragung in Rohren. VDI-Verlag, Düsseldorf.Gough, M.J. & Rogers, J.V. (1987). Reduced Fouling by Enhanced Heat transfer UsingWire Matrix Mixing Elements, A.I.Ch.E. Symp. Ser. 257, 83,p16-21.Hausler, R.H. & Thalmayer, C.E. (1975). Fouling and Corrosion in Feed EffluentExchangers- Discussion of a New Test Method. Proc. API Refining 40th MidYear Meeting, Chicago, API.Hazlett, R.N., Hall, J.M. & Matson, M. (1977). Reactions of Aerated n-dodecaneLiquid Flowing Over Heated Metal Tubes, md. Eng. Chem. Prod. Res. Dev., 16,p171 -177.Hess, P.S. & O’Hare, G.A. (1950). Oxidation of Linseed Oils, md. Eng. Chem., 42,p1424-1431.286ReferencesHillyer, J.C. (1963). Organic Coolant Research Summary Report : July 1960-March 1963,Report DO- 16911, pubi. Phillips Petroleum.Hopkins, R.M. (1973). Fouling of Heated Stainless Steel Tubes with Ferric Oxide fromFlowing Water Suspensions. Ph.D. thesis, University of British Columbia.Horswill, E.C. & Ingold, K.U. (1966).The Oxidation of Phenols 1. The Oxidation ofDMP, 2,6-di-t-butylphenol and 2,6-dimethyiphenol with Peroxy Radicals, Can. J.Chem., 44,p2638.Howard, J.A. & Ingold, K.U. (1962). The Inhibited Autoxidation of Styrene. I. TheDeuterium Isotope Effect For Inhibition by 2,6-Di-tert-butyl-4-methylphenol, Can.J. Chem., 40, 1851-964.Howard, J.A. & Ingold, K.U. (1966). Absolute Rate Constants for HydrocarbonAutoxidation III. x-Methylstyrene, 13-Methyistyrene & Indene, Can. J. Chem., 44,p1113-1118.Huntrods, R.S., Nighswander, A., Mehrotra, A.K. & Behie, L.A. (1989). Modelling ofCoke Formation in Gas Quenchers of Industrial Ethane Cracking Furnaces, Chem.Eng. Res. Des., 67,p632-8.Jackman, A.P. & Aris, R. (1971). in Proc. 4th European Symp. on Chemical ReactionEngineering, publ. Pergamon, Oxford.Kay, J.M. & Nedderman, R.M. (1985). Fluid Mechanics and Transfer Processes. publ.CUP, CambridgeKern, D.Q. (1950). Process Heat Transfer. publ. McGraw-Hill, NYKern, D.Q. & Seaton, R.E. (1959). A Theoretical Analysis of Thermal SurfaceFouling, Brit. Chem. Engng., 4,p258-62.Kiaren, D.G. (1983). The Fluid Bed Heat Exchanger: Principles and Modes of Operationand Heat Transfer Results Under Severe Fouling Conditions.Fouling Prey. Res.Dig., 5(1), p3-17.Knudsen, J.G. & Katz, D.L. (1958) Fluid Dynamics and Heat Transfer, publ. McGraw-Hill, NYKotjabasakis, E. & Linnhoff, B. (1987). Compensating for Fouling in Heat ExchangerNetworks. in Fouling Science and Technology. pubi. Kiuwer AcademicPublishers, p529-556.Lai, R.C. (1992). Chemical Reaction Fouling Involving Model Oxidative Solutions.B.A.Sc. Summer thesis, University of British Columbia.Lal, R.C. (1993). Autoxidation and Antioxidation: Fighting Fouling Chemically. B.A.Sc.thesis, University of British Columbia.Lambourn, G.A. & Durrieu, M. (1983). Fouling in Crude Oil Preheat Trains, in HeatExchangers- Theory and Practice. eds. Taborek, Hewitt & Afgan, publ.Hemisphere, NY. p841-852.287ReferencesLatos, E.J. & Franke, F.H. (1982). Relative Thermal Fouling Rates of Petroleum andShale Oils, in Proc. Corrosion 82, NACE.Lawler, D. (1979). in Proc. Inst. of Corrosion Sci.fI.Chem.E. Conf, University ofSurrey, Guildford, UK, p155Link, W.E. & Formo, M.W. (1962). Analysis of Autoxidation Mixtures. in Autoxidationand Antioxidants. publ. J.Wiley & Sons, NYLloyd, W.G. & Zimmerman, R.G. (1965). Antioxidant Inhibition by Substituted Phenols,md. Eng. Chem. Prod. Res. Dev., 4,p180-5.Loo, C.E. & Bridgewater, J. (1981). Theory of Thermal Stresses and Deposit Removal.in Proc. Progress in the Prevention of Fouling in Industrial Plant, NottinghamUniversity, UK, Inst. of Corrosion Sci. & Technology Conf.Mair, R.D. & Graupner, A.J. (1964). Determination of Organic Peroxides by IodineLiberation Procedures, Analytical Chem., 36, p 194-204.Mayo, F.R. (1953). J. Am. Chem. Soc. , 75,p6133-41.Mayo, F.R. & Miller, A.A. (1956a). The Oxidation of Unsaturated Compounds. I. TheOxidation of Styrene J. Am. Chem. Soc., 78, p1017-1022.Mayo, F.R. & Miller, A.A. (1956b). The Oxidation of Unsaturated Compounds. II.Reactions of Styrene Peroxide J. Am. Chem. Soc., 78, p1023-1034.Mayo, F.R., Miller, A.A. & Russell, G.A. (1958). Oxidation of UnsaturatedCompounds IX, J. Am. Chem. Soc., 80,p2500-7Mayo, F.R. (1958). The Oxidation of Unsaturated Compounds. V. The Effect of OxygenPressure on the Oxidation of Styrene J. Am. Chem. Soc., 80,p2465-80Mayo, F.R. & Lan, B.L. (1986). Gum and Deposit Formation from Jet Turbine andDiesel Fuels at 130°C, md. Eng. Chem. Prod. Res. Dev., 25,p333-48.Mayo, F.R., Stavinoha, L.L. & Lee, G.H. (1988). Source of Jet Fuel Thermal OxidationTester Deposits from an Oxidised JP-8 Fuel, md. Eng. Chem. Res., 27, p362-363.Mazius, Z.K. (1965). in The Oxidation ofHydrocarbons in the Liquid Phase. pubi.MacMillan, NY p32.McNab, G.S. & Meisen, A. (1973). Thermophoresis in Liquids, J. Colloid Inteif Sci.,44, p339-349.Melo, L.F. & Pinheiro, J.D. (1988). Particle Transport in Fouling Caused by Kaolin-Water Suspensions on Copper Tubes, Can. J. Chem. Eng., 66,p36-41.Metzner, A.B. & Friend, W.L. (1958). Theoretical Analogies between Heat, Mass andMomentum Transfer and Modifications for Fluids of high Prandtl or SchmidtNumbers, Can. J. Chem. Eng., 36,p235-40.288ReferencesMonrad, C.C. & Pelton, J.F. (1942). Heat Transfer by Convection in Annular Spaces,Trans. A.I.Ch.E., 38, p593-608.Morris, R.E., Hazlett, R.N. & Mdllvaine, C.L. (1988). The Effects of StabilizerAdditives on the Thermal Stability of Jet Fuel, md. Eng. Chem. Res., 27, p1524-1528.Morris, R.E. & Hazlett, R.N. (1989). Methods for Quantifying IFTOT Heater TubeDeposits Produced from Jet Fuels, Energy & Fuels, 3,p262-7.Morris, R.E. & Mushrush, G.W. (1991). Fuel Instability Model Studies: the Liquid-Phase Co-Oxidation of Thiols and Indene by Oxygen, Energy & Fuels, 5, p744-748.Muller-Steinhagen, H., Watkinson, A.P. & Epstein, N. (1986). Subcooled Boiling andConvective Heat Transfer to Heptane Flowing Inside an Annulus and Across aCoiled Wire - Parts I and II, J. Heat Trans., 108, p922-933.MUller-Steinhagen, H.M., Reif, F., Epstein, N. & Watkinson, A.P. (1988). Influence ofOperating Conditions on Particulate Fouling, Can. J. Chem. Eng., 66,p42-50.Muller-Steinhagen, H.M. (1990). in Proc. G.V.C. Conf., MunichMushrush, G.W., Watkins, J.M., Hazlett, R.N., Hardy, R.R. & Eaton, H.G. (1988).Liquid Phase Oxidation of Thiophenol and Indene by t-butyl hydroperoxide andOxygen, Fuel. Sci. & Tech. Int’l., 6,p165-83.NACE (1970). N.A.C.E. Report, Mat. Prot. 1970Nathan, C.C. (1970). Materials Protection and Peiformance .9,p15-8.Neiman, M. (1964). Russ. Chem. Rev., 33, p13.Nelson, W.L. (1934). Fouling of Heat Exchangers, Refinery & Nat. Gasol. Mfr., 13,p271.Nelson, W.L. (1958) Petroleum RefInery Engineering, 4th edn. publ. McGraw-Hill, NY,p549Nicholson, A. (1991). The Effect of Oxygen Concentration on Methylacrylic AcidStability, Plant/Oper. Prog., 10,p171-83.Nijsing, R. (1964). Diffusional and Kinetic Phenomena Associated with Fouling. EuratomReport EUR 543.eNixon, A.C. (1962) ‘Autoxidation and Antioxidations of Petroleum’. Ch. 17 inAutoxidation and Antioxidants publ. Interscience, NYNorton, C.J. & Drayer, D.E. (1968). Liquid-Phase Oxidation of High Molecular Weight1-Alkenes. Adv. Chem. Ser. 75,p78-92.Oufer, L.S. (1990). Fouling Characteristics of Organic Fluids. Ph.D thesis, Oregon StateUniversity.289ReferencesOufer, L.S., Breber, G. & Knudsen, J.G. (1990) A Study of Chemical Reaction Foulingunder Sub-Cooled Boiling Conditions. A.I.Ch.E. National Meeting, San Diego,CA, Paper #35ePalen, J.W. & Westwater, J.W. (1966). Heat Transfer and Fouling Rates During PoolBoiling of Calcium Sulfate Solutions, Chem. Eng. Frog. Symp. Ser. 64 , 62, p77-82.Panchal, C.B. & Watkinson, A.P. (1993). Chemical Reaction Fouling Model for Single-Phase Heat Transfer. AJ.Ch.E. Symp. Ser. 295, 89,p323-34Parkins, W.E. (1961). Surface Film Formation in Reactor Systems. Atomic Energy ofCanada Ltd. Report AECL-1265Paterson, W.R. & Fryer, P.J. (1988). A Reaction Engineering Approach to the Analysisof Fouling, Chem. Eng. Sci., 43,p1714-7.Pinhero, J.D.R.S., Ed. (1981). Fouling of Heat Transfer Surfaces. in Heat Exchangers -Thermal Hydraulic Fundamentals and Design., publ. McGraw-Hill, NYPritchard, A.M. (1979). Fouling - Science or Art? An Investigation of Fouling andAntifouling Measures in the British Isles, in Fouling ofHeat Exchange Equipment.pubi. Hemisphere, NY.Pritchard, A.M. (1988). The Economics of Fouling. in Fouling Science and Technology.pubi. Kiuwer Academic Publishers. p3 1-48.Pritchard, A.M. (1990). The Economics of Fouling, in Proc. G.V.C. Conf, MunichPuttagunta, V.R., Miyadoyne, A. & Singh, B. (1992). Viscosity-Temperature Correlationfor Prediction of Kinematic Viscosity of Conventional Petroleum Liquid, Chem.Eng. Res. Des., 70, 627-63 1.Rabas, T.J. & Chenoweth, J.M. (1991) Eds. Fouling and Enhancement Interactions, HTDVol. 164, publ. ASMEReddy, K.T., Cernansky, N.P. & Cohen, R.S. (1988). Modified Reaction Mechanismof Aerated n-Dodecane Liquid Flowing Over Heated Metal Tubes, Energy & Fuels,2,p205-13.Reich, L. & Stivala, S.S. (1969). Autoxidation ofHydrocarbons and Polyolefins -Kineticsand Mechanisms. publ. Marcel Dekker mc, NY.Reid, R.C., Prausnitz, J.M. & Poling, B.E. (1987). The Properties of Gases and Liquids.pubi. McGraw-Hill, NY.Rosenhow, W.M. & Hartnett, J.P. (1973) Hanbook of Heat Transfer, pubi. McGrawHill, NYRuckenstein, E. & Prieve, D.C. (1973). The Rate of Deposition of Brownian Particlesunder the Action of London and Double Layer Forces, J. Chem. Soc. Faraday IL69,p1522-36.290ReferencesRussell, G.A. (1955). The Competitive Oxidation of Cumene and Tetralin, J. Am. Chem.Soc., 77,p4583-8 90Russell, G.A. (1956a). Oxidation of Unsaturated Compounds. III. Products of theReaction of Indene and Oxygen, J. Am. Chem. Soc., 78, p 1035-1040.Russell, G.A. (1956b). Oxidation of Unsaturated Compounds. IV. Kinetics intheReaction of Indene with Oxygen, J.Am. Chem. Soc., 78, p 1041-1044.Russell, G.A. (1956c). The Effect of Oxygen on the Decomposition of AIBN and BenzoylPeroxide in Aromatic Solvents, J. Am. Chem. Soc., 78, plO4.’7-S4..Scarborough, C.E., Cherrington, D.C., Diener, R. & Golan, L.P. (1979). Coking ofCrude Oil at High Heat Flux Levels, Chem. Eng. Prog., 75,p41-6.Schwartz, F.G., Whisman, M.L., Alibright, C.S. & Ward, C.C. (1964). StorageStability of Gasoline - Fundamentals of Gum Formation. Bull. 626 Bureau ofMines, Washington, D.C.Scott, G. (1965). Atmospheric Oxidation and Antioxidants. pub!. Elsevier, AmsterdamSieder, E.N. & Tate, G.E. (1936). Heat Transfer and Pressure Drop of Liquids in Tubes,md. Eng. Chem., 28, p1429.Shell Technical Bulletin - Shell Thermia Oils, (1990). Personal Communication, PetroCanada.Small, P.A. (1953). Some Factors Affecting the Solubility of Polymers, J. Anal. Chem.,3, p73-80.Smith, J.D. (1969). Fuel for the Supersonic Transport - Effects of Deposits on HeatTransfer to Aviation Kerosene, md. Eng. Chem. Prod. Res. Dev., 8,p299-308.Stephenson, W.K. & Rowe, C.T. (1993). Fouling Control in Ethylene Recovery Systems.A.I.Ch.E. Symp. Ser. 295, 89,p335-40Steele, G.L., Brinkman, D.W. & Whisman, M.L. (1981), Deposit Formation in LiquidFuels - 1. Effect of Coal Derived Lewis Bases on Storage Stability of Jet A TurbineFuel, Fuel, 60,p477-80Sundaram, K.M. & Froment, G.F. (1979). Kinetics of Coke Deposition in the ThermalCracking of Propane Chem. Eng. Sci., 34,p635-44.Taborek, J., Aoki, T., Ritter, R.B., Palen, J.W. & Knudsen, J.G. (1972). Fouling -the Major Unresolved Problem in Heat Transfer, Chem. Eng. Prog., 68, p59-67.Taylor, W.F. & Wallace, T.J. (1967). Kinetics of Deposit Formation from HydrocarbonFuels at High Temperatures- Genera! Features of the Process, md. Eng. Chem.Proc. Res. Dev., 6, 258-62.Taylor, W.F. & Wallace, T.J. (1968a). Kinetics of Deposit Formation from HydrocarbonsFuels at High Temperatures- Effect of Trace Sulfur Compounds, md. Eng. Chem.Prod. Res. Dev., 7, 198-202.291ReferencesTaylor, W.F. (1968b). Mechanism of Deposit Formation in Wing Tanks. Proc. SAEAeronautic and Space Engng., Los Angeles, p2811-2822Taylor, W.F. (1969a). Kinetics of Deposit Formation from Hydrocarbons - IV. Additiveand Surface Coating Effects, J.Appl. Chem., 19,p222-6.Taylor, W.F. (1969b). Kinetics of Deposit Formation from Hydrocarbons - FuelComposition Studies, md. Eng. Chem. Prod. Res. Dev., 8,p375-80.Taylor, W.F. (1974). Deposit Formation From De-Oxygenated Hydrocarbons- 1. GeneralFeatures, md. Eng. Chem. Prod. Res. Dev., 13,p133-8.Taylor, W.F. (1976). Deposit Formation From De-Oxygenated Hydrocarbons- 2. Effectof Trace Sulfur Compounds, md. Eng. Chem. Prod. Res. Dev., 15,p64-8.Taylor, W.F. (1978). Deposit Formation From Dc-Oxygenated Hydrocarbons- Effect ofTrace Nitrogen and Oxygen Compounds, md. Eng. Chem. Prod. Res. Dev., 17,p86-90.Taylor, W.F. (1980). Deposit Formation From Dc-Oxygenated Hydrocarbons- Studies inPure Compound Systems, md. Eng. Chem. Prod. Res. Dev., 19,p65-70.Tobolsky, A.V., Metz, D.J. & Mesrobian, R.B. (1950). J. Am. Chem. Soc., 72, p1942.Turner, C.W. (1993). Rates of Particle Deposition From Aqueous Suspensions inTurbulent Flow : A Comparison of Theory With Experiment, Chem. Eng. Sci.,48,p2185-9.Ueno, Y., Kotetsu, N. & Yoshida, T. (1974). Aromatikkusu, 26,p117-20.Van de Vusse, J.G. (1961). Mass Transfer With Chemical Reaction, Chem. Eng. Sci.,16, p21-30.Van Nostrand Jr., W.L., Leach, S.H. & Haluska, J.L. (1981). Economic PenaltiesAssociated with the Fouling of Refinery Heat Transfer Equipment, in Fouling ofHeat Transfer Equipment. publ. Hemisphere, NY. p619-643.Van Sicide, D.E., Mayo, F.R. & Arluck, R.M. (1965a). Liquid Phase Oxidations ofCyclic Alkenes J. Am. Chem. Soc., 87, p4824.Van Sickle, D.E., Mayo, F.R. & Arluck,.R.M. (1965b). The Liquid Phase Oxidation ofCyclopentene, J. Am. Chem. Soc., 87, 4832.Van Sickle, D.E., Mayo, F.R., Gould, E.S. & Arluck, R.M. (1967). Effects ofExperimental Variables in Oxidations of Alkenes J. Am. Chem. Soc., 89, p977.Van Sickle, D.E., Mayo, F.R., GArluck, R.M. & Syz, M.G. (1969). Oxidations ofAcyclic Alkenes J. Am. Chem. Soc., 89, p967.Vatistas, N. (1989). The Effect of Adhesion Time On Particle Deposition, Chem. Eng.Sci., 44,p1603-8Vranos, A., Marteny, P.J. & Knight, B.A. (1981). Determination of Coking Rate in JetFuel, in Fouling ofHeat Transfer Equipment. pubi. Hemisphere, NY p489-499.292ReferencesWatkinson, A.P. (1968). Particulate Fouling of Sensible Heat Exchangers. Ph.D. thesis,University of British Columbia.Watkinson, A.P. & Epstein, N. (1969). Gas Oil Fouling in a Sensible Heat Exchanger,Chem. Eng. Frog. Sym. Ser., 65, p84-90.Watkinson, A.P. (1988). Critical Review of Organic Fluid Fouling : Final Report #ANL/CNSV-TM-208 Argonne National Laboratory, Ill.Watkinson, A.P. (1991). Interactions of Enhancement and Fouling, in Fouling andEnhancement Interactions, eds. Rabas & Chenoweth, HTD Vol. 165, p1-8Watkinson, A.P. (1992). Chemical Reaction Fouling of Organic Fluids, Chem. Eng.Technol., 15, p82-90.Weiland, J.H.J., McCay, R.C. & Barnes, J.E. (1949). Rates of Fouling and Cleaning ofUnfired Heat Exchanger Equipment, Trans. A.S.M.E., 71,p849-53.Weinland, B.W. (1967). Some New Antifoulant Applications, Can. Petroleum , March,p58-60.White, C.M., Jones, L. & Li, N.C. (1983). Aging of SRC-II Middle Distillates fromIllinois No.6 Coal Part II, Fuel, 62,p1397-403White, F.M. (1991). Viscous Fluid Flow, publ. McGraw-Hill, NYWiegand, J.H. (1945). Discussion of Annular Heat Transfer Coefficient for TurbulentFlow, Trans. A.I.Ch.E., 41,p147-52.Wilson, D.I. & Watkinson, A.P. (1992). Mechanism and Solvent Effects in ChemicalReaction Fouling, in Proc. 3rd UK National Ht Tfr Conf., Birmingham, UK, publ.I.Chem.E., Rugby, UKYung, B.P.K., Merry, H. & Bott, T.R. (1989). The Role of Turbulent Bursts in ParticleRe-entrainment in Aqueous Systems, Chem. Eng. Sci., 44,p873-82.Zhang, G. (1990). Progress Report Appendices, Contract #11982401, UBC/ANLZhang, G. & Watkinson, A.P. (1991). Modelling of Autoxidation Fouling, presented atC.S.Ch.E. Conf., Vancouver, B.C.Zhang, G., Watkinson, A.P. & Wilson, D.I. (1992). Final Progress Report, Contract#11982401, Argonne National Laboratory/University of British Columbia, publ.ANL, ILL.Zhang, G., Wilson, D.I. & Watkinson, A.P. (1993). Fouling of a Cyclic Olefin on Plainand Enhanced Surfaces, A.I. Ch.E. Symp. Ser. 295, 89, p3 14-322293Appendfx A.]Table A. 1.1 Experiment NomenclatureExperiment Type Apparatus Used Reference NumberInitial Fouling Experiments PFRU 001,Batch Kinetic Experiments SCR 101,Batch Antioxidation/Solvent Studies SCR 301,(performed with R.Lai)Ageing Experiments Ageing Oven, TGA 401,Tube Fouling Experiments TFU 501,294Appendix A.IPumpLow Level AlarmPressure AlarmTank HeaterPressure Relief ValvesTransformerPower VariacVoltage/Current MetersCurrent TransformerPCData Logging SoftwareTemperature ConversionAnalog/Digital ConversionLiquid Flow RotameterGas Flow RotameterCooling Water RotameterOrifice AP TransducerTube AP TransducerPabsolute TransducerPressure GuagesAP Power supplies/guagesFluid ThermocouplesTemperature DisplayTank Temperature ControllerHigh Temperature AlarmCooling Water Failure AlarmSurface ThermocouplesSihi model CAO 3102KE (ss)3 h.p. 30 HSA motorOmega LV- 1102Omega PSW Model 1213 kW 2O8Vac 10 (low flux)Nupro RL3 PRV208-19 Vac 10Powerstat 1 156D-3Y30 2O8Vac: 0-50 Aac0-2OAac0-2OAac800-5AacPackard Bell PB 8810LABTECHEXP- 16/Gain 100DAS-8 boardBrooks R9M-25- 1Brooks R-6- 15-ABrooks R-1OM-25-1Omega PX820-O1ODVOmega PX820- O1ODVOmega PX1O2-100Omega Bourdon guagesOmega DP-350Omega ‘K’ typeOmega Model 650Omega CN 911Omega CN 9000Omega CN 9000‘K Type thin foil sensors2 chambered, self priming pumpWestinghouse, explosion proofBall-Cock Device205-2000 kPaChromalox MTS-230, ss25-600 psig rangeRIG ElectronicsSuperior Electric Co.Novatron AgenciesHammond ElectricIBM XT CompatibleLaboratory Technologies Corp.Metrabyte Corp.Metrabyte Corp.ss float: mm scaless float: % scaless float: mm scale0-10 psid0-10 psid0-100 psig0-100 psig0-2V analog output1/8,1/16”: ss, groundedThermocouple Selector BoardlmV/°C analog ouputController/Trip SwitchController/Trip SwitchRdF Corp. ‘Stikon’ #201 12tinned copper overbraidTable A. 1.2 TFU Equipment SpecificationsEquipment Specification Comment295Appendix A.1Figure A.1.1 TFIJ Orifice and Rotameter Flow Calibrations.1)3-II1401201008060402000. . Plates DP =Rotameter Reading = -0.05 0.10 0.15 0.20Mass Flow Rate(kg/s)- 0.02648 + 6.0902 [kg/sI14.614 + 617.60[kg/s] RA2RA2= 0.996= 0.997296Tj I OCIRCUITBREAKERred___________________II_fØI’ll150AHHr-i-JILLOAD-l9VacTRANSFORMER600A‘VARIACblackExchangerSectionResetSwitchH3-phase23OVac30AWallTempAlaimRelay#1 CoolantTempAlannRelay#2PumpStartStopSwitchORGANICCHEMICALREACTIONFOULINGHEATEXCHANGERElectrical ControlLayout -MainHeater ContolandInterlinkedTripstoSystemIanWilsonCHML229November 1992Appendix A.1Figure A 1.3 Schematic Diagram of TFU Signal Processing System: Meter OutputsSurface thermocouple signals were monitored by an EXP-16 Multiplexer board which wasconnected directly to the DAS-8 board in the PC.2980,E>C0U0a0U0Appendix A. 1Figure A1.4 Schematic Diagram of TFU Heated Section Voltage Processing CircuitC-)0>(NiEC-)C-)-Q>cC(Noci(NEDC-,I I299Appendix A.1Figure A. 1.5 Schematic Representation of Mass Transfer Effects in Gas/Liquid Reactionsgas phasemass transportfluid t I’9renewalo bubblesbulk mixingO 4..liqwd phase impellerFilm Theory Interface Behaviour in the Fast Reaction Regimegas film liquid film cC ‘ I BA :ooo mol/)25.5 mol/m3 /5.4 mourn3 (not to scale)o oD,G D,L(concentrations for O.4M indene in Paraflex at 100°C; air at 377kPa)300Appendix A.2Table A.2.1 Comparison of indene Autoxidation Kinetics under TFU ConditionsExperiment Tsurf,o Tbulk kR Initial Gum Yield g*Rate____________(°C) (°C) (moI/L)/hr (g/L.hr) ±0.1 (gIg) (gIL)502 101 0.0321 2.60 10 ±1503 199.7 102.8 0.0299 2.55 0.61 11 ±1504 198.3 101.3 0.0292 2.96 0.51 12±1505 222.6 101.4 0.0330 2.96 0.51 12.5±1506 187.4 100.2 0.0228 2.53 0.47 12 ±1507 211.5 100.8 0.0316 2.62 0.57 11 ±1508 199.4 102.3 0.0358 2.78 0.48 12 ±1509 212.4 103.6 0.0390 2.45 0.42 12 ±2510 211.3 101.9 0.0359 2.82 0.52 11 ±1511 209.1 101.5 0.0357 2.83 0.52 11 ±2512 211.3 101.5 0.0356 3.2±0.15 0.79 12±20.405 M indene in Paraflex; air flow rate 300 mLIL.min (ntp)301Appendix A.2Figure A.2. 1 Effects of Surface Temperature on Chemically Initiated Fouling Solutions ofIndene in Paraflex0.0010B0.0009 B 210 °C• 224°C0.9008 239.5°CB0 0007 o 247 4°CBo 252.4°C o B0.0006 0 254.6°C . 000005 00C 0.000400.0003 / I0[14 0.0002 /0.000 100.0000___,,.l....l..,.l..,,I I....0 50 100 150 200 250 300 350 400 450 500Time(minutes)Tbulk 100°C; Poxygen 79.2 kPa; 0.41 mol/L indene; 2.5 mM benzoyl peroxide302Appendix A.2Figure A.2.2a Effect of Flow Velocity on Initiated Fouling at 222°CFouling Resistance Profiles0.00100.00090.00080.00070.00060.00050.00040.00030.00020.00010.0000Time(minutes)Figure A.2.2b Effect of Flow Velocity on Initiated Fouling at Tsurf 220°CGum Analysis ResultsGum(gIL)40353025 -2015 -105-QL0 1 2 3 4 5 6 7Time(hours)Tbulk 100°C: oxygen 79 kPa; 1.0 mM bP initiation; 0.41 M indene in Paraflex0 50 100 150 200 250 300 350 400A• Re=3300Re=4920 A• AA Re=6513AA••A AA303Avvendix A.2IThuIk 100°C; 0.41 mol/L indene in Paraflex; 2.5 mM bP initiation; pair 79 kPFigure A.2.2c Effect of Flow Velocity on Initiated Fouling at 248°CGum Analysis Results0.30 -0.250.20Gum(xO.O1)[gIL]0.150 Re=3300, 4 Re=40200 Re=6513• Re=4020.00.10 • •00.05 000I0.00 I. ..I.. .. .. .1...0 1 2 3 4 5Time[hours]Figure A.2.2d Effect of Flow Velocity on Initiated Fouling at Tsf 248°CFouling Resistance Profiles0 Re = 33000.00080.00070.00060.00050.0004U0Re = 4020Re = 4920Re = 6513 04.000.00030.00020.00010.U 0aa0000 50 100 150 200 250 300Time(minutes)304Appendix A.2Figure A2.3a-c20181614121086420’0.400.350.300. Number, Soluble Gum and Indene Concentration Resultsin TFU Fouling Runs-T8Vañedo 501• 502‘ 503A 504505• 5065076055504540353025201510500l • 10o-‘I — I — I — I — I— I - I - I1 2 3 4 5 6 7 8 9a. POxdb. [gum]II0 501 A•502t 503A504 9r A- 13505, E5() •:- 507—I . • I • I •) 1 2 3 4 5 6 7 8 9*Q‘‘( ;0 501A 50413 505 :-•506r 507503502. I • I • I • I • I • I • I •c. [indene]0 1 2 3 4 5Time (hours)6 7 8 9305C)344..UiQQiOU,QUiOLOUiQ0t’r’——.—TjC) C)O)CDC)CD 0 C) 0 C)[iene](mo]IL)9p9p9pppp00--r\.)r’.iww-0Ui0C-fl0Ui0UiQC-flC-fluiUiC-li—--“--.00r’JQLQODC) CDCD IPOx(meqL)••0DO00 03 U,9) C)Gum(giL)034aOC-fl(fT010101--Q0OrO oc 000.(J,.g).ID0..•,.I.I.,.I...h.,!...I...I,,,00....I...4..i....L....I....I.UiU,U,LflCJi-—--.-‘.00-a•aø01Appendix A. 3Appendix A.3 Sample CalculationsA. 3.1 TFU Operating ConditionsData from Run 503 (initial period); W = 0.11 kg/s; Tb,jn = 100.21 C;Tb,0t= 115.49 C; Tsuff,0 = 199.7 CTb,mean = (100.21+115.49)/2 107.85 CTfilrn = (199.7+107.85)/2 = 153.8 CPhysical Properties at averaged temperaturesDensity: [indene] = 0.41 rnolJL P’PPfex = (852.7+14.96.[0.41])1852.7= 1.0072density (107.85 C) = 1.0072 (864.04 - (0.588)(107.85)) = 806.24 kg/rn3density (153.8 C) = 1.0072 (864.04 - (0.588)(153.8)) = 779.03 kg/rn3Kinematic viscosity - calculated using Equation [3.5] and an indene correction, 0.79, for0.41 M indene in Paraflexv(107.8C) = 1.9247x106m/sv(153.8C) = 1.0235x10/sReynolds numberurn = 4W/pd12using a tube thickness of 12 thou, dg,j = 8.9154x103murn = 4(0.110)/ (806.24)(8.9 1 543(103)2= 2.186 mIsRe = umdj/v= 2.186 (8.9154x103)/(1 .9247x106)= 10124turbulent flowFriction factor f = 0.007674 (experimental) (Fanning)Blasius f = 0.079 Re-025 [Re > 4000]= 0.079 (10124)0.25= 0.00787 good agreementGnielinski f = 0.25(1.82 logioRe - 1.64)2 [Re > 2300]= 0.25(1.82 (4.005) - 1.64)-2= 0.00783 good agreementFriction velocity u = um (f/2)= 2.186 1(0.007674/2)= 0.1354m1s307Appendix A.3A.3.2 Heat Transfer Calculations in the TFUTable 6.3 gives the initial heat transfer results for Run 504 (W 0.110 kg/s; Remean =9976). The heat flux was based on a tube thickness of 10 thou; voltage = 16.7V, current =220.4 A. The heat loss through the insulation (diameter 20cm, heated length 762 mm)was estimated using a free convection coefficient of 10 W1m2.K, givingConvective heat loss = 10 (0.762) (0.2)- Tambient)= 4.78 (38.07 -21.16)= 81WHeating Power = (16.7)(220.4) - 81 = 3681- 81= 3600WHeat transferred to fluid = WCPmean (Tb, out - Tb,in)Tb, mean = (114.6 + 99.4)/2 = 107 CCpmean = 1874.7+107 (3.7559) = 2277 J/kg.K= 0.11 (2277) (114.6- 99.4)= 3807 W +5% discrepancyHeat Flux q = 36001( (0.762)(0.009017))= 166.8 kW/m2The bulk temperature at axial position x is estimated asTb(x) = 99.4 + (114.6-99.4) x 1(0.762)= 99.4 + 19.947xThermocouple 6 was located at x = 0.435m;Tb(0.435) = 99.4 + (19.947) (0.435)= 108.08 CThe surface thermocouple reading was 196.2 C. With d = 0.009017 and d0 =0.009525m, Equation [3.28] givesT,(x) = T,0(x) + (3600/2) (5.76775 x103)met = 14.214 + 0.014819 (196.2) - 2.6224 x106 (196.2)2= 17.02 W/m.K= 197.4 CThe local heat transfer coefficient ish(x) = 166 800/(197.4- 108.08)= 1868 W/m2.KThe value in Table 6.3 was taken directly from the spreadsheet. at 1869 W1m2.K.The overall heat transfer coefficient was calculated using thermocouples 4-10, asU0 = (1/573.06) 1 h(j) &x(j) j = 4,10similarly,T,0 = (1/573.06) 1 T,0 (j) &x(j)308Appendix A.3The weighting factor &x(j) was calculated using&x(j) = 1/2 (x(j+1) + x(j)) - 1/2 (x(j)-x(j-1))forj = 10 x(j+1) = 0.762 in this Run.The results are given in Table 6.3.The overall Nusselt number for heat transfer is calculated from Nu = U0d/;2. = 0.1311-0.00014123 (107)= 0.116W/m.KNu = 0.009017 (1873)/0.116= 145.6The Gnielinski correlation value is given by Equation (3.21)f = 0.25 (1.82 logio(9976) - 1.64)2= 0.007864f/2 = 0.003932The Prandtl number Pr (806.24)(1.93 xlO-6)(2277)/(0.116)30.54Equation (3.21) Nu = 0.003932 (8976) 30.54/[1+12.7(1 0.003932) (30.50.67 4)]= 133.62The prediction is 8% lower than the observed value - a surface temperature correction factorsuch as that of Seider and Tate has not been included.A.3.3 Particle Velocities under TFU Conditions (Run 503)The dimensionless particle relaxation time, t, is given by Equation [7.13]. Assuming thatp, = 990 kg/rn3,t,+ = (0.1354)2(990/779.03)d/18(1 .0235x 10-6)2= 1.2356x109dGiving d7 m 0.060518 m 0.400these values are at the lower bound of the inertial coasting regime. The mass transfercoefficients are given byInertial coasting kim = [(0.1354)(779)/(5.23)(990)] exp(0.48 t+)= 0.0204 t exp(0.48t)Diffusion kD = (0.1354)/(11 .8)(Sc0667D = 1.38 x1023 (153.8+273.15)/3 (pv)d= 7.8405x10’9/dSc = v d 1.2754 x1018= (1.3054x10’2)d309Appendix A.3Group ‘A’ globules in Table 6.4d tp kim Sc kD7 m 0.0605 1.27 x103 mis 9.14 x106 2.61x10-7mJs18 m 0.400 9.89 x103 mis 23.5 x106 1.39x10m1sA.3.4 Calculation of Convective Mass Transfer CoefficientsThe convective mass transfer coefficient under the conditions in Run 503 is also given byMetzner and Friend’s analogykD = (0.1354)/(11.8)(Sc667) (mis)where Sc is calculated using the diffusion coefficient estimation method of Hayduk andMinhas (Reid et al. (1987). Epstein (1993b) used this analogy in his calculation of styrenemass transfer coefficients in kerosene, but replaced the isothermal constant (11.8) with afitted constant (502.3) and fluid properties evaluated at the surface temperature. Haydukand Minhas’ method requires an estimate of the molal volume of the component, whichwas based on indene (116.86 cm3/mol) and oxygen using the Le Bas contribution methoddescribed in Coulson et al. (1985). For indene,= 10.2/(116.86)-0.791= -0.7037viscosity (cP) = 0.001 (1.0235)(779.3)= 0.797 cPD = 13.3 x1012 (273. 15+153.8)’ (0.797)°°I(1 16.86)0.713.905 x109m2/sSc = 1.0235 x106/3.9 5 x109= 262k = 2.8x104m1sEvaluating fluid properties at the surface temperature gives for indeneviscosity 0.65 19 x106m2/sD = 3.905 x109m2/sSc = 101.9km = 1.225xl05m1sthus the change in Sc067 (53%) does not influence km as much as the change in constant(420%). The values for the first three hydroperoxides and oxygen were calculated as310Appendix A. 3VA Dx109 Sc kmXlO4 D1,x109 Sc kmxlO5(cm3/mol) (m2Is) (mis) (m2Is) (mis)indene,MH 116.9 3.905 262 2.8 6.4 102 1.22MOOH 135.4 3.528 290 2.6 5.8 112 1.15(MOO)2H 267.1 2.196 466 1.9 3.7 177 0.85(MOO)3H 432.1 1.566 653 1.5 2.6 246 0.68oxygen 25.6 10.698 95.6 5.5 10.7 96 2.17TfjlmOne can estimate a maximum fouling rate involving bulk mass transport and completeadhesion (sticking probability 1) by assuming that the gum can be approximated asMOOMOOH. At a concentration of 10 gIL (approaching g*), the fouling mass flux is givenbymass flux = 0.000 19 (10) 60 kg/m2.min= 0.114 kg/m.minpff= 184.6, giving dRf/dt = 0.114/184.6= 6.18 x 10 m2.KIW.minusing Epstein’s approach,dRf/dt = 10 (60) 0.85 x 10-/184.6= 2.78 x iO m2.K/W.minObserved rate = 8.1 x 10-8 .KIW.minBoth estimates overpredict the initial fouling rate by over two orders of magnitude.311Appendix A.4Appendix A.4 Fouling Model CalculationsEstimation ofFriction FactorsThe fouling model in Equation [7.25] indicates that the fouling rate is proportionalto u2, which is given by Um2.f/2. The PFRU friction factor had to be estimated aspressure drop data was not available and the skewed alignment of the probe caused someconcern in using a correlation. Equation [7.27] requires a knowledge of the velocitydependence in the friction factor, which was estimated using the heat and mass transferanalogy represented in Equation [6.11.Nu = Re Pr033 f12 [A.4.1]A plot of logNu against log Re will yield a gradient of 1-n, where n is the velocitydependence in the friction factor (depicted asfaRefl). The data from the fouling runs isplotted in Figure A.4. 1 and shows that the results from both fouling probes fall along astraight line, as predicted by [A.4. 1]. The variation in Pr was not considered to besignificant. The TFU data gave n = 0.203 ±0.03 (R2 = 0.989), which is in good agreementwith existing correlations (0.2<n<0.25). At Pr 30, [A.4. 1] givesf (TFU) 0.059 (±0.0 19) Re020 [A.4.2]This expression is in reasonable agreement with the Blasius expression [3.34]. The PFRUdata gave n 0.344 ±0.017 (R2 = 0.99), which is larger than expected and may explainwhy the heat transfer results in Section 3 showed significant deviations from existingcorrelations. Figure A.4. 1 shows that the value of Nu obtained at a given Re in the PFRUwas always larger than in the TFU, which is not predicted by Equation [A.4.1]. Thisdifference is due to the annular configuration and was accounted for in the Weigand andMonrad and Pelton correlations by a configuration factor, (da,Jda,j)Z, where z was 0.45and 0.53 respectively. The PFRU ratio is 2.50/1.07 2.336; assuming z 0.5 gives a312Appendix A.4correction factor of 1.53, which agrees with the differences seen in Figure A.4. 1. This thengives an approximate expression for the friction factor in the PFRUf (PFRU) = 0.295 (±0.03) Re034 ±0.02 [A43]The error in this expression is difficult to estimate as the velocity dependence differssignificantly from the values in the literature. Equation [A.4.3] gave estimates off for Re3000-7000 as 0.019-0.014, which are larger than the values calculated using Equation[7.18], 0.012-0.009. The Reynolds numbers involved belong to the transition regionbefore fully developed turbulent flow, which features higher friction factors and a strongerdependence on Re (i.e. more negative exponent n). Pressure drop data are required forfurther verification. The lack of reliable friction velocity estimates thus introduces furthererrors in to the model calculations.Clog (Re)Tb 100°C: 0.40 mol/L indene in Paraflex : Pr 30Figure A.4. 1 Estimation of Friction Factor Velocity Dependence using the j-Factor Heatand Mass Transfer Analogy: Variation of Nusselt Number with Re2. 3.6 3.8 4.0 4.2 4.4313Appendix A.4Appendix A.4 Fouling Model Calculation SpreadsheetFouling Model Calculation Spreadsheet viscosity correction = 0.79PFRU 1 mM bp65 3979.2 5037 210.0 100.20 155.1 1.010-06 0.1256 151.4 2.076-03 0.0400 1330 0.7740 0.0695 3.363e-07 5.SeOe-o3thrnl, Ta 2100 3047 210.0 80.68 145.3 1.136-06 0.1263 104.7 2.076-03 0.0050 925.2 0.6870 0.0663 4.219e-07 calcd 5.562e-03PFRUbPTs21OC 3295 210.0 100.39 155.2 1.010-06 0.1256TFUbPTs2IOC 9669 210.0 109.03 159.5 9.600-07 0.125390.6 2.076-03 0.0271 796 0.5060 0.0492 1.322a-06 calcd 7.660s-031 .366a-06135.1 2.076-03 0.0293 1977.2 2.0130 0.1362 1.317e-07 calcd 2.301e-03Ran 9 Re (Flea) Re tras T surface Tb.rneas TtiIrn viscosity lsrnda film Nsssslt 1/Ts,s (K) hR Uret Urn u Initial Foal rats Oats sit r isa(C) ( C) (C) (rn2ia) (W/rn.K) (1/K) I (rncl/L)/h )W/rn2.K) (rn(s) (rn/a) (rn2.K/W.rnin) cabs K/Wthermal Ivitiation25 2497 3046.9 180 90.40 130.2 1.386-06 0.1275 100.6 2.210-03 0.0079 897 0.6870 0.0663 1.000e-07 1.2115-9331 2407 3046.8 200 80.90 140.4 1.200-06 0.1267 104.1 2.116-03 0.0089 922 0.6870 0.0663 1.900a-07 2.633e-0328 2407 3046.8 210 91.80 145.9 1.125-06 0.1263 103.6 2.075-03 0.0085 915 0.6870 0.0663 5.000a-07 7.419a-0332 2407 3046.8 225 80.70 152.9 1.035-06 0.1258 106.9 2.016-03 0.0076 940 0.6870 0.0663 5.700e-07 9.184e-0329 2407 3046.5 240 79.70 159.9 9.565-07 0.1253 108.7 1.956-03 0.0072 952 0.6870 0.0663 1.440e-06 2.SlOe-02chemical inifiatiss40 2603 3294.9 207.1 100.40 153.8 1.026-06 0.1257 88.7 2.086-03 0.0287 780 0.5060 0.0482 1.065e-06 1.734e-0243 2603 3294.9 224.6 100.40 162.5 9.296-07 0.1251 90.3 2.016-03 0.0333 790 0.5060 0.0482 3.333a-06 5.980e-0244 2603 3294.9 239.5 100.20 169.9 8.605-07 0.1245 90.9 1.955-03 0.0330 792 0.5060 0.0482 5.140e-06 9.859e-0249 2603 3294.9 254.6 100.60 177.6 7.970-07 0.1240 92.3 1.896-03 0.0234 800 0.5060 0.0482 5.328s-06 1.1 lSe-0160 2603 3294.9 247.4 100.30 173.9 8.266-07 0.1242 93.7 1.920-03 0.0221 814 0.5060 0.0482 6.400e-06 1.291e-0156 2603 3294.9 252.4 100.40 176.4 5.060-07 0.1240 92.2 i.oog-oa 0.0223 800 0.5060 0.0482 7.500e-06 1.SSle-01chemical - velocity43 2693 3294.9 224.6 190.49 162.5 9.290-07 0.1251 90.3 2.016-03 0.0333 790 0.5060 0.0482 3.330e-06 5.974e-0245 3887 4920.3 222.1 100.60 161.4 9.41 0-07 0.1252 116.3 2.020-03 0.0253 1018 0.7560 0.0672 9.300v-07 1 .648s-0247 5145 6512.7 218 99.90 159.0 9.660-07 0.1253 142.0 2.046-03 0.0274 1245 1.0000 0.0847 5.600a-07 9.666e-0350 2603 3294.9 247.4 100.30 173.9 8.265-07 0.1242 93.7 1.920-03 0.0221 814 0.5060 0.0482 1.520a-06 3.066e-0253 5145 6512.7 246.7 100.20 173.5 8.300-07 0.1243 145.2 1.920-03 0.0250 1262 1.0000 0.0847 7.400e-07 1.487a-0254 3887 4920.3 247 100.20 173.6 8.286-07 0.1243 121.2 1.920-03 0.0227 1053 0.7560 0.0672 1.030a-06 2.072a-0255 2603 3294.9 252.4 100.40 176.4 8.066-07 0.1240 92.2 1.906-03 0.0223 800 0.5060 0.0482 2.190e-06 4.530e-0256 3176 4020.3 246.7 100.20 173.5 6.306-07 0.1243 106.1 1.920-03 0.0248 922 0.5170 0.0568 1.209a-06 2.411e-02TFU rasalts503 7987.4 10111 195.73 110.52 155.1 1.010-06 0.1256 136.5 2.116-03 0.0299 1901 2.0660 0.1392 8.OSOa-08 1.336e-03504 7511 9507.6 195.3 108.90 153.6 1.030-06 0.1257 134.3 2.126-03 0.0292 1873 1.9600 0.1329 7.750e-08 l.259e-03505 7758 9920.3 222.6 11 0.65 166.6 8.596-07 0.1248 136.3 2.020-03 0.0330 1886 1.9980 0.1350 2.340e-07 4.386e-03506 7311 9254.4 187.4 105.90 146.7 1.116-06 0.1263 131.6 2.176-03 0.0225 1543 2.0310 0.1381 5.800e-0S 8.685e-04507 7625 9651.9 211.5 109.20 160.4 9.510-07 0.1252 135.6 2.066-03 0.0316 1883 2.0120 0.1362 1.300e-07 2.279a-03508 12899 16328 199.38 109.80 154.5 1.020-06 0.1257 206.9 2.126-03 0.0358 2853 3.3824 0.2170 1.500e-08 2.463a-04509 10497 13287 212.4 112.10 162.3 9.320-07 0.1251 178,5 2.066-03 0.0390 2476 2.6399 0.1730 6.6009-08 1.181s-03510 4229 5353.2 211.3 108.80 1 60.1 9.540-07 0.1253 83.0 2.060-03 0.0359 1 153 1.0990 0.0790 1 .800e-07 3.145a-03511 6340 8025.3 209.1 110.30 159.7 9.580-07 0.1253 126.5 2.070-03 0.0357 1758 1.6690 0.1151 2.209a-07 3.829e-03512 4227 5350.6 211.3 110.20 160.8 9.470-07 0.1252 85.6 2.060-03 0.0356 1188 1.0970 0.0788 2.400e-07 4.225e-03314Appendix A.4Appendix A.4 Fouling Model Calculation Spreadsheet (continued)Eact= 96.6 rho(t) Iarn(t)= 184.600 (m4.K/W.kg)nit r/nu at thml [indene] Rxn(kg/m3.s) Rxn/rho.Iam(f) 0 0/5 (daviatlonj210.0 (m) (mol/m3) kR(lCb)Mind (m.K/s.W) (s/rn) (s/m2) (%)5.95E-03 1.42E-04 370 1.556e-04 8.427e-07 7058.5 4.967e÷07 71.34.38E-03 1.37E-04 370 1.689e-04 9.148a-07 4784.1 3.481e÷07 57.67.42E-03 1.38E-04 370 1.666e-04 9.025e-07 8220.3 5.955e÷07 75.34.45E-03 1.34E-04 370 1.500e-04 8.125e-07 5481.1 4.096e,07 63.06.16E-03 1.32E-04 370 1.418e-04 7.6820-07 8015.5 6.091e÷07 74.7Eact 97.22.O1E-02 1.61E-04 370 5.628a-04 3.0490-06 6580.9 4.083e+07 42.02.949-02 1.58E-04 370 6.523e-04 3.5340-06 8321.6 5.256e÷07 54.22.47E-02 1.576-04 370 6.467e-04 3.503e-06 7064.2 4.493e÷07 46.01.44E-02 1.55E-04 370 4.595e-04 2.489e-06 5796.1 3.741e*07 34.22.27E-02 1.53E-04 370 4.3268-04 2.3448-06 9683.7 6.3458+07 60.62.20E-02 1.55E-04 370 4.367e-04 2.366e-06 9309.2 6.0040+07 59.02.94E-02 1.589-04 370 6.523e-04 3.5348-06 8314.1 5.251e+07 54.19.129-03 1.236-04 370 4.955e-04 2.6848-06 3399.4 2.765e+07 41.96.52E-03 1.019-04 370 5.377e-04 2.913a-06 2237.5 2.223e+07 44.25.39E-03 1.539-04 370 4.326e-04 2.344e-06 2299.9 1.507e+07 65.82.69E-03 9.85E-05 370 4.896e-04 2.652e-06 1015.7 1.032a+07 22.93.71E-03 1.18E-04 370 4.454e-04 2.413e-06 1536.2 1.302e*07 28.56.43E-03 1.559-04 370 4.367e-04 2.366e-06 2718.3 1.7530+07 40.34.37E-03 1.35E-04 370 4.875e-04 2.641e-06 1654.3 1.227e,07 66.6Eact= 88.62.169-03 6.61E-05 370 5.865e-04 3.177e-06 678.9 1.027e,07 31.12.18E-03 6.71E-05 370 5.728e-04 3.103e-06 701.7 1.0458+07 27.02.50E-03 6.62E-05 370 6.474e.04 3.507e-06 713.9 1.079e.07 30.42.569-03 6.85E-05 370 4.473e-04 2.423e-06 1058.0 1.544e+07 55.12.139-03 6.65E-05 370 6.199e-04 3.358a-06 633.8 9.5298.06 23.04.04E-04 4.36E-05 370 7.023e-04 3.804e-06 106.3 2.438a+06 82.51.06E-03 5.059-05 370 7.651e-04 4.1448-06 255.5 5.057e+06 19.02.96E-03 1.099-04 370 7.042e-04 3.815e-06 776.9 7.151e+06 84.83.990-03 7.13E-05 370 7.003e.04 3.794e-06 1051.6 1.476e*07 35.33.980-03 1.05E-04 370 6.984e-04 3.783e-06 1052.6 9.9888+06 36.95.56E-03 9.45E-05 430 8.451e-04 4.578e-06 1214.5 1.286e,07 56.45.56E-03 1.37E-04 370 1.5668-04 8.481e-07 6558.3 4.802e+07 69.17.665-03 1.58E-04 370 5.3180-04 2.8818-06 2658.9 1.685e,07 43.52.309-03 6.67E-05 370 5.748e-04 3.114e-06 739.1 1.1078+07 34.0315Appendix BAppendix B Data SummaryThe following tables summarise the operating conditions in each experiment, as excessivespace would be required to reproduce all the experimental data. The experimental datacompiled during this study are available on request from Prof. A.P. Watkinson, do theDepartment of Chemical Engineering at the University of British Columbia.Concentrations are quoted as mol/L at room temperature (20°C)316Avvendix BTable B. 1 Summary of PFRU Fouling Experiments #001-016,030Run # Solvent Alkene Initiation Tbulk Tsw-f Re Heat Flux U0[mol/L] (°C) (°C) (kW/m2) (W/m2K)hexadecene[0.466]thermalthermalthermal84.885.878.279.9202.3203.7280.8200hexadecene thermal[0.388]83.2 200001• 002003004005006007008009010011012013014015016030ParaflexParaflextetralintetralinParaflexParaflextetralintetralinParaflexParaflexkerosenekerosenekerosenekerosenekerosenetrichlorobenzenetrichlorobenzene83.083.081.483.078.282.7100.285.284.382.083.2101.7102.725028021120023420025022825018118018024030503050305010721107214318431830501072110721107213050,decreased1000?106901069014080140801069014080140801360013600108.7113.6197.9172173161197119.6170212.8167157.3110145.6137.2112.4221.5138227227.2120.6121hexadecene[0.388]indene[0.502]hexadecene[0.2051indene[0.507]hexadecene[0.388]restart 009hexadecene[0.388]indene[0.878]hexadecene[0.388]hexadecene[0.388]indene[0.95 1]913937937137513989439439101404140414701076829668136813681521152113611640166012501250thermalthermalthermalthermalthermalthermalthermalthermalthermal6 mM bPthermalthermal10mM bPthermal82.7 183106.1104.982.3245242180indene thermal 80.0 180[0.41]All experiments performed at 377 kPa air saturation: PFRU317Appendix BTable B.2 Summary of Thermally Initiated PFRU Experiments 11017-036Run # Solvent [indene] Re Tbujk Tswf Heat Flux Uo Comments(wt%) (°C) (°C) (kW/m2) (W/m2K)017 kerosene 10 10690 81.7 180 131.6 1379 Contaminated018 kerosene 5 10690 81.4 180 129.8 1348 Contaminated019 kerosene 2 10690 81 180 130 1348 Contaminated020 kerosene 10 10690 81.4 180 132.4 1366021 kerosene 5 10690 81 180 131.6 1345022 kerosene 2 10690 81 180 - Pump leakage023 kerosene 2 10690 81 180 - - Pump leakage024 Paraflex 9.1 3050* 80.6 180 90.3 920.7 Unusual Rf-t025 Paraflex 5 3050 80.4 180 90.2 897026 Paraflex 2 3050* 82.4 180 90.2 886027 Paraflex 10 3050* 80.8 180 89.4 926028 Paraflex 5 3050 81.8 210 115.8 915029 Paraflex 5 3050 79.7 240 149.5 952031 Paraflex 5 3050 80.8 200 108.8 922.1032 Paraflex 5 3050 80.7 225 136.1 940033 Paraflex - 3050 81.0 210 118.5 918 [0.32MIDCP034 Paraflex - 3050 81 210 118.5 918 [0.139MJDCP[0.41Mj indene035 Paraflex 5 3300 101.4 210 84.8 768036 Paraflex 5 3300 100.3 240 115 808*- Re based on indene/Paraflex correction factor of 0.79: thermal initiation: 377 kPa air saturation318Appendix BTable B.3 Summary of Chemically Initiated PFRU Fouling ExperimentsRun # Tbulk Tsfa Re Heat Flux U0 Comments______(°C) (°C)____________(kW/m2) (W/m2K)______________037 80.7 210 3050 115.9 900038 82.8 210 3050 116.2 910 ABN initiation039 80.8 210 3050 119.4 918040 100.4 210 3300 83.2 780041 89.3 210 3011 89.1 830 Unusual Rf-t042 90.0 212 3011 98.6 830 Unusual R1-t043 100.4 225 3300 110.3 792044 100.2 240 3300 110.3 792045 100.6 225 4920 123.7 1018046 100.7 225 6513 152.4 1244047 99.9 225 6513 149.6 1245 Repeat of 046048 90.3 209 3300 100.5 837 Unusual Rf -049 100.6 255 3300 123.1 800050 100.3 250 3300 120 814051 100.2 248 4920 155.6 1060052 100.1 250 6513 191.3 1284053 100.2 247 6513 184.9 1284 Repeatof 052054 100.2 248 4920 154.6 1053055 100.4 251 3300 121.6 800 Repeat of 050056 100.2 248 4020 135.1 9220.405 mol/L indene in Paraflex : 2.5 mM bP initiation unless marked: 377 kPa air overpressure319Appendix BRun # Initiation [BMPI(ppm)Tbulk(°C)T5(°C)057 thermal 0 100.2 240 3300058 thermal 200 101.6 240 3300059 thermal 100 101.2 240 3300060 thermal 200 98.8 239 3300061 thermal 50 99.8 239 3300062 thermal 0 100.3 240 3300063 thermal - 100.0 240 3300064 thermal - 100.1 240 3300065 1 mM bP - 100.2 210 50370.405 mol/L indene in Paraflex: 377 kPa air saturationRun # [BMPj Tbulk Solvent______(ppm) (°C)_______________301 0 100 toluene302 0 100 trichlorobenzene303 0 100 toluene304 0 100 n-octane305 0 100 kerosene306 0 100 Paraflex307 50 100 Paraflex308 50 100 Paraflex309 50* 100 Paraflex310 0 100 Paraflex311 100 100 Paraflex312 200 100 Paraflex313 400 100 Paraflex314 100 80 Parallex315 100 80 Paraflex316 100 110 Paraflex317 100 120 Parallex318 100 90 Paraflex0.405 mol/L indene: 377 kPa air saturation: thermal initiationCommentsBlank - no indeneContaminationRepeat of 058500 minutes430 minutes300 minutesTFU ComparisonTable B.4 Summary of Final PFRU Fouling Experiments;Antioxidant, Interrupted and Comparison StudiesRe Heat Flux 1J0 Comments(kW/m2) (W/m2K)Unusual Rf -110.8111.4111.2109.8110.8110.3109.9110.0145.27937907937927978038008061330Table B.5 Summary of SCR Experiments Performed by Lai and WilsonTBA* Equimolar BMP, TBALiquid exhausted320Avvendix BRun # Tbulk(°C)Table B.6 Summary of SCR Experiments[indene] solvent ajr Initiation Comments(mol/L) (kPa)100 81 0.71 kerosene 329103 91 0.71 kerosene 329104 101 0.71 kerosene 329105 101 0.71 kerosene 329106 111 0.71 kerosene 329107 121 0.71 kerosene 329108 91 0.71 kerosene 329109 81 0.71 kerosene 329110 101 0.71 kerosene 188*111 101 0.71 kerosene 75*112 100 0.71 kerosene 377113 101 0.71 kerosene 377114 101 0.71 kerosene 377115 101 0.71 kerosene 377116 101 0.0 Parallex 377117 100 0.70 Paraflex 377118 100 0.70 Paraflex 377119 100 0.405 Paraflex 377120 100 0.405 Paraflex 377121 100 0.20 Paraflex 377122 80 0.405 Paraflex 377123 90 0.405 Paraflex 377124 120 0.405 Paraflex 377125 80 0.405 Parallex 377126 80 0.405 Paraflex 377127 80 0.405 Paraflex 377128 80 0.405 Paraflex 377129 80 0.405 Paraflex 377130 80 0.405 Paraflex 377131 80 0.405 Parallex 377132 80 0.405 Paraflex 377133 80 0.405 Parallex 377134 100 0.405 Paraflex 377135 90 0.405 Paraflex 377136 110 0.405 Paraflex 377137 100 0.405 Paraflex 287138 100 0.405 Paraflex 152139 100 0.405 Paraflex 473140 100 0.405 Paraflex 377141 100 0.405 Parallex 377142 120 0.405 Parailex 377143 140 0.405 Paraflex 377144 157 0.405 Paraflex 377145 100 0.405 Paratlex 377146 100 0.405 Paraflex 377147 100 0.405 Paraflex 377GZ- performed by G. Zhang: SCR - bubbled air unless specifiedthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermalthermal5.0 mM bP7.5 mM bP7.5 mM bP2.5 mM bP5.0 mM ABN5.0 mM ABN2.5 mM ABN7.5mMABN2.5 mM bP2.5 mM bP2.5 mM bP2.5 mM bP2.5 mM bP2.5 mM bP1 mM bP1 mM bP1 mM bP1 mM bP1 mM bP1 mM bP1 mM bP1 mM bPGZ - evaporationGZ - evaporationGZ - evaporationGZGZGZGZGZGZ - 329 kPa PtotalGZ - 329 kPa PtotalAir blanketBubbled airRepeat of 113PFRU apparatusParaflex BlankPFRU apparatusPFRU apparatusCompressor FailureRepeat of 127Compressor Problems1000 mL air/L soin/minCompressor Problems321Appendix BTable B.7 Summary of Tube Fouling Unit ExperimentsRun # Tbuj