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 FOULING by David Ian Wilson M.A. (1988) Jesus College, Cambridge M.Eng. (1989) Jesus College, Cambridge A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES Department of Chemical Engineering  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA March, 1994  © D.I.Wilson,  1994  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission.  Department  of  The University of British Columbia Vancouver, Canada  Date  DE-6 (2188)  fl  t9-  Abstract  Chemical reaction fouling of heat exchangers is a severe problem in the petrochemicals industry, where deposits can be formed by a wide range of undesirable reactions. Autoxidation has been identified as a prime source of deposit formation in oxygenated process streams and fuel storage systems but the fouling mechanism has not been fuily investigated. The fouling of heat exchangers subject to autoxidative fouling was studied using model solutions of an active alkene, indene, in inert solvents saturated with air. The heat exchangers were operated at moderate surface temperatures (180-250°C) and at turbulent flow velocities. The experiments featured air pressures of 342-397 kPa and heat fluxes of 90-280 kW/m . The effects of chemical reaction rate, surface temperature and flow velocity 2 were investigated and compared with existing chemical reaction fouling models. Chemical initiators were used to eliminate the chemical induction periods observed under ‘natural’ thermal initiation and permitted the study of the chemical reaction rate and surface temperature as separate variables. The chemistry of the physical system and the complex reaction mechanism prevented extensive model development. Two fouling probes were used: an annular probe which allowed visual inspection of deposit formation and a novel tubular heat exchanger constructed during this work which allowed inspection of the deposit in situ after an experiment. The same fouling mechanism was found to generate deposit in both probes. Chemical analyses were developed to monitor the autoxidation reaction during the batch fouling experiments. The results confirmed that fouling was caused by the deposition of insoluble polyperoxide gums generated by the reaction of indene and oxygen. The gums  il  aged on the heat exchanger surface to form complex oxygenated solids which were not easily removed. These results confirmed the hypotheses of Asomaning and Watkinson (1992). The kinetics of indene autoxidation was studied in a separate series of semi-batch stirred tank experiments. The rate of formation of polyperoxides was influenced by the solvent nature, temperature, oxygen concentration and mode of initiation. The aromatic polyperoxides exhibited limited solubility in aliphatic solvents. The kinetics of indene autoxidation could not be described by the schemes reported in the literature and were found to be subject to oxygen mass transfer effects. The fouling resistance behaviour was controlled by conditions in the bulk fluid. The initial, linear fouling rate decreased with increasing flow velocity and increased with bulk reaction rate and surface temperature. A simple fouling model, involving generation of deposit in the reaction zone next to the heat transfer surface and an attachment factor related to the mean fluid residence time, was fitted to the experimental data. Once the solubility limit was reached, the fouling resistance showed increasing rate behaviour, caused by the deposition of globules of insoluble gum. The effect of an antioxidant on the fouling process was studied. The efficiency of the antioxidant, di-t-butyl-4-methylphenol, was found to be severely reduced under the enhanced thermal conditions in the heat exchanger. Simulated ageing experiments were performed to investigate the fate of polyperoxides exposed to the enhanced temperatures on the heat exchanger surface. The studies confirmed that the insoluble polyperoxide gums undergo ageing processes after deposition.  U’  Table of Contents Abstract  ii  Table of Contents  iv  List of Tables  vii  List of Figures  x  Acknowledgement  xvi  2.  INTRODUCTION  1  LITERATURE REVIEW  7  2.1 2.2  2.3  2.4  2.5  3.  The Role of Autoxidation in Chemical Reaction Fouling Experimental Studies in Autoxidation Reaction Fouling 2.2.1 Fuel Stability Studies 2.2.2 Thermal Fouling Studies Autoxidation Chemistry 2.3.1 Autoxidation Mechanisms and Kinetics 2.3.2 Solvent and Additive Effects in Autoxidation 2.3.3 The Autoxidation of Indene, C 8 H 9 2.3.4 Antioxidation Mechanistic Modelling of Fouling 2.4.1 Transport and Adhesion Models 2.4.2 Transport and Reaction Models 2.4.3 Reaction Engineering Objective  7 11 13 18 21 21 25 28 30 31 33 36 43 44  EXPERIMENTAL MATERIALS AND METHODS  46  3.1  46 46 50 50 51 55 56 62 64 66 68 69 74 78  3.2 3.3 3.4  Materials and Physical Properties 3.1.1 Model Solutions 3.1.2 Physical Properties Kerosene Properties Paraflex Properties Concentration of Dissolved Gases Portable Fouling Research Unit (PFRU) Apparatus 3.2.1 PFRU Heat Transfer 3.2.2 PFRU Experimental Procedure Stirred Cell Reactor (SCR) Tube Fouling Unit (TFU) 3.4.1 Tube Fouling Unit Apparatus 3.4.2 Surface Temperature Measurement 3.4.3 Data Collection and Processing  iv  3.5 3.6  4.  AUTOXIDATION OF MODEL SOLUTIONS 4.1 4.2  4.3 4.4 4.5 4.6 4.7 4.8 5.  Solvent Effects in Indene Autoxidation Autoxidation Kinetics 4.2.1 Mass Transfer Effects in Autoxidation Kinetics 4.2.2 Kinetics of Autoxidation in Mass Transfer 4.2.3 Gas Phase Resistance Effects in Mass Transfer 4.2.4 Oxygen Effects in Autoxidation Chemically Initiated Autoxidation of Indene Temperature Effects in Autoxidation Autoxidation in the Presence of Antioxidants A Kinetic Model of Indene Autoxidation Ageing of Polyperoxide Gums Summary of Autoxidation Studies  82 86 88 90 91 92 95 99 100 100 106 108 112 117 121 124 127 132 135 146 151  INITIAL FOULING EXPERIMENTS  153  5.1  153 154 156 160 162 164 167 173 177  Model Solution Selection 5.1.1 Tetralin as Solvent 5.1.2 Hexadec-1-ene as Dopant 5.1.3 IndeneasDopant 5.1.4 Deposit Characterisation 5.1.5 Initial Mechanistic Insights Effects of Dopant Concentration Fouling in Model Solutions with Two Dopants Temperature Effects in Thermally Initiated Fouling Velocity and Surface Temperature Effects in Chemically Initiated Fouling Stages in Autoxidation Fouling Fouling in the Presence of Antioxidants Summary of Initial Fouling Experiments  186 191 197 203  FOULING EXPERIMENTS IN THE TUBE FOULING UNIT  206  6.1 6.2  206 209 211 214 218  5.2 5.3 5.4 5.5 5.6 5.7 5.8 6.  3.4.4 TFUHeatTransfer 3.4.5 TFU Operating Procedures GumAgeingOven Chemical Analysis 3.6.1 Hydroperoxide Analysis- Peroxide Number (POx) 3.6.2 Polyperoxide Analysis (Gum in Solution Assay) 3.6.3 Indene Concentration- Gas Liquid Chromatography 3.6.4 Further Analysis- FTIR, SEM  Autoxidation of Indene in the TFU Initial Fouling Studies 6.2.1 TubePressureDrop 6.2.2 Local Fouling Behaviour 6.2.3 Deposit Distribution and Morphology v  6.3 6.4 6.5 6.6 7.  Surface Temperature Effects in TFU Fouling Velocity Effects in TFU Fouling Comparison of TFU and PFRU Fouling Probes Summary of TFU Fouling Studies  MODELS OF ASPECTS OF AUTOXIDATION FOULING 7.1 7.2 7.3  Autoxidation Kinetics in the Fouling Apparatus 7.1.1 Model Development 7.1.2 Model Performance Fouling Mechanisms in the Initial Fouling Regime 7.2.1 Particulate Fouling Models 7.2.2 Chemical Reaction Fouling Models Analysis of Fouling Rate and Behaviour 7.3.1 Formulation of a Lumped Parameter Fouling Model 7.3.2 Analysis of Experimental Fouling Data  226 229 234 237 240 242 242 246 249 249 254 257 258 260  8.  CONCLUSIONS  270  9.  RECOMIVIENDATIONS FOR FURTHER STUDY  275  Abbreviations  277  Nomenclature  278  References  283  APPENDICES A. 1  Experimental Apparatus and Configuration  294  A.2  Experimental Summaries  301  A.3  Sample Calculations  307  A.4  Fouling Model Calculations  312  B.1  DataSummary  316  vi  LIST OF TABLES 2.1  Examples of Chemical Reaction Fouling in Refineries  8  2.2  Summary of Autoxidation Related Fouling Studies  12  2.3  Fouling Results of Taylor (1969b) and Asomaning and Watkinson (1992)  14  2.4  Activation Energies Reported in Chemical Reaction Fouling  16  2.5  Velocity Effects Reported in Chemical Reaction Fouling  17  2.6  Summary of Chemical Reaction Fouling Models  34  3.1  Alkenes Used in Model Solutions  48  3.2  Properties of Solvents Used in Model Solutions  49  3.3  Physical Properties of Initiators and Antioxidants  48  3.4  Estimates of Dissolved Oxygen Concentration in Paraflex and Kerosene  55  3.5  Comparison of Nusselt Numbers in PFRU Heat Transfer  63  3.6  TFUAlarmMatrix  72  4.1  Solvent Effects in the Autoxidation of Model Solutions of Indene  104  4.2  Mass Transfer Effects in Batch Autoxidation of Indene  110  4.3  Regression of Thermally Initiated Autoxidation Data to Mass Transfer Models  118  4.4  Gas Phase Resistance Effects in Autoxidation Kinetics  120  4.5  Comparison of Autoxidation Kinetics in the SCR and PFRU  120  4.6  Oxygen Effects in the Initiated Autoxidation of Indene in Paraflex  123  4.7  Chemically Initiated Autoxidation of Indene in Paraflex  126  4.8  Activation Energies in the Autoxidation of Indene in Model Solutions  129  4.9  Autoxidation of Indene in Paraflex in the Presence of Antioxidants  134  4.10  Elemental Analysis of Aged Polyperoxide Gums  150  vii  5.1  Summary of Initial Fouling Experiments  155  5.2  Chemical Analysis of Deposits From Initial Fouling Runs  165  5.3  Effects of Indene Concentration in Fouling Experiments  169  5.4  Fouling from Solutions of Indene and Dicyclopentadiene in Paraflex  174  5.5  Temperature Effects in Thermally Initiated Fouling Experiments  178  5.6  Elemental Analysis of Fouling Deposits  185  5.7  Effects of Bulk Chemical Parameters on Fouling and Reaction Behaviour  185  5.8  Reaction Diagnostics from Chemically Initiated PFRU Fouling Experiments 189  5.9  Effect of Flow Rate on Initial Fouling Rate in Initiated PFRU Experiments 192  5.10  Reaction Diagnostics and Fouling Summary of Interrupted Fouling Runs  194  5.11  Fouling in the Presence of an Antioxidant (2,6,di-t-butyl 4-methylphenol)  199  6.1  Comparison of Indene Autoxidation Kinetics Under TFU Conditions  208  6.2  Initial Conditions and Tube Pressure Drops in TFU Fouling Runs  215  6.3  Local Heat Transfer and Fouling Behaviour in Run 504  216  6.4  Comparison of Sand Roughness Criteria with Gum Globule Dimensions in TFU Fouling Run 503  224  6.5  Surface Temperature and Flow Velocity Effects on Initial Fouling Rate  227  6.6  Elemental Analysis of Deposits from TFU Runs  230  7.1  PFRU Reaction Coupling Model Parameters  243  7.2  Experimental Data from Thermally Initiated Fouling Experiments 0.41M Indene in Paraflex; 79 kPa Oxygen Saturation  248  Particle Relaxation Times and Mass Transfer Coefficients in Particulate Fouling Calculated at Conditions Used in Fouling Runs  252  Comparison of Initial Fouling Rate Activation Energies  261  7.3 7.4  vm  7.5  Comparison of Activation Energies of Adjusted Fouling Rate (Rf7V)  262  A. 1.1 Experiment Nomenclature  294  A. 1.2 TFU Equipment Specifications  295  A.2. 1 Comparison of Indene Autoxidation Kinetics under TFU Conditions  301  A.4  Fouling Model Calculation Spreadsheet  314  B. 1  Summary of PFRU Fouling Experiments #001-016, 030  317  B .2  Summary of Thermally Initiated PFRU Experiments #017-036  318  B .3  Summary of Chemically Initiated PFRU Fouling Experiments  319  B .4  Summary of Final PFRU Fouling Experiments Antioxidant, Interrupted and Comparison Studies  320  B.5  Summary of SCR Experiments Performed by Lai and Wilson  320  B .6  Summary of SCR Experiments  321  B .7  Summary of Tube Fouling Unit Experiments  322  lx  LIST OF FIGURES 1.1  Heat Transfer in the Presence of Fouling Deposits  3  1.2  Types of Fouling Behaviour  5  2.1  Autoxidation of Indene  29  2.2  Antioxidant Action of 2,6,di-t-butyl 4-methylphenol (BMP)  29  2.3  Schematic of Processes Involved in Chemical Reaction Fouling  32  2.4  Comparison of the Experimental Fouling Data of Crittenden et al. from the Polymerization Fouling of Styrene in Kerosene and the Predictions from Epstein’s Fouling Model (1993b) 39  2.5  Panchal and Watkinson Autoxidation Fouling Models  42  3.1  Kinematic Viscosities of Solvents Used in Model Solutions  52  3.2  Prandtl Number of Solvents Used on Model Solutions  52  3.3  Viscosity of Mixtures of Indene in Paraflex at 20°C; Comparison of Correlation Predictions and Experimental Data  54  3.4  Schematic Diagram of PFRU Apparatus  57  3.5  PFRU Orifice Plate Calibration  58  3.6  Schematic Diagram of PFRU Fouling Probe  60  3.7  Schematic Diagram of Stirred Cell Reactor (SCR) Apparatus  67  3.8  Schematic Diagram of Tube Fouling Unit Apparatus  70  3.9  Schematic Diagram of Heated Section Construction  75  3.10  Details of TFU Surface Thermocouple and Mounting Designs  77  3.11  Comparison of Predicted TFU Nusselt Number with Experimental Data  83  3.12  Thermal Entry Length Effects in TFU Heat Transfer  83  3.13  Comparison of Experimental Fanning Friction Factor Data with the Correlations of Blasius and Gnielinski  85  x  3.14  Schematic Diagram of Gum Ageing Oven Apparatus  89  3.15  Photograph of Gum Assay Filters Showing Change in Gum Appearance  94  3.16  Gas Chromatogram of Solution of Indene in Paraflex  96  3.17  Gas Chromatograms of Indene in Kerosene from Flame Tonisation and Photo-lonisation Detectors in Series  98  4. la-c Solvent Effects in Indene Autoxidation: Analysis Results  102  4.2  Kinetic Plots of Indene Concentration Data: Comparison of Kinetic Models 107  4.3  Comparison of Mass Transfer Kinetic Models of the Autoxidation of Indene in Paraflex  118  4.4  Temperature Effects on the Gum Solubility Limit, g*, in Paraflex  128  4.5  Variation of the Gum Solubility Limit, g*, with Indene Concentration  128  4.6  Temperature Effects in the Thermally Initiated Autoxidation of lOwt% Indene in Kerosene: Peroxide Number Behaviour  129  4.7  Kinetic Model of Indene Autoxidation: Effect of K on Mass Concentration of Gum and Comparison with Data from Runs 140, 141 (e = 1) 140  4.8  Kinetic Model of Indene Autoxidation : Effect of K on Peroxide Number and Comparison with Data from Runs 140, 141 (e = 1)  140  4.9  Kinetic Model of Indene Autoxidation: Effect of e on Mass Concentrations (K = 0.025 m /mol.min: Data from SCR Runs 140, 141) 3 141  4.10  Kinetic Model of Indene Autoxidation : Effect of e on Peroxide Number (K = 0.025 m /mol.min: Data from SCR Runs 140, 141) 3  141  4.11  Kinetic Model of Indene Autoxidation : Revised Mass Concentration Curves at K 0.025 m /mol.min with Data from SCR Runs 140, 141 3 142  4.12  Kinetic Model of Indene Autoxidation: Revised Peroxide Concentrations at K = 0.025 m /mol.min with Data from SCR Runs 140, 141 3 142  4.13  Kinetic Model of Indene Autoxidation: Comparison of Revised Mass Concentration Curves and Data from SCR Run 142 at Tb = 120°C  144  Kinetic Model of Indene Autoxidation: Comparison of Revised Mass Concentration Curves and Data from SCR Run 142 at Tb = 120°C  144  4.14  )  Kinetic Model of Indene Autoxidation: Comparison of Revised Mass Concentration Curves with Data from TFU Runs 501, 503, 504  145  Kinetic Model of Indene Autoxidation : Comparison of Revised Peroxide Concentrations with Data from TFU Runs 501, 503, 504  145  4.17  FTIR Spectra of Samples from Gum Ageing Experiments at 200°C  147  4.18  Reduction in Gum Mass During Ageing Oven Experiments  149  4.19  Reduction in Gum Mass During TGA Ageing Experiments  149  5.1  Peroxide Number Analyses from Initial Fouling Experiments using Model Solutions with Tetralin as Solvent  157  Fouling Resistance and Hydroperoxide Concentration Profiles from Fouling of Hexadecene in Kerosene in Run 014  159  Fouling Resistance and Hydroperoxide Concentration Profiles from Fouling of Indene in Kerosene in Run 013  161  Photograph of Fouled PFRU Probe Following Fouling Run 014. Thermally Initiated Autoxidation of 0.388 M Hexadecene in Kerosene.  163  Deposit Thickness Profile of PFRU Probe Fouled by the Autoxidation of 0.388 M Hexadecene in Kerosene  163  FTIR Spectra of PFRU Deposit and Soluble Gum from Autoxidation Fouling of Indene in Kerosene in Run 013  166  4.15 4.16  5.2 5.3 5.4 5.5 5.6 5.7  Fouling Resistance and Peroxide Number Profiles from Indene in Paraflex 169  5.8  Kinetic Fit of Peroxide Data from Initial Indene Fouling Runs to Eqn.[4.36] 172  5.9  Fouling Resistance Profiles in Indene/DCP Studies  175  5.10  Peroxide Number versus Time in Indene/DCP Studies  175  5.11  Temperature Effects in Thermally Initiated Fouling Experiments in Paraflex 180  5.12  Effects of Surface Temperature on Fouling Resistance Profiles in Thermally Initiated Solutions of Indene in Paraflex Runs 025, 028, 029, 031, 032 180  5.13  Gum Concentration and Fouling resistance in Thermally Initiated Fouling  x  182  Surface Temperature Effects on Initial Fouling Rate in Thermally and Chemically Initiated Fouling Runs  189  5.15  Log-log Plot of Velocity Effects in Chemically Initiated Fouling  192  5.16  Photomicrographs of PFRU Surface Deposit from Interrupted Runs  195  5.17  Autoxidation Fouling in the Presence of Antioxidant: Gum Concentrations  196  5.18  Fouling Resistance Profiles in the Presence of an Antioxidant  196  5.19  Effect of an Antioxidant (BMP) on Reaction and Fouling Induction Periods 202  5.14  6. la-c Comparison of Indene Autoxidation in SCR, TFU and PFRU  207  6.2  Fouling Resistance Profiles Showing TFU Reproducibility  210  6.3  Overall Fouling Resistance and Soluble Gum Concentration in Run 503  212  6.4  Tube Pressure Drop and Soluble Gum Concentration in Run 505  213  6.5  Local Fouling Resistances in Run 504  217  6.6  Photograph of Fouled TFU Tube from Run 507 after Division into Sections 219  6.7  TFU Test Section Deposition Regions  219  6.8  Photograph of Thermal Entry Length Deposition in Run 507  219  6.9  Optical and Scanning Electron Micrographs of Deposit from Run 507  223  6.10  Variation in Deposit Coverage with Axial Position in TFU Runs 502-504  224  6.11  Arrhenius Plot of Initial and Final Fouling Rates in TFU Experiments  227  6.12  Plot of Final Local Fouling Resistance Against TFU Deposit Coverage for Estimation of Deposit Thermal Conductivity 230  6.13  Tube Pressure Drop and Overall Fouling Resistance in TFU Run 510  232  6.14  Comparison of Fouling Resistance Profiles in TFU and PFRU  235  6.15  Fouling Resistance Profiles From Initial PFRU Runs at High Indene Concentrations showing Unusual Fouling Behaviour  238  Compartmental Kinetic Model of PFRU System  243  7.1  xiii  Regression of First Order Indene Rate Constant from Data from SCR Experiment 122  247  PFRU Reaction Coupling Model Predictions of Volume Effects in Kinetic Parameters in the PFRU System  247  Comparison of Reaction Coupling Model Predictions and Experimental Induction Periods in Thermally Initiated Fouling Runs in Paraflex  248  7.5  Schematic of Fouling Model Mechanism  259  7.6  Arrhenius Plot of Initial Fouling Rates  261  7.7  Arrhenius Plot of (RfYV) Against Surface Temperature  262  7.8  Arrhenius Plot of 0 versus Inverse of Surface Temperature for PFRU Runs 265  7.9  Fouling Model Analysis: Variation of 0(210°C) with Friction Velocity  267  7.10  Paterson and Fryer Fouling Model Analysis: Variation of 0(210°C)ISth with Friction Velocity  269  7.2 7.3 7.4  A. 1.1 TFU Orifice and Rotameter Flow Calibrations  296  A. 1.2 Schematic Diagram of TFU Power Circuitry  297  A. 1.3 Schematic Diagram of TFU Signal Processing System: Meter Outputs  298  A. 1.4 Schematic Diagram of TFU Heated Section Voltage Processing Circuit  299  A. 1.5 Schematic Representation of Mass Transfer Effects in Gas/Liquid Reactions 300 A.2. 1 Effects of Surface Temperature on Chemically Initiated Fouling Solutions of Indene in Paraflex  302  A.2.2 Effect of Flow Velocity on Initiated Fouling TSurf 222°C a. Fouling Resistance Profiles; b. Gum Analysis Results Tsurf 248°C c. Fouling Resistance Profiles; d. Gum Analysis Results  303 304  A.2.3 Peroxide Number, Soluble Gum and Indene Concentration Results in TFU Fouling Runs Tsurf Varied  305  -  xiv  A.2.4 Peroxide Number, Soluble Gum and Indene Concentration Results in TFU Fouling Runs Flow Velocity Varied  306  A4. 1 Estimation of Friction Factor Velocity Dependence using the j-Factor Heat and Mass Transfer Analogy: Variation of Nusselt Number with Re  313  -  xv  Acknowledgement  My sincere thanks are due to all members of the Department of Chemical Engineering for their encouragement and company in the course of this study. I am particularly indebted to my supervisor, Prof. Paul Watkinson, for his advice, patience and support over the last five years. This study is a testament to the skill and experience of the workshop, stores and technical 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, Samuel Asomaning and Dr. Guohong Zhang are gratefully acknowledged. The financial support of the University of British Columbia and the Natural Sciences and Engineering Research Council of Canada are gratefully acknowledged, as is the support of the Royal Ulster Constabulary Benevolent Fund.  This volume is dedicated to William James Wilson, the man who encouraged me to seek but did not live to see me find.  xvi  1. Introduction  1.  Introduction Heat exchanger fouling is generally defined as the unwanted accumulation of  deposit on heat transfer surfaces (Taborek et at., 1972). Deposit accumulation increases the operating costs of process plant heat exchangers as fouling reduces the efficiency of heat transfer and increases frictional energy losses. Fouled heat exchangers have to be cleaned or replaced, which involves significant capital and operating costs associated with cleaning and plant down time. The economics of heat exchanger fouling are discussed in detail by Van Nostrand et at. (1981), Garret-Price et at. (1985) and Pritchard (1990), who estimated fouling 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 exchangers afforded by advances in heat transfer science and technology can be negated in industrial practice by fouling, which is a widespread and poorly understood phenomenon. The goal of engineering fouling research is to understand the mechanisms and rates of fouling so as to 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 by Kotjabasakis and Linnhoff (1987) and Fryer (1987), as the effect of fouling in networks is often observed later than desirable.  Heat exchanger designers currently compensate for fouling by overdesign or by using specialised antifouling equipment. The latter devices are usually expensive and involve cleaning-in-place strategies, such as the Taprogge system (Bott, 1988) or the fluidised bed heat exchanger (Kiaren, 1983). The use of turbulence promoters in conventional shell and tube units (Gough and Rogers, 1987; Crittenden et at., 1993) and other surface geometries (Rabas and Chenoweth, 1991) is increasing as the increased  1  1. Introduction  pressure drop is offset by enhanced heat transfer capacity and superior antifouling performance under certain fouling conditions. These devices feature enhanced surface shear rates to retard deposition and are often used without a complete understanding of the fouling mechanism involved. Zhang et at. (1993) related the antifouling behaviour to the lower heat transfer surface temperature necessary to achieve a given heating duty in one such device.  Heat exchangers for fouling service are designed with excess heat transfer capacity in order to offset the losses in efficiency caused by fouling. Equation [1.11 is the general design equation for a simple heat exchanger transferring  Q Watts over an area A with  temperature 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 contributions by 1/U  when A 1  =  =  1 1/h  + ôf,1/?\.f,1 + ômetlXmet + óf,2!?f,2+  2 1/h  [1.21  2 and subscripts are defined in Figure 1.1. As deposit thickness Of] or f A 02  increases over time, U decreases. The usual design strategy is to compensate for the eventual reduction in the overall heat transfer coefficient, U, by increasing the heat transfer surface area, A. The decrease in heat transfer performance is often described by a fouling resistance, Rj, defined as Rf  =  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 of conditions (fluid, velocity, temperature) is usually a correction factor based on previous experience published by the Tubular Exchangers Manufacturers’ Association (TEMA). This value is often used in design (erroneously) as an asymptotic value of fouling resistance which assumes that fouling will reach a steady state after which no further 2  1. Introduction  Figure 1.1  Heat Transfer in the Presence of Fouling Deposits  Heating Medium 11 T  I  Flow  , 0 Ti  20 T  oil  6  met  6f2  1 refers to the processed fluid, 2 refers to the heating medium, ômet is the tube thickness, and of 2 the foulant thicknesses. The heat transfer driving force is the log mean temperature difference, lmz\T, defined as  1mAT  , 2 (T =  ) 1 , 1 T 11 ) 0 (T Ti, 0 , 2 , 2 (T 1 in) T in 0 T , 2 (T ) 0 , 1 -  -  -  -  -  for a cocurrent heat exchanger shown here  3  1. Introduciion  deposit 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 dynamic problem (Bott and Walker, 1971) which ignores the role of optimal operating conditions in decreasing fouling rates. The commonly observed types of fouling resistance behaviour and associated terms are shown in Figure 1.2. The improper use of the TEMA standards can exacerbate fouling, and some examples are given by Bott (1990) in The Fouling Notebook, a nomograph intended to improve engineering practice. The scope for improved heat exchanger design is significant and requires reliable understanding of the mechanisms involved in deposit formation, removal and cleaning.  Fouling occurs in diverse media and so Epstein (1983) classified fouling by the mechanism responsible for deposit generation. Epstein identified five classes of fouling as biofouling, particulate fouling, crystallisation fouling, corrosion fouling and chemical reaction fouling. Chemical reaction fouling was defined as the formation of deposit at the heat transfer surface by chemical reaction in which the surface itself is not a reactant; surface participation is included in the corrosion fouling category. Chemical reaction fouling is an extremely widespread phenomenon as many process fluids undergo foulant generating reactions when heated. Examples of chemical reaction fouling are common in the chemical, nuclear and food industries. Bohnet (1987) identified organic fluid fouling as one of the areas of fouling lacking conclusive study as the diversity of possible reactions presents a complex matrix of possible or concurrent mechanisms. This study addresses one area of chemical reaction fouling in the processing of organic fluids, where deposits are formed by the autocatalytic oxidation (autoxidation) of compounds in the processed stream. The first engineering analysis of refinery fouling was performed by Nelson in 1934 but the level of understanding is still relatively sparse. Industry attention to chemical reaction fouling has focussed on fouling mitigation through the overdesign of process plant and the use of chemical additives to inhibit reaction, disperse foulant precursors or weaken 4  Fouling Resistance  sawtooth fouling increasing rate fouling  Rf  linear fouling falling rate fouling asymptotic fouling  Rf*  0 Time  5  1. Introduction  deposits for efficient cleaning (Cowan and Weintritt, 1978; Nathan 1970). The mechanistic understanding of chemical reaction fouling remains relatively unclear despite the significant costs of chemical reaction fouling and the operating expense of antifouling programmes. This work describes an investigation of chemical reaction fouling caused by autoxidation reactions in oxygenated organic solutions being heated in conventional heat exchanger geometries operating at moderate surface temperatures (100-250°C) in the turbulent flow regime. A known chemical reaction system was chosen using model solutions of an active component, indene, in relatively inert solvents. Characteristics of the chemical reaction which generated fouling precursors were studied separately in a series of batch reactor experiments and used to develop a model of indene autoxidation and to explain the fouling resistance behaviour observed in the fouling experiments. Section 2 is a summary of relevant published material describing autoxidation and chemical reaction fouling. Section 3 describes the experimental methodologies used. The autoxidation chemistry of the model solution is described in Section 4. Initial fouling experiments performed to study the effect of operating parameters are summarised in Section 5; Section 6 describes model fouling experiments performed in a novel fouling unit constructed for this research (the Tube Fouling Unit, abbreviated to TFIJ). Section 7 is a summary of fouling model investigations.  6  2. Literature Review  2.  Literature Review  2.1  The Role of Autoxidation in Chemical Reaction Fouling  The study of organic fluid fouling is complicated by the diversity of the medium involved. Refinery feedstocks and process streams contain species which can undergo a range of chemical reactions depending on composition and the operating conditions involved. Lawler (1979) and Scarborough et at. (1979) concluded that any fluid’s fouling behaviour under given conditions of temperature and pressure is dominated by its chemical composition. Organic fluid fouling often involves chemical reaction fouling in conjunction with other fouling mechanisms, particularly corrosion fouling, particulate fouling and precipitation fouling. Heat exchangers downstream of crude oil desalters are subject to all four processes so that deposits from these units often contain rusts, sand and organic material. Table 2.1 shows the operating conditions and feedstocks involved in various refinery heat exchangers subject to señous fouling and their likely deposition mechanisms. Alkenes and asphaltenes occur most often and these chemical groups have been identified as 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 form insoluble particulates in solution. Studies of fuel storage stability which focus on chemical reactivity rather than heat transfer, such as that by Schwartz et at. (1964), identified alkenes as the major source of gums during fuel storage. Watkinson’s extensive review of organic reaction fouling (1988) confirmed the primacy of asphaltenes and alkenes in organic reaction fouling. Asphaltenes are large, complex ring structure molecules found in heavier crudes and bitumens which form deposits both by reaction to form coke-like deposits and by precipitation. Asphaltene fouling is discussed by Dickakian and Seay (1988), Scarborough et at. (1979) and Crittenden et at. (1993). The present work is concerned primarily with alkene fouling,  7  2. Literature Review  Table 2.1  Examples of Chemical Reaction Fouling in Refineries  Process Stage Heater  Fluid Stream  Maximum Wall  Deposition Source  Reference  Temperatures [°C] Crude Preheat  Crude/residues  374  Asphaltenes and FeS particulates  Lambourn & Durrieu (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 reflux/gas oil  357  Oxygenated alkenes Butler et a!. (1949)  Catalytic Reformer  Gas oil /catalyst slurry  350  Viscous sludge from alkenes  Weiland et at. (1949)  Catalytic 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 NH C 4 1  N.A.C.E. (1970)  Coker-Visbreaker  Residues  540  Organics  NA.C.E. (1970)  240  Organics/rust  N.A.C.E. (1970)  Alkylation Unit Hydroformer  Naphtha  180  Oxygenated alkenes Weiland et at. (1949)  Desulfurizer  Naphtha  360  Metal ions, FeS  8  Crawford & Miller (1963)  2. Literature Review  which 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, the pyrolysis of hydrocarbons generate a range of free radicals which combine to form cokes and long chain polymers. Condensation of alkenes via Diels-Alder reactions can also produce longer chain compounds. At temperatures greater than 700°C, alkanes undergo thermal pyrolysis to form alkenes and other products which foul the cracking reactors extensively. Alkane pyrolysis reactor coking has been studied extensively by Froment and co-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 radical reaction is determined by the operating conditions. In the absence of dissolved oxygen, alkenes can undergo addition polymerisation to form medium to long chain polymeric species which are insoluble in solution or undergo decomposition on heated surfaces. Free radical initiation occurs via metal ions or heteroatomic species present in solution, as well as thermal generation of free radicals in solution. Polymerisation fouling has been studied by 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 alkoxy or peroxy radicals which form peroxides and other oxygenated products via autoxidation reactions. Alkenes undergo autoxidation readily and can form polymeric peroxides which decompose to form insoluble deposits in heat exchangers. Section 2.3 is a review of autoxidation chemistry. Watkinson’s review (1988) and that of Crittenden (1988b) identified autoxidation as a mechanism meriting further study owing to the relative lack of heat transfer studies involving autoxidation fouling and its significance in industrial applications. The key role of oxygen and autoxidation in hydrocarbon plant fouling has been reported by Canapary (1961) and Butler et al. (1949), who achieved lower fouling rates by 9  2. Literature Review  stripping the process feedstocks of air. Braun and Hausler (1976) reported reduced fouling rates from crude oil feedstocks at reduced oxygen concentrations and showed that the deposit morphology also changed with oxygen concentration. Eaton and Lux (1984) reported complicated oxygen effects in their crude oil fouling studies. Alkenes are found in higher concentrations in cracked feedstocks: Gillies (1979), Canapary (1961) and Crawford and Miller (1963) all traced the heavy fouling in these alkene-rich streams to the alkene components. The effect of autoxidation initiators such as metal ions was demonstrated by Crawford and Miller, who observed higher fouling rates when copper ions were added to their feedstocks. The deposits formed in plant fouling indicate an autoxidative source; Vranos etal. (1981) found oxygen concentrations of 10-20 wt% in jet fuel deposits and infra-red analysis indicated that the solids contained significant levels of hydroxyl groups and ketones generated by hydroperoxide decomposition. The formation of insoluble gums in hydrocarbon fuels during storage has been studied due to the undesirable effects of such materials on end use performance. Schwartz et al. (1964) identified the autoxidation of alkenes as a primary source of gums and subsequent studies have confirmed this (Mayo and Lan 1986). Reaction mechanisms and fuel composition effects have been studied extensively, as well as the role of oxygen, initiators (metal ions, ultra violet light), heteronuclear species and inhibitors (antioxidants and chelating agents). Recent experiments investigating specific aspects of fuel stability have been performed using ‘model’ solutions of compounds in relatively inert base stocks used as a control. Morris et al. (1991) have thus used solutions of indene (a reactive alkene) 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 polymeric peroxides. Although not strictly fouling studies, these low temperature experiments (20120°C) contain valuable information about hydrocarbon reactivity and autoxidation. Watkinson (1988) commented that if the extensive library of information available from fuel stability studies were to be connected to fouling behaviour, it would greatly reduce the 10  2. Literature Review  experimentation required to understand the interaction of autoxidation, solution composition and fouling. Asomaning and Watkinson (1992) studied the autoxidation of selected alkenes in kerosene from this basis and found considerable agreement with Taylor’s jet fuel stability studies (1969b). Autoxidation has thus been identified as a significant source of deposits in industrial heaters and of gums in fuel storage. There is a wealth of previous basic chemical research in autoxidation. The mechanism of autoxidation fouling is still poorly understood, however, so most industrial research has focused on the mitigation of such fouling by chemical treatment rather than by optimising design and operation. Antioxidant technology is valuable commercial property and so the literature is seldom current.  2.2  Experimental Studies in Autoxidation Reaction Fouling  Few studies extending the understanding of fuel storage stabilities to thermal fouling behaviour have been performed, although suggested in reviews (Watkinson, 1988; Crittenden, 1988b). Similarly, autoxidation has been suggested as the primary source of deposition in oxygenated feedstock fouling but few results have been published linking the effects involved quantitatively. This is needed for the design of robust antifouling heat exchangers, such as described by Bradley and Fryer (1992) using an existing fouling model to assess the scope of various design and operating strategies. Table 2.2 is a summary of experimental studies of fouling where autoxidation has been identified as a primary source of deposition. These studies can be divided into reaction mechanism studies and thermal fouling studies. Little work has been performed to quantitatively link autoxidation chemistry to foulant generation and thermal effects, which is a goal of the current study.  11  2. Literature Review  Table 2.2  Summary of Autoxidation Related Fouling Studies  Feedstock  Jet fuels, doped alkanes [oxygenated] Jet fuels, doped alkanes [deoxygenated] Oil Refinery Feedstocks Crude oil, kerosene, shale oils Crude Oil Refinery cuts with pitch, resin  Fouling Probe,  Temperature  (Measurements)  Range  HeatedTube  Pressure  Experiment Period  (°C)  (kPa)  (hours)  150 <Tb <260  20.6  t <  4  Taylor eta!. (1967, 68a,b, 69a,b)  150 <Tb  <  560  6990  t <  4  Taylor et a!. (1974, 76, 78, 80)  200  <  Tw  <  340  1480  t <  50  Tb  <  100  Hausler and Thalmayer (1975), Braun and Hausler (1976)  Pressurised  t <  3  Latos and Franke (1982)  .  (Mass Deposition) Heated Tube (Mass Deposition) Hot Wire Probe .  (Thermal Fouling)  175 <T <400  Hot Wire Probe .  Reference  90  (Thermal Fouling)  Tb  Heated Probe, Rotating Cylinder  T <287 71 <Tb  2000  t  20  Eaton and Lux (1984)  16  Vranos eta!. (1981)  287  <  (Thermal Fouling) Jet Fuel  Heated Tube  120 <Tb  <  355  6990  t <  190  <  538  5597  t  <  2.5  Hazlett eta!. (1977)  T <310  5597  t <  2.5  Morris et al. (1988, 1989)  190 <T <538  5597  t  <2.5  Morris and Mushrush (1991)  106 253  t  100  Crittenden and Khater (1984)  480  t <  50  Asomaning and Watkinson (1992)  1034  t <  68  Oufer (1990),  (Mass Deposition) Aerated n-dodecane  Modified JFT’OT  <  Tw  (Mass Depsoition) Jet fuels, Stabilizers  Modified JFTOT (Mass Deposition)  Model Fuels, Additives  Modified JFTOT .  .  (Mass Deposition) 160  <  T  Hot Wire Probe,  140  <  Tw <204  Annular Probe  Tb  Aerated kerosene  Tubeside  <  380  -  Vapounsation (Thermal Fouling) Alkenes in aerated kerosene  80  (Thermal Fouling) Styrene in heptane  Annular Probe  151 <T  (Thermal Fouling) 87  Tb  -  bulk temperature (°C);  T  -  <  Tb  <  <  190  100  maximum wall temperature (°C)  12  2. Literature Review  2.2.1  Fuel Stability Studies  The fuel stability studies of Taylor et al. (1967  -  1980), Hazlett et at. (1977) and  Hausler and Thalmayer (1975) involved passing hydrocarbon streams over heated surfaces under evaporative conditions and measuring the formation of deposits and of oxidation products in solution. The effect of surface temperature is reported in terms of activation energies for deposit formation rates based on an Arrhenius type expression, i.e. Deposition rate  dmf/dt  =  B exp  [- Eact!R Tsurt]  [2.11  or as a threshold temperature below which little deposition occurs. The effect of liquid flow rate and thus hydrodynamics was not reported. Taylor initially studied deposition from jet fuel blends doped with various additives under oxygenated and deoxygenated conditions. Deposition rates were significantly higher in the presence of dissolved oxygen, and solution and solid analyses were consistent with an autoxidation mechanism. Deposition from blends doped with organic oxygen, nitrogen and sulphur species varied with compound structure; peroxides and their analogues increased deposition while phenol analogues reduced or maintained deposition rates. Hazlett et al. (1977) reported similar results for sulphur derivatives using the Jet Fuel Thermal Oxidation Tester (JFTOT), a commercial device used to assess the fouling tendency of process liquids. Taylor (1968b) studied deposition on various metal surfaces and found enhancement with metals that could initiate radical formation. Steels were relatively inert compared to copper, the metal used by Crawford and Miller (1963) in their crude oil tests. Taylor (1968a,b, 1969b) used ‘model solutions’ of 10 wt% alkanes, alkenes and aromatics in n-dodecane to study the effect of jet fuel component structure on deposition rates and found heavy deposition from alkenes. Table 2.3 is reproduced from Asomaning and Watkinson (1992) and shows Taylor’s results and the measured thermal fouling results from their thermal fouling experiments. Hexadec- 1 -ene, dicyclopentadiene and indene are known to form polymeric peroxides and these show very large deposition rates. This 13  2. Literature Review  Table 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 to  Formula  Point  Deposition to  Resistance  Rf (indene)  (101 kPa)  Deposition of  after 40 hrs,  indene  .KJW] 2 [m  t  oct-1-ene  CH=CH 5 ) 2 Me(CH  C 1 H 8 6  121.6°C  0.0  0.0  0.0  dec-l-ene  CH=CH 7 ) 2 Me(CH  H 1 C 2 0  170.9°C  0.1  0.015*  0.002*  hexadec-1-ene  12 ) 2 Me(CH CH=CH 3  1 6-32 C  284.4°C  0.25  0.33  4-vinylcyclohexene  C 1 H 8 2  127.0°C  0.16  0.21  1,5-cyclooctadiene  C 1 H 8 2  150.8°C  0.0  0.0  indene  8 H 9 C  182.6°C  0.75  1.0  dicyclopentadiene  H C 1 0 2  170.0°C  0.45**  2.2**  -  0.55 -  1.0 -  Reproduced from Taylor (1969b) [] and Asomaning and Watkinson (1992) [tI. -  t  -  10 wt% alkene in dodecane evaporated at 135°C under 20.6 kPa oxygen saturation. 10 wt% alkene in aerated kerosene at 40.6 kPa oxygen saturation, Re  5 T  180°C, Heat flux  300 kW/m . 2  * -  Heat Flux 250 kW/m ; 2  14  9600, Tbulk  ** -  80°C,  Heat Flux 350 kW/m 2  t  2. Literature Review  variation with compound structure confirms Canapary’s hypothesis on the complex nature of organic reaction fouling. Hazlett et al. (1977) analysed the fuels for hydroperoxides, the initial product of autoxidation, and detected significantly higher levels post heating. They also employed on-line gas chromatography (GC) to measure dissolved oxygen levels in solution and found that deposition rates and deposit patterns were determined by oxygen concentration. Taylor and Wallace (1967) obtained an activation energy of 42 id/mo! for jet fuel deposit formation at temperatures <300°C; above this temperature the activation energy decreased, indicating that different mechanisms were involved (or that mass transfer was significant). This work confirms that it is important to identify the range of any fouling studies before comparisons are made. Table 2.4 shows the activation energies reported in autoxidation 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) in their study of jet fuel coking. Table 2.5 lists the velocity effects reported in the literature and shows conflicting trends. Many early studies did not report whether constant interface temperatures were used, which would decouple temperature effects from any velocity effect. Epstein (1993a,b) proposed a model for chemical reaction fouling which included complex velocity effects as reported by Crittenden et al. (1987b). Vranos et al. found the coking rate to be proportional to Re° ; this is one of the few instances of fouling rate 6 increasing with flow rate under autoxidative conditions and suggested that coking was controlled by mass transfer. Deposit studies using FTIR and elemental analysis indicated that the deposits were formed by liquid phase oxidation reactions, while chemical analysis of solution samples detected hydroperoxides and other liquid phase oxidation products. Transmission electron microscope studies showed that the foulant consisted of particulates of 15A diameter; Vranos et al. concluded that solubility processes as well as oxidation reactions were involved in the formation of deposits. This reaction and precipitation  15  2. Literature Review  Table 2.4  Activation Energies Reported in Chemical Reaction Fouling  Temperature Range  Activation Energy Eact  (°C)  (kJ/mol)  Crude Oil (heavy) (light)  160 280 160 280  21 33  Refinery Feedstocks  200 340  Gas Oil  146 204  120  Watkinson and Epstein (1968)  Crude Oil Feedstocks  365- 447  53  Scarborough etal. (1979)  Used Lubricating Oils  343  Jet Fuels  93  Crude Oil Feedstocks  71  Shale Oil  175  Jet Fuel  149  Kerosene (Vapourisation)  160  Feedstock  Dilute Polymerisation of styrene in kerosene Styrene in heptane  -  -  40  -  -  -  -  -  -  -  -  -  Crittenden et at. (1992) Braun and Hausler (1976)  455  74 97  260  40  Taylor and Wallace (1967)  287  36  Eaton and Lux (1984)  -  Steele et al. (1981)  400  15 20  260  42  Vranos etal (1981)  380  70  Crittenden and Khater (1984)  56  Crittenden etal. (1987a)  -  80-150 164  120  -  Reference  201.1  179-3 19  16  Latos and Franke (1982) .  Oufer (1990)  2. Literature Review  Table 2.5  Velocity Effects Reported in Chemical Reaction Fouling  Reference  Feeds tock  Flow Velocity  (mis)  Effect of Flow Rate Increase on Fouling Rate  Comments  decreases  Industrial plant data  Nelson (1958)  Desalted, wet and corrosive crude oils  Chantry and Church (1958)  Forced Circulation Reboilers  Charlesworth (1963)  Organic Coolants  Parkins (1961)  Terphenyls  Canapary(1961)  Gas Oil  Hillyer (1963)  Terphenyls  Smith (1969)  Jet Fuel (kerosene)  4500  Re  10 000  increases  Watkinson and Epstein (1969)*  Sour Gas Oil  9800  Re  41 900  decreases  Scarborough et at. (1979)  Crude oil  Vranos et al. (1981)*  Jet Fuels  Dickakian and Seay (1988)  Crude Oil  Fetissoff (1982)*  Styrene in heptane  8600  Oufer (1990)  Styrene in heptane  1  0.3 <Urn <2.1 3 <Urn  <  10  decreases decreases  -  decreases 1  Urn  3  decreases  decreases 600  <  Re  <  10 000  increases  1 Re  Re  1.77  Re 0.6  increases  <  Re  <  <Urn <  28 800  no effect  2.5  decreases  boiling studies 35 Re(  Cnttenden et al. (1987a)*  Styrene in kerosene  Crittenden et at. (1992)  Crude Oil Feeds  urn mean fluid velocity:  * indicates that  1100 <Re <5200  wall temperatures were fixed  17  5.8)  increases or decreases  Maximum in Re relationship  inconclusive  Industrial plant data  2. Literature Review  hypothesis arises in other cases of chemical reaction fouling, notably milk fouling and polymerisation fouling.  2.2.2  Thermal Fouling Studies  Thermal fouling studies involve the measurement of heat transfer coefficients and are performed to assess the significance of fouling on heat transfer. Many of the fouling studies listed in Table 2.2 did not include the effects of deposition on heat transfer. Most of the 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 of autoxidation fouling. Eaton and Lux (1984) found the fouling behaviour of their hydrodesulfurizer feedstock to depend on the feedstock nature. Fouling rates increased significantly with dissolved oxygen concentration, whereas Braun and Hausler (1976), who used a series of petroleum feedstocks, found oxygen important in some cases and not in others. The reactants or fouling precursors were not identified. Hausler and Thalmayer used feedstocks doped with suiphides and found that in the presence of oxygen, the suiphides 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 fouling studies. Oufer (1990) reported shorter induction periods in oxygenated solutions of styrene in heptane and generally larger rates, though no conclusive trend was established. Oufer also investigated the effect of oxygen on thiol-doped styrene solutions and again reported shorter induction periods. Crittenden and Khater (1984) studied fouling in a horizontal, single tube kerosene vapouriser under oxygenated conditions and attributed fouling to a liquid phase autoxidation mechanism. Their results showed strong variation of fouling with axial and  18  2. Literature Review  circumferential location due to the complicated distributions of oxygen concentration and bubble density in an evaporative system. The complexity of chemical reactions in chemical reaction fouling has spurred the use of model solutions in experimental studies so as to simplify the system and trace deposit 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 foulant generation process. More reliable physical property and kinetic parameters can then be used in testing fouling models. The first study of chemical reaction fouling using model solutions was performed by Palen and Westwater (1966) using solutions of styrene in toluene. Thermally initiated solutions of 35-100 v/v% styrene were used to study fouling under pool boiling conditions. Fouling was reported to be extremely rapid after an induction period which decreased with increases in styrene concentration and heat flux. Crittenden et al. (1987a,b) reported the first systematic study of chemical reaction fouling with associated reaction kinetic studies in their study of polymerisation fouling using  solutions of 1 v!v% styrene in kerosene. Chemical reaction parameters were obtained separately from the fouling experiments and used in a numerical model to compare the  fouling data with the model predictions. This study did not involve autoxidation and reported complex velocity effects which are discussed further in Section 2.4. The same approach was used by Oufer (1990) to study fouling in boiling solutions of styrene in heptane. Oufer varied styrene and oxygen concentrations, surface temperature, flow velocity and the effect of sulphur additives. No chemical analysis was performed to establish solution reaction chemistry; the large activation energies (>179 kJ/mol) and strong velocity effects (dR /dt ceRe’ where n ranged from -3.5 to -5.78) reported do not correlate 1 well with previous results from Tables 2.4 and 2.5, many of which are for non-boiling solutions.  Asomaning and Watkinson’s (1992) study of fouling from oxygenated model solutions of alkenes in kerosene has been mentioned as a thermal fouling extension of 19  2. Literature Review  Taylor’s jet fuel work (1969b). Their results reproduced in Table 2.3 show that the alkenes which Taylor found to produce most gum also generated the heaviest fouling deposits. Deposit analyses indicated that the foulant was derived from polymeric peroxides and indene polyperoxide was obtained as a precipitate from the resultant solutions of indene in kerosene. No solution chemistry analysis was performed to establish the controlling reaction. Fouling runs of indene and dicyclopentadiene under deoxygenated conditions gave reduced fouling consistent with differences in autoxidation and polymerisation rates, while runs at higher heat fluxes produced heavier fouling. No activation energies or flow rate effects were reported. Asomaning and Watkinson’s study identified indene, dicyclopentadiene and hexadec-1-ene as alkenes for further model solution studies in autoxidation fouling. The current work is part of an extended project using such solutions described in part by Wilson and Watkinson (1992), Zhang et al. (1992), Lai (1992), Lai (1993) and Zhang et al. (1993). The experimental study of autoxidation fouling has thus progressed from explaining results in terms of a general n-th order chemical reaction with unidentified reactants to model solution studies investigating the processes in the fouling mechanism. The relationship between reaction rates and deposition still remains unclear, however. The apparatus used in autoxidation fouling studies consists of compact heat exchangers with well defined geometries and flow patterns. Electrical heating is commonly used as the power and heat fluxes are readily controlled and monitored. In most of the fuel deposition studies the fluid is heated using a once-through arrangement at high temperatures in order to accelerate the fouling rate. In thermal fouling studies recirculation of the process liquid is generally employed in order to use reasonable quantities of fluid as experimental periods are frequently long. The selection and design of fouling probes has been reviewed by Fetissoff (1982) and Chenoweth (1988). The latter also describes commercially available fuel oxidation testers. Researchers have tended recently to use 20  2. Literature Review  simpler flow geometries (annulus, tube) for thermal fouling in order to facilitate modelling studies. Many of the analytical techniques available to the modern organic chemist have been used in solution analysis. Determination of dissolved oxygen concentrations is complex but has been performed by Hazlett et al. (1977). Modern surface analysis methods have 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 Chemistry  2.3.1  Autoxidation Mechanisms and Kinetics  Autoxidation is the general term for reactions involving the autocatalytic oxidation of hydrocarbons by a free radical mechanism where the dominant radical species is the peroxy radical, ROO. Autoxidation occurs in paint drying, fuel storage and rubber degradation, with the latter phenomenon spurring the original research into autoxidation mechanisms. A review of the extensive material describing autoxidation and its mechanisms is given by Reich and Stivala (1969). The discussion which follows is a summary of autoxidation mechanisms related to alkanes and alkenes under the conditions used in the current study. Hydrocarbons can undergo addition polymerisation in the presence of alkyl radicals, R, as in the chain (vinyl) polymerisation of styrene R  +  R  -  R-R  +  R  —  R-R-R  [2.21  Alkenes usually undergo addition polymerisation more rapidly than alkanes due to the ease of addition of R to the C=C double bond. In the presence of significant concentrations of dissolved oxygen, however, the alkyl or allyl radical combines with molecular oxygen to form the peroxy radical, R0 , via the fast step 2 R  +  02  —  [2.3]  2 RO  21  2. Literature Review  Autoxidation chemistry is determined by peroxy radical behaviour and the mechanism of autoxidation was summarised in the ‘modified basic autoxidation scheme’ below by Van Sickle et al. (1965a, 65b, 67) as follows. The labels are those found in the literature. Initiation  ) 1 (R  Propagation  R  —  02  +  —  2 R0  +  RH  2 R0  +  RH  R0 R 2 H  +  —  —  02  [1]  2 R0  [2]  H 2 RO  —  R0  +  RH  RO  +  RH  -  -  2 R0 2  —  2 R0  —  2R  —  R  +  RO R 2 H -  R 2 R0  Termination  R  ROH ROM  (H abstraction)[3a]  (E  R)  R0 2 R0 RO +  +  (addition) ) 2 ( RO  MO  (epoxidation)  R  (E  [3b] [2’]  [71 [8]  R•)  [9]  inert products  [6]  inert products  [5]  inert products  [4]  where MO denotes epoxide, M denotes a monomer unit, RH, and R0 is the alkoxy radical. The reaction scheme is quite complex and the values of the respective kinetic rate constants are determined by hydrocarbon structure and solvent nature. In the absence of significant steric crowding, the rate constants k 2 and Ic ’ are equal. 2 Steps [2,2’] are relatively rapid in significant oxygen concentrations so the dominant radical at high oxygen concentrations is the peroxy radical and steps [4, 51 can be ignored. Steps [3a,3b] are thus rate determining and the reaction rate can be written as d[0 ] 2 /dt  =  d[RH]Idt  =  -  ] [RH] 2 (k a 3 +k3b) [R0  [2.4]  The steady state approximation is applied to the peroxy radical concentration. Equating the free radical initiation rate, R , to termination via step [6] gives the general result for long 1 kinetic chain lengths: d[0 ] 2 /dt  =  d[RH]/dt  =  -  )° [RH] 6 (RI2.k (k3a+k b 3 )5  The reaction rate is thus determined by the mode and rate of free radical initiation. 22  [2.5]  2. Literature Review  As the oxygen concentration decreases, the alkoxy radical and eventually the alkyl radical become dominant, altering the mechanism appropriately. The conditions at which autoxidation (via the peroxy radical) becomes an alkoxy radical process are determined by the monomer structure. Mayo et al. (1956a,b) thus observed a minimum in styrene reaction rate at 50°C as the oxygen pressure was increased from 0 kPa to 100 kPa, corresponding to the region between addition polymerisation and autoxidation where R& is the dominant radical species. The transition in mechanisms explains the complex effects reported for oxygen concentration in chemical reaction fouling. The oxygen concentrations involved in the transition between radical mechanisms are determined by the alkene structure and are thus difficult to predict for mixtures. Nicholson (1991) thus reported the use of air blanketing of methylacrylic acid in order to inhibit polymerisation and did not report significant autoxidation problems. Asomaning and Watkinson (1992) found their aerated and deoxygenated fouling resistance data to mirror the expected polymerisation rates for indene 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 polymeric peroxy radical. The competition between these steps is determined by the hydrocarbon’s structure. Alkanes generally favour hydrogen abstraction and the reaction rate is governed by the energetics of hydrogen abstraction and the ability to stabilise the resulting radical. The tendency of alkenes to undergo addition (oxidative copolymerisation) is also determined by radical stability, with aromatic alkenes and terminally double-bonded long chain alkenes reported as favouring this pathway over abstraction. The addition step generates polymeric peroxides, which Mayo and Lan (1986) identified as the primary source of gum in fuel storage and which Asomaning and Watkinson (1992) considered to be the source of deposits in their fouling studies. Heavy deposition was observed in indene, 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. Van 23  2. Literature Review  Sickle et at. (196Th) studied the tendency of alkanes and alkenes to form epoxides and alkols by radical rearrangement (steps [7], [8]) and reported values of > 8 1 2 k k 1000 for alkenes which favour addition at high oxygen pressures. They explained this in terms of resonance energy losses and emphasised the role of ring strain in reducing the tendency to form epoxides inS- ring alkenes such as cyclopentene. Steps [7,81 are thus unlikely to be significant in the autoxidation of compounds such as indene. These results mirror the fouling tendencies observed by Asomaning and Watkinson, who concluded that autoxidation fouling was caused by the formation of insoluble polymeric peroxides which subsequently decomposed on the hot heat exchanger surface. The propagation step products can undergo further reaction as peroxides and polyperoxides are known to be unstable at high temperatures and decompose to form carbonyls, alkols and various free radical species. Tert-butyl-hydroperoxide, (Me) COOH, 3 is thus a common free radical initiator. The formation of these oxygenated secondary reaction 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 of hydroperoxides has been reported to be unimolecular at low concentrations but bimolecular at high concentrations, especially where hydrogen bonding is likely to generate peroxide dimers in non-polar solvents. Equation [2.5] indicates that autoxidation rates are controlled by the initiation rate, . Hydroperoxide decomposition has been reported to be the primary initiation source in 1 R many studies, which would involve a kinetic expression of the form; 1 R  H] 2 1 [RO , 11 {k  =  +  H] 2 [RO }  [2.6]  Proposed reaction schemes are; Unimolecular RO H 2 Bimolecular  +  2 RO H 2  RH  —‘  RO  —  2 RO  + +  R R  +  0 (Neiman, 1964) 2 H  [2.7]  +  0 (Toboisky et al. 1950) 2 H  [2.8]  24  2. Literature Review  Toboisky 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]. Most chemical kinetic studies have considered the initial stages of autoxidation, at low extents of conversion and consequently small peroxide and side product concentrations, where R 1 is relatively well defined. An alternative approach to complex reaction kinetics was used by Norton and Drayer (1968) in their kinetic model of hexadec-1-ene autoxidation to form hydro- and poly-peroxides. The data were fitted to an empirically based scheme which described the distribution of products observed in the experimental data hexadec-1-ene -4 hydroperoxide  —  polyperoxide  —*  insoluble peroxides [2.9]  The model did not reflect the mechanism of the free radical chemistry involved but did offer some rationalisation of the processes involved. The rate constants obtained were thus lumped kinetic constants; such an empirical approach may be more appropriate in cases with complex initiation sources.  2.3.2  Solvent and Additive Effects in Autoxidation Autoxidation kinetics and product distributions are influenced by temperature,  product reactivity, oxygen concentration and solvent nature. Increases in temperature generally increase reaction rates but can introduce side reactions of labile products. Reaction products such as acids can inhibit further autoxidation. The oxygen dependence of the dominant free radical has been discussed previously but analysis of Eqns. [l]-[9] has been shown to predict the higher epoxide yields at lower oxygen concentrations reported from experiments (Reich and Stivala, 1969). Solvent effects are observed in rate constants, hydroperoxide behaviour and product solubility. Polar solvents tend to enhance the propagation rate constants (k3a, k3b)  25  2. Literature Review  while non-polar solvents have retarding effects. Solvents with abstractable protons available can interrupt the solute reaction via chain transfer to the solvent (5). Chain Transfer  R  S-H  +  R-H  —*  +  S  [2.101  The 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 and can also affect the product distribution. Solvent effects in hydroperoxide decomposition involve hydrogen bonding of the polarised hydroperoxide molecules to form dimers in non-polar solvents, while enhanced hydroperoxide decomposition rates have been reported in polar solvents. Cage mechanisms, where solvent viscosity is significant, have also been postulated for peroxy radical termination. Product solubility depends on solvent nature, where the maxim ‘like dissolves like’ applies. Most autoxidation products are soluble in the original hydrocarbon but particular exceptions are polymeric peroxides, such as reported by Norton and Drayer (1968). Polymeric indene peroxide was soluble in pure indene but not in solutions with more aliphatic, less aromatic and less polar solvents so Russell used solvent precipitation to identify the reaction products of indene autoxidation (1956a). The solubility of reaction products in solutions depends on the solvent nature, often expressed in terms of the Hildebrand solubility parameter,  The solubility parameter is related to  the free energy of solution and is used in the estimation of the activity coefficient of a dissolved solute in equilibrium with the solvent. The value of  aH  is usually found by  experiment, but Small (1953) proposed an additive approach and estimation methods are given in ASTM D3827-86. The solubility of a solute (2) containing no solvent (1) is given by  22 my where  lny2  where  2 c  = =  -  [AH/RTJ (1 T/T)  2 1 [V t ! 2 RTI  +  -  (ãH,1  -  2 8H,2)  is the mol fraction of solute in solution and  V2  F (Cp(1,2))  [2.11] [2.121  is the molar volume of solute.  This approach is not valid for long chain polymers where Flory-Huggins solution thermodynamics apply. The solubility of foulant precursors is an important yet poorly 26  2. Literature Review  understood aspect of chemical reaction fouling as many researchers have reported that deposits originated as insoluble materials precipitated from solution rather than generated by simple wall reactions. The roles of sulphur, nitrogen and oxygen heteroatomic species in autoxidation have been discussed extensively in the autoxidation literature. The structure of a compound determines the C-X bond energy, which dictates its reactivity towards the peroxy radicals and hydroperoxides formed during autoxidation. Changes in mechanisms are reported at enhanced temperatures due to fission of the C-X bond to form free radicals or condensation products. Disuiphides and diazo compounds contain the weakest C-X-X bonds and react in similar fashion to their oxygen analogues, the peroxides, which are recognised free radical initiators. The thermal disociation of disuiphide, for example, is given by R-C-S-S-C-R  -  [2.13]  2 R-C-S  Thiols similarly resemble phenols in acting as antioxidants or inhibitors at low temperatures, above which they contribute to deposit formation. White et al. (1983) found phenols to be a source of deposits via phenolic coupling reactions, while Taylor (1976) reported enhanced deposition rates of sulphur-doped fuels above threshold temperatures which he linked to C-S bond strengths. Taylor (1970) observed enhanced deposition rates when various oxygenated dopants were added to aerated jet fuels at 100 ppm, depending on dopant structure. Watkinson (1988) reviewed the role of oxygen, sulphur and nitrogen heteroatomic species in fuel stability studies and fouling work and found similar patterns to exist in both. The significance of these species in fouling is shown by the increased concentrations of sulphur, oxygen and nitrogen reported in deposit analyses when compared to the fluid (Watkinson and Epstein, 1969). Dissolved metal ions are well known sources of radical initiation and can also function as hydroperoxide decomposers. Crawford and Miller’s study of metal ions in oxygen uptake showed significant increases at concentrations as low as 1 ppm (1963). The role of metal ions in autoxidation was reviewed by Reich and Stivala (1969). 27  2. Literature Review  8 H 9 The Autoxidation of Indene, C  2.3.3  The present work concentrates on model solutions of the aromatic alkene, indene, in solvents which are relatively inert to autoxidation. Indene autoxidation was studied by Russell (1956a,b), who reported significant polymeric peroxide formation and an addition/abstraction ratio [k3a/k3b] of 4:1 at 50°C. The major products were polymeric O), where n ranged from 2-10), carbonyls, and indene ((C O 8 H indene peroxides 9 hydroperoxide but not epoxide. Studies were performed in pure indene or in solutions in bromobenzene. Ueno et al. (1974) analysed the products from the extended autoxidation of indene and reported a product slate of various carbonyls and aldehydes consistent with polyperoxide formation and degradation. Morris et al. (1988) and Morris and Mushrush (1991) have also used indene in model solution studies of sulphur compounds in fuel storage stability. Figure 2.1 is a summary of the indene autoxidation mechanism. Russell also investigated the mechanism of thermal initiation in indene and reported it as a bimolecular reaction with oxygen to form the peroxy radical; 8 H 9 C  +  02  -  0• C 0 8 H 9  [2.14]  This mechanism has been disputed on energetic grounds in favour of one involving the unimolecular dissociation of hydroperoxide. Carisson and Robb (1966) investigated the autoxidation of indene in excess antioxidant at 70-95°C and proposed a termolecular initiation step involving two indene molecules and an oxygen molecule with an activation energy of 78.7 kJ/mol. They reported that traces of hydroperoxide increased autoxidation rates significantly. Indene polyperoxides were formed by the cooxidation of indene with oxygen via the indene peroxy radical. Russell (1956b) studied thermal and AJEN initiated oxidation and concluded that the reaction mechanism did not change under a different initiation source. Howard and Ingold (1962) measured the rate constants for the initial propagation  28  2. Literature Review  Autoxidation of Indene  Figure 2.1  initiation  00 indene  indene peroxy radical  0  7  abstractioV  ocr hydroperoxide  addition  H  abstraction  hydroperoxide  Figure 2.2  0  polyperoxy radical Z  _0. addition  polymeric peroxides  Antioxidant Action of 2,6 t-butyl,4 methyl phenol (BMP) tBi  ROO  +  ROOH  +  R00+ tBu  tBu  29  2. Literature Review  and termination steps in indene autoxidation at 50°C but, like Russell, did not report any activation 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  [2.151  Howard and Ingold reported an indene additionlabstraction ratio of 9:1 at 30°C.  Antioxidation  2.3.4  Additives are often used to prevent the formation of permanent deposits on heat transfer surfaces. Stephenson and Rowe (1993) discussed the mitigation strategies employed 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 form complexes with the metal ions present in solution. Antioxidants are primarily preventative in action and Scott (1965) describes the two modes of antioxidant action as hydroperoxide decomposition and radical deactivation. In the former, hydroperoxides are decomposed to form stable products without generating new radicals; this can be achieved by selected metal 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 chain reaction; examples of radical scavengers are hindered phenols, amines and certain thiophenes. 2 RO  +  A-H  2 RO  +  A  —  —>  H 2 RO  +  A 2 RO  A  (stable)  [2.16]  (stable)  [2.17]  Figure 2.2 shows the reaction of a common gasoline antioxidant, 2,4,di-butyl-4methylphenol (BMP), with peroxy radicals to form stable products. The efficiency of hindered phenols as antioxidants is structure dependent. The chemistry of antioxidation has been reviewed by Reich and Stivala (1969). Antioxidation remains an important research 30  2. Literature Review  area 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 amine varied with operating temperature. BMP was relatively ineffective at temperatures above a ‘ceiling temperature’ of 100°C. Ceiling temperatures refer to the temperature above which an antioxidant rapidly loses efficiency.  2.4  Mechanistic Modelling of Fouling  The aim of mechanistic modelling is to be able to predict and explain the behaviour of a fouling system when parameters such as hydrodynamics, temperature and chemistry are varied. A reliable and robust fouling model could be used both in the design of antifouling heat exchangers (Fryer et al., 1990) or in the development of control strategies for heat exchangers in fouling service (Fryer and Slater, 1985). The accuracy of a mechanistic model is determined by the understanding of the physical phenomena involved. The complexity and number of processes involved in chemical reaction fouling has prevented the development of a universal model, so that certain cases, such as coking, are relatively 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 bulk fluid. In situ generation can occur at the surface itself, in the ‘reaction zone’ near the surface where conditions are favourable, or in the bulk fluid (i.e., conditions are favourable everywhere). The foulants andlor reactants may have to be convected to the reaction zone and the surface, and likewise reaction products or the foulant (if still mobile) may be returned to the bulk fluid and may cause deposition downstream. The foulant precursor can form deposit either by surface layer growth or by the precipitation and/or adhesion of insoluble agglomerates. Froment (1981) reported the morphology of coke deposits in ethane crackers as a mixture of amorphous coke and regular, dendritic regions similar to 31  2. Literature Review  Figure 23 Schematic of Processes Involved in Chemical Reaction Fouling  Hot Wall Aged Deposit New Deposit  precursors  I  REACTION ZONE  foulant  4 products (including foulant)  Bulk Fluid Flow  Heat Flow  -Ii,  Adapted from Crittenden ci al. (1987b)  32  2. Literature Review  metal crystal growth. Most studies of deposit morphology in the medium temperature range describe 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 fouling involving two different materials milk proteins and polystyrene, respectively. Dickakian -  (1990) has also emphasised the role of asphaltene insolubility in crude oil fouling. The mechanism of forming solid deposits from the liquid and gaseous phase is the controlling step in chemical reaction fouling models. Once formed, the deposit may undergo further reaction as described by Atkins (1962) or be removed by shear forces. Existing models involve various combinations of these steps and Table 2.6 is a summary of chemical reaction fouling models in the literature.  Transport and Adhesion Models  2.4.1  Many fouling models describe the phenomenon as a series of transport, reaction, adhesion and removal steps. Kern and Seaton (1959) formulated an empirical model to correlate plant data of particulate fouling, which fitted an expression of the form Rf(t)  =  Rf*( 1  -  [2.18]  exp[-b.t])  where Rf* is the asymptotic fouling resistance and 1/b a time constant. Kern and Seaton modelled fouling as a competition between a deposition flux,  proportional to the  dep  concentration of precursors C and the mass flow rate W, and a removal flux,  rem  proportional to the deposit thickness Xf and the wall shear stress, ‘r. dRf/dt  =  [1I2] dxf/dt  =  Pdep  Rf*.b  =  -  [2.19]  ørem  This model gives the initial fouling rate as (dRf/dt)  t=O  =  ødep  =  kdep Cp W  Taborek et al. (1972) included a deposit shear strength factor,  iyD,  [2.20]  in the removal term  which Pinhero (1981) suggested should be a function of velocity. The increase in  lrem  with  deposit thickness has been rationalised by Loo and Bridgewater (1981) in a theory of  33  Cs  <.5  C  0  —I  n  a  Vapour phase pyrolysis Ultrapyrolysin of ethane  Fernandez-Bsujin and Solomon (1976)  Ssndaram and Froment (1979)  Autoxidation Fouling in Model Solutions  Panchal and  Epstein (1993a,b)  Watkinson (1993)  Model solution evaporation styrene  Oufer, Brebcr and Knudsen (1993)  Hydrocarbons in seneral  -  Milk Fouling  Paterson and Fryer (1988)  Hydrocarbons in Crittenden, Kolaczkowski and general Hout (1979)  Vapour phase pyrolsis  Liquid phase gas oil fouling  Jackman and Aris (1971)  Watkinson and Epstein (1970)  Nijsing (1964)  Deposition Term Rate is directly dependent upon thermal boundary layer thickness Constant monthly increase in coke resistance for various refinery streams  Removal Term  Kinetics and mass transfer control with nth order reaction and an attachment factor involving mean residence time at the surface  lb Mass Transfer from bulk liquid Ia Mass Transfer with Chemical Reaction 2 Boundary Layer Generation of Foulanl  Kinetics and mass transfer control with second order reaction  -  Boundary tayer as differential chemical reactor kinetic/adhesion control (I) Foulant generation in boundary layer (2) Deposition via sticking probability as described by Watkinson and Epstein  Fitted to each set of experimental data la,2 showed trends observed in Rf profiles No sticking probability; batch kinetics  Foulant Back Diffusion invoked in lb,2  Predicted velocity and temperature effects in styrene polymerisation fouling  Unknown parameters in boiling mass transfer. Large deviations from observed results. (1) First order Kern and Seston shear removal term (2) Negligible foulant back diffusion  none considered  Correct prediction of initial rate dependence on velocity None considered. Fryer and Slater (1985) simulation uses first order Kern and Seaton shear removal term  -  Limited testing Complex many parameters Extended to two-layer concept proposed Atkins  (1) Quasi-stesdy state assumption (2) Good agreement between industrial data and numerical simulation (1) (2) (3) by  None considered  Kinetics control (1) at surface temperature (2) first order in concentration of cracking products  Solution with mass transfer control fits plant run-time data  (1) Quasi-steady state assumption (2) Untested  (1) Correct prediction of initial rate dependence on velocity (2) Incorrect prediction of asymptotic resistance on velocity  (1) Diffusion of foulant back into bulk fluid (2) First order Kern and Seaton shear removal term  None considered  Kinetics and/or mass transfer control with first order reaction  First order Kern and Seaton shear removal term  None considered  Kinetics and/or mass transfer control with first order reaction  -  Remarks Fouling Rate can be reduced by increasing fluid velocity Two layer concept porous coke layer adjacent to mid and hard coke next to wall  Product diffusion back to the (I) Solution with diffusion control fits plant liquid is an integral part of data. Fouling rate predicted to increase with velocity the differential equations (2) Extended to consider colloidal transfer to hot surface  None considered  None considered  Kinetic control two reactions (1) First order dissociation of A into products (2) Zeroth order coke formation -  Mass transfer and adhesion of suspended particles (1) sticking probability proportional to exp (-E/RT) (2) sticking probability inversely proportional to shear stress at wall  Hydrodynamic boundary layer and diffusion partial Organic coolants differential equations in nuclear (1) instantaneous first order reaction in zone close to the wall reactors (2) very rapid crystallisation at hot surface  Oil refining Fired refinery heaters  Nelson (1934)  Atkins (1962)  Application  Authors  C,,  C  C  ri  I  LID  2. Literature Review  deposit shattering by thermal stresses. The removal flux has been assumed to be proportional to ‘r in many cases and some of the possible interpretations are reviewed by Epstein (1988). For asymptotic fouling, Equation [2.19] gives (dRfldt)  =  0  =  ødep  Prem  -  kciep Cp W  =  -  Rf*  tw/14JD  [2.2 1]  and thus Rf*  =  kdep  [2.22]  Cp W NJD Itw  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 different behaviour to that predicted by Kern and Seaton’ s model; the initial fouling rate decreased asW increased and Rf* did not obey [2.22]. They proposed a model featuring mass transfer of particulates to the surface followed by adhesion to form deposit, and removal;  where  [1I2’] dxfldt  =  1 S a  J  =  km  dep  Jdep  2 a  -  (Cp,bulk  -  [2.23]  tw Xf  Cp,surf  [ 0])  [2.24]  and S, the ‘sticking probability’ was used as defined by Parkins (1961) as  deposition rate particle flux  =  S  oc  exp  [2.25]  Watkinson and Epstein’s Arrhenius type term represented the formation of an adhesive bond, while the shear stress dependence correlates the removal of material by shear force processes. Equation [2.23] was successfully fitted to their initial rate data but not the asymptotic fouling resistance. The ‘sticking probability’ concept describes fouling as a competition between adhesion and removal phenomena in the process of deposit formation rather than removal of foulant material once it has been incorporated in the deposit surface. Asymptotic fouling is rarely reported in cases of chemical reaction fouling, so removal would seem to be less significant than in cases of particulate fouling. Removal processes in particulate fouling have been investigated by Yung et al. (1989). The sticking probability has also been interpreted by Epstein (1981) and Paterson and Fryer (1988) as a competition between  35  2. Literature Review  characteristic residence times, between that necessary for bond formation and that for the periodic renewal of fluid at the surface. Cleaver and Yates (1975,1976) obtained a similar expression 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 [100 ] disrupting the laminar sublayer and thus renew the fluid at the surface. u’ is the 2 v/u* friction velocity, defined as  =  Um/(f/2).  Vatistas (1989) considered the process in terms  of removal probabilities, defining a dimensionless adhesion time based on removal and adhesion processes in particulate fouling. Paterson and Fryer (1988) similarly described S as proportional to the time for which the surface and foulant precursor are exposed to each other, when adhesive processes operate. Most work on sticking probability has involved particulate fouling rather than chemical reaction fouling. Turner (1993) reviewed the published data and models for particulate sticking probabilities. He suggested that the commonly used Arrhenius term in S for particulates arose from the activation energy of the double layer potential rather than from some adhesive chemical reaction stage. Turner also discussed the forces and processes affecting particle deposition at a solid surface.  Transport and Reaction Models  2.4.2  Neither Kern and Seaton nor Watkinson and Epstein incorporated chemical reaction effects to describe the common case where foulant precursors are generated in situ. Crittenden et al. (1987b) reviewed chemical reaction fouling and described a general fouling model which incorporated most of the cases reported in the literature. The flux of fouling precursor to the reaction zone is given by =  km (C,bu1k  -  Cp,suri)  [2.261  The flux is balanced by the consumption of precursor in the deposit generating reaction, rxn  where  =  kr Cp,surf ‘  kr  =  [2.27]  k° exp ( Eact IR Tsurf) -  36  [2.281  2. Literature Review  The fouling rate is obtained by equating [2.19], [2.26] and [2.27], then rearranging to give dRf/dt  =  l/ d (Xf)/dt  =  {l/pf Xf  }[ Cp,bullc /{1/km  +  1/kr Cp,surt’’}1  [2.291  Equation [2.33] shows that surface reaction controls the fouling rate at high mass transfer rates (large k) and conversely mass transfer is the controlling step at high surface temperatures. The latter case corresponds to conditions in pyrolysis reactors modelled by Femandez-Baujin and Solomon (1976), where fouling increases with mass flow rate. Froment (1981) describes successful simulations of pyrolysis reactors using simple deposition models and a detailed kinetic model to predict C,b lk for each component. 14 Huntrods et al. (1990) similarly simulated a propane pyrolysis condenser where cracked products 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 mixed reactionlmass transfer control. The comprehensive model of Crittenden et al (1987b) was used to explain the results from an experimental study of polymerisation fouling from solutions of 1 v/v% styrene in kerosene. The model can be summarised as dRf/dt  The deposition term, back,  ødep  =  dep  rem 4  -  P ageing  [2.30]  was based on [2.29] but included a ‘back convection’ flux,  of foulant generated in the reaction zone transported back into the bulk solution rather  than forming a deposit; ødep  where  =  (1Jpj) (J  km (Cp,bulk  Jm back T  -  =  t (Cf,f k  [2.31]  iback) -  -  krCp,surt deposition  Cp,surf)  =  Cf,b1k  [ 0])  [2.32]  removal offoulant [2.33]  This gives an effective interfacial foulant concentration, Cf surf. which was found to be strongly dependent on temperature and was interpreted as the solubility limit of the foulant in solution. Direct calculation of Cf surf involved too many parameters and in its absence the model overpredicted fouling rates at higher temperatures and did not show the complex velocity effects observed. 37  2. Literature Review  The rem term took the form suggested by Kern and Seaton and the  çbageing  term  incorporated a deposit ageing correction first suggested by Atkins (1962). The deposit was modelled as two solid phases of different thermal conductivities. Recently generated foulant was observed as an amorphous, tarry material with lower thermal conductivity than the coke-like material found after extended exposure to surface temperatures.  ageing 4  thus  includes the reduction in Rf associated with deposit ageing. Epstein (1993a,b) included a sticking factor in the reaction term in equation [2.291 to explain the velocity effects rather than invoke foulant back convection; the reaction term was written as the product of an Arrhenius term and the fluid residence time near the surface. This yielded a reaction plus attachment term of the form kr  =  34 v exp k  2 [- EactfR Tsuff] / u*  [2.34]  This term introduced a velocity dependence into the reaction term in Equation [2.29] which predicted the observed extrema in fouling rates at given surface temperatures as the flow rate increased. Epstein (1993a) initially assumed that the transport term, km. could be estimated using the simplified equation of Metzner and Friend (1958) in calculating . 34 k km  =  u*I1l.8 Sc 213  [2.35]  The calculated values of k 34 obtained from the experimental data increased with temperature, which Epstein suggested could be due to thermophoresis, the diffusion of particles down a temperature gradient, or another force resisting attachment which increases with 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. The model predictions are shown along with the data in Figure 2.4 and show excellent agreement with the data with an average absolute deviation of 14.2%, which was within the reported experimental error. The denominator in Equation [2.35] was calculated to be 502.3 rather than 11.8; Epstein attributed this to the non-isothermal conditions present, the  38  2. Literature Review  Figure 2.4  Comparison of the Experimental Fouling Data of Crittenden et al. from the Polymerisation Fouling of Styrene in Kerosene and the Predictions from Epstein’s Fouling Model (1993b) (preprinted with permission)  2  -)  2 ‘F’  100  200  400  300  500  G/(kg /m s) 2 39  600  700  2. Literature Review  relatively low Reynolds numbers involved and the fact that  was based on fluid  properties evaluated at the surface rather than the bulk fluid temperature. Oufer (1990) formulated a transport and reaction model to describe fouling from model solutions of styrene in heptane under boiling conditions. Back convection of the polystyrene product was considered to be negligible. The polymerisation of styrene was found to be second order with respect to styrene from independent kinetic studies so Equation [2.29] was solved by setting Equation [2.26] equal to [2.271. The model did not give good agreement with the observed initial fouling rate, which was attributed to errors in estimating 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 before attempting 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 and reaction model similar to Equation [2.29] with n = 1.5 but found that there were too many unknown parameters involved to make valid conclusions. This work was extended by Panchal and Watkinson (1993) to compare three fouling models with fouling data obtained from annular and tubular fouling monitors, operating under identical hydrodynamic and chemical conditions. These devices operated in batch reactor mode so that bulk concentrations of reactant changed over time. The fouling models used an indene autoxidation model similar to that described by Norton and Drayer (1968) to simulate the bulk and surface reactions; Reactants A 2 , 1 (soluble)  ‘  Precursor B (sparingly soluble)  Foulant C (insoluble)  [2.36]  The foulant precursor, B, was assumed to consist of 2-6 polyperoxide units and the insoluble foulant, C, was described as a 16-polyperoxide unit material. The first reaction used Russell’s kinetics,  40  2. Literature Review  RA1,A2B  =  37 CAl k  1.5  0.5  [2.37]  while the formation of foulant was taken to be first order in foulant precursor, RB_,C  =  38 CB k  [2.38]  where k 38 was regressed from fouling data. This autoxidation model was used to introduce the time dependent bulk concentrations into the respective fouling models shown in Figure 2.5. The models differ in the foulant generation step. Case 1 describes the scenario where precursor generation occurs primarily in the bulk solution; precursor formation near the hot surface is ignored so that bulk reaction control applies. Case lb corresponds to mass transfer control, where foulant generated in the bulk is transported to the wall where it is assumed to adhere completely. Case la involves mass transfer of precursor from the bulk to react at the wall to form foulant. Any foulant found at the wall was assumed to form deposit, ignoring any sticking probability or attachment factor arguments. Case lb was rejected as it did not predict the increase in deposition with surface temperature found in the experimental data. The Case la model followed the trends observed in the data for variations in surface and bulk temperature, though less successfully than Case 2. The second model described foulant generation as occurring primarily in the thermal boundary layer. Back diffusion of precursor as descibed by Crittenden et al. (1987b) was included and the equations solved numerically. This model was found to give the best agreement with the experimental data. The Case 3 model is similar to that described by Crittenden et al. (1987b) but seemed to ignore reaction in the bulk, which was an intrinsic part of the Case 1 model. This model did not predict the effects of bulk temperature, unlike Cases la and 2. No flow velocity effects were reported and the differences in fouling data from the two fouling monitors were not discussed. Each model included an unknown parameter which was calculated by fitting the model predictions to the data at a selected point. This is a common procedure in fouling modelling and the values obtained were reasonable. The  41  2. Literature Review  Panchal and Watkinson Autoxidation Fouling Models  Figure 2.5  Case 1: Precursor Generation in Bulk Solution la  A  lb  A  molecular transfer >  B  on wall  >  B  >  C sticks on wall  --->  C  particulate trtEmSpOrt -->  B  -->  C  Case 2: Precursor Generation in Boundary Layer A  >  A  >  B  >  C  molecular transfer  B  B  <  C sticks on wall  Case 3 : Precursor Generation on Wall Surface molecular  transfer  A  >  B  <  bulk solution  thermal boundary layer >  >  <  Reaction Mechanism Reactant A  —-->  Precursor B  ---->  Foulant C  From Panchal and Watkinson (1993)  42  A  --->  B  on wall  B  --->  C  on wall  2. Literature Review  success of the boundary layer formulation concurs with the reaction engineering approach of Paterson and Fryer (1988).  2.4.3  Reaction Engineering  Paterson and Fryer (1988) adopted a ‘reaction engineering’ analysis to explain the form 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 enhanced temperatures give rise to the maximum foulant generation rates. In the absence of significant bulk concentrations of foulant precursor, the conditions in this region of the fluid boundary layer dominate the fouling process. The zone is modelled as a differential chemical 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 based on residence time arguments dRf/dt  a Ar exp  dRf/dt a A 0 exp  . IR T] { 1I’/t} As exp [- Es IR Tsurf] { 1N’t} [2.40] 1 [- E  [- Ef/R Tff] { l/-r}  [2.41]  Paterson and Fryer correlated their fouling data in the form of a fouling Biot number, Bif Rf U 0 and thus found the initial rate to vary as i/urn. This approach is a simplification of ,  the surface/fluid processes involved but gives some insight into the relationship between the primary factors involved. Fryer et at. (1990) also discussed the result of large concentrations of foulant precursor in the bulk fluid, when mass transfer of material from the bulk swamps the generation of foulant in the reaction zone. The reaction engineering analysis is then invalid and fouling is best described as a generation-agglomeration-adhesion process.  43  2. Literature Review  Mechanistic modeling of chemical reaction fouling in autoxidative conditions is less developed than that of pyrolysis reactors, for example, due to (a)  The range and complexity of liquid phase free radical reactions, which involve complex kinetics;  (b)  The difficulty in performing reliable autoxidation fouling experiments;  (c)  The operating conditions involved feature mixed reactionlconvection control of the fouling 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 as well as its thermal performance in order that reliable mechanistic models can be developed.  2.5  Objective Autoxidation has been identified as a major source of fouling deposit in heating  oxygenated hydrocarbon streams from ambient temperatures to 400°C, above which condensation and pyrolysis reactions become dominant. Fuel stability studies have identified the effect of chemical structure, conditions and initiating species on the formation of polymeric peroxide gums in fuel storage. Alkenes, particularly conjugated alkenes, tend to form such gums more readily. Asomaning and Watkinson (1992) proved the hypothesis drawn from plant observations that autoxidation fouling is linked to the formation of such gums and also found that fouling was determined by the alkene structure, as described in fuel stability studies. Autoxidation fouling is still poorly understood and no mechanistic model exists to explain the interaction of temperature, hydrodynamics and chemical reaction in such systems. Studies have identified the significance of dissolved oxygen, temperature and feedstock composition but reported velocity effects contradict, partly because of the diversity of experimental methods. Crittenden et al. (1987b) reported a fouling model for  44  2. Literature Review  polymerisation fouling using a well defined chemical system and this approach was used to investigate autoxidation fouling.  The objective of this study is therefore to investigate the mechanisms involved in autoxidation fouling and the links between the series of reaction and deposition steps which generate foulant. The process is poorly understood and requires experimental verification of the mechanisms involved. This knowledge could then be used to formulate and verify a mechanistic model for autoxidation fouling at moderate temperatures in turbulent flow, single phase heat transfer to hydrocarbon liquids. Intermediate objectives were 1.  Identify a suitable mOdel solution for studies of autoxidative fouling;  2.  Determine the model solution reaction mechanism and significant kinetic parameters;  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 reaction complexity so that the formation of deposit and deposit structure can be compared with the alkene reaction kinetics and products. Model solutions also facilitate the development of mathematical models of fouling as physical and chemical parameters can be estimated reliably to test the reliability of such models.  45  3. Experimental Methods and Materials  3.  Experimental Materials and Methods This chapter describes the experimental materials, apparatus and procedures used in  the current study of autoxidation fouling. The model solution used in the fouling studies was selected after a search of potential alkenes and solvents described in Sections 4.1 and 5.1. The experimental apparatus and chemical analyses were developed during the  experimental program on the basis of operating experience. Experiments were numbered according to type as described in Table A. 1.1.  3.1  Materials and Physical Properties  3.1.1  Model Solutions  The model solutions used in the fouling and kinetic studies of indene autoxidation consisted of an alkene which undergoes autoxidation readily under the experimental conditions and a solvent which should remain relatively inert. Candidates for solvent liquid were selected on the basis of their chemical inertness to autoxidation, chemical nature (aromatic, polar, aliphatic, miscibility with the alkene), safety in handling, boiling point (to maintain 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.  ), based on the experience of 8 H 9 The alkene used was primarily indene (C Asomaning and Watkinson (1992). This was obtained as a technical grade liquid (92 wt%+) from Aldrich Ltd. and was stored frozen until required. No attempt was made to purify the indene due to the large quantities involved. Gas Chromatography  -  Mass  Spectrometry (GCMS) analysis indicated that the impurities present were those expected from indene manufacture. Peroxide analysis showed that relatively little autoxidation had occurred during production and storage. Different batches of indene did cause variations in  46  3. Experimental Methods and Materials  the rate of autoxidation and so series of experiments were performed using indene from the same 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, 10 the dimer of the cyclic diene pentadiene. C , 8 12 C 3 H , 6 and dicyclopentadiene, (DCP), H Hexadec-1-ene was obtained as a technical grade liquid from Aldrich Ltd. (94 wt%) and used as obtained. DCP was also obtained as a technical grade liquid from Aldrich Ltd. (95 wt%) but contained an oxidation inhibitor, p-t-butylcatechol, which had to be removed by distillation. This was performed by G. Zhang using a Penn State column and the distillate was stored frozen under nitrogen. DCP was used in two fouling experiments to study the interaction of two active alkenes in a model solution. DCP was not used subsequently due to problems in the peroxide analysis and its bad odour.  Table 3.2 summarises the properties of the various solvents used in the model solutions. Trichlorobenzene (Aldrich Ltd.), toluene (BDH) and n-octane (Fisher) were obtained as high purity liquids. The kerosene used was a commercial kerosene supplied by Imperial Oil Ltd. and contained an unknown oxidation inhibitor. 13C nuclear magnetic resonance (n.m.r.) spectroscopy showed that the kerosene was mainly paraffinic in nature but also contained some alkene and aromatic compounds. Paraflex HT1OTM is a process oil marketed by Petro-Canada Ltd. It consists of a lubricating oil basestock hydrotreated twice to saturate all components to their alkane equivalents, giving a very stable, inert liquid. 13C n.m.r. did not show any unsaturated species.  Two free radical initiators were used to study the effects of chemical initiation on the autoxidation of indene. Benzoyl peroxide (bP) (Aldrich Ltd.) and 2-2’-azo-bis-2methylpropionitrile (ABN) (Pfaltz and Bauer) are both thermally activated initiators which decompose on heating to from free radical fragments. These materials were obtained as 47  3. Experimental Methods and Materials  Table 3.1  Alkenes used in Model Solutions  Compound  Indene  Hexadec- 1 -ene  Dicyclopentadiene (DCP)  Formula  8 H 9 C  12 C 3 H 6  2 ) H 5 (C  Formula Mass  116.1  224.3  130.21  aromatic alkene  aliphatic alkene  diene dirner  4.0 °C  -1.0°C  Activity Melting Point  -  2.0 °C  Density (20°C)  996 kg/rn 3  783 kg/rn 3  986 kg/rn 3  Boiling Point  181.6 °C  274°C  170.0 °C  Structure  Table 3.3  Compound  Physical Properties of Initiators and Antioxidants  Benzoyl peroxide (bP)  Azobis-2-methyl propionitrile (ABN)  2,6-t-butyl-4methyiphenol (BMP)  Tributylamine (TBA)  CO) 5 H 6 (C 0 2  C(CN)N] ) 3 [(CH 2  [(CR Cl ) 3 . 3 . CR 2 OH 2 H 6 C  (CH [CH N j 3 ) 2  242.2  164.0  220.36  185.4  Initiator  Initiator  Antioxidant  Antioxidant  Melting Point  104-106°C  102-103°C  69-70°C  -70°C  Density (20°C)  solid  solid  solid  778 kg/rn 3  -  -  -  Formula  Formula Mass Activity  Boiling Point Structure -  o  48  216°C  3. Experimental Methods and Materials  Table 3.2  Properties of Solvents used in Model Solutions  n-octane  toluene  Paraflex HT 10  kerosene  trichiorobenzene  tetralin  Supplier  Fisher  BDH  Petro Canada Ltd  Imperial Oil Co.  Aldrich Ltd.  Aldrich Ltd.  Density (15.6°C)  705 kg/rn 3  873.4 kg/rn 3  3 855.4 kg/rn  808.5 kg/rn 3  1459 kg/rn 3  973 kg/rn 3  Solvent Property  Colour [D1500j Flash Point [D92] Viscosity (15.6°F) Pour Point [D97] Aromatics [PCM 435] Prandtl Number (100°C) (80°C) Chemical Activity Formula Mass GC Distillation [D28871 lwt% 10 30 50 90 99 Boiling Pt.  +  30  ÷ 30  13.3 °C  4.4°C  166°C  45°C  105.58 °C  71.13 °C  0.57 rnPa.s  0.62 mPa.s  23.9 mPa.s  2.26 mPa.s  3.20 mPa.s  2.3 rnPa.s  -21°C 0 wt%  100 wt%  0.06 wt%  1-2 wt%  100 wt%  100 wt%  5.97 6.33  4.57 4.98  37.3 53  12.4 14.1  10.1 12.3  10.5 12.2  aliphatic  aromatic  saturated compounds  ahphatic, some unsaturates  chlorinated aromatic  aromatic  114.2  92.1  145.5  132  213.0°C  207.6°C  310  87°C 314°C 333 °C 347°C 369°C 464°C 125.6°C  110.6°C  Structure  /VV  References in parentheses refer to ASTM or API testing methods  49  190°C  230°C 250°C  3. Experimental Methods and Materials  pure solids from the manufacturers and stored frozen after opening. No attempt was made to purify the initiators. The effect of antioxidants on indene autoxidation was studied using di-t-butyl-4-methylphenol (BMP) and tri-t-butylamine (TBA), obtained as 99 wt% pure compounds from Aldrich Ltd. Table 3.3 summarises the relevant properties of these materials. All but TBA exist as solids at room temperature and were added to the solution by dissolving the powder in the indene before this was added to the reactor or holding tank to start an experiment.  3.1.2  Physical Properties  The physical properties of trichlorobenzene, indene, tetralin, n-octane and toluene were readily available from the literature (Coulson and Richardson 1983; Daubert and Danner 1989), whereas the Paraflex and kerosene were obtained in 205L bulk shipments with only general property data. The boiling point, viscosity and density were measured in order to obtain accurate data for use in the general correlations for specific heat capacity and thermal conductivity. The kerosene data was collected and analysed by G. Zhang (1990). The Paraflex analysis was performed in conjunction with R. Lai.  Kerosene Properties  Kerosene viscosity was measured using a Fenske viscometer immersed in a constant temperature bath; kerosene density was measured using a hydrometer and the data regressed to give the following functions of temperature. The thermal conductivity and heat capacity were then estimated by Zhang using an API method. The data were collected over the range 45-100°C. Density  ) 3 p (kg/rn  Viscosity  i.  (Pa.s)  =  =  820.15-0.7 T(°C) in {1I(T(°C) 50  +  44.737)  [3.1] +  8.5142} 2.1419 -  [3.2]  3. Experimental Methods and Materials  Heat Capacity  Cp (J/kg.K)  Thermal Conductivity ?. (W/m.K)  =  =  1874.7 + 3.7559 T(°C)  [3.3]  0.13 11  [3.41  -  1.4123 x 4 10T (°C)  Figures 3.1 and 3.2 show the kinematic viscosities and Prandtl numbers of the solvents used in the model solutions along with that of indene, the alkene selected for fouling studies. The difference in physical properties between indene and Paraflex is much greater than that between indene and the other solvents, so the density and viscosity of solutions of indene in Paraflex were also measured and compared with existing correlations.  Paraflex Properties  Paraflex viscosity was measured using a Haake VT500 rotary viscometer equipped with an NV coaxial cylinder sensor system for liquids of low viscosity. The NV system is mounted in a controlled temperature bath and the unit is interfaced with a computer for data analysis. The results were fitted to a number of viscosity-temperature expressions and the best correlation was obtained using the form derivd by Puttagunta et al. (1992) for  conventional petroleum liquids; Kinematic Viscosity  =  logy 0 (m / s) (v 2  [1 where KVb  =  1.95; KVc  =  -0.8696; KVs  + =  KVb (T(°C) 37.78)!310.93]s -  +  KVc [3.5]  2.1642. The densities of both Paraflex and  solutions of indene in Paraflex were measured using specific gravity bottles. Petro Canada recommended the use of the Fisher Dens fly/Temperature Tables to calculate Paraflex density at other temperatures. The values obtained for the range 15-100°C were then fitted to the expression Density  ) 3 p (kg/rn  =  864.04  -  0.588 T(°C)  [3.6]  Petro Canada also supplied expressions from a Shell Thermia Oils Technical Bulletin for the 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.  51  3. Experimental Methods and Materials  Kinematic Viscosities of Solvents used in Model Solutions  Figure 3.1  7 icr  40  20  0  60  100  80  120  Temperature (°C) Prandtl Number of Solvents used in Model Solutions  Figure 3.2  I  •  I  I  I  ——  I  indene  2 io  101  -  100  0  20  40  60  Temperature (°C)  52  80  100  120  3. Experimental Methods and Materials  (W/m.K)  Thermal Conductivity  Heat Capacity  Cp (J/kg.K)  =  0.13711  =  1821.6  -  +  7.40524x iO T(°C)  [3.7]  3.6676T(°C)  [3.8]  The heat capacity calculated at 20°C compared favourably with the predictions of the Chueh and Swanson method (see Coulson et al. 1983) and the results of some simple calorimetry experiments using hexane and warm Paraflex in a vacuum flask. The mean molecular mass of the Paraflex oil was estimated at 310, which was confirmed by the viscosity estimation procedure described in ASTM D-2502-87. The densities of solutions of 0.0-1.0 mol/L indene in Paraflex were measured at 20°C and fitted a linear function of indene concentration; Density  ) 3 p (kg/rn  The viscosities of 0.0  -  152.28  =  +  14.96 [indene] (mol/L)  [3.91  1.0 mol/L solutions of indene in Paraflex were measured at 20°C  using 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. The values of Paraflex and indene viscosity reported at 20°C agreed within the limits of experimental error. Figure 3.3 shows the data and the predictions of the viscosity mixing rules of Kern (1950), Chhabra (1992) and Souders (see Coulson et al., 1983). Kern  =  /pt w 1  =  ) 2 1 {iI(jt  +  [3.10]  Iu w 2 }O.5 + (1)2/(1422)  Chhabra  1/Jtmix  Souders  o (10 ptmjx)) 1 logio(log  =  Prnix Is,mix/Mmix  [3.11] X  i0  2.9  [3.12]  where w 1 is the mass fraction of component i, ct is the mole fraction of component i;  ‘S,i?x  and M 11 the mole averaged Souders viscosity constant and molecular mass respectively. None of the mixing rules exhibits the observed viscosity behaviour so the viscosity of a solution was estimated using the viscosity of Paraflex and a ratio obtained from the experimental data (e.g. 0.79 for 0.405 M indene in Paraflex). The change in viscosity was evident in the TFU fouling experiments as the heat transfer coefficients reported for solutions of indene in Paraflex were larger than for Paraflex alone at the same mass flow rate. The viscosity dependence in Re (—0.8) is larger than that in Pr 53  (— -0.33), so a larger  3. Experimental Methods and Materials  Viscosity of Mixtures of Indene in Paraflex at 20°C; Comparison of Correlation Predictions and Experimental Data  Figure 3.3  30  25  20  0.0  0.2  0.4  0.6  [indene] (mol/L)  54  0.8  1 .0  3. Experimental Methods and Materials  value of Nu is expected. The difference in thermal conductivities was negligible, so was ignored.  Concentration of Dissolved Gases  Autoxidation does not occur in the absence of oxygen, the solubility of which in organic liquids is thus extremely important. A direct method for determining dissolved oxygen concentration was not available during the current work; the on-line gas chromatography method described by Hazlett et at. (1977) was beyond the scope of this study. The saturated concentration of a dissolved gas can be calculated using ASTM methods 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 shows  the values calculated for the solubility of oxygen in kerosene and Paraflex at 377 kPa air saturation. The values for kerosene vary markedly and neither method shows the expected decrease in oxygen solubility with liquid temperature. It is evident however, that oxygen ,  is sparingly soluble in the solvents used in the current study. The ASTM D2779-86 method was used in subsequent calculations; this requires an estimate of the Ostwald coefficient for oxygen, L , obtained from a plot. The graphical data were fitted to Equation [3.13] with a 0 coefficient of variation (R ) 2 =  0.15876  +  0.996.  5.9342x10 T 4 (°C)  -  2 T 7 7.3367x10  [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]  Solvent  80°C  100°C  80°C  100°C  Paraflex  214 310  215  186  188  314 ASTM D2779-86  213  213  kerosene Method  55  ASTM D3827-86  3. Experimental Methods and Materials  3.2  Portable Fouling Research Unit (PFRU) Apparatus  This apparatus was constructed by Fetissoff (1982) and modified by subsequent workers. The device features two fouling monitors; a Hot Wire Probe (HWP) designed to study 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 the initial fouling experiments and in the studies of fouling in the presence of antioxidants. The HWP 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 contact with the liquid were constructed from stainless steel as this was not thought to be a significant source of metal ions. Seals were made from Teflon where possible as this offered greatest resistance to the fouling solutions and cleaning liquids. Rubber-based seals and 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 through a flow control valve, a set of orifice plates, the fouling probe and a rotameter before being returned 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 bucket and a stopwatch. The pressure drop was recorded using a mercury manometer. The rotameter was later removed in order to increase the maximum flow rate when investigating flow velocity effects. Figure 3.5 shows the orifice plate calibration for Paraflex at 80°C and 100°C using the expression W  =  CorAt {2 p APorI[(dt,ildor) 4  -  1j}0.5  where W is the mass flow rate, A is the tube cross sectional area  [3.14]  /4), APor is the 2 (= 3tdt  pressure drop across the orifice and Cor the office discharge coefficient. Similar calibrations were performed for tetralin and kerosene.  56  3. Experimental Methods and Materials  Figure 3.4 Air or  Schematic Diagram of PFRU Apparatus  Cooling Water  Orifice Meters Pump  57  3. Experimental Methods and Materials  Figure 3.5  PFRU Orifice Plate Calibration for Paraflex  0.40 0.35 0.30 0.25 0.20 114  0.15 0.10 0.05 0.00  0  50  100  150 (a)  58  200  250  300  3. Experimental Methods and Materials  Liquid samples were taken downstream of the fouling probe using a sequence of ball valves arranged so that sampling did not disturb the flow. Samples had been initially taken from the holding tank but this did disturb the flow. Air or gas mixtures were supplied to the airspace in the holding tank, providing a gas blanket above the solution. Mixing in the holding tank was provided by recirculation of the liquid. Relief valves were fitted to avoid overpressure conditions at the pump and in the tank. The system was insulated throughout and cooling was supplied by mains water passing through a coil in the holding tank, controlled by a needle valve. Heating tapes on the tank exterior and lagging were used to bring the liquid to its operating temperature and were usually left on during an experiment. Temperatures were measured using ‘J’ type thermocouples located at various points in the system. Pressure gauges and mercury manometers were used to monitor absolute and differential pressures respectively. Temperatures and pressures were displayed on the instrument panel. Figure 3.6 shows the design of the PFRU probe. Liquid flows up the annulus between the metal core and the outside wall. An electrically heated section 102mm in length is located 20 equivalent diameters downstream of the probe entrance. The outer wall of the annulus is constructed of glass at this point to allow visual inspection of the heated surface during the experiments. The PFRU probe was originally supplied by Heat Transfer Research Inc. (HTRI, College Station, TX) and had developed an unrepairable bend which prevented 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 and sheathed with a stainless steel tube. Imbedded in the sheath are four ‘J’ type thermocouples which measure the temperature at a distance x below the surface; the surface temperature is then calculated using calibrations supplied by 1-ITRI T surf  =  T  measured  -  (Xs/?net)  59  q  [3.15]  3. Experimental Methods and Materials  Figure 3.6  Schematic Diagram of PFRU Fouling Probe  to Digitrend Display andDoric235Datalogger  to Variac  150 mm  A 78mm I  .102 mm I  294 mm  10.7 mm  *  -  -  25.4 mm  t  PLOW  60  3. Experimental Methods and Materials  The thermocouples are located 80 mm from the start of the heated section. A high surface temperature trip was installed to prevent the probe burning itself out at temperatures greater than 350°C. The heater is powered from a 110 Vac supply via a variac; the voltage is measured and used to calculate the power and thus the heat flux at the surface. The voltage was rectified 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 order polynomial ) 2 q (kW/m  =  135.67- 1129.2 v (volts)  +  4 2 4914.6 v 3 + 2438 v 4143.7 v -  [3.16]  where v is the datalogger voltage reading. The PFRU heating voltage and system temperatures were recorded remotely on paper tape using a Done 235 datalogger. These data were transferred to a Microsoft ExcelTM spreadsheet for further processing and presentation. The variac setting is kept constant during an experiment so that the device operates at a constant heat flux. If the deposit does not modify the solid surface/liquid heat transfer coefficient then this condition also gives a constant depositlliquid interface temperature. Any increase in metal surface temperature is then due to the thermal resistance of the deposit. The heat transfer coefficient was calculated from; q  =  [3.17]  U(t) (Tsurf Tb) -  The fouling resistance could then be calculated using the expression  Rf(t)  =  1/U(t)  -  0 i/u  [3.18]  The clean heat transfer coefficient, U , was based on the average value calculated at the 0 start of the experiment.  61  3.  Experimental Methods and Materials  PFRU Heat Transfer  3.2.1  The heat transfer characteristics of the PFRU were studied using clean solvents under experimental conditions. One of the thermocouples was inoperative and the skewed alignment of the annulus core produced non-uniform surface temperatures. The heat transfer coefficient, U, in Equation [3.171 was calculated using a mean surface temperature and expressed as a Nusselt number, Nu =  da,o  -  =  11 is the hydrodynamic diameter, Udh/A, where d  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 predictions from the following heat transfer correlations. Monrad and Pelton (1942)  Nu  =  08 Pr 0.020 Re  53 (da,o/da,i)°  [3.19)  Wiegand (1945)  Nu  =  0.023 Re 08 Pr 0.4 (dao/dai)° 45  [3.20]  Gnielinski (1986)  Nu  =  1000) Pr (f18) (Re 213 1] 1 + 12.7 ./(f/8) [Pr  [3.21]  -  -  where the Moody friction factor,f is given by Knudsen’s expression for annuli (1958); f  =  025 0.304 Re-  [3.221  The Sieder-Tate (1936) surface temperature correction,  4 (pc/ji)°’  was also used in  calculating the final values. The experimental results in Table 3.5 showed some agreement  with the annuli correlations. Asomaning (1990) reported similar findings in the PFRU using kerosene alone. Systematic error sources lie in the reliability of the physical property data and the fouling probe itself. The geometry of the PFRU probe is such that thermal entry length effects will still be significant at the thermocouple location, which corresponds  to x/dj  =  5.6. This is particularly true for Paraflex, where the high solvent viscosity results  in a low Reynolds number.  Kay and Lambourn (see Fetissoff, 1982) calculated the  thermal entry length effects for constant heat flux and fully developed flow between parallel plates 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 to  62  3. Experimental Methods and Materials  Table 3.5  Comparison of Nusselt Numbers in PFRU Heat Transfer kerosene  tetralin  trichioro benzene  Paraflex  Paraflex  Bulk Temperature  80°C  80°C  80°C  80°C  100°C  Pr  14.1  12.2  12.3  53.0  37.3  Re  10690  10720  12090  2550  5145  Nu(PFRU)  163  155  172  97-106  137  Nu(Wiegand)  181  144  166  109  166  Nu (Monrad-Pelton)  142  134  154  79  122  Nu (Gnielinski)  119  113  132  41  90  Solvent  63  3. Experimental Methods aiwl Materials  reduce this figure. Heat transfer data from the initial fouling experiments showed that Nu ce  ’ which does not agree with the correlations. The PFRU was thus concluded to be 065 Re  unsuitable 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 with heat flux was studied in order to determine the onset of subcooled boiling. The maximum operating surface temperature for kerosene at 378 kPabs and Re  =  10690 was found to be  around 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). The Nusselt number for Paraflex at Re  =  2550 and Tblk  =  80°C increased as q ; this 3 ’ 0  corresponds to a surface temperature correction such as that of Sieder and Tate.  3.2.2  PFRU Experimental Procedure  The annular probe and the apparatus were cleaned thoroughly before each experiment. The tank was filled with the charge of solvent and run for several hours at the required bulk temperature and air overpressure in order to ensure that the liquid was saturated with air and the system warmed up. Compressed air was usually drawn from the building supply although this often fluctuated due to demand from other users. The datalogger was started before adding any alkene as it was located in an adjoining room. To start a fouling experiment, the system was quickly depressurised and momentarily shut down so that the alkene could be added and rinsed in with some of the solvent charge. The pump was turned on, air pressure restored and the filling line converted to the sampling line while the unit returned to the required operating conditions. The PFRU heat flux and flow rate were set and the first sample was then taken. This starting procedure rarely took more than twenty minutes. Attention was paid to adjust the cooling water flow rate to maintain a steady bulk liquid temperature. 64  3. Experimental Methods and Materials  Solution samples were taken as required during the experiment and visible changes in solution nature and heated surface appearance were recorded. The following aliquots were quickly pipetted into suitable sample containers 2 mL (indene concentration), 10 mL -  (gum assay), 2 x 2OmL (Peroxide Number). The shut-down procedure involved depressurising the system whilst gradually reducing the PFRU power level in order to avoid disrupting any deposit by a thermal shock. Once the pressure had reduced to atmospheric and the PFRU power to zero, the pump was turned off and the probe quickly removed. 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 a small amount of acetone, primarily to cool the apparatus down. The probe was replaced by a dummy probe, the coolant flow increased and the rinse circulated until the apparatus was cool enough to allow safe loading of acetone. Two acetone rinses in succession were used to remove any residual indene polyperoxide material; methyl-iso-butyl-ketone (MIBK) and tetrohydrofuran (THF) were also used to clean the system where the unit showed signs of contamination. 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 volumes were coordinated in such a way as to recycle as much solvent as possible without affecting the 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 was measured using vernier calipers but this was complicated by the softness or brittle nature of the deposit. Photographs and diapositives of the magnified deposit surface were obtained using either a 35 mm camera or a Polaroid camera fitted to a Nikon HFX-II optical microscope system. The magnification was calibrated using an etched vernier scale. The deposit was often scraped off the probe and kept for further analysis, taking care not to damage the metal surface. The probe was then cleaned using acetone and household cleaner before the next experiment. 65  3. Experimental Methods and Materials  3.3  Stirred Cell Reactor (SCR)  The autoxidation of indene in solution was studied in a stirred batch reactor constructed by G. Zhang. Figure 3.7 is a schematic drawing of the apparatus. The 3L cylindrical reactor and all fittings in contact with the liquid were constructed from type 316 stainless steel; Teflon seals were used where possible. Air was admitted via a gas line which 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 former corresponds to the TFU arrangement. Gas flows were monitored using rotameters and calibrated using graduated cylinders immersed in a water bath. Air was supplied from the building supply or from gas cylinders; a one way stop valve was fitted to the building supply line to prevent suckback when the feed pressure cycled. The system pressure was read from an Omega 0-100 psig pressure guage. A pressure relief valve was used to avoid overpressure as Alexander (1990) describes various explosions caused by runaway autoxidations 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 controller connected to the reactor’s ‘K’ type thermocouple. The oil was circulated by a GKH 1/40 h.p. electric motor, controlled by a GKH GT-21 speed controller. The reaction liquor was agitated by an impeller driven by a top-mounted electric motor (GKH, 1/8 h.p.) through mechanical 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 needle valve. The sample volume was collected in a 100 mL jar and the aliquots described in Section 3.2.2 pipetted into the respective containers. The sample line and pipettes tended to get blocked by insoluble orange/red gum towards the end of an experiment; no solution was found to this problem. This gum also settled out rapidly in the collection jar and so 66  3. Experimental Methods and Materials  Schematic Diagram of Stirred Cell Reactor (SCR) Apparatus  Figure 3.7  Rotameter Motor  I  Thermocouple  I  I  —  Relief valve —  to vapour trap pot  Gas Supply One Way Valve  stiffer ..\  SS  SS’.\  S\\\\  S”.  \  Heating Oil Bath - / / / / // / / // / / // / / / / / / / / / / / / / / / / 7, / / /7, /  67  3. Eperimenta1 Methods and Materials  gum analysis during this period showed considerable scatter. The pipettes were rinsed with acetone and hexane and left to dry before re-use. An experiment was started by charging the reactor with a volume of solvent and running the reactor at the required conditions for several hours. The reactor was then quickly 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 sample was taken, ensuring that the solution was fully mixed. Samples were taken at regular intervals until the experiment was completed, at which time the reactor was depressurised and the heater turned off. The reactor was removed from the oil bath and the remaining liquor 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 and cleaned with acetone to remove any residual products. Severe contamination occurred in the trichlorobenzene 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 in the design of the PFRU fouling monitor; examination of the PFRU deposit is limited and thermal entry length effects are significant along its relatively short heated section. The TFU provided a longer heated length with a known axial temperature profile, and allowed the deposit to be examined in situ as the fouled sections were designed to be removed after use. Operational experience from the PFRU proved invaluable in designing and operating the TFU.  68  3. Experimental Methods and Materials  Tube Fouling Unit Apparatus  3.4.1  The apparatus is a modified version of Watkinson’s design (1968) adapted to use removable test sections. The test section consisted of a 1.83m long 304L stainless steel tube (d, 0  9.525 mm,  9.017 mm) heated across its central section by alternating  electric current. The power to the heated section was kept constant during a run so that the device operated at constant heat flux, as in the PFRU. Figure 3.8 is a schematic diagram of the apparatus. All materials in contact with the liquid were constructed from stainless steel or Teflon. Liquid was recirculated from a 65 L holding tank through a pump, an orifice plate, the heated section, a series of coolers and a rotameter, then returned to the tank as a jet to promote mixing. Hoke globe valves were used to split the flow between the test section and a return line, which ensured good mixing when lower flow rates were used in the heated section. A mesh strainer could be switched on line as desired. The specifications and origins of the equipment are listed in Table A.1.2 in Appendix 1. The tank included filling and drain ports, a low level alarm float device and a 3 kW low heat flux heater to warm the liquid to the operating temperature before an experiment. The tank heater was controlled by an Omega CN 911 controller and was not required during an experiment. The system was pressurised by gas which was bubbled into the liquid from the base of the tank and vented through the top. Gas flows could be drawn from the building compressed air supply or a cylinder source and were controlled by rotameters on the control panel. The rotameters were calibrated as described previously. The tank and all system piping was insulated by fibreglass insulation. The piping section between the pump and orifice plate was also fitted with heating tapes connected to a variable transformer in case insulation losses were significant but the pump heat output was found to counter any such losses. Liquid flow rates were monitored using a rotameter and a set of orifice plates connected to a differential pressure transducer. The orifice plates (d, 1 69  =  12.2 7mm, d ,. 0  =  3. Experimental Methods and Materials  Schematic Diagram of Tube Fouling Unit Apparatus  Figure 3.8  PRV  primary  TAH sample line/drain  gas to vent  gas line  heating tape lines sample line/drain  PAR  -  Pressure Alarm High; LAL  -  Level Alarm Low; TAR  -  Temperature Alarm High; PRV  Relief Valve; T Thermocouple; P Pressure Transducer; AP Differential Pressure Transducer -  -  -  70  -  Pressure  3. Experimental Methods and Materials  9.525 mm, Cor  =  0.571  ) were calibrated using pure Paraflex at 100°C following the  procedure described in Section 3.2. The calibration plots for the rotameter and orifice plates are given in Figure A.1.l. The orifice plate pressure drop was recorded for data processing while the rotameter was used to monitor the flow rate and provide a visual record of the liquid. The temperature of the bulk liquid entering and leaving the heated section was monitored 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 control purposes. A series of alarms were built in to the system wiring to minimise risks and are described in Table 3.6. The pressure drop across the heated section was monitored by a differential 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 range of the differential unit. The differential pressure transducers were calibrated using a Marsh 0-15 psig test gauge. The coolers removed the heat supplied in the heated section. The primary cooler consisted of an 2.44 m long x 1 inch diameter copper tube around the 1/2 inch stainless steel tube carrying the process liquid. Mains cold water passed countercurrent to the process 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 and maintained the bulk temperature constant during all Paraflex heat transfer tests. The primary cooler capacity decreased during fouling experiments due to fouling so two auxiliary coolers were added downstream of the primary unit. These were of similar construction to the primary cooler but were shorter (0.5 m, 0.4 m) and used cooling water from a separate tap supply, so that they could be operated independently of the primary unit. Cooler fouling  proved to be a serious problem in operating the TFU at high heat fluxes and imposed limits on the maximum surface temperature which could be maintained at a given flow rate over a  71  3. Exverimentai Methods and Materials  Table 3.6  TFU Alarm Matrix  Hazani  Cause  Trip  Action  Overpressure  Blockage/Explosion  PAR: PRV  System Shut Down  Fluid Leak/Runs Dry  Leakage/Rupture  LAL  System Shut Down  Overheated Liquid  Cooilng Water Failure  TAH on cooler outlet  System Shut Down  Tube Burn Out  Flow Stopped  TAR on tube surafce  Heating Off  Thermocouple Damage  Tube too hot  TAR on tube surface  Heating Off  Power Surges  Power Failure/Restored  Reset relays  Power stays off  72  3. Experimental Methods and Materials  given heated length. Cooler fouling was not observed in the PFRU due to the proportionately 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 needle valve and a ball valve in series. Similar lines were located at the primary cooler exit and the pump. The sample line after the test section consisted of two electrically operated valves connected by a 100 mL volume of 3/8 inch piping. This safety arrangement was used as the process fluid line was beyond normal reach and ran close to the heated section. The manual 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 tube lengths were constructed of drawn T304 stainless steel (ASTM A269-80A) and were supplied by Greenville Tube Corp, Clarksville, Arkansas. The tubes had a nominal thickness of 10 thousands of an inch, but tube thickness measurements using a ball micrometer and calculations from tube masses gave values ranging between 11 and 14 thou. The tube was connected to the fouling loop by Teflon Swagelok TM fittings which ensured that the test section was electrically isolated from the rest of the apparatus. Current was supplied to the test section from a mains 208 Vac supply via a power variac and a 20819V step-down transformer. The step-down transformer was connected to the test section by pairs of #3 welding cables bolted to 10mm thick copper busbars. These busbars were isolated from the frame by Tufnol holders and were designed in two parts which screwed together to clamp the tube in place. The first busbar was located 500 mm (55 tube diameters) from the upstream fitting to ensure well developed duct flow in the heated section. The second busbar marked the end of the heated section and was positioned as dictated by the power requirement. The resistance of the stainless steel tube was underestimated at the design stage and the step-down transformer oversized for long test  73  3. Eperiinental Methods and Materials  lengths; high heat fluxes required larger currents and thus a shorter test section. Figure 3.9 shows the arrangement of the test section in schematic form. Figure A.1.2 describes the electrical network in detail. The voltage across the test section, v, was measured using an a.c. panel meter and displayed on the control panel. The current, I, was measured using a current transformer on one set of welding cables. The current transformer output was measured using an ac ammeter and this was calibrated using an Amprobe ACD-9 Current meter: I (Amps)  =  0.06733  +  142.85 (ammeter reading)  2 R  =  1.000  [3.23]  The power to the heated section was calculated assuming a power factor of unity (after Watkinson, 1968) as equal to vI, then converted to a heat flux using an estimated surface area based on a tubing thickness of 10 thou; Heat Flux  ) 2 q (kW/m  =  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 was estimated using a tube thickness of 14 thou; this increased the value of q by 1.7%. Tube thicknesses varied along each tube and the means were closer to 11 thou than 14 thou; the nominal thickness was thus considered a reasonable figure to use in the calculations.  3.4.2  Surface Temperature Measurement  Watkinson (1968) and Hopkins (1973) used similar electrically heated test sections where the tube was reused after each experiment. These workers used silver soldered and glued thermocouples respectively to measure the tube’s outer surface temperature; these methods were not appropriate for the current system of readily removable test sections. An alternative was sought which would provide accurate surface temperature measurement with electrical isolation and durability. The method selected involved thin film temperature sensors which were compressed against the side of the tube using clamping blocks  74  3. Experimental Methods and Materials  Schematic Diagram of Heated Section Construction  Figure 3.9  Tufnol  Steel  Copper  Lava Ceramic  PTFE  Fibreglass insulation  []  Tube %%%%%‘s ,,_,_ %%‘.%\%  Ceramic Fibre  F-s F5  ,s F-’  %%%%%% ,,_,_  F-s F-..  ,,_,,‘ ‘5%’’’’ ‘‘‘F.,,  Thermocouple wire  ‘5%’’’’ F.,,’,, ‘‘‘‘‘‘5 ‘‘F.,,.  ‘‘5%’’’ ‘‘1,,, ‘‘‘‘5’’ ,,,,,. ‘‘‘‘‘‘5 ‘‘F.,,. ‘‘5’’’’ ‘‘‘‘F.. ‘‘‘‘‘5’ ‘‘‘‘,. ‘‘‘‘‘‘5 F.,’,’, ‘‘‘‘‘‘5 ‘‘‘‘‘F. ‘‘‘5’’’ ‘‘‘‘F.. ‘‘‘‘‘‘5 ‘‘‘‘‘F. ‘‘‘F.,. ‘‘‘‘‘‘5 ‘‘‘‘F., ‘‘‘‘‘‘5 ‘‘F.’,, ‘‘‘5’’’  ‘‘F.,,, F.  F  /  /  /  f  F. F. / F. ‘‘‘‘‘‘5 ,,,,,.  —  f  “5  ‘‘‘‘‘5’  ‘‘‘‘‘F. ‘‘‘‘‘‘5 ‘‘‘‘‘‘5 ‘‘‘‘‘F. ‘‘5’’’’ ‘5’’’’’ ‘‘‘‘‘F. ‘‘‘‘‘‘5 ‘‘‘‘‘‘5 ‘‘‘F., ‘‘‘‘‘‘5 ‘‘‘F., ‘‘‘‘‘‘5 ‘f_f, ‘‘‘‘5’’ ,,‘,‘ ‘‘‘‘‘‘5 F.,,,,,  ‘5,%,%ø,%,%, 5’’’’’  Ns;s;; -s’ ——%  f_f,,,,,  (5  F ‘5  ,,,,,,,F. %\\%%%%%% ,,,,,,,F. %%%\%\\\% ,_,f,,,F. \\%\%%\%% f__f,,,, f_f,,,,, \\\\\%%\% ,,,,,,,, %%\\\%\\\ ,,_,,,F., ,,,,f,F.,  ‘  ‘5,  ‘  ‘5.  3 “5  F.  ‘5  1 ‘5  F ‘5  F  S. F ‘5  ,,,,fF.F., \\\%%%%%% f_f__F.,, \%\%\%\%\ ,,,,,,,, %\\%\%\%% ,,,_,F.,F. \\%%%%\\% ,,,_,F.,F. \%\\\%‘5%%  F. F. S. F. 5. >3 S_ >3 ‘5  5..  a  ‘‘‘‘‘F.,, ‘55,5,  75  3. Experimental Methods and Materials  fabricated from heat resistant materials. The method was developed during the current work and updated regularly so that some of the earlier data is less reliable. The thin film sensors used were essentially ribbon thermocouples mounted in a thin polyimide (Kapton ) carrier which provided electrical resistance up to 270°C, above TM which 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 block used in the fouling experiments. Both designs consist of solid blocks of material which could be screwed or fastened together to compress the sensor against the tube. The sensor and its copper overbraided lead had to be mounted normal to the length of the tube as the alternating current’s magnetic field could cause large fluctuations in the thermocouple signal. This is discussed further in Section 3.4.3. The original design was constructed entirely of Tufnol but this polymeric composite was found to char after prolonged exposure to temperatures  >  200°C. Design I was developed where the Tufnol mount was retained  for its strength and flexibility but featured a compression section fabricated from Lava , TM an insulating ceramic material kindly donated by AlSiMag Ltd. This material is machinable until it is fired at 1050°C, when it becomes a hard, insulating ceramic. The Lava saddles  appeared to function well but the Tufnol holders charred after continual high temperature exposure, so Design II was developed where the clamping block was machined entirely from Lava. Design I featured a clip/screw fastening system whereas Design II used offset screws to provide the clamping action. Ten thermocouple mountings were located along the heated section and could be moved as required. The intervening sections were insulated as shown in Figure 3.9; a Lava collar 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 fibreglass blanket which were secured by removable straps. These straps and the screw tightened thermocouple mountings made removal of the tube straightforward. The heated section and 76  3. Experimental Methods and Materials  Figure 3.10  Details of TFU Surface Thermocouple and Mounting Designs 6mm ‘2 ‘I’  23mm  19 mm  i-I V 1,  c-I  60 mm  17 mm  10mm Design II  Design I  (not to scale)  thermocouple glued into a grooved channel in lower clamping block  Stikon 20112 Thin Foil Thermocouple  BUTT BONDED JUNCTION  (Dimensions in inches)  MATRIX  ——.50  .75  r°°  I  I 77  3. Experimental Methods and Materials  the electrical contacts were enclosed by a steel cover which prevented accidental contact with any live surfaces. The unheated sections of the tube were insulated using a wrap of ceramic fibre woven tape and fibreglass piping insulation fastened in place by removable straps. A surface sensor was taped on to the outside of the insulation at the heated section to provide an estimate of heat losses to the surroundings. A high temperature alarm sensor, consisting of a thermocouple mounting similar to Design I, was positioned at the hotter end of the heated section. The Tufnol support was square in section and did not extend to the apparatus frame. This sensor was dedicated to the alarm only so that PC failure would not create a hazard.  3.4.3  Data Collection and Processing Experimental data were collected using the LabTechTM datalogging software  package on a PC equipped with a DAS-8 analog/digital interface board. Data were saved to a floppy disk and then transferred to a faster PC for data processing using spreadsheeting and graphical software packages. Fluid and surface temperatures were measured using a multiplexer with cold junction compensation calibrated at 0°C and 100°C. The error in the multiplexer temperature output was given as ±0.7°C. The thermocouples’ accuracy was checked up to 300°C using a Hewlett Packard Model 2801 quartz thermometer and an oil bath. The difference in recorded temperatures did not exceed the manufacturer’s specifications. The tank bulk liquid 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 heater temperature errors were ±0.5°C. Similar error limits applied to the TAH alarm sensors but these signals were isolated from the datalogging system and not recorded.  78  3. Experimental Methods and Materials  The orifice plate and tube section pressure drops were obtained as analog outputs from the DP-350 pressure sensors. These signals were filtered through operational amplifiers to remove background noise then connected to the DAS-8. The absolute pressure transducer outputs ranged from 0-50 mV so were amplified and filtered before connection to the DAS-8; the DAS-8 is sensitive to noise when input signals are less than 0.5 V. All amplified and filtered signals were calibrated within the LabTech software to account for offsets caused by signal processing. The current through the heated section was measured by adapting the current meter to give an analog output as well as a digital reading. The meter LED control signal was rectified, amplified and filtered then fed to the DAS-8. The PC and meter readings were then both calibrated using the Amprobe current meter. The voltage across the heated section was measured using a 1:1 isolating transformer to isolate the DAS-8 from the heating current; the voltage from the transformer was then rectified and ratioed using two resistors in series (10 Q, 20 Q). The DAS-8 output was calibrated using the Novatron panel meter after loading effects had been eliminated. The signal processing arrangements are shown schematically in Figures A.1.3 and A.1.4. All system data were logged and displayed on the PC screen so that the system could be easily monitored. The surface thermocouples were subject to interference from the heating current magnetic field, which caused waves in the temperature readouts of amplitude  ±  4°C at maximum power. An oscilloscope showed that the noise had a  frequency of 50 Hz. This noise was filtered out by sampling every six seconds and calculating the moving average over the previous minute. The value obtained each minute was then logged against the midpoint of the averaging period. This procedure reduced thermocouple fluctuations to ±1.5°C. A sampling period of one minute was considered reasonable during experiments lasting five hours or longer. All process variables were smoothed in a similar fashion. Data collection was often interrupted to save the data in case a power failure occurred. 79  3. Experimental Methods and Materials  The heat flux in the heated section was assumed to be uniform and calculated using Equation [3.241. The heat transferred to the liquid (Qe vi) was compared to the enthalpy gain of the fluid (Qi);  Q (Watts)  W Cp (Tb,out  =  [3.25]  ) 0 Tb,j  -  where Cp was evaluated at the mean bulk temperature, Tb  (Tb,ot  + Tb,jfl)/2.  The heat  transfer coefficient h(x) measured by a thermocouple at axial location x m from the first busbar was calculated as: h(x) (W/m .K) 2  q![T,(x)  =  -  Tb(x)]  [3.261  The increase in bulk temperature along the heated section was assumed to be linear under conditions 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 the tube was assumed to be given by the analytical solution of the steady state heat conduction equation for a long hollow cylinder with uniform internal heat generation and an adiabatic outer wall; (x)  =  (x) 0 T,  +  Qe  i.  27lL?.met 2  r  -  r  -  in 2 r  r1 (s-)  [3.281  This correction ranged from 1 to 2°C and was around 1% of the temperature difference between the wall and the fluid. The value of Amej used was calculated using T,(x). Watkinson (1968) used Qirather than to  Qe in Equation [3.28] as insulation losses amounted  3% Qe. Insulation losses calculated using Qi were less than 2%  Qe and were subject to  errors in flow rate calculation, temperature measurement and heat capacity estimation. Insulation losses were also estimated by considering free convection from the insulation surface to the ambient atmosphere. 11 q  =  Uconv  -  [3.29]  Tanib)  where T 5 and Tomb were the insulation surface and ambient temperatures respectively. The free convection heat transfer coefficient to the quiescent air was estimated at 10 80  3. Exverirnental Methods and Materials  .K. This value was a conservative estimate in order to account for radiative heat 2 W/m transfer losses. The values of  Q (calculated using  obtained were smaller than those given by Qe calculations for completeness and Qe  -  -  Q.  an insulation diameter of 100 mm)  This correction was included in the  Q was thus used in Equation 3.28.  Theovera1l heat transfer coefficient was calculated by a simple integration of the heat transfer coefficients along the heated section after the thermal entry length. The surface entry length was judged to have ended when surface temperatures increased in step with the bulk fluid temperature. Any thermocouples registering unusual temperatures were also omitted from the calculation. Watkinson (1968) calculated the overall heat transfer coefficient, U, (based on the inside of the tube) by fitting a quadratic through 15 data points and calculating the integral U=  [3.30] The method used here was to integrate the following integral using the trapezoid rule; fXL  (Qe  =  L  J  dx  -  2trL(T,(x)  -  Tb(X))  [3.31]  This approach was deemed more suitable for the fewer data points with non-uniform spacing. The two methods were compared in a number of the heat transfer tests and showed good agreement. The simpler method was maintained as it was easier to incorporate into a spreadsheet and also generated the local heat transfer coefficients which were required to study local fouling behaviour. The fouling resistance was then calculated using Equation [1.3].  81  3. Experimental Methods and Materials  3.4.4  TFU Heat Transfer  The performance of the TFU as a heat exchanger was studied using pure Paraflex at atmospheric pressure and a bulk liquid temperature of 100°C. A test section length of 770 mm was used in these trials. The Paraflex flow rate, cooling water flow and heating power setting were set at the required conditions and the system left to reach thermal equilibrium for 15 minutes before data collection was started. The data were then processed as described in Section 3.4.3. The insulation heat losses were studied in the absence of heating across the complete range of flow rates. No discernible drop in fluid temperature was recorded across the tube in these tests and surface temperatures ranged from 1-5°C below the fluid temperature. Heat transfer tests were performed with Tb1k = 100°C, flow rates from 1.0  -  2 and 228±3 kW/m ), 2 4.5 mIs and the variac set at 80% and 90% capacity (185±5 kW/m respectively. These tests also provided operating characteristics for future fouling experiments. Figure 3.11 shows the correlation between the experimental data and Nu predicted by the Gnielinski correlation (Equation [3.21]) using Gnielinski’s expression for the 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 Section 7. The agreement between the sets of data and with the correlation was good and the  1 deviation lay within the estimates of experimental error for the apparatus (5 % at Re  =  1 increased slightly with heat 7760). The heat transfer coefficients from runs at the same Re flux, as reported in Section 3.2.1. Figure 3.12 shows the axial variation in the local heat transfer coefficient at four values of Re . The local heat transfer coefficient decreases from a high inlet value and 17 increases towards the exit of the tube as the bulk temperature rises. The thermal entry  > 10 000, in 1 , but almost insignificant at Re 1 length is significant at lower values of Re 82  3. Experimental Methods and Materials  Comparison of Predicted TFU Nusselt Number with Experimental Data  Figure 3.11 300  -  .  .  .  .  250 200  1:  150  100 •  Fouling Runs  o  Paraflex Tests  50  I  00  100  50  150  .  .  .  200  .  I  .  250  300  Nu (Gnielinski) Figure 3.12  Thermal Entry Length Effects in TFU Heat Transfer  1.30 1 .25 A  1.20 1.15 1.10  A  A  1.05  a  —  -  1.00  ...—  .-  ......  I  A  0.95 0.90 0  10  Re  =  20  5454  Heat Flux  =  30  •  Re  =  40 xID 7703  50  ; Paraflex at Tbulk,jn 2 183 kW/m  83  Re  I =  =  60  70  10805  •  80  Re  100°C; Re calculated at inlet  14127  3.Experiinental Methods and Materials  agreement with the literature. Various correlations were tested against these data but most failed in the first few tube diameters where the heat transfer coefficient approaches infinite values. The profiles could be generated by a numerical solution of the governing enthalpy and momentum equations but this was beyond the scope of the current study. A further unknown is the heat loss by conduction through the copper terminals and cables. The heat loss to the terminals is thought to eliminate conduction along the tube as the busbars present a surface area 40 times greater than the stainless steel tube cross section. , in the insulation and heat transfer 1 The pressure drop across the tube section, AP  tests was compared with the pressure drop calculated using Equation [3.33] and an estimated tube thickness of 12 thou. TFU conditions were Tblk,t 377 kPa, Re  5000  -  100°C, Pair = 101 and  16000. End effects in the fittings (Ke, K) were estimated using  velocity losses given by Kay and Nedderman (1985) and were small relative to the tube length contribution. =  2  where the p and  Urn  +  Ke  +  1 4f  ) 1 2r  [3.331  are the average fluid density and flow velocity along the test section, L  is the separation between the pressure taps, and  f is  the Fanning friction factor. The  calculated values of the friction factor from the heat transfer and fouling runs were plotted against literature correlations in Figure 3.13. Blasius’ friction factor correlation applies to smooth tubes at Re f  =  >  4000;  0.079 Re 025  [3.34]  The figure shows that the experimental values of f tend usually to be larger than the correlations’ predictions. The discrepancy was considered to be within reasonable experimental error given the uncertainties in tube thicknesses, variations between tubes, tube roughness and fluid properties. The agreement in pressure drop and heat transfer data with existing correlations established the reliability of the TFU as an experimental heat transfer section.  84  3. Experimental Methods and Materials  Figure 3.13  Comparison of Experimental Fanning Friction Factor Data with the Correlations of Blasius and Gnielinski  0.01 1 0.010 114  11.4  0  Gniehnski  •  Blasius  •  e18  .••  0.009  -  9,.3  -  0.008 •o  0.007  •o  0.006 0.005 0.005 -  0.006  0.007  0.008  0.009  Calculated Friction Factor  85  0.010  0.011  3. Experimental Methods and Materials  3.4.5  TFU Operating Procedures  The TFU operating procedures were similar to the PFRU and benefitted significantly from the experience gained from this apparatus. The experimental methods differed 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 and the terminals themselves were thoroughly cleaned with contact cleaning solvent and left to dry. The tube was then installed and the unit pressurised to the operating pressure with air in order to detect any leaks. The tank was then charged with solvent and the solvent circulated at the operating pressure to check for leaks again. If no leaks were detected, the terminal and thermocouple locations were marked using a permanent marker before the clamps were tightened in place and the test section insulated. The data collection system was then activated and the solvent circulated at atmospheric pressure. The thermocouple readings were monitored at a variac setting of 50% in order to check that they had been properly installed. Any other adjustments were performed at this stage. To start a run, the system was pressurised and the solvent circulated while the tank heater brought the liquid to the required temperature. Once thermal equilibrium was reached at the given flow rate, data were collected for ten minutes to provide insulation performance information. The variac was then turned to the required setting and the system left to reestablish thermal equilibrium. Adjustments in variac setting were often necessary to achieve the required surface temperatures as the change in viscosity on adding indene to Paraflex caused the surface temperature to drop. Air bubbles were cleared from all cooler and pressure lines at this time. Pure solvent heat transfer data was collected for 5-10 minutes and the system was then depressurised. Indene and any additives were added whilst running via the tank filling port. The system was pressurised, flow conditions checked and data collection started before the first sample was taken to start the experiment. 86  3. Experimental Methods and Materials  Experimental procedures (sampling and analysis) during a run were similar to those in the PFRU except that the ëooling water flow had to be monitored once gum formed and caused fouling in the primary cooler. Once this became significant the auxiliary coolers were 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 logger was stopped before simultaneously reducing the heated section power and depressurising the system. The pump was then stopped and all process liquid drained from the system. It was important to maintain the gas flow at this stage in order to avoid gum blockage of the gas feed line. The test section was carefully removed and replaced by a dummy sectIon for cleaning. The cleaning procedures described in Section 3.2 were used with scaled up rinse quantities. The test section was rinsed with hexane and left to dry. The terminal and thermocouple locations were recorded from the drIed test section before the tube was cut up into labelled 50.8 mm sections using a pair of steel snips. The snips dislodged a minimum of material without generating any steel dust. Significant losses were 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 as desired. After analysis, any remaining foulant was scraped off gently and the metal cleaned using acetone, chromic acid solution and distilled water. The dried sections were then washed with hexane, dried under vacuum and weighed again. The mass of foulant was expressed as mg/cm 2 by assuming a tubing thickness of 10 thou and calculating a mean section length from its mass and the mass of the clean tube. The thin metal tubing was ideal for optical and electron microscope analysis.  87  3. Experimental Methods and Materials  Gum Ageing Oven  3.5  The chemical literature describes the homolysis of peroxides in solution but there is scant coverage of the degradation of peroxide gums following deposition on a solid surface. A series of gum ageing experiments was performed to investigate the fate and kinetics of the pyrolytic degradation of gums generated during indene autoxidation. The experiments were performed in an inert (nitrogen) atmosphere under isothermal conditions and were intended to trace physical and chemical changes in the gum during the ageing pIu.  Figure 3.14 is a schematic of the apparatus, which consists of an adapted Varian enes 14(31) UC with a top opening tic!. Ihe oven is purged with nitrogen and all other connections are blanked off. The temperature distribution inside the oven was checked with a manouvreable K type thermocouple and found to be close to isothermal. The unit was located 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 the mass recorded. The crucible was quickly placed in the centre of the oven and heated for the required period then removed and left to cool before weighing. The product was easily removed for further analysis and testing. Blank samples showed zero weight loss and repeat samples showed good reproducibility. The gum ageing experiments were also duplicated using Thermal Gravimetric Analysis (TGA). Approximately 10 mg of finely ground gum was equilibrated at 30°C then heated at 100°C/mm to the ageing temperature in a nitrogen atmosphere. The experiments were performed using a Perkin-Elmer TGS-2 device interfaced to a Perkin Elmer TAGS data station.  88  Figure 3 14  Schematic Diagram of Gum Ageing Oven Appara  Rotajner  nitrogen Supply  89  3. Experimental Methods and Materials  Chemical Analysis  3.6  The difficulties in analysis of autoxidation mixtures have been discussed in detail by Link and Formo (1962). Analysis of the model solutions was significantly simpler than industrial fluids but it cannot be described as a complete monitor of the reaction system. The products of indene autoxidation range from hydroperoxides and polyperoxides to epoxides, carbonyls and rearrangement products which are not available for use as reference compounds or standards. The chemical analyses developed for solution samples described here represent a set of reasonably accurate, simple methods which monitor the significant steps involved in the generation of fouling precursors. Methods such as High Performance Liquid Chromatography could ostensibly have been used to separate individual polyperoxide fractions but the costs were deemed unreasonable. The solid analysis methods represent a similarly pragmatic approach to the study of a complex organic material. The solution analyses were designed around an autoxidation scheme adapted from that of Norton and Drayer (1968): A  --->  indene Method  GLC  B hydroperoxide  C ---> D [3.35] polyperoxide insoluble peroxides  --->  Peroxide Number Gum Assay  -  This scheme is not based on the chemical mechanisms described in Section 2 and represents a simplification of the reaction system to a form which can be described by the available analytical methods. No method was found to determine the amount of insoluble peroxides formed during an experiment. The amount of insoluble gum remaining after an SCR experiment depended on the number and timing of samples so that a simple kinetic analysis is not feasible.  90  3. Experimental Methods and Materials  3.6.1  Hydroperoxide Analysis  -  Peroxide Number (POx)  A titration method for hydroperoxide determination was chosen over precipitation from base or polarography due to the simplicity and reliability of the iodine titration for autoxidative systems. The method employed is based on ASTM E298. Hydroperoxide is reduced by iodide, yielding iodine which is then titrated against standardised sodium thiosuiphate. The redox reactions are; R-OOH 12  +  +  21-  +  3 O S 2 2Na  2H  ---->  R-OH  --->  21-  +  +  12  +  0 2 H  6 O 4 S 2 Na  +  2Na  [3.36] [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 of thiosuiphate. The titration is specific to the hydroperoxide end group and thus the peroxide number includes contributions from both monomeric and polymeric hydroperoxides. Polyperoxide linkages could be included in the Peroxide Number if stronger oxidants such as boiling hydriodic acid (Russell, 1956a) were used. A volume of sample corresponding of 1-4 milliequivalents of oxygen was dissolved in 20 mL glacial acetic acid in a stoppered Erylenmeyer flask which had been deaerated using a carbon dioxide purge. The exclusion of air prevents air oxidation of iodide in solution. The mixture was deaerated, 5 mL saturated aqueous sodium iodide solution added, deaerated again and then immersed in a water bath at 60°C for one hour. The flask was then cooled, 50 mL deaerated distilled water added and titrated against 0. 1M sodium thiosulphate. 1-2 mL starch solution (5 gIL) was used as indicator, which lost its dark blue colour at the end point. The sodium thiosulphate was standardised against potassium permanganate and stored under carbon dioxide. The Peroxide Number was calculated as  POx (meqlL)  =  1000 x 0.1 x (volume Na 2S ) 3 O 2 (volume sample)  91  [3.38]  3. Experimental Methods and Materials  Samples were run as duplicates with a control to monitor reagent quality. The method was tested using t-butyl hydroperoxide and found to be satisfactory. The analysis did not work in the DCP experiments as the DCP seemed to react directly with the iodide or iodine generated. Hydroperoxides are acidic and alkali precipitation was tried using aqueous sodium hydroxide on a sample from an SCR experiment. A brown solid was obtained but the technique was not developed further.  Polyperoxide Analysis (Gum in Solution Assay)  3.6.2  The early PFRU fouling studies showed that the Peroxide Number was not a direct measure 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 solution so a gum assay was developed, based on RusseWs separation of indene polyperoxides using hexane. 10 mL of solution was pipetted into a 50 mL Erylenmeyer flask, stoppered and stored 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 mL was found to be sufficient excess for kerosene, Paraflex and n-octane, but 30 mL was needed for toluene and trichlorobenzene. A Whatman #1 filter paper was dried under vacuum and weighed. The filter paper was placed in a Buchner funnel, rinsed with hexane and the sample solution filtered through it. lOmL cold hexane was used to rinse the flask and paper before both were dried under vacuum for 30-60 minutes. The flask and filter paper were weighed, the paper discarded and the flask cleaned with acetone, dried and weighed. The gum content was calculated by difference and expressed as g/L. Initial trials showed reasonable accuracy (±0.2 g/L) and repeatability but later experiments indicated that duplicate sampling was preferrable.  92  3. Experimental Methods and Materials  The gum obtained varied with the solvent and the stage of the reaction. The gum recovered from the aromatic solvents was a darker orange colour than the yellow or amber gum produced by Paraflex and n-octane. The insoluble gum found in the SCR unit after a Paraflex 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 Paraflex experiments and shows that the gum changes form during the reaction. Section 4.1 describes how a gum solubility limit exists in aliphatic solvents; before this concentration is reached, the gum is suspended in solution and is collected on the filter paper. The filter paper 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 about gum agglomerate sizes. The mean molecular mass of the gum produced during selected indene in kerosene experiments was estimated using the benzene melting point method. The gum precipitated by hexane was dissolved in 2 mL pure benzene (Fisher, 99.99mo1%) and frozen in a boiling point tube fitted with a stirrer and a ‘J’ type thermocouple. This approach was used because the sticky gum is difficult to weigh out accurately. The freezing point was noted and the molecular mass calculated via: M where  =  [ATe pt m(gum)]/[z.Tf., 0 m(benzene)]  [3.39]  is the molal freezing point depression constant (5.1 K for benzene) and the freezing point depression (K). The estimate is subject to errors in mass and  temperature measurement, gum hexane content and temperature accuracy, but was considered a helpful guide in the analysis programme. Estimates of M for the gum produced from indene in kerosene varied from 350 to 450, corresponding to 2-4 indene polyperoxide units of mass (116.1  +  32  =  148.1).  93  3.  Figure 3.15  Experimental  Methocts’ and Materials  Photograph of Gum Assay Filters showing Change in Gum Appearance  94  3. Experimental Methods and Materials  3.6.3  Indene Concentration  -  Gas Liquid Chromatography  Gas-Liquid Chromatography (GLC or GC) is the standard method for the quantitative analysis of organic liquid solutions. The Bromine Number Test for unsaturates (ASTM Dii 59-57T) and ultraviolet spectroscopy were also evaluated but neither of these techniques proved to be as specific or as accurate as GC methods. The disadvantages of GC methods are the relativley long run times associated with the solvents used and the destruction of any peroxides in the injector, preventing their detection. 2 mL samples of solution were stored frozen until analysis, when they were mixed with 5 mL of GC solvent. The GC solution used varied with each solvent but was based on C1 which dissolved all samples and is quickly eluted from the CH , dichloromethane, 2 column. Paraflex, n-octane, toluene and trichlorobenzene used a 200:1 v/v solution of 2 and an internal standard, n-decane. The majority of the samples were anlaysed using MeC1 one of two machines, using identical columns and temperature programming. A Varian Vista 6000 GC fitted with a thermal conductivity detector and manual injector was used initially, with helium as the gas phase. A Varian CDS 401 data station performed all chromatogram acquisition and integration. Subsequent analyses were performed on a Varian 3400 Star GC fitted with a liquid autosampler, flame ionisation (FID) and photionisation (PID) detectors. Nitrogen was used as the carrier gas. Separation employed a 1.83 m x 1/8 inch stainless steel packed column with 80/100 Supelcoport as the stationary phase and 10 wt% OV-lOl as the liquid phase. These columns were repacked and re conditionned as required. Some early analyses were also performed using a Hewlett Packard Series 2 GC fitted with a Mass Spectrometer detector which allowed peak identification. These analyses used a 12 m HP-i capillary column with helium as the gas phase. The pure solvents eluted as single peaks while Paraflex and kerosene eluted as a series of peaks. Figure 3.16 is a chromatogram generated by the Varian 3400 unit from a  95  3. Experimental Methods and Materials  Figure 3.16  Gas Chromatogram of Solution of Indene in Paraflex  If)  0  Go  If) 0i  I  I  I  C.) C)  Column Temperature 250°C ‘40°C  40  100°C I  I I I I  0  6  I  I I I I I I I  I I I I I  29.5  95  96  I I I I I I I  32  Time (minutes)  C)  3. Experimental Methods and Materials  0.41M solution of indene in Paraflex and shows the temperature program used. The Paraflex components elute after indene and the internal standard so that quantitative analysis is straightforward. The indene concentration is calculated using the internal standard as a reference. The ratio of indene peak area to that of the n-decane is calculated and compared with a calibration curve prepared using standard solutions. The sample/solvent ratio was chosen to ensure a linear calibration curve. Indene analysis in kerosene solutions was complicated as indene elutes among the kerosene 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’ the indene peak. The GC solution was a 200:1:1 v/v mixture of 2 C1 indene and octadec- 1CH , ene, a reference which elutes after the kerosene components. Peak height was used in the calculations and an accuracy of ±5% was observed in the calibration. Added standard analysis usually employs a series of samples with different amounts of added standard but this approach was considered to be too time intensive in the absence of an autosampler. A different approach was used later with the Varian 3400 system. This unit featured a PID with a 10 meV UV lamp, making the detector more sensitive to aromatics and alkenes than alkanes. Figure 3.17 is a comparison of PID and FID chromatograms from a solution of 10 wt% indene in kerosene; the detectors are arranged in series and the selectivity for indene is apparent. Internal standard analysis at 1:20 v/v dilution in 2 C1 with bromobenzene as CH reference yielded ±2% accuracy over the calibration range. The PID selectivity for aromatics and alkenes over alkanes was also used in the studies of antioxidation described in Section 5.7. The antioxidant, BMP, eluted among the Paraflex peaks at low concentration (ppm). The BMP peak was visible in the PID chromarogram but high background noise prevented quantitative study and the qualitative results were deemed sufficient. Solvent extraction was not considered.  97  3. Experimental Methods and Materials  Gas Chromatograms of Indene in Kerosene from Rame lonisation and Photo-lonisation Detectors in Series  Figure 3.17  PD  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  N 0 0  •  a  N  mnJ  0  RD  N  co 10  co aD  Cl  N  0 ‘0  r  co  I  W c () C) ‘d  I  I  I  I  cv ci)  r  $  98  I  I  I  I  I  I  I  I  3. Experimental Methods and Materials  3.6.4  Further Analysis  -  FTIR,  SEM  A variety of analytical methods was used to provide further information about the chemical and physical nature of liquid and solid samples. Fourier Transform Infra-Red Spectroscopy (FTIR) is used to identify which chemical functional groups are present in gas, liquid or solid samples. Two microprocessor controlled devices were used; a Perkin Elmer Model 1710 device and a Midac FTIR with an MCT detector. Early solids analysis used mulls of the solid sample spread between two KBr disks. Subsequent solids analysis used the KBr pellet technique. Approximately 5 mg sample 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 for analysis. Liquid analysis was performed using a,0. 1 mm pathlength liquid cell fitted with KBr windows. Elemental analysis (C, H, Cl, 0 by difference) was performed by Mr. P. Borda of UBC Chemistry and by commercial laboratories (Canadian Microanalytical, Guelph Analytical Laboratories). TFU tube sections were examined using a Hitachi S-2300 Scanning Electron Microscope located in the UBC Department of Metals and Materials Engineering. This device was equipped with a Polaroid Camera and a Quartz PCI Image Capture System for photographic and digital recording of images. A coupon approximately 8 mm x 8 mm was cut from the tube section and gold sputtered before mounting in the SEM. Gold sputtering was necessary to improve the thermal conductivity of the sample as the 20 kV beam quickly destroyed the unprotected surface at higher magnifications.  99  4. Auto xidation of Model Solutions  4.  Autoxidation of Model Solutions of Indene  Model solutions of hexadecene or indene in four solvents (Paraflex, kerosene, tetralin and tiichlorobenzene) were tested in the PFRU system to establish which would be good candidates for fouling experiments. These preliminary fouling studies are described in Sections 5.1-3. The chemical analysis results from these runs indicated that the autoxidation of model solutions of alkenes in the PFRU system differed from the mechanisms and kinetic forms reported in the chemical literature. These initial studies also confirmed 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 stirred cell reactor in order to identify and understand the important reaction parameters. Indene was studied as this was the alkene selected for further fouling studies. The work was performed in parallel with the fouling studies and is described first in order to introduce the reaction analysis used in the fouling runs.  4.1  Solvent Effects in Indene Autoxidation  The choice of solvent is important in model solutions for fouling studies as the solvent can affect both the reaction kinetics and the physical behaviour of the reaction products. The initial fouling studies in Sections 5.1-3 reported both of these effects and so the 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 which undergoes autoxidation less quickly than indene, and trichlorobenzene, a polar aromatic solvent. The relative inertness of the aliphatic solvents was confirmed by control experiments 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 of 100  4. Autoxidation of Model Solutions  hydroperoxide 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 this was also found in the kinetic experiments in the SCR. A heavy black solid was formed as well 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 studied in the SCR under the conditions described above. The experiments were run until the Peroxide Number showed little change or sampling was hindered by the formation of sticky red/orange gums. Figures 4.1 a-c show the Peroxide Number, soluble gum and indene concentration results in each solvent. The Peroxide Number climbs steadily and then decreases, except in the case of Paraflex where it maintains a ‘plateau’ value. The chemical induction period, ‘r, corresponds to the start of the rapid increase in POx. The rapid increase was often observed after the POx reached a value of 5 meq/L. More precise determination of r was limited by the maximum sampling rate (llhr) in the analytical method. The kerosene data all show a much longer delay (larger t), which is attributed to the presence of a commercial inhibitor in the solvent. The inhibitor is presumably exhausted after the induction period and the kerosene data behaves similarly to the other data after time  Solvent effects are most apparent in the soluble gum concentration data. Figure 4. lb shows a pronounced difference in gum behaviour between aromatic and aliphatic solvents. In aliphatic solvents the gum concentration increased linearly with time after the induction period until reaching a maximum, referred to as the solubility limit, g  ,  at time t*.  The value of g* is higher for kerosene, which ‘ C n.m.r. analysis showed to contain some 3 unsaturated components. The solubility limit is thus a function of solvent nature, which could be explored further by using mixtures of solvents. Comparison of Figures 4. lb and  101  4. Autoxidation of Model Solutions  Figure 4.la-c Solvent Effects in Indene Autoxidation : Analysis Results Figure 4.la  Hydroperoxide Concentration 80  70  0  60 zI-  0 0 00  40  A  30  ‘1)  0  0  50  20 10  0  1...I...  30  20  10  0  50  40  60  Figure 4. lb Gum Concentration 50 40  00 o0  C  30 0  4:1 0  20  k,E 0 ; A  10  ...  •••  AiOA  000  ,  0  20  10  0  I  I..  40  30  50  60  50  60  Figure 4. ic indene Concentration 0.50  AA  0.45  0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00  A  1 oc 0 0 0* A  00 -  0  -  -  10  0  j 1 £O .an....,  -  20  30  .  .  40  .  Time (hours) 0  Paraflex  •  kerosene  A  Solutions of 0.41 M indene at Tbulk  =  n-octane o  toluene  trichlorobenzene  I  100CC: 79.2 kPa oxygen saturation: thermal initiation  102  4. Autoxidation of Model Solutions  4.lc shows that indene consumption continues after t. The reactor internals were coated with a sticky, red insoluble gum after the experiments using aliphatic solvents. This insoluble gum was not found with the aromatic solvents, suggesting that all the oxidation products 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 addition to abstraction. This ratio would be exact if the gum were a primary oxidation product; the kinetic models described in Section 4.6 show that the gum formation rate is a complex function of the rate of indene consumption, but the yield still provides a useful parameter for comparing sets of analytical data. The yield was calculated as the mass of gum produced in the early stage of autoxidation (time t to t*) divided by the maximum mass of polyperoxide that would be generated if all the indene consumed formed equimolar quantities of indene polyperoxide linkages, of mass 148.1, i.e. Yield  k3b a 3 k3b + k  {gum(t)} {gum (t*)} [indene(t)]) x 148.1 ([indene (t*)j -  -  [41]  where [indene] is given in (molIL) and {gumj is in (gIL). This equation assumes 1:1 stoichiometry between gum and indene, which is discussed further in Section 4.6. Table 4.1 summarises the calculated yields and the other kinetic parameters obtained from these experiments. 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 than the 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 of autoxidation 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 gum rate in trichlorobenzene is thought to be linked to the contamination problem as indene is being consumed at similar rates (shown by kND). All the kinetic parameters from the kerosene experiment were low and prompted the study of antioxidants discussed in Section  103  4. Autoxidation of Model Solutions  Table 4.1 Solvent  Solvent Effects in the Autoxidation of Model Solutions of Indene Solvent Nature  Initial POx Rate Constant kM, POx (‘J[mmoIIL]/hr) ± 0.02 0.34  Initial Gum Rate (g/Lbr) ± 0.1 1.5  Indene rate Solubility constant Limit g* kND (g/L) (1/br) ± 0.5 ±0.005 0.096 soluble  Insoluble Gum Analysis  (g/g) 0.715  C 65.9 wt% H 4.2 wt%  soluble  0.689  799 H 9 C 1 0 88  0.139  10.0  0.529  8 7 0 5 . 7 H 9 C 3 , 2  1.1  0.077  14.0  0.694  7 0 9 . 7 H 9 C 1 87  2.2  0.118  11.0  0.567  0 6 0 8 H 9 C 0 , 2  trichlorobenzene  polar, aromatic  toluene  aromatic  0.81  2.5  0.092  n-octane  aliphatic  1.04  2.23  kerosene  mainly aliphatic  0.19  Paraflex  aliphatic  0.85  0.41 M indene under thermal initiation : Tbulk  Gum Yield  100°C : 79.2 kPa oxygen saturation  104  4. Autoxidation of Model Solutions  4.5. Other experiments in kerosene indicated that indene reacted at similar rates to those in  Paraflex, which confirmed that the autoxidation reaction is not influenced strongly by inert solvents 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 rates in all solvents except kerosene, which featured an extended induction period and lower rate of reaction. The data in the toluene experiment, however, showed a marked reduction in rate as the reaction proceeded. The inhibition corresponded to the appearance of a new GC peak which eluted later (i.e. more polar) than the toluene peak. This peak had the same retention time as benzaldehyde; this secondary oxidation product of toluene is known to be self-inhibiting in autoxidation and thus appears to be inhibiting toluene autoxidation in this case. This result and the fouling trials in tetralin (Section 5.1) confirm the complexity of solvent 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 hydroxyl  and carbonyl activity. Carbonyl groups could be formed by peroxide decomposition or condensation. Gum formed later in the experiments was orange-red in colour and contained higher 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 is hindered by its agglomerating properties, trapped solvent and difficulty in handling this very tactile substance. The gums were analysed for C, H and 0 (by difference) and Table 4.1 shows that the gums were similar in composition to indene polyperoxide, the expected primary oxidation product.  105  4. Autoxidation of Model Solutions  42  Autoxidation Kinetics  The chemical analysis results from the autoxidation experiments were fitted to a set of kinetic models and the parameters obtained were used to compare the reaction behaviour between individual runs. Simple models were used as no single model was found which could fit all the observed features of the data. The data from Section 4.1 were used here to describe the selection of the diagnostic parameters.  Figure 4.lc shows that the indene concentration decreases at similar rates in the uninhibited solvents, confirming that the reaction rate is not strongly affected by the solvent nature. Van Sickle et al. (1965b) reported that the rate constant for cyclopentene oxidation increased with the solvent polarity, but this is not reflected in the trichlorobenzene data. The indene 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 and indene. Combining Equations [2.5] and [2.141 gives d[0 ] 2 /dt  =  [ 5 -kpR[RH]’ ] 2 °• 0  [4.2]  where kpR is a lumped propagation rate constant. Equating the rates of oxygen and indene consumption, assuming constant dissolved oxygen concentration and integrating gives [pJ{]-O.5  =  5 0 [j?J.{j O.  +  0.5 kpR [02] t  [4.3]  Figure 4.2 shows the data from the Paraflex test plotted using Equation [4.3] and the first order 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 is  essentially a parameter fit to explain the observed experimental results. Neither equation fitted 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 stops increasing rapidly. This point could be interpreted as the point where hydroperoxide 106  4. Autoxidation of Model Solutions  Kinetic Plots of Indene Concentration Data: Comparison of Kinetic Models  Figure 4.2  20  0 -1  -1  10  -3  4  -6 0  5  10  15  20  25  0 30  Time (hours) Thermal initiation of 0.41 mol/L indene in Paraflex, TbUlk Russell  (Equation 4.3)  n  =  1.5  Norton and Drayer  (Equation 4.4)  n  =  1.0  107  100°C, 79.2 kPa oxygen saturation  4. Autoxidation of Model Solutions  decomposition overtakes the thermal initiation step. The values of kND in Table 4.1 show that indene is being consumed at similar rates in all solvents except kerosene, confirming the observations in Section 4.1. These two models do not fit the data well and neither provides a satisfactory description of hydroperoxide behaviour. Russell did not measure hydroperoxide concentrations, while Norton and Drayer’s model does not predict the variations in POx observed with the solvents in these experiments.  Mass Transfer Effects in Autoxidation Kinetics  4.2.1  Russell’s experiments were performed on a small scale and featured agitation such  that “rates were not complicated by diffusion controlled processes”. Mass transfer limitations in reaction rate are more likely to arise at higher temperatures and on a larger scale. The oxygen for reaction with indene must be absorbed from the vapour phase and if the liquid reaction rate exceeds the mass transfer capacity of the reactor, the reaction rate will show a mass transfer limitation. Oxygen is a sparingly soluble gas in petroleum oils and mass transfer effects were consequently shown to control the reaction kinetics in the SCR and fouling experiments. A more complete treatment of mass transfer with chemical reaction can be found in the text by Danckwerts (1970).  The reaction between oxygen and indene can be represented as the reaction between sparingly soluble reactant A and involatile reactant B: A +  Products  B  14.5]  Assuming that no other reactions occur, -  dNA/dt  =  -  [4.6]  dNB/dt  where N 1 is the total number of mols of component  j in  solution can be described using mass transfer coefficients; 108  solution. The flux of A into  4. Autoxidation of Model Solutions  JAAMIMA  =  1 (CA,int AM k  CA,bulk)  =  AM kG(PA,g  -  PA,int)  [4.71  where AM is the surface area available for mass transfer, k 1 and k 0 the liquid and gas phase mass transfer coefficients, A the partial pressure of A and CA the concentration of A in solution. The subscripts g, i nt and bulk refer to the bulk gas, vapour-liquid interface and bulk solution respectively. Figure A. 1.5 is included as a schematic representation of absorption of A followed by chemical reaction with B. Equating the reaction in solution to the mass transfer flux, dNA/dt  -  =  -  dNB/dt  =  -  V dC/dt  =  1 (CA,jflt AM k  -  CA,b1k)  [4.81  where V is the volume of solution. Equation [4.8] holds when mass transport is by physical (i.e. diffusional) processes alone and the chemical reaction is termed slow with respect to mass transfer. The parameter.ki is determined by the physical properties and mixing conditions of the liquid phase alone and a maximum diffusive flux occurs when reaction in the bulk reduces CA,bUlk to zero; the maximum flux condition can then be written -  V dCB/dt  =  1 CA,jt AM k  [4.9]  The gas phase resistance to mass transfer is usually minimal in the case of sparingly soluble components as in this case and is ignored. Equation [4.9] was used to estimate the values of k 1 from the data in Section 4.1 as CB was monitored and  was maintained constant  during the experiments. A maximum flux argument was used as the dissolved oxygen concentrations could not be measured in the reactors and the physical mass transfer 1 was not known. The surface area Ap coefficientk 1 was estimated as the quiescent vapourliquid interface area in the SCR (0.01327 m , ignoring bubbles); a mean value of V was 2 used in the calculation of V dCBIdt. Table 4.2 shows the initial consumption rate of indene in the solvents above and the calculated value of k . Typical values of k 1 1 in this configuration 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 system can be expressed as a function of the diffusivity DA and a diffusional contact time, tD, by  1 k  =  [4.10]  5 (DjtD)°  109  4. Autoxidation of Model Solutions  Table 4.2  Mass Transfer Effects in Batch Autoxidation of Indene -dNB/dt  Oxygen  Estimated  6 x10 (mol/s)  Concentration ASTMD2779 CA,jnt ) 3 (mourn  tnchlorobenzene  7.32  toluene  Solvent  Diffusion time tD  5 x10 (mis)  Diffusion time tD lW-Cl (s)  5.56t  9.615  0.343  0.915  11.96  4.62  10.89  0.206  0.375  n-octane  14.35  12.10  8.66  1.164  1.84  kerosene  4.18  7.37  4.14  5.596  5.578  Paraflex  10.54  5.42  14.19  0.153  0.289  (s)  t estimated as the density of trichlorobenzene> ASTM range W-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 saturation -  110  4. Autoxidation of Model Solutions  Table 4.2 includes the estimated values of tD calculated using the diffusivity correlations of Wilke and Chang (W-C) and Hayduk and Minhas (H-M) (see Reid et at., 1987). Astarita found that 0.004  <  tj  <  0.040 s for physical mass transfer in well agitated systems; the  calculated values of tD all lie above this range and indicate that autoxidation is limited by mass transfer in these experiments. The main source of error is the estimated contact surface area, AM, but this is unlikely to be an order of magnitude too small. There was no direct method available for measuring bubble sizes (and thus surface area) at the temperature and pressure conditions involved in the reactors. Subsequent experiments at higher gas flow rates (and presumably larger bubble surface areas) did not show any significant increase in the indene reaction rate. All the SCR and fouling experiments inspected using this analysis gave a similar result. The mass transfer effect is inferred from the observed reaction rates as it cannot be confirmed without a reliable method of measuring the dissolved oxygen concentration in the bulk liquid. Dissolved oxygen probes used in aqueous media rely on electrochemical principles and are unsuitable for use in organic solutions. A simple method was not available during the course of this work. The observed chemical reaction rate was markedly larger than the estimated maximum rate due to physical absorption alone. This result infers that the autoxidation reaction is eitherfast or instantaneous with respect to mass transfer. The reaction could belong to the intermediate regime (i.e. reaction rate similar to the mass transfer rate) if the estimate of AM was too conservative, but this would give very complex overall kinetics and was 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. Later experiments using chemical initiators showed that the reaction rate could be increased by chemical means, which ruled out the instantaneous regime. The autoxidation reactions in this work were thus inferred to belong to the fast regime, where most of the chemical 111  4. Autoxidation of Model Solutions  reaction occurs in the diffusive film next to the vapour-liquid interface and the bulk concentration of oxygen (and thus the bulk reaction rate) is negligible.  Kinetics of Autoxidation with Mass Transfer  4.2.2  Mass transfer effects were shown to be significant in the previous section using relationships from the surface renewal theory of mass transfer with chemical reaction. The film model was used in the following analysis of kinetics in the SCR in order to obtain simpler results which could be compared with the experimental data. The film model is applied to the reaction in the idealised liquid diffusion film and the overall result written in terms of concentrations in a well mixed batch reactor.  The absorption of gaseous reactant A (here, oxygen) into a liquid is governed by Ficks law for dilute solutions CAJdx -DAd 2  =  [4.11]  RA  where RA is the volumetric reaction rate of A. A similar equation can be written for reactant B (here, indene). The film model of Lewis and Whitman describes the mass transfer resistance in terms of an effective diffusion film of thickness ÔD. The boundary conditions for mass transfer with chemical reaction, where all reaction occurs in the diffusion film, are then dCA/dx  =  0  at x = ÔD  [4.121  CA  =  t 11 CA,j  at x =0  [4.13]  The solution of [4.11] depends on the form of RA. Two cases will be considered here; the kinetic scheme proposed by Russell and a diffusionlreaction model first proposed by Van de Vusse (1961) and described in further detail by Kay and Nedderman (1985).  For n-th order destruction of diffusing species A, RA 112  =  -  kCA”, giving [4.111 as  4. Autoxidation of Model Solutions  DA 2 CA/dx d  [4.14]  CA 1 k  =  Kay and Nedderman show that the solution of [4.14] is facilitated by multiplying across by dCA/dx and integrating;  DA 2 CAIdx (dCA/dx) (d )  =  kflCA (dCAIdx)  DA/2 (dCA/dx) 2  =  1(n+1)J CA’ 11 [k  +  B  [4.15]  Applying boundary condition [4.13] gives 13  -  “A,int  —  2 \ dx x = 0  n  +  1  [4.161  The solution of Equation [4.15] thus involves the evaluation of the following integral (0  I  dx  I  =  Jo  2k11 C’’ A dC (n+1)DA  +  213 DA  -0.5  —  [4.17]  Numerical integration of [4.17] involves an estimate of J3, which includes the (unknown) concentration gradient at the gas-liquid interface.  Kay and Nedderman described a simplified result for fast reactions, where the reaction is effectively complete within the diffusion film and both CA and dCAIdx are zero at x  ÔD.  Writing this as one boundary condition  dCA/dx givesj3  =  where CA  0  =  =  0  [4.18]  0. [4.16] thus gives the concentration gradient (and hence the absorption rate) at  the gas-liquid interface  (dCA dx  —  =  0  —  1 r  ,-‘n-i-l ]0.5  V.’l’fl ‘—A,int (n+1) DA  L  i  [4.19]  This result was used to formulate an expression for mass transfer with Russell’s kinetic scheme.  Russell’s kinetic scheme is expressed by Equation [4.2], giving [4.11] as -  CA/dx d DA 2  =  -  5 CA° 15 kpyCB  113  [4.20]  4. Autoxidation of Model Solutions  This can be simplified by assuming that the indene concentration in the diffusive film is effectively constant, which is true when indene is in excess. The current experiments involve oxygen solubilities of 0(5 mourn ) and initial indene concentrations of 0(410 3 ) so this assumption is justified. Equation [4.20) is thus half order in oxygen 3 mourn concentration, giving [4.191 as 0 (dCA/dx)X  =  -  (4 5 /3Dj° CA,1t°’ kpyCB’ 75  [4.211  The absorption rate of A and hence the reaction rate of B is then given by -  -  MA  (dC “ dx I, =  D  -  -  —  dN 8 AN4 dt  =  -  fv  dC LAMJ dt  [4.22]  Applying [4.21] gives a kinetic model for the batch reaction dCB/dt  =  5 CA,lt° -(AM/V)(4kDA/3)° CB° 75  [4.23]  Assuming that constant conditions apply to all quantities except CB, integrating yields the result for Russell’s kinetics with mass transfer  or  25 (t) C°  =  25 (0) CB°•  25 (t) CB°  =  25 (0) CB°  -  -  5 CA,t° (AM/V) (4 kpDA/3)° 75 (t14)  [4.24]  kpj.,j t  [4.25)  where kRAI is a lumped rate constant. The dependence on oxygen and indene concentrations is 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  111 ( Eprop + (Ei  -  ) 2 Eterm)/  [4.26]  This treatment assumes that the ratio AM/V remains constant during the experiment. The surface area is unknown but liquid sampling reduces the volume significantly during a long SCR experiment because of the large aliquots needed for POx analysis. This did not appear to affect the results, however, as the SCR data fitted the models as well as the data from the TFU and PFRU, where sampling losses were considerably less significant. Experiments were not performed where the solvent volume alone was varied, but initiated runs of 0.41 mol/L in Paraflex were run at 100°C using different volumes and initiator concentrations. The values of kR /hr [1 mM bP] cf. 0.0 175 [2.5 mM bP]) for 25 1 obtained (0.020 (molfL)°  114  4. Autoxidation of Model Solutions  the different volumes (1.05 L cf. 1.75 L.) show a correlation with AM/V when the effect of the difference in initiators is considered. The lumped kinetic constants were used in comparing reaction behaviour as complete decoupling of the individual components is impossible without a correct estimate of A,, /V. 1  Van de Vusse considered the oxidation of hydrocarbons and proposed a reaction scheme which was zeroth order in oxygen. The rate limiting step in indene autoxidation is the reaction of the peroxy radical with an indene molecule, giving RB  RA  =  =  -  3 (k  +  ) CB [RO 3 k ] 2  =  -  0 R  [4.27]  where [RO 2 J is the concentration of peroxy radicals and R 0 the rate of disappearance of  A. This is zeroth order in oxygen concentration but oxygen is the limiting reactant so R 0 must obey the conditions  0 R  =  0 R  where  CA  0  [4.28]  =  0  where  CA =0  [4.291  >  This approach is the basis of Norton and Drayer’s model (i.e. without mass transfer). Van de Vusse considered the slow and intermediate regimes but did not include the effects of oxygen depletion. The diffusion equation [4.11] can be integrated assuming excess CB and constant [R0 ] to give 2 DA CA  0x R /2 2  =  +  1x B  +  [4.30]  82  Kay and Nedderman considered the general case of absorption with chemical reaction and showed that zeroth order (i.e. simple) chemical reaction kinetics yield complex absorption kinetics for slow or intermediate reactions.They described a simplified result for fast chemical reactions, involving the boundary conditions CA  =  dCA/dx  CA,flt =  0  at  x=0  where CA =0  [4.13] (x  =  xF)  [4.18]  where XE 1S the point of oxygen exhaustion and in a fast reaction lies within the diffusion film (XE  OD). The solution of Equation [4.30], where XE  115  ôj, gives  4. Autoxidation of Model Solutions  -  DA (dCA/dx)o  (2 RODA)° 5 CA,t°• 5  =  [4.311  Applying [4.22] and [4.27] gives the kinetic model for zeroth order reaction with mass transfer; dCB/dt  =  -  ]D)° CA,flt° 2 kjRO (AM/V) (2 5 5 CB° 5  [4.32]  Assuming constant radical concentration, this yields the result  or  5 (t) = CB°  5 (0) CB°  5 (t) = CB°  5 (0) CB°  -  -  5 j 2 D° 5 0 [RO (AM/V) (2 k CA,jflt°• tJ2  [4.33j  kRt  [4.341  where 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 contributions and the overall kinetics become complex. In the fast regime the concentration of B is described by a function in cosh(x) which does not vary significantly across the diffusive layer. The concentration of free radicals is unlikely to be constant at the start of a thermally initiated experiment. If bimolecular decomposition of hydroperoxide is assumed to generate free radicals in this phase, when the concentration of B is relatively constant (low conversion), Equation [4.32] can be written in terms of hydroperoxide; dCB/dt =  -  HI/dt = 2 d[RO  -  (AM/V) (2 5 [RO 0 k H 2 ]DA)° CA,jflt° 5 CB° 5 [4.35]  This yields a hydroperoxide kinetic expression H]° (t) = 2 [R0 5  H]° (0) 2 [RO 5  +  kM,pox t  [4.36]  where kM,p0 5 is a lumped kinetic constant which varies as CA,tflt° 5 CB° . This form of 5 expression was found to fit the thermally initiated POx data very well in the initial phase after the induction period and also provides an explanation for the observed hydroperoxide behaviour. The thermally initiated experiments follow Equation [4.36] initially but as the reaction proceeds, hydroperoxide will be consumed in order to generate radicals and to form polyperoxides which the assay registers as single hydroperoxide units. The rate of increase will thus tail off and once the free radical reaction is fully established the hydroperoxide kinetics become too complex to model. Russell’s kinetics, by comparison, 116  4. Autoxidation of Model Solutions  predict that the hydroperoxide concentration will increase linearly until the consumption of indene becomes significant. An expression similar to [4.361 can be deduced in the absence of mass transfer by assuming that initiation occurs via unimolecular dissociation of hydroperoxide. In this case kM,pOX would vary as 1 5 and the rate of CB .0 rather than CB° indene 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 data were adjusted to exclude the chemical induction period. The regressed rate constants are shown in Table 4.3 and the regression coefficients show that the models fit the data reasonably well. The values of kRM and kR are similar between solvents and show the same trends observed in the gum rate in Section 4.1. The values of kM,poX mirror the trend in kR except for trichlorobenzene, where the contamination side reaction also affected the gum yield 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 shown in Figure 4.2. The values of the regression coefficients in Table 4.3 for the Russell-based model are larger in this set of data, but this trend is reversed in most of the other thermally initiated experiments. The zeroth order model gave a better fit in nearly all chemically initiated experiments. This model also provides a basis for the observed hydroperoxide behaviour and was thus used to compare the indene kinetics in all SCR and fouling experiments.  4.2.3  Gas Phase Resistance Effects in Mass Transfer  The gas phase resistance to mass transfer has been assumed to be negligible in the above treatment of autoxidation kinetics. Two sets of experiments were performed to confirm this and to study the effect of the different modes of oxygen transfer in the SCR and the PFRU reactors. In the SCR, gas is bubbled into a stirred liquid volume while in the 117  4. Auloxidation of Model Solutions  Figure 4.3  Comparison of Mass Transfer Kinetic Models of the Autoxidation of indene in Paraflex  0.8  0.8  0.7  0.7  0.6  0.6  -I .  ‘3)  0.5  0.5  0.4  0.4  0.3  0.3  0.2  0.2k  0.1  0.1  C  0.0  Table 4.3  Solvent  0  5  10 Time (hours)  15  20  0.0  Regression of Thermally Initiated Autoxidation Data to Mass Transfer Models  Rate Constant (Russell)  Regression Fit  Rate Constant (zeroth)  Regression Fit  kRM (mol/L) 25 0 /hr  2 R  kR kmol/L)/hr  2 R  trichlorobenzene  0.01294  0.992  0.01595  0.991  302  0.336  toluene  0.01275  0.960  0.01703  0.945  303  0.812  n-octane  0.01952  0.986  0.02583  0.970  304  1.045  kerosene  0.00408  0.885  0.00706  0.793  305  0.192  Paraflex  0.01924  0.999  0.01958  0.984  306  0.846  TbUlk  100°C; Pair  377 kPa; 0.41 M indene  118  Experiment #  POx Rate Constant (Eqn [4.36]) (v”[meq/L]/hr)  4. Autoxidation of Model Solutions  PFRU gas is passed over a liquid surface agitated by the recirculation of the liquid through the fouling loop. The effect of gas flow rate in the SCR was studied using solutions of 0.41 mol/L indene in Paraflex at Tblk  100°C and Pm,.  377 kPa with 1 mM bP as initiator. Air at  400 mLlmin and 1000 mL/min (ntp) was used, corresponding to 1.5 and 5.25 excess over  that consumed. Table 4.4 shows that the gas flow rate does not have a significant effect on the 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 were performed at gas flow rates of 150-300 mL/min so the surface area so the variation due to bubble surface area should be small. The results indicate that the gas phase resistance to mass transfer is not significant here. Kerosene was not tested at higher flow rates as the gas flows stripped the lighter components of the kerosene out of solution.  The effect of gas delivery configuration was studied using thermally initiated solutions of 0.70 mol/L indene in kerosene at Tblk  100°C and  377 kPa. The  results are shown in Table 4.4 and are relatively similar given the degree of experimental error. The indene concentration data used in calculating the kinetic constants was obtained using the added standard method and involves greater error margins. R 2 for the kR values here is 0(0.95) whereas the Paraflex runs involve R 2  0.998. The similarity in reaction  diagnostics despite the inherent difference in gas delivery is not completely understood but does 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 SCR and the PFRU in order to compare the performances of the two reactor configurations. The PFRU was operated isothermally in these cases, as a batch reactor with mixing provided by the pump recirculation, Thermally initiated solutions of 0.70 mol/L indene in Paraflex and kerosene were used at Tb1k  100°C and P 119  377 kPa. The results are summarised in  4. Autoxidation of Model Solutions  Table 4.4  Gas Phase Resistance Effects in Autoxidation Kinetics  Solvent  Run #  Configuration  0 [indene] (molIL)  Maximum Initial Gum Rate Rate (g/L.hr) ) 5 (x10 ± 0.2 (moWs)  (g/g)  Rate Constant kR J(molIL)/hr  Gum Yield  400 mL/min (bubbled)  140  Paraflex  0.372  1.421  2.8  0.53  0.0361  l000mL/min(bubbled)  141  Paraflex  0.371  1.617  2.6  0.44  0.0399  200 mL/min (blanket)  112  kerosene  0.645  0.942  2.0  0.78  0.0158  200 mL/min (bubbled)  113  kerosene  0.653  0.905  2.3  0.54  0.0143  200 rriL/min (bubbled)  114  kerosene  0.646  1.125  3.0  0.75  0.0149  100°C : 79.2 kPa oxygen saturation  Tbulk  Table 4.5  Comparison of Autoxidation Kinetics in the SCR and PFRU  Experiment #  Reactor  Solvent  (mol/L)  Induction Period (br)  [indene]o  Initial Gum Rate (g/L.hr)  Yield (g/g)  Indene Rate Constant, kR ‘I(moL(L)/br  117  SCR  0.622  1.0  Paraflex  1.9  0.723  0.0161  118  PFRU  0.624  1.0  Paraflex  2.8  0.778  0.0155  114  SCRt  0.646  7.0  kerosene  2.0  0.750  0.0149  112  SCR  0.645  8.5  kerosene  3.0  0.780  0.0158  115  PFRU  0.646  7.0  kerosene  1.4  0.550  0.0113  t bubbled air in SCR: -  ThUlk  =  -  induction period for appearance of gum  100°C: 79 kPa oxygen saturation  120  4. Autoxidation of Model Solutions  Table 4.5 and show significant differences with reactor configuration. The Paraflex results indicate that indene is being consumed at similar rates in the SCR and PFRU but the gum formation rate is significantly larger in the PFRU. This increased gum yield was also observed in the TFU and was thought to be due to the reactor configuration. The SCR features dead zones below the stirrer which allow gum to precipitate out while the fouling rigs are largely free from dead zones. This hypothesis was also supported by the precipitation of insoluble gum in the SCR and the occasional overshoots in gum concentration above g* observed in the fouling rigs after  t’.  The kerosene results show  that the reaction rate in the SCR was significantly faster than in the PFRU, which is inconsistent with the other results. Autoxidation otherwise proceeds at similar rates in Paraflex and kerosene. The two solvents do differ, however, in the gumlPeroxide Number behaviour. The gum concentration in Paraflex and the other solvents in Section 4.1 increases in step with the Peroxide Number until  tK.  The gum concentration in kerosene  lags the increase in POx and thus the induction periods in Table 4.5 are different even though 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 suggests that some oxidation products are more soluble in the slightly more aromatic kerosene.  4.2.4  Oxygen Effects in Autoxidation  The kinetic models in Section 4.2.2 predict that the autoxidation rate varies with dissolved oxygen concentration. This was studied in the SCR using chemically initiated solutions of 0.41 mol/L indene in Paraflex and by Asomaning (1993) using thermally initiated solutions of 10 wt% solutions of indene in kerosene. The Paraflex runs used 2.5 mM bP as initiator at Tblk  100°C and the dissolved oxygen concentration was varied by  adjusting the air pressure in the reactor. The kerosene runs were performed at Tblk and the oxygen partial pressure varied by diluting air with nitrogen. 121  85°C  4. Autoxidation of Model Solutions  The reaction diagnostics from the Paraflex experiments are summarised in Table 4.6. The reaction rates increase with oxygen partial pressure but the rates in run #139 are lower than expected. This run was performed using a different batch of indene and emphasised the need to use the same indene source in series of experiments. The yield of gum increased with oxygen pressure as reported in the literature (Davies, 1961) but the value 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 to Initial gum rate  =  -0.660  +  0.041 P 02 (kPa)  [4.37]  This linear relationship is the product of indene and yield both being functions of oxygen concentration. The zeroth order rate constant, kR, varied as CAjflt  where n  0.69 (R 2  =  0.999) while the rate constant for mass transfer with Russell’s kinetics, kRAJ, varied as n 0.89 (R 2  =  = =  0.999). The individual zeroth order rate constants involved larger regression  coefficients 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 these autoxidation reactions. This result was re-confirmed by Asomaning’s study (1993), where the values of n for the zeroth and Russell models were found to be 0.55 (R 2  =  0.995) and  0.58 (R 2 =0.998) respectively. Asomaning’s study did not yield sufficient gum data to perform a similar analysis on gum behaviour. The Peroxide Number data was fitted to Equation [4.36) and the values were found to depend on CA,EflF 47 which is in good agreement with Equation [4.351. A separate study of thermally initiated autoxidation of 10 wt% indene in kerosene by Zhang et al. (1992) found the initial POx data to follow kR,po  =  x10 P 1.219 10 05 0 2 exp (79 300IRT)  [4.38]  where 0 P 2 is in kPa. This result is in good agreement with Equation [4.32). The indene data fitted the zeroth order kinetic model reasonably well but varied as CA,  ,1 .1 6  The  Peroxide Number in chemically initiated experiments did not obey Equation [4.36], as expected. 122  4. Autoxidation of Mode! Solutions  Table 4.6  Oxygen Effects in the Initiated Autoxidation of Indene in Paraflex  Experiment #  Oxygen Partial Pressure  Oxygen Concentration CA, mt ) 3 (mourn  Initial Gum Rate  Yield  (g/L.hr)  (g/g)  Indene Rate Constant kR (1iV’(molJL).hr)  134  79.2  5.42  2.51  0.706  0.0198  137  50.2  3.43  1.27  0.63 3  0.0153  138  28.4  1.94  0.49  0.343  0.0098  139  99.4  6.80  3.07  0.962  0.0185  0.41 M indene in Paraflex: Tbulk  =  100°C: 2.5 mM bP initiation  123  4. Autoxidation of Model Solutions  The oxygen results confirm that the autoxidation in model solutions is subject to significant mass transfer effects. The experimental apparatus was not designed for a rigorous study of mass transfer with chemical reaction, which was beyond the scope of the current project.  It is worth comparing the rates in the current study with those reported by Russell at 50°C in pure indene. Russell’s kinetic expression gives an indene consumption rate of 0.00384 mol!L.hr for 0.41 mol/L inclene in an inert solvent at 0 P 2  =  79.2 kPa at 50°C.  Russell’s procedures did not report any surface areas which could be used in an estimate of as 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 of 0.0468 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 indene  The thermally initiated fouling experiments involved significant chemical induction periods and scatter in the kinetic parameters so chemical initiators were considered as a means to eliminate the induction period and promote repeatable chemical reaction in the bulk liquid in fouling experiments. Two initiators, benzoyl peroxide (bP) and ABN, were studied in the SCR using solutions of 0.40 mol/L indene in Paraflex at 1 Tb 1 k Pair  =  80°C and  377 kPa at concentrations of 2.5, 5.0 and 7.5 mmol/L. The chemical initiators eliminated the chemical induction period completely and the  Peroxide Number behaviour changed, as expected. The initial POx in bP was close to the initial concentration of initiator and was zero in ABN. The POx then increased almost linearly until it approached a similar plateau value as observed in the thermally initiated experiments, after which it remained relatively constant. The POx did increase slowly after 124  4. Autoxidation of Model Solutions  this in later initiated experiments at higher temperatures, but both patterns mirror the thermally initiated trends and confirm RusselUs result that autoxidation involves the same mechanism under both initiation regimes. The value of g* did not vary with the mode of initiation and the gum analysis results were almost identical. The use of initiators increased the rate of autoxidation significantly and the kinetic parameters are summarised in Table 4.7. Both the gum rate and rate constant kM were more sensitive to the concentration of benzoyl peroxide than of ABN, although the yield of gum was 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 to generate other oxidation products. Russell (1956b) reported that the rate of indene oxidation was proportional to  +  [AJBN]), suggesting that R 0 should vary with the  square root of initiator concentration. This yields a mass transfer model where kR 4 is proportional to initiator concentration; the data in Table 4.7 do not follow this trend. If R 0 is instead assumed to be linear in initiator concentration, Equations [4.33, 4.34  ] predict  that kR 2 should be linear in initiator concentration. This gave (benzoyl peroxide)  2 1 k  =  5 (1 6.888x10  +  461 [bPl)  (ABN)  2 kR  =  5 (1 7.314x10  +  225 [ABN]) R 2  2 R  =  =  0.990  [4.39]  0.988  [4.40]  The initial gum rate did not fit a simple kinetic model, particularly at high initiator concentrations where the gum concentration overshot g* before approaching the saturation limit. The relationship between gum rate and indene kinetics is complicated by the gum being 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 by using lower initiator concentrations in the fouling runs.  125  4. Autoxidation of Model Solutions  Table 4.7  Experiment #  Chemically Initiated Autoxidation of Indene in Paraflex  Initiator  g*  (mmolJL)  (g/g)  Indene Rate Constant kR (lhJ(mol/L).hr)  Rate Constant Fit 2 R  Initial Gum Rate  Yield  (L.hr)  125  0.0  5.5 ±0.5  0.44  0.372  0.00890  0.992  122  0.0  6.0 ±1.0  1.02  0.498  0.00804  0.960  129  2.5 [bP]  6.5 ±0.5  0.94  0.411  0.01 148  0.993  126  5.0 [bP]  7.0 ±1.0  1.38  0.408  0.01506  0.993  128  7.5 [bP]  6.5 +1.0  2.6  0.509  0.01769  0.992  1321-  2.5 [ABN]  7.0 ±1.0  1.18  0.562  0.01742  0.980  131  5.0 [ABNJ  6.0 ±0.5  1.59  0.629  0.01213  0.989  133  7.5 [ABN]  7.0 ±1.0  1.78  0.638  0.01421  0.995  1- problems in SCR air pressure: 0.41 M indene in Paraflex 79 kPa oxygen saturation : Tbulk -  126  100°C  4. Autoxidation of Model Solutions  4.4  Temperature Effects in Autoxidation  The effects of bulk temperature on reaction rates and gum properties were studied using thermally and chemically initiated model solutions of indene in kerosene and in Paraflex.  Figure 4.4 shows the variation of the polyperoxide gum solubility limit, g*, with liquid 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 monatonically with temperature. Equation [2.24] describes the solubility of a solvent free solute as a function of temperature, solvation energy and activity coefficient Y2 The g* data did not ft the expression for an ideal solution (Y2  =  1) very well, as expected for a polar solute in a  non polar solvent. The data could be fitted to [2.11] by manipulating the magnitude of Y2 The estimated value of Y2 necessary to fit the data was large compared to the solvation energy term; the analysis was thus abandoned given the uncertainty in gum and solvent properties (in particular, dipole interaction parameters). The polyperoxide chain lengths are likely to decrease with increasing temperature, which introduces further uncertainty in the calculation of the gum mol fraction. Figure 4.5 shows the variation of g* with indene concentration in Paraflex at 100°C. The gum solubility limit increases linearly with indene concentration and thus the  aromaticity of the solvent. The values of g* in kerosene in Figure 4.5 are larger than in Paraflex 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 by Zhang et at. (1992) using thermally initiated solutions of 0.71 mol/L (10 wt%) indene in kerosene with  =  329 kPa at temperatures ranging from 80-120°C. The Peroxide  Number behaviour shown in Figure 4.6 is similar to that observed by Hess and 1 O H are 127  4. Autoxidation of Model Solutions  Temperature Effects on the Gum Solubility Limit, g*, in Paraflex  Figure 4.4  30 b.O  25  *  20  &  1  C  thermal initiation 2.5mMABN  010 5  0  80  60  100  120  0  2.5mMbP  0  PFRU  A  1.OmMbP  I  I  140  160  180  Temperature (°C)  Figure 4.5  Variation of the Gum Solubility Limit, g*, with Indene Concentration  30 0. 25 *  i  0  Paraflex  •  kerosene  20  0!2  0.8 [indenej (molJL)  128  4. Autoxidation of Model Solutions  Figure 4.6  Temperature Effects in the Thermally Initiated Autoxidation of 10 wt% Indene in Kerosene : Peroxide Number Behaviour 80 70  !:  0..-.  : ‘—. 0  .  0  ‘-...  0  40  -0.-  30  “.-.  81°C 91°C  A-.  121°C  --0--  111°C  20 10  ‘-a’- 101°C 0 0  20  40  60  80  100  120  Time (hours) 10 wt% indene in kerosene; Pair  Table 4.8  Solvent  329 kPa; performed by Zhang and Wilson  Activation Energies in the Autoxidation of Indene in Model So’utions  Initiation  (°C)  Eact Initial Gum Rate (kJ/mol)  Range  Indene Overall  Peroxide  (kJ/mol)  Indene Reaction ER (k.J/mol)  M,POx (kJ/mol)  Gum Yield Range  kerosene  thermal  80-120  31.5 ±3  26.4±3  57.7  73.9 ±4  051-0.82  Paraflex  thermal  80-120  32.7 ±2  30.5±2  51.2  35.0 ±3  0.50-0.66  Paraflex  2.5 mM bP  80-110  36.2 ±3  30.0±2  48.1  0.41-0.71  Paraflex  1.0 mM bP  100-140  47.2 ±4  28.3±2  56.6  0.44-0.71  5 wt% indene in Paraflex: 10 wt% indene in kerosene: 79 kPa oxygen saturation  129  4. Autoxidation of Model Solutions  (1950) in the autoxidation of linseed oil. The POx induction period varied with temperature and 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 peroxy radicals and is unlikely to be subject to mass transfer effects (low reaction rate). The chemical induction period r can then be written as  oc R . An Arrhenius plot of ‘r against 1  inverse temperature gave an activation energy of 79.3 (±5) kJ/mol, which is less than the range 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 activation energy, of 73.9 ±4 kJ/mol for kM,po; this value suggests that the initial generation of hydroperoxide in kerosene is not mass transfer limited. The formation of gum was also seen 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 transfer kinetic models; the indene started to decrease significantly once gum formation started and fitted the model until the POx value declined. The regressed values of kR were used to estimate two activation energies; the overall activation energy, EM, 0 and the reaction activation energy ER which are related by EM,o  =  0.5 Ediffusion  +  0.5 ER  ( Eprop  +  -  ) 12 Eterm)  [4.41]  ER/2 was calculated by plotting kRA/(CA,1 DA) in the Arrhenius plots. The value of ER was found to depend strongly on the diffusivity correlation used; the Hayduk and Minhas correlation was used in the Paraflex calculations and the Wilke-Chang correlation in kerosene. A pseudo activation energy was also obtained from the initial gum formation rates. The estimated values are given with those obtained from the Paraflex experiments in Table 4.8. The yield of gum increased slightly with temperature across the range 0.50-0.72 but there was considerable scatter in this trend. The activation energies of gum formation and indene consumption in Paraflex were obtained using solutions of 0.41 mol/L indene in Paraflex with Pair = 379 kPa and three different levels of initiation. The thermally initiated series was used to compare autoxidation  130  4. Autoxidation of Model Solutions  in Paraflex with the above results in kerosene. A series with 2.5 mM bP initiation was used to compare thermal and chemical initiation while a series using 1 mM bP was used to study the model solution selected for use in the TFU experiments. The activation energies are summarised 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 not behave as in Figure 4.6 but increased rapidly to a plateau value; any increase after this was relatively slow. Gum formation did not lag the increase in POx and regression of the indene data showed that mass transfer effects were evident once gum formation had started. The activation energy for kM,po in Paraflex is almost half that found in the kerosene, which confirms that the initial increase in hydroperoxide in Paraflex is mass transfer limited. The decoupled activation energy was calculated to be 60 kJ/mol, which is similar to the kerosene value. The activation energies shown in Table 4.8 show good agreement within the limits of experimental error. The initial gum rate and kR activation energies in the same temperature range do not vary significantly with initiation mode and solvent nature and suggest that a common autoxidation mechanism applies. The gum yield tended to increase with temperature in the Paraflex experiments and this contributed to the higher gum rate activation energy reported using 1mM bP initiation, where temperatures ranged up to 140°C. The decoupled reaction activation energy ER  0.  + Eprop  -  0. SEterm was  found to be approximately 52 kJ/mol. Howard and Ingold (1962) assumed that the bimolecular peroxy radical termination step activation energy Eterm 0.  + Eprop.  0, giving ER  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 of ER do not reflect the differences in initiation method. Literature values for the propagation activation energy range around 45 kJ/mol (Scott (1965)) giving an estimate of ER as 100  131  4. Autoxidation of Model Solutions  kJ/mol. Howard and Ingold (1962) thus reported an ER value of 96.7 kJImol for styrene oxidation, which is significantly larger than the values in Table 4.8. The source of the discrepancy in activation energies is likely to be the initiation step, which is neither well defined nor understood under the conditions in these experiments. The activation energies were used to calculate mean values of kR for solutions of 0.41M indene in Paraflex at 100°C and Pair = 379 kPa. The ratio of the thermally initiated mean (0.01532 v’(molJL)/hr) and the 2.5mM bP initiated mean (0.0 1921 -./(molIL)/hr) was 1.26, which is close to the value of 1.4.6 predicted by Equation [4.39]. The ratio of the 2.5 mM 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 transfer model as the volume of liquid in the latter experiments was 0.59 that in the former, giving an estimated ratio of 0.7. This ratio was in reasonable agreement within the bounds of experimental error. The mean values of kR in the SCR were compared with the results from the fouling reactors. 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 rn 1 and 5.98 m. The 1 estimation of TFU surface area was considered conservative as it excluded the jets of recirculated liquid inside the holding tank. The 2.5 mM bP initiated SCR mean was smaller than the PERU value under similar conditions (0.0264 /(molIL)!hr)) while the thermally initiated values at 80°C were 0.00863 and 0.008 17 /(molIL)/hr respectively. The estimated 1 (SCR) and 3.85 m (PFRU)) are not consistent with these results AM/V ratios (7.7 m and suggest that the PERU system is more complex than expected.  4.5  Autoxidation in the Presence of Antioxidants  The effect of antioxidants such as the unknown additives in the kerosene experiments was studied using thermally initiated model solutions doped with the 132  4. Autoxidation of Model Solutions  substituted phenol, BMP. The aim was to establish the effect of BMP on the formation of polyperoxide gum, which had been identified as the fouling precursor in the initial fouling experiments. Dopant concentrations of 0, 50, 100,200 and 400 ppm BMP were added to model solutions of 0.41 mol/L indene in Paraflex at 100°C and the results are summarised in Table 4.9. The antioxidant extended the chemical induction period but the Peroxide Number, gum and indene concentrations all behaved as in the thermally initiated base case after the induction period. The kinetic parameters in Table 4.9 confirm that the antioxidant did not affect the autoxidation reaction once it was established. Howard and Ingold (1%2) studied the BMP-inhibited autoxidation of styrene initiated by AIBN and also found that the oxidation rate was constant after the induction period. The concentration of BMP was monitored using qualitative GC-PID analysis and confirmed that the end of the induction period corresponded to the exhaustion of BMP. BMP acts as a peroxy radical scavenger, consuming two peroxy radicals per molecule in consecutive reactions. Assuming that radical generation in the induction period proceeds via reaction [2.14], the steady state assumption for [R0 ] gives 2 1 R  =  2k [RH] [021  =  2 kAox [R0 ] [BMPJ 2  [4.421  The consumption of BMP is proportional to R , which is effectively constant and zeroth 1 order 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 k 1 as 1.22 x10 9 m /mol.s 3 with R 2  =  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 unaffected by the oxidation products of BMP. A similar result is expected for fouling; Asomaning and Watkinson(1992) similarly reported that the fouling behaviour of DCP in kerosene did not seem to be affected by the presence of t-but-catechol apart from the extended induction 133  4. Autoxidatjon of Model Solutions Table 4.9  [BMP]  Autoxidation of Indene in Paraflex in the Presence of Antioxidants  Experiment #  (ppm)  Induction Period (hr)  Initial Gum Rate (g/L.hr)  Gum Yield (g/g)  Tbulk  kR  (°C)  (mo1IL)/hr  Gum Analysis 81 H 9 C 2 0 0  0  310  3.5  0.72  0.490  100  0.01264  501’  307  10  1.41  0.690  100  0.01446  100  311  30  0.90  0.470  100  0.01108  70 H 9 C 2 0 5  200  312  45  0.90  0.444  100  0.01123  72 H 9 C 2 0 3  400  313  86  0.85  0.440  100  0.01480  23 0 5 , 7 H 9 C  100  315  120  -  -  100  318  85  0.30  100  316  11  100  317  4  0.41 M indene in Paraflex: Pair = 377 kPa;  -  8 1 0 5 , 7 H 9 C  80  -  0.511  90  0.01207  82 H 9 C 2 0 1  1.01  0.391  110  0.03895  73 H 9 C 2 0 5  1.99  0.458  120  0.01793  1’ different indene batch  134  -  4. Autoxidation of Model Solutions  period observed in the inhibited solution. Autoxidation fouling in the presence of antioxidants is discussed in Section 5.7. The effect of temperature on doped solutions was studied under similar conditions in order to provide activation energies for comparison with uininhibited autoxidation. The Peroxide Number and gum concentration behaviour did not differ from previous behaviour, but the gum data contained considerable scatter and the pseudo activation energy, 67±10 kJ/mol, is larger than the values in Table 4.8. The indene data was subject to similar scatter and a least squares fit of kR gave an activation energy of 28.2 kJ/mol with 2 R  =  0.315. The results indicate that the autoxidation mechanism is not altered by the  presence of BMP after the induction period. The activation energy of the initiation rate constant was calculated from the induction pçriod data to be 120±3 kJ/mol. This is larger than Zhang’s value for indene initiation in kerosene but is in the range (92-117 kJ/mol) reported by Lloyd and Zimmerman (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 thus estimated as 2.8x10 7 mollm s, which is 3x that calculated from Russell’s data. This was 3 reasonable 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 close  to zero or negative. This confirms that the initiation step changes once the autoxidation reaction is under way.  4.6  A Kinetic Model of Indene Autoxidation  The SCR studies confirmed that the autoxidation mechanism in the fouling experiments differed from the literature due to the scale (mass transfer effects), extended conversion (zeroth order kinetics) and solvent effects. The mass transfer effects limit the direct application of the SCR results to the fouling reactors but the trends observed in the 135  4. Autoxidation of Model Solutions  SCR were all followed in the TFU and PFRU. The SCR results could be summarised into  overall kinetic model for indene autoxidation in Paraflex; 5 CB(t)°  5 CB(0)°  =  -  KM CA, 069 kR([bP) exp (-EM,JRT) t 1  [4.44]  where kR(bP) 2 is given by Equations [4.39) and [4.40) and KM is a lumped kinetic constant. The value of KM at 1 mM bP initiation was calculated as 1.309 x10 6 (m Imol)3 /hr; at 2.5 mM bP initiation, 5.74 x10 069 Imol)-° under thermal initiation, 3x10 3 (m 1hr; 4 69 4 /mol)- 69 3 (m 0 1hr. The concentration of polyperoxide gum in solution cannot be predicted by a simple equation such as  [4.441 as the gum is a secondary oxidation product. The gum  concentration profiles lag the conversion of indene as the primary oxidation product, indene hydroperoxide, is soluble in solvents such as Paraflex. Norton and Drayer modelled this as A  —  B  —b  C  —  D, where B was the soluble hydroperoxide. Such models are  simplifications of the complex chemical reaction mechanisms involved and cannot be related directly to the free radical kinetics.The models are formulated in order to explain the product distribution observed experimentally. The Norton and Drayer model would describe the reaction kinetics in the SCR by the following series of differential equations, which can be solved analytically dC / 1 dt  =  dC / 2 dt  =  C 1 k  dC / 3 dt  =  C 2 k  dC I 4 dt  =  C 3 k  -  C 1 k  [4.45] -  -  C 2 k  [4.46)  C 3 k  [4.47] [4.48]  where C 1 is indene, C 2 indene hydroperoxide and C 34 polyperoxides. This model did not explain the exhaustion of indene observed in some SCR experiments and ignored the effect of decreasing indene concentration over an extended period. An alternative model is proposed which incorporates the autoxidation mechanism as understood from the SCR experiments. This model assumes a series of bimolecular reaction steps, which is a departure from the free radical kinetics involved. The probability 136  4. Autoxidation of Model Solutions  of forming a peroxide of m units is assumed to be proportional to the concentrations of rn-i peroxide and the monomer. This assumes that the concentration of free radicals of species i is in equilibrium with the concentration of species i. Ignoring oxygen effects, the semibatch kinetics are described by a set of series-parallel reaction terms dC / 1 dt  =  dC / 2 dt  =  ek 1C 1  dC / 3 dt  =  1 C 2 ek  dC / 4 dt  =  ek 1 C 3  -  C 1 k  -  -  -  1 C 2 k  -  1 C 3 k  [4.49]  CC 2 k 1  [4.50]  1 C 3 k  [4.51] [4.52]  where e is a propagation efficiency, (1- e) representing the fraction of reacted peroxide that dissociates to form inert products. Chain transfer does not affect the concentration of rn-er polyperoxide so is not included here. Since C 4 are all peroxides, it is assumed that k . 2 2  =  . The equations do not have an analytical solution but the rates can be expressed in terms 3 k  of the observed indene reaction rate. dC 2 1 dC 3 dC dC 1 4 dC 1 dC  (ek 1 C ) C 2 k (-k k C 1 C -k 2 ) C 3 (ek 1 C 2 ) C 3 k C k 2 k ) C 3 1 (-k C 1 C 3 1 ek C ) C 3 (-k k C 1 C -k 2  —  -  —  -  -  -  —  -  —  —  —  where K  -(ek 1 ) C 2 k (C +C k 1 + 2 k ) 3 (eC C -k 2 ) 3 (C + C k 1 + 2 k ) 3 C 3 -ek (C +C k ) 3 1 + 2 k  —  =  -  —  —  —  —  —  —  —  —  —  (KC 2 e) 1 + K(C +C ) [453] 3 -eC 3 K(C ) 2 2 +C 1 + K(C ) [454] 3 3 -eKC 1+ 3 +C 2 K(C ) [4.55]  1 / 2 k . k The reaction stoichiometry where e =1 (no dissociation) is  0 , C 1 a where a =a 1 2  =  =  C 1 a 3 1, a  + =  C 2 a  2, a 4  =  +  C 3 a  C 4 a  +  [4.56]  3. In the presence of dissociation the stoichiometry will  include dissociation products CD2, CD3, C 1 whose concentrations will be governed by equations of the form d CD3 1 dC  —  —  e)k (1 2C 1 C 2 1 (-k k C 2 -k ) C 3 1 C -  -  —  —  (e 1) k C 2 + ( k 2 C 1 k ) 3 +C -  The mass concentration of productj is M .C 1 , where 1 , 2 j. The molecular weights M  —  —  (e 1)KC 2 1 + K(C +C ) [457] 3 -  is the molecular weight of product  3 and M M 4 were assumed to be 148.1, 296.1 and 444.3  respectively. 137  4. Autoxidation of Model Solutions  This kinetic scheme could be expanded to include further features but this would increase the number of parameters to be fitted to the experimental data. As proposed it involves three parameters which are to be regressed from three sets of concentration data per experiment (indene, gum and Peroxide Number). The rate of indene disappearance is described by Equation [4.34], giving dC / 1 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 C 2 2 when C  =  elK; this feature was useful in  selecting combinations of parameters as described below. The model assumes that the reaction scheme does not change during the experiment and was thus unsuitable for thermally initiated studies, where the initiation step changes until the zeroth-order kinetic scheme is established. The model was compared with the results from chemically initiated autoxidations of indene in Paraflex where all three sets of data were available. A FORTRAN 77 program was used to calculate the concentration profiles using a Runge-Kutte-Fehlberg subroutine to solve the set of ordinary differential equations. The routine was tested using Norton and Drayer’s model and gave results in good agreement with the analytical solutions published in the same paper. Concentration units of mourn 3 were used throughout. Mass concentrations were expressed as g/m 3 and the Peroxide Number gave a concentration of (POx)12 mourn . The values of kR in Section 4 were 3 converted 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 Number gave the concentration of the monomeric hydroperoxide alone (i.e POx hydroperoxide concentration (C 2  +  3 C  +  C ) 2 , as the total  ) exceeded POx after a short period. The gum 4 C  results were mass concentrations and were assumed to be equal to C (mass) 3  +  (mass). 4 C  Numerical difficulties arose if K1 m /mol.min as the asymptotic value of C 3 2 was so small 3 cf. [indene] 370 mollm ). 3 (- 1 mourn 138  4. Autoxidation of Model Solutions  Figures 4.7 and 4.8 show the variation in mass and peroxide concentrations as K /mol.min at e 3 was varied from 0.001-0.1 m runs 140 and 141 (Tbulk  =  =  1. The figures show the data from SCR  100°C, Pair = 79 kPa, 5 wt% indene in Paraflex, 1mM bP  initiation) 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 is rapidly 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. 2 at high K is lower and reached very Figure 4.8 also shows that the asymptotic value of C 2 curve shape quickly, whereas at high K the asymptote is never reached. The change in C with K is significant, as the experimental curves resemble those seen at low K; the low values of POx involved, however, belong to the high K region. The gum mass results show that the gum concentration increases in a pseudo-linear fashion, as observed in the experimental data. By inspection, K  /mol.min for these data. 3 0.005 m  Figures 4.9 and 4.10 show the effect of the efficiency, e, on the product spectrum. /mol.min was used which gave values of POx and gum close to 3 A value of K of 0.025 m the data values. Figure 4.10 shows that the model POx curve does not change shape with e ) is not 3 and is different from that observed in the data. The initiator concentration (1 molIm a significant factor in the curve shape. The model POx values are high, which suggests that a larger value of K be used; this would increase the gum formation rate, which Figure 4.9 shows is already slightly high. Figure 4.8 also shows that the gum data rises after an initial lag, which was common to all the initiated experiments at moderate temperatures. The lag in the model gum curves is shorter and thus the model falls to account for this experimental observation. 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 polyperoxide dimer was soluble in hexane and escaped detection in the gum assay, so that C(gum)  =  4 C  alone. The curves in Figure 4.9 were revised and are plotted in Figure 4.11. The figure  139  4. Autoxidation of Model Solutions  Kinetic Model of Indene Autoxidation: Effect of K on Mass Concentration of Gum and Comparison with Data From Runs 140, 141. (e = 1)  Figure 4.7  50000  40000  K  30000  •‘1. -:-o  ---‘-I  20000  .-  -  .4  1,  10000  •  _-....1  --,-----  °  .  0  K  indene 0—gum  =  ‘A ..  0.001  .  -  0  .1  0  100  •  200  Time  300  400  500  (minutes) Kinetic Model of Indene Autoxidation: Effect of K on Peroxide Number and Comparison with Data from Runs 140, 141. (e = 1)  Figure 4.8  :: 50 0  K=0.05  -  0  100  200  Time (minutes)  140  300  400  500  4. Autoxidation of Model Solutions  Kinetic Model of Indene Autoxidation : Effect of e on Mass Concentrations = 0.025 m /mol.min: Data from SCR runs 140,141 3  Figure 4.9  50000  40000  30000  20000  10000  0  0  100  200  400  300  500  Time (minutes) Figure 4.10  Kinetic Model of Indene Autoxidation: Effect of e on Peroxide Number K = 0.025 m /mol.min: Data from SCR runs 140,141 3 40  I.  30  I  I  --0.7 -  -.-.,.-.-.-.-..-.  I  20  .  I  I  300  400  0  I  10  0  0 0  0  100  200 Time (minutes)  141  500  4. Autoxidation of Model Solutions  Figure 4.11  Kinetic Model of Indene Autoxidation: Revised Mass Concentration Curves at K = 0.025 m /mol.min with Data from SCR Runs 140, 141 3  50000  40000  30000  20000  10000  0  0  100  200  400  300  500  Time (minutes)  Figure 4.12  Kinetic Model of Indene Autoxidation: Revised Peroxide Concentrations at K = 0.025 m /mol.min with Data from SCR Runs 140, 141 3 40 e=  30  -  e=Q.9  I  I  .0  6  =08  20-  10  -  ..  -.  -  .s  0  .,.  0  -. -.  I  I  100  200  •  I  300 Time (minutes)  142  400  500  4. Autoxidation of Model Solutions  shows that the data fits the curve for e  =  0.9 within experimental error up to the solubility  limit. The solubility limit has been shown to be a physical parameter and would not be predicted by a kinetic model without suitable modification. The POx curve was treated similarly and Figure 4.12 shows the C 3 curves from the same set of results plotted as POx. The curve at e  =  0.9 shows good agreement with the  experimental data and thus suggests that the Peroxide Number method is measuring the hydroperoxide dimer rather than the monomeric hydroperoxide. This result was found to apply to other sets of experimental data. Figures 4.13 and 4.14 show the revised mass and peroxide concentrations predicted by the model for SCR run 142 (0.41 mol/L indene in Paraflex, Tb1k = 120°C, 79 kPa oxygen saturation), where kM,R was calculated as 0.0734 J(mol/m)/min. The revised concentrations predicted by the model at the mean conditions in the TFU fouling runs (0.41 mol/L indene in Paraflex, Tblk saturation, kAl,R  =  =  100°C, 72 kPa oxygen  0.073403 /(mol/m3)/min) are plotted in Figures 4.15 and 4.16 along  with the data from runs 501, 503 and 504. Both pairs of figures show that the data exhibits the revised trends surprisingly well; the same value of K was used in all the figures, at 0.025 m /mol.min. 3 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  60  minutes and t, taken as 280 minutes. The value obtained, 0.66, was larger than the experimental values (0.4,0.53) but was significantly smaller than 0.9, the value of e which generated the curves. The model thus suggests that the yield is not a true measure of the additionlabstraction ratio. The model gives good agreement with the experimental results when the original set of assumptions have been revised to include a ‘missing’ mono-hydroperoxide and a soluble polyperoxide dimer. The model mechanism may, however, be incorrectly formulated and the observed agreement could be a fortunate coincidence of the numerical solutions and the  143  4. Autoxidation of Model Solutions  Figure 4.13  Kinetic Model of Indene Autoxidation: Comparison of Revised Mass Concentration Curves and Data from SCR Run 142 at Tb1k = 120°C  50000  40000  •  30000  C  20000  0 rIj  10000  0  50  0  100  150  200  250  300  400  350  Time (minutes) Figure 4.14  Kinetic Model of Indene Autoxidation: Comparison of Revised Mass Concentration Curves and Data from SCR Run 142 at Tb1k = 120°C  40  30  e  20  I  0.8  C3  10  0  =  K = 0.025 m3/mol/min SCR run # 142 Tbulk = 120°C: 1mM bP 0.41 M indene inParaflex  0  50  100  150  200  Time (minutes)  144  250  300  350  400  4. Autoxidation of Model Solutions  Figure 4.15  Kinetic Model of Indene Autoxidation: Comparison of Revised Mass Concentration Curves with Data from TFU Runs 501,503, 504  50000  40000  .  a 0  30000  20000  10000  0 500  Time (minutes) Figure 4.16  Kinetic Model of Indene Autoxidation: Comparison of Revised Peroxide Concentrations with Data from TFU Runs 501,503,504 40  30  I  I  20  K  10  0  =  0.025 m3/moLmin  TFU Data from runs 501, 503, 504 0.41 M indene in Paraflex 1.0 mM bP; Tb = 100°C  0  100  200  300  Time (minutes)  145  400  500  4. Autoxidation of Model Solutions  data. The ‘missing’ hydroperoxide hypothesis suggests that the Peroxide Number test was not sensitive to the initial oxidation product and requires experimental verification.  4.7  Ageing of Polyperoxide Gums  Peroxides and polyperoxides are known to undergo degradation at enhanced temperatures 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 the primary foulant precursor in autoxidative systems and that the deposits were composed of polyperoxide degradation products. The polyperoxide gums thus appeared to undergo deposit ageing as described by Nelson (1934). The rates and mechansim of the ageing process were not described in the literature, so the thermal degradation of polyperoxide gums was studied using TGA and the ageing oven apparatus. Gum samples were heated for prescribed periods in a nitrogen atmosphere to avoid combustion and because the deposit phase was likely to be deficient in oxygen. A reliable method of ageing the gums in solution proved elusive so dry samples were used in these studies. The indene polyperoxide gum was obtained from an SCR experiment using a model solution 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 and contained both peroxide and carbonyl groups. On heating, the gum rapidly melted to form a sticky, 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 cooler sections of the PFRU probe. On prolonged heating, the red gum hardened to a dark brown tar which cooled to form a brittle solid which was insoluble in acetone. The colour changes thus 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 experiment at 200°C. The scale is a software construction and the absorbence values are written against 146  —400  —200  0  T=200C  Res= 1cm—i  Wavenumbers (cm—i)  GainOOl  08/26/92 15:40  C)  C)  C) C)—  —.  “-i  C)  C)  tTl  C)  C)  0  ci  C)  0  -t  C) C)  C)  -t  rj  0•  —  4. Autoxidation of Model Solutions  the major peaks. The spectra were generated by R. Lai (1992). The hydroxyl peak (not shown,  3400 cm) in the base gum disappeared rapidly on heating and the carbonyl 1  region becomes more crowded as heating continues. Quantitative analysis of the carbonyl region (1700-1800 cm) gave inconsistent results. Thermal degradation here seems to 1 follow the mechanism reported at lower temperatures, where peroxides react to form carbonyl 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 of the rate of ageing. The loss of material is due to volatile products of condensation and scission which are removed by the nitrogen purge. Figures 4.18 and 4.19 show the reduction in gum mass observed in the ageing oven and TGA at temperatures ranging from 160-240°C. The initially rapid loss of material is followed by a slower decay which continues until the end of the experiment. If ageing corresponded to the exclusion of simple molecular species such as water or C0 , the mass would approach well defined asymptotes 2 (i.e. C 2 H 0 8 H 9 0 2 -  =  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 any trends in C:H:O composition during ageing, confirming this observation. The chemical mechanism of ageing is likely to be quite complex and further characterisation was not attempted. The data from the TGA include an induction period caused by the initial ramping to the analysis temperature. The rates from the TGA are consistently lower than in the ageing oven, which was thought to be due to the differences in mass transfer between the apparatus; the ageing oven sample atmosphere is more vigorously agitated. The initial rate of devolatilisation in the ageing oven was plotted against inverse absolute temperature to give Initial Degra&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 similar expression 148  4. Auloxidation of Model Solutions  Reduction in Gum Mass During Ageing Oven Experiments  Figure 4.18  100 160°C  95 90 85  0  IL  180°C  •  200°C  0  220°C  0  240°C  0  o  0 •• 00 •  8O 75 70  •  0 .00 0 0 0  o  o  6  55  0  50 45 40  .1  0  10  I  20  30  I  I  40  50  60  Time (minutes)  Figure 4.19  Reduction in Gum Mass During TGA Ageing Experiments 100  95 90 85  80 75 70 65 60 55 50 45 40  0  10  20  30 Time (minutes)  149  40  50  60  4. Autoxidation of Model Solutions  Table 4.10  Experiment #  Elemental Analyses of Aged Polyperoxide Gums  Temperature (°C)  Time (minutes)  Carbon (wt%)  Hydrogen (wt%)  Oxygen (wt% t)  117  100  0  74.92 74.63  5.55 5.55  17.53 19.73  401  200  1  75.30  5.58  19.12  401  200  10  76.18  5.18  18.64  401  200  20  76.54  5.20  18.26  401  200  90  77.90  4.72  17.38  402  160  60  76.68  5.13  18.19  403  240  4.5  77.84  4.64  17.52  404  180  20  77.62  5.11  17.27  405  220  18  75.43  4.70  19.87  t  -  oxygen wt% calculated by difference:  * -  Source of gum for ageing experiments  Ageing experiments performed in conjunction with R. Lal  150  4. Autoxidation of Model Solutions  FinalDegradationRate  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 bond energy, 146.5 kJ/mol. This result indicates that other reactions are involved or that mass transfer is involved in the ageing experiments. The activation energy for the initial degradation rate is also considerably lower than the activation energy for the initial fouling rate (84 kJ/mol). The experiments did show that polyperoxide deposits will age under temperatures typical of the fouling runs to produce the deposits reported in Section 5, thus confirming the ageing hypothesis.  4.8  Summary of Autoxidation Studies The batch kinetic studies confirmed that the oxidation of model solutions of indene  is an autocatalytic process and identified several characteristics which influence the performance of batch fouling experiments performed under these conditions. Indene reacts with oxygen under these conditions to form yellow-amber polyperoxide gums 9 [(C ) 2 0 8 H H] as a major product. The solvent can interfere with the indene cooxidation reaction and prevent the formation of polyperoxides, or reduce the yield of these products via chain transfer. The solubility of the polyperoxide products in aliphatic solvents was found to be limited by the solvent nature; the solubility limit, g*, increased with solution aromaticity and temperature and is thought to be a physical parameter of the system. The soluble gum consisted of 2-4 peroxide units and FTIR indicated hydroxyl and carbonyl activity. Polyperoxide was precipitated as globules of darker orange gum after the solubility limit was reached. The thermal degradation of these gums was studied in an inert atmosphere at 160-240°C to simulate the ageing process on a heat transfer surface. The gums melted to form dark redlbrown solids with increased carbonyl activity. The rate of  151  4. Autoxidation of Model Solutions  mass loss increased with temperature; activation energies of 40 kJ/mol and 16 kJ/mol were obtained for the initial rate and final rate respectively. The disappearance of indene featured a chemical induction period which corresponded to the accumulation of hydroperoxides in solution. This induction period was eliminated 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 not influence the subsequent rate of reaction significantly. The activation energy of indene initiation 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 oxygen mass transfer to solution. The process was modelled as mass transfer followed by a fast chemical reaction which was zeroth order in oxygen; the experimental data fitted this scheme reasonably well. Temperature, initiation method, indene and oxygen concentrations were 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 was studied in further detail. This model solution gave reasonably reproducible results but the kinetics were found to depend on the apparatus involved. A model of indene autoxidation was proposed which predicted the trends observed in the experimental data. Further experimental work is required to verify that the model is a true representation of the mechanism of indene autoxidation in these model solutions.  152  5. Initial Fouling Experiments  5.  Initial Fouling Experiments  Asomaning’s study of autoxidation fouling (1990) used model solutions of alkenes in kerosene and demonstrated the usefulness of using chemically simpler systems to investigate the role of different chemical species. Asomaning did not investigate the effect of 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 and operation so a range of model solutions was studied in order to select a candidate for use in the TFU experiments, where flow velocity and surface temperature would be the parameters varied. These initial experiments were performed in the PFRU and the information 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 of chemical initiators, antioxidants, flow velocity and surface temperature. The PFRU studies thus constitute the major part of the current work.  5.1  Model Solution Selection  Asomaning’s experiments used 10 wt% solutions of alkene in kerosene and identified three alkenes which produced significant fouling under the experimental conditions (Tblk  70-85°C, Tsurf = 180-205°C, Re  =  9830); indene, dicyclopentadiene  and hexadec-1-ene. Indene and hexadec-1-ene presented fewer problems in operation than DCP, so DCP was not considered further at this stage. Hexadec- 1-ene was a strong candidate for model solution studies as its autoxidation had been modelled by Norton and Drayer (1968). The four solvents selected were kerosene and three alternatives: an aliphatic (Paraflex), an aromatic (tetralin) and a polar aromatic (trichlorobenzene). The interaction of the different solvents and alkenes provided significant insights into the chemical processes involved in fouling in autoxidative systems. 153  5. Initial Fouling Experiments  The initial model solution experiments are summarised in Table 5.1.  The  experiments were usually started under the thermal conditions used by Asomaning; solution samples were taken and frozen for chemical analysis after the experiment. The soluble gum assay 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% in kerosene) was used in the first experiments. The Reynolds number also varied so the flow rate used by Asomaning was used where possible. The values of Re in Table 5.1 are calculated using the expressions described in Section 3.1 and feature a smaller kerosene dynamic viscosity than that used by Asomaning. Blank runs of solvent were run for 48 hours in order to establish the thermal stability of kerosene, Paraflex and tetralin. No fouling was observed in Paraflex or kerosene and the POx values showed minimal increases after 48 hours. No fouling was observed in tetralin but chemical analysis indicated that significant autoxidation had occurred. Trichlorobenzene was not run as a blank and this solvent caused contamination problems in the PFRU which were repeated in the SCR. Trichlorobenzene developed a grey colour after equilibrating at 80°C overnight and the fouling runs generated deposits which were very different from indene experiments in the other solvents. GC analysis from run 030 indicated that little indene had been consumed although heavy fouling had occurred. The deposit contained some chlorine, even though trichlorobenzene was thought to be inert to autoxidation under these conditions. Extensive cleaning was needed after the trichlorobenzene experiments and thus trichlorobenzene was abandoned as a solvent. The nature of the contamination reaction was not considered further.  5.1.1  Tetralin as Solvent  No fouling was observed in solutions of indene or hexadecene in tetralin, even after the surface temperature was raised to 234°C in run 007. A very thin gum was noticeable on 154  Ui Ui  hexadecene (0.388) hexadecene (0.388) indene (0.878)  kerosene  kerosene  kerosene  Paraflex  Parallex  Paraflex  Paraflex  012  014  013  002  005  009  006  hexadecene (0.466) hexadecene (0.388) hexadecene (0.388) indene (0.840)  yes 8.3  3050 119.6 910 211  81.4  no yes no no yes 48 72 24 45 122 3050 3050 4318 4318 3050 113.6 197.9 161.0 197.0 157.3 937 937 1398 943 1076  204 281 250 280 250  85.8 78.2 83.0 83.0 100.2  yes  1361  183  82.7  33  1368 1521 1521 1640  10690  no no yes yes  yes  138.0  7.5  72 120 178 41  13600  10690 14080 14080 14080  121  137.2 112.4 221.5 227.1  1250  180 180 240 245  180  83.2 101.7 102.7 106.1  80.0  All experiments performed at 377 kPa air overpressure  030  yes  27  13600  120.6  1250  180  tnchloro benzene trichloro benzene  016  82.3  indene (0.95 1) indene (0.41)  tetralin  008  no no no  24 24 48  10721 10721 10721  170.0 212.8 185.0  1404 1404 1470  200 234 205  83.0 78.2 82.8  • tetralin  007  no  48  10721  no no no  Fouling  173.0  1398  204  83.2  hexadecene (0.388) hexadecene (0.205) indene (0.507)  tetralin  004  -  (hours)  Duration  48 48 93  Re  3050 10721 10690  108.7 172.0 145.6  913 1375 1368  202 200 181  84.8 79.9 82.0  -  -  Paraflex tetralin kerosene  ) (kW/m ’ 2  ) (W/m K 2  (°C)  (°C)  (mol/L)  001 003 011  q  Uo  f 5 T  Thulk  Alkene  Summary of Initial Fouling Experiments  Experiment Solvent  Table 5.1  Flow Rate loss due to sampling  Datalogger failure  initiator added initiator addded  Contamination  Contamination  thin matte film  thin gum layer  Colour change  Comments  N  I  5. Initial Fouling Experiments  the PFRU heated section after these experiments but it did not represent a significant fouling resistance. The GC analyses of the samples from these runs indicated that bOth solvent and alkene were undergoing reaction. This was evident from the solution colour which changed from being colourless to a dark orange/red colour. The chromatographs showed that the reduction in area of the tetralin peak (tres accompanied by the growth of a new product peak  =  =  3.1 mm) over time was  5.6 miii), whilst FTIR analysis of  solution samples showed a corresponding increase in carbonyl activity. These results are consistent with the autoxidation of tetralin reported in the literature, where tetralin is first oxidised to tetralin hydroperoxide and thereafter to the ketone (tetralone) or other side products. Figure 5.1 shows the peroxide analysis data. The Peroxide Number increased rapidly to an equilibrium level; this behaviour is consistent with the autoxidation of tetralin reported in the literature. The equilibrium POx value and the rate of tetralin disappearance both decreased when the concentration of alkene was increased. The plateau hydroperoxide concentrations do not correlate with the initial concentrations of tetralin and suggest that the alkenes inhibited the autoxidation of tetralin. This is probably due to chain transfer between the more reactive tetralin peroxy radical and the alkene; Russell (1955) investigated the copolymerisation of tetralin and cumene and reported similar results. Tetralin, as a more readily oxidised species, thus acts as a fouling inhibitor as it interrupts the formation of the alkene polyperoxide chains thought to cause fouling. The tetralin results illustrate the complexity of real systems where many compounds can interact depending on their relative concentrations and activities.  5.1.2  Hexadec-1-ene as Dopant  Asomaning (1992) reported significant fouling from a solution of hexadecene in kerosene but this result could not be reproduced. Asomaning’s method differed from the current work in that the alkene was added to the solvent before the system was warmed up 156  5. Initial Fouling Experiments  Figure 5.1  Peroxide Number Analyses from Initial Fouling Experiments using Model Solutions with Tetralin as Solvent  2000 .  . ‘C  ‘C  15OO x  1000  n  x ‘C  500•  x 0 0  U—  0  •  -  0  0 0  0  0  0 I  10  20  30 Time (hours)  40  50  60  pure tetralin; o 0.507 M indene; x 0.388 M hexadecene; 0.205 M hexadecene -  -  157  -  5. Initial Fouling Experiments  and pressurised, whereas in the current work the alkene was added to the solution after the system had equilibrated at the operating conditions. Fouling was observed in kerosene and Paraflex 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 needed to remove it. The end of the fouling induction period (when thermal fouling became significant) corresponded to a maximum in the Peroxide number, as shown in Figure 5.2 for hexadecene in kerosene (run 014). The figure shows that the fouling induction period was controlled by the bulk chemical reaction. Norton and Draye?s hexadecene autoxidation model links the maximum in Peroxide Number to the generation of significant concentrations of polyperoxide. The chemical analysis and fouling results thus support Asomaning and Watkinson’s hypothesis that fouling in autoxidation systems is caused by polyperoxides. Further hexadecene experiments were performed using an enhanced bulk temperature, (100°C), and a chemical initiator, 0.OIM benzoyl peroxide, to shorten the induction period. Chemical reaction control of the induction period was demonstrated after run 009, where the fouled PFRU probe was quickly cleaned, replaced in the system and the experiment restarted. A much shorter induction period (10 hours) was observed and the probe fouled completely in 20 hours. Extended runs such as run 009 often featured reductions in flow rate as liquid sampling eventually depleted the liquid level in the holding tank; this factor and the physical strain of extended experiments highlighted the need for shorter experiments. Quantitative analysis of hexadecene concentration was not available at this stage.  158  5. Initial Fouling Experiments  Figure 5.2  Fouling Resistance and Hydroperoxide Concentration Profiles from Fouling of Hexadecene in Kerosene in Run 014  0.0004  10  9 8 5-i  0.0003  7 6  .  5 zd i2  c  0.0002  3 0.0001  2  0 0  500  1000  1500  2000  0.0000 2500  Time (minutes) Run 014: 0.388 M hexadec-1-ene in kerosene: Tbulk  159  106.1°C, Tsurf  245°C: Pair= 373 kPa  p  5. Initial Fouling Experiments  Indene as Dopant  5.1.3  Fouling 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 smell during the course of the run. The red/orange coloured gum reported by Asomaning was observed on glass surfaces in contact with the liquid when fouling was occurring on the PFRU heated section. A fouling induction period was observed in both solvents, which corresponded to a maximum in the Peroxide Number. The fouling resistance-time plots then followed an accelerating profile. Figure 5.3 shows the Peroxide Number and fouling resistance results from run 013 in kerosene. The induction period for indene in kerosene was significantly longer than that observed in Paraflex, but the Peroxide Number and fouling resistance profiles were similar. Measurement of indene concentration in kerosene was not qualitative at this stage; quantitative indene analysis in Paraflex was under development and showed a decrease in indene concentration during the run.  ETIR analysis of liquid samples from indene and hexadecene solutions in kerosene and Paraflex showed many common features. Broad hydroxyl peaks of low to medium strength were visible in the 3200-3600 cm 1 range once the Peroxide Number started to increase. These could be caused by hydroperoxides, acids (i.e. hydroperoxide dissociation products) or water generated by condensation steps. The most noticeable absorbances were found in the 1600-1800 cm 1 range, corresponding to carbonyl groups formed from the decomposition of hydroperoxides. These absorbances were complex and difficult to assign, but confirmed that the product slate was consistent with autoxidation occurring in solution.  Indene was selected as the dopant for further study as lower surface temperatures were needed to generate significant fouling resistance measurements. The autoxidation of 160  5. Initial Fouling Experiments  Figure 5.3  Fouling Resistance and Hydroperoxide Concentration Profiles from Fouling of Indene in Kerosene in Run 013  30  0.0012  25  0.0010  20  )  4 c:’  I  15  0.0006  ,  . — ‘-ci  10  .  0.0004  5  0.0002 -------  0  0  200  400  600  800 1000 1200 1400 1600 1800 Time (minutes)  Run 013: 0.878 M indene in kerosene: TbuIk  82.7°C, Tsurf= 183°C: pair  161  373 kPa  0.0000  r  5. Initial Fouling Experiments  indene was also more extensively reported than that of hexadec-1-ene. Hexadec-1-ene presented greater problems for quantitative chemical analysis than indene and indene methods were being further developed for a parallel study of indene fouling in kerosene performed by G. Zhang.  5.1.4  Deposit Characterisation  The deposits formed on the PFRU heated section were photographed then removed for further analysis. Figure 5.4 is a photograph of the fouled PFRU probe after run 014 with hexadecene in kerosene which shows the general pattern of deposit found in the PFRU 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 on glass surfaces. Under these circumstances the unheated sections featured a streaky coating of 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 deposit thickness profile from run 012 (0.388 M hexadecene in kerosene) measured using vernier calipers. The deposit thickness increases as the local surface temperature increases and the thermal boundary layer becomes fully developed. The deposit profile was compared with thermal entry length data at similar Re; the experimental thermal entry length seems to extend further along the probe than expected. The PFRU probe is limited in this case as all the thermocouples are located at the same axial position. The nature of the deposit also changes with axial position; as the distance from the thermal entry point increases and the surface temperature increases, the deposit changes from a veneer to a cracked, powdery solid. Indene deposition profiles were harder to measure as the deposit was generally softer and was compressed or removed by the vernier caliper’s action. Indene deposits were usually formed at lower surface temperatures and were not as carbonaceous as the hexadecene deposits. A thickness measurement probe (Elektro-Physic Köln Minitest 162  5. Inilia! Fontin Experiments  Figure 5.4  Photograph of Fouled PFRU Probe Following Fouling Run 014 Thermally Initiated Autoxidation of 0.388 M Hexadecene in Kerosene __  AIlDWJIJJIII1IJ1}II1llujIITW  V  E  F-  itNV 1tOl  ‘  AtEca4 IN  Figure 5.5  Deposit Thickness Profile of PFRU Probe Fouled by the Autoxidation of 0.388M Hexadecene in Kerosene. 400  ,  350  .  300  I  Healed Section Inlet  FLOW  1  —> 250  I  200 150  100 End of Heated Section  50 I  0  0  10  20  •  I  1  •  30  •  40  I  50  •  •  60  I  70  •  I  80  •  I  90  Axial Distance (mm) Run 012: 0.388 M hexadec4-ene in kerosene: Tbujk  =  163  102.7°C, Tsurf = 240°C: Pair = 373 kPi  100  5. Initial Fouling Experiments  2000) based on the eddy current principle was tested but a suitable calibration was unavailable. The composition of some of the initial deposits is given in Table 5.2 along with an estimate of their mean thermal conductivity based on the approximation Rj=  where  Of is the deposit thickness. The indene deposit analyses in both solvents were in reasonable agreement with the values reported by Asomaning and Watkinson (1992) and the calculated values for indene polyperoxide. The analysis results for the hexadecene deposits show considerable scatter and do not compare well with the calculated values for the 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 are in the range reported for amorphous organic deposits by Watkinson (1988). Values of Aj greater 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 deposit from run 013 and a sample of the yellow/amber gum precipitated from solution in the same experiment. The broad band at 3100-3600 cm 1 in the gum confirms the presence of the OH group which could be caused by -OOH. Both spectra show significant carbonyl activity. The deposit carbonyl distribution is complex and indicates various C=O sources. The other peaks are consistent with the breakdown of indene polyperoxide. The composition and FIIR results indicate that the foulant is an aged product of polyperoxide, as suggested by Asomaning and Watkinson.  5.1.5  Initial Mechanistic Insights  The study of candidate model solution combinations highlighted the importance of the chemical reaction in chemical reaction fouling. The initial fouling experiments demonstrated that a compound likely to undergo autoxidation to form fouling precursors in one solvent could be inhibited by another solvent or competing species. 164  5. Initial Fouling Experiments  Table 5.2 Run  Chemical Analysis of Deposits From Initial Fouling Runs Solvent  Alkene  C  H  0  Rf final  (wt%)  (wt%)  (wt%)  .K/W) 2 (m  (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  0.19  012  kerosene  hexadecene (gum)  65.16  3.20  31.64  0.0044  0.19  014  kerosene  hexadecene  -  -  -  0.0051  0.18  015  kerosene  hexadecene  -  -  -  0.0028  0.23  -  -  hexadecene polyperoxide  75.0  12.5  12.5  -  -  -  006  Paraflex  indene  76.09  5.32  18.59  0.0010  0.19  013  kerosene  indene  82.19  6.33  11.48  0.0011  0.19  73.0  5.4  21.6  indene polyperoxide  165  5. Initial Fouling Experiments  Figure 5.6  FT’IR Spectra of PFRU Deposit and Soluble Gum from Autoxidation Fouling of Indene in Kerosene in Run 013  .ilt  if•’  r1-fl-  P 1  1 ri  III  ‘1  i1  ft  4-  i’ll  I  ‘  -  m]f11}4  I  :  ii  PeroxideGum  I  JI J14  ,J ff  —  —  —  -  r  4  1  -  1688  2800  2400  32G  Frequency (cm-i) -  1  1  U]  j4irJ  I;in  Deposit 1f  [I  11i  194  -‘  jl  :9 [  —  [f  I  ‘-  -  Wfl  F1TrII m C) U  —  S  -  :  1T  r  - -  3288  -1  ) 1 Frequency(cm  -  2400  1688  2000  166  -  1208  800  40  5. Initial Fouling Experiments  The use of chemical analysis in the PFRU fouling experiments provided qualitative evidence that autoxidation occurs in solution during the fouling runs. The analyses confirmed that the fouling induction periods observed in Asomaning’s experiments were caused by the bulk chemical reaction. The onset of fouling is linked to a stage in the bulk reaction associated with a maximum in the Peroxide Number. Norton and Drayer’s model of hexaclecene autoxidation (1968) suggests that the maximum in hydroperoxide concentration corresponds to the appearance of polyperoxide as the major product. The delay between the end of a chemical initiation period, marked by the increase in the Peroxide Number, and the increase in fouling resistance confirms that the fouling precursor is not a primary product of indene or hexadecene autoxidation. These observations support Asomaning and Watkinson’s hypothesis that fouling was caused by polyperoxides rather than hydroperoxides. The deposit analysis results were consistent with a polyperoxide foulant precursor model. The use of a chemical initiator introduced a possible method to eliminate lengthy induction periods if the reaction mechanism can be preserved; Russell (1955b) reported the same reaction pathway for thermally and AIBN initiated indene autoxidation. The use of initiators is discussed further in Sections 4.3 and 5.5. The lack of an aromatic solvent in which to pursue further fouling studies is unfortunate as this would provide a solvent in which indene (the selected dopant) autoxidation products would be most soluble. The role of insoluble species in chemical reaction fouling was reviewed in Section 2.3.2 and an aromatic solvent would have provided valuable information in this area.  5.2  Effects of Dopant Concentration  A series of experiments was performed in order to study the effect of indene concentration on the fouling process. These runs were performed under identical thermal 167  5. Initial Fouling Experiments  and chemical conditions using both kerosene and Paraflex as solvent; this facilitated further comparison with Asomaning’s experiments and later work at UBC and ANL using the indene/kerosene system. Four concentrations of indene in kerosene (0.84, 0.74, 0.40, 0.15 molfL) were run at a bulk temperatures of 80°C, a surface temperature of 180°C and Re=10700 under 79.2 kPa oxygen saturation (as air). All concentrations produced significant fouling deposits and showed similar behaviour to Figure 5.3. The results from both solvents are summarised in Table 5.3. The chemical induction period was defined as the time at which the Peroxide Number started to increase. The Peroxide Number showed sequential features in common with prior autoxidation research, i.e. induction period, linear increase, acceleration, maximum concentration, decrease in concentration. The maximum value varied between 25-35 meqfL (12-18 mmol/L hydroperoxide) but did not show a strong dependence on initial indene concentration. The maxima were accompanied by the appearance of an orange gum in solution which adhered to glass surfaces and the rotameter float. The chemical induction period preceded the fouling induction period and decreased as indene concentration increased. The fouling induction period coincided with the maximum in Peroxide Number; the fouling resistance curves followed an accelerating profile after the maximum and did not show any evident variation in form with initial indene concentration. Four concentrations of indene in Paraflex were studied (0.15,0.41,0.68,0.71 mol/L) under the same thermal conditions, at Re  3050. The chemical induction periods  were shorter than in kerosene and again depended on initial indene concentration. The Peroxide Number and fouling resistance data from run 025 (0.41 molJL indene) are plotted in Figure 5.7 and show similar features to the kerosene runs except that the Peroxide Number shows a gradual rise to a plateau level rather than a maximum. The plateau Peroxide numbers (25-40 meqlL) did not vary significantly with initial indene concentration and were similar to those reported in kerosene. The orange gum seen in the kerosene runs was also evident in the Paraflex experiments. The fouling resistances measured at higher  168  5. Initial Foulinf Exveriments  Table 5.3  Effects of Indene Concentration in Fouling Experiments Solvent  Run  (mol/L)  Chemical Induction Period (lv)  /(meq/L)/hr  jndenej  [indenel Rate constant rate constant fit kR ‘(mo1/L)/hr ) 2 (R  d(VPOx)/dt  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  0.986  024  Paraflex  0.68  10 ±1  0.325 (0.30,0.34)  0.01 152  0.922  025  Paraflex  0.41  26 ±2  0.220 (0.20,0.23)  0.00793  0.995  026  Paraflex  0.15  20 ±1  0.155 (0.14,0.16)  0.00546  0.985  All runs at Tblk  Figure 5.7  80°C, Tsuff  180°C; Re (Paraflex)  =  3050; Re (kerosene)  =  -  -  -  —  -  —  -  —  11 000; P 02  =  79.2 kPa  Fouling Resistance and Peroxide Number Profiles from Indene in Paraflex  40  0.0018  F I  35  —  —.  —-•----  30  Peroxide Number Fou’ing Resistance 0.0014 0.0012  25 0.0010  20 0.0008 15 0.0006 10  0.0004  5 0  0.0002 _:.  0  1000  2000  4000  3000  0.0000  Time [minutes] Run 025: 0.41 M indene in Paraflex; Tbulk  =  80°C; Re  169  =  3050; oxygen  =  79.2 kPa; Tsurf = 180°C  I  a  5. Initial Fouling Experiments  indene concentrations did not increase uniformly but fluctuated after an initial uniform increase; later inspection showed that considerable deposition had occurred. This is discussed further in Section 6.5. The similarity in results suggested that the same mechanism controls fouling in these non-polar, aliphatic solvents.  The accelerating fouling resistance profiles under constant flow and surface conditions must be due to an increase in the concentration of foulant precursor. This was not reflected in the Peroxide Number, however, and prompted the development of the gum assay.  The hydroperoxide induction period data was compared with a simple kinetic model. Indene cooxidation occurs in the absence of significant hydroperoxide concentration, giving the rate expression; ]Idt 2 -d[0  =  [RHj1.5 [O2]° 1 k  [5.11  where k 1 is an overall initiation rate constant. If the end of the chemical induction period is due to the buildup of hydroperoxide to a critical concentration, [ROO1I]*. where unimolecular decomposition becomes dominant; kD[ROOH]*  =  [RH][0 1 k ] 2  [5.2]  where kD is an initiationldecomposition rate constant. Assuming that all oxygen reacted appears as hydroperoxide (i.e. zero polyperoxide or epoxide formation and low conversion), Equation [5.1] gives [ROOH]*  =  1 [RHI1.5 k  t 5 [02]O  [5.3]  Equations [5.2] and [5.3] suggest that the chemical induction period, r, should be proportional to ./([O ]/[RH]), which is counter intuitive in its oxygen dependence. 2 Assuming a constant critical value of [ROOHJ M gives x inversely proportional to [RH]1.5 and a more realistic oxygen dependence. A plot of the chemical induction period data in Table 5.3 did not fit either of these simple models; the induction period decreased linearly 170  5. Initial Fouling Experiments  with indene concentration. The departure of the fouling system kinetics from the literature models prompted the study of indene autoxidation described in Section 4. The initial increase in Peroxide Number was found to obey Equation [4.36] and the plots 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 the SCR at 100°C showed the same trend as observed in the PFRU. The Peroxide Number diverged from Equation [4.36] as it approached the maximum, as reported in the SCR studies in Section 4.1. Figure 5.8 indicates that the rate of indene autoxidation in Paraflex and kerosene is effectively equal following the induction period. The indene concentration data from the Paraflex experiments showed that indene consumption increased significantly once the Peroxide Number had started to increase, confirming that thermal initiation gives way to hydroperoxide-initiated oxidation. The data fitted the zeroth order kinetic model reasonably well but the values of kR in Table 5.3 increase slightly with indene concentration. This variation was not seen in the rate constants for first order indene consumption, kND, (around 0.035 hr’). The discrepancies between the results from the indene concentration runs and the literature prompted a rigorous study of autoxidation in these model solutions. The autoxidation of indene proved to be similar in kerosene and Paraflex, with differences in the chemical induction period. The extended induction periods observed in kerosene were caused by inhibiting additives or components in the kerosene. Paraflex was thus chosen for further fouling studies on the basis of shorter induction periods, a wider range of operating temperatures and ease of chemical analysis. The larger viscosity resulted in smaller Reynolds numbers for given flow rates but this is less important for ‘heavier’ petroleum fluids. The variation in fouling behaviour with indene concentration prompted the use of one concentration, namely 5 wt% indene (0.41 mob’L).  171  5. Initial Foulin2 Experiments  Figure 5.8  Kinetic Fit of Peroxide Data from Initial Indene Fouling Runs to Eqn [4.36]  0.50 0.45 0.40 0.35 0.30  oI-  0.25 0.20 0.15 0.10 0.05 0.00 0.0  0.1  0.2  0.3  0.4  0.5  0.6  0.7  ‘Ilnitial indene concentration, /[RHJ  V’(mo1/L)  172  0.8  0.9  1.0  5. Initial Fouling Experiments  5.3  Fouling in Model Solutions with Two Dopants  The tetralin solvent studies indicated that an active solvent could alter the fouling process by inhibiting the generation of fouling precursor. There was little published work describing the opposite case, namely the interaction of two active species which were known to cause significant deposition individually. It is very difficult to study the synergism of different reaction mechanisms as autoxidation requires chemical conditions which preclude vinyl polymerisation (i.e. high oxygen concentrations). Indene and dicyclopentadiene (DCP) were thus used to study the interaction of two active species under autoxidative conditions. Table 5.4 summarises the results from the three experiments performed under identical flow conditions and thermal initiation. The gum analysis method was developed after run 028 but the gum concentration in Paraflex could be deduced from the Peroxide Number results; other indene runs in Paraflex at 80°C indicated that the increase in gum concentration lagged the Peroxide Number by about an hour. The DCP experiments proved difficult to perform owing to its very strong odour, and caused several difficulties in chemical analysis. The gum produced in runs 03 3,4 was not readily soluble in dichloromethane, which compromised the accuracy of the GC analysis method; DCP also caused a side reaction in the Peroxide Number method. Figures 5.9 and 5.10 show the fouling resistance and Peroxide analysis results from the runs described in Table 5.4. The chemical induction period in all three runs is similar but the fouling characteristics differ greatly. The DCP-only run fouling resistance is almost linear against time and started to increase once gum was generated in the bulk solution. The final increase in fouling rate corresponds to the Peroxide Number reaching a plateau level at 38 meq/L. The gum concentration had reached 4.3 gIL at this point, which was less than g* observed for indene at the same temperature. GC analysis indicated that 17% of the DCP had reacted by the end of the experiment, which is considerably less than  173  5. Initial Fouling Experiments  Table 5.4 Alkene  Fouling from Solutions of Indene and Dicyclopentadiene in Paraflex Concentration  Run  (mol/L)  First order rate constant kND (1/hr)  Chemical Fouling induction period induction period (hi-)  (hi-)  Deposit Analysis  indene  0.410  028  0.0354  5-8  17 ±0.5  954 H 9 C 7 0 4  DCP (polyperoxide)  0.321  033  0.0094  6.5 ±1  9 ±0.5  0 H 9 C 1 28 67 12 H 9 (C ) 2 0  indene DCP  0.415 0.139  034  0.0340  5.5 ±1  19 ±0.5  154 H 9 C 0 0 28  (soluble gum)  PFRU ; Tsurf  H 9 (C 1 0 ) 03 87  80°C; Tsurf  210°C; Re  3050; thermal initiation; 79.2 kPa oxygen overpressure  174  5. Initial Foullnxpeiments  Figure 5.9  Fouling Resistance Profiles in Indene/DCP studies  I  f 200  0  400  600  1000  800  1200  1400  1600  1800  2000  Time (minutes)  Figure 5.10  Peroxide Number versus Time in Indene/DCP studies 50  40 4 TJ  I  o  indene  •  DCP  •  indene/DCP 30  •  20  A  8°  •  •  A  •  A  8  6 A  10 A A  0  .A  0  200  r  400  -  -  600  -  I  800  i  1000 Time (minutes)  175  1200  -  -  I  1400  -  -  i  -  1600  -  -  -  1800  -  -  2000  5. Initial Fouling Experiments  the consumption during the indene run (50%). Asomaning (1990) also observed linear fouling curves with DCP in kerosene. The indene fouling profile shows a transition between an initial, linear fouling rate and the accelerating fouling profile observed in Section 5.2 at lower surface temperatures. The start of the linear fouling profile coincides with the Peroxide Number increasing less rapidly and ends when the Peroxide Number has reached its plateau value of 24 meq/L. The DCP linear fouling rate is significantly larger than the indene rate, despite a smaller initial concentration, in agreement with Asomaning’s results. The differences in the fouling characteristic profiles of indene and DCP indicates that 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 were similar to the DCP result observed in Paraflex, where deposition was detected when the soluble gum concentration in the bulk started to increase. These results suggest that the fouling mechanism for indene in kerosene is linked to bulk deposition rather than a surface reaction process. Solvent effects on fouling behaviour require further study. Figure 5.9 shows that the deposition from the alkene mixture was different to the single alkene solutions. 0.139 M DCP was added to 0.41M indene to observe the effects on the indene kinetics rather than trying to maintain a constant wt% or mol fraction of alkene. The hydroperoxide and gum concentration profiles and the first order rate constant for indene disappearance were similar to those observed in the indene run. The DCP concentration did not change significantly during the experiment. The fouling resistance did not 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 increase in gum concentration, contrary to both single alkene run results. Micrographs of the fouling deposits showed the foulant to be composed of two types of material; a closely packed matrix of small orange/yellow particulates (6-20 jim) 176  5. Initial Fouling Experiments  nearer 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 on the cooler sections; when removed and heated in an oven at 200°C, these blobs resembled the red/orange analogues observed on the hotter section. The material nearest the heated surface was harder, most resistant to removal and insoluble in the acetone which dissolved almost all the other deposit. FuR analysis of the deposits all featured the chemical activity associated with polyperoxide decomposition discussed in Section 5.1. DCP and indene both caused significant fouling in Paraflex. The foulant nature and the sequencing of the fouling resistance and chemical reaction indicate that the fouling process is dominated by the generation of insoluble polyperoxide gums in the bulk solution. The fouling mechanism differed between the two alkenes and their combined fouling behaviour was not the sum of individual contributions. The two species have to compete for the oxygen available in solution, which would affect the chemical reaction generating the fouling precursor. Further synergism studies were postponed until a reliable fouling model was defined in order to interpret the results of such experiments.  5.4  Temperature Effects in Thermally Initiated Fouling  The effects of surface temperature were studied in the PFRU once a suitable model solution had been selected. Fouling experiments were performed under identical chemical and flow conditions (0.4] M indene in Paraflex; Tb1k overpressure; thermal initiation; Re  80°C; 79.2 kPa oxygen  3050) and the initial surface temperature was varied  between 180-240°C. The results are summarised in Table 5.5. The chemical analysis methods provided a monitor of the bulk reaction and Table 5.5 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 results were similar in the two runs involved. The solubility limit, g*, is considered to be a 177  5. Initial Fouling Experimnets  Table 5.5 Initial Surface Temperature  Temperature Effects in Thermally Initiated Fouling Experiments Run  [°CJ  Indene Rate Constant kR [‘J(molfL)/hr]  Chemical Induction Period [hours]  Initial Fouling Rate  Final Fouling Rate  .KJW.min] 2 [m  .K!W.min] 2 [m  180  025  0.00793  41.0 ±2  i.0xi0  10.5x  200  031  0.00861  24.0 ±2  7 1.9x10  5 7.3x10  210  028  0.00849  8.5 ±1.5  7 5.0x10  5 6.9x10  225  032  0.00766  6.5 ±1  7 5.7x10  5 8.7x10  240  029  0.00723  6.0  PFRU Runs with Re  3050; Tbulk  ±  0.5  14.4x 10  -  80°C; 0.41M indene in Paraflex; 79.2 kPa oxygen overpressure  178  Gum Solubilty Limit, g* [gIL] -  4.8 ±0.7 -  5.0  ±  -  0.7  5. Initial Fouling Experiments  solvent-related property and did not vary with surface temperature. The other reaction parameters show that the bulk autoxidation reaction is coupled to the heat exchanger operating conditions. This would be expected under thermal initiation, where the hotter PFRU 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 in the absence of clearly defined rate constants; the rate of radical formation is represented by il-u and gives a pseudo-activation energy of 53 kJ/mol. Zhang et at. (1993) reported a value 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 activation energies for the respective induction period obtained from the SCR experiments in Section 4.5. The variation of induction period with PFRU surface temperature is due to the coupling of reaction and heat transfer in the PFRU apparatus and is considered further in Section 7.1.  An alternative to thermal initiation for further studies was needed which avoided the coupling of the bulk reaction with PFRU operating conditions and the extended induction periods at lower surface temperatures. This prompted the investigation of free radical initiators described in Section 5.5.  The fouling resistance profiles are plotted in Figure 5.12 and indicate the two fouling regimes discussed in Section 5.3. The abscissa, adjusted time, refers to the period after evident chemical reaction has started (given by the chemical induction period, -u). The transition between the two regimes is dramatic; the fouling resistance increases initially in a linear fashion then accelerates to a rapid rate, when the run was terminated to save the probe. The initial fouling rates in Table 5.5 were obtained by drawing a line through the data points following the time when the fouling resistance had become significant 179  (‘-j  4x  5. Initial Fouling Experiments  Temperature Effects in Thermally Initiated Fouling Experiments in Paraflex  Figure 5.11  o  0 Initial Fouling Rate -6  A •  -8  Final Fouling Rate 1/Induction Period  -10  A  I-12 -  A  A  14  -18 0.0019  0.0020  0.0021  0.0022  0.0023  llTsurface (K) Figure 5.12  Effects of Surface Temperature on Fouling Resistance Profiles in Thermally Initiated Solutions of Indene in Paraflex Runs 025, 028, 029, 031, 032 -  c).0020 Surface Temperature  0.0018 0.00 16 .1)  225°C  0.0014  a  200°C  0.0012  +  180C  1 + +  210°C  0.0010  :  240°C  •  +  B B  0.0008 0  1  + + + +  a  * a  •  0. 0006 0.0004  0.0002  +, I  0.0000  0  —-  250  I +  500  750  1000  1250  Adjusted Time (minutes)  180  1500  1750  2000  5. Initial Fouling Experiments  i0- m .KJW in Figure 5.12). The data in this region was reasonably linear and estimates 2 of fouling rate could vary by  ±  20%, or more at lower temperatures where the fouling rate  was small. The final fouling rates are calculated from the last data points in the accelerating stage. Figure 5.11 shows that the final fouling rates do not show a strong dependence on surface 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 of the range reported for chemical reaction fouling in the literature. The value indicates that chemical 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 10 wt% indene in kerosene under similar conditions, while Crittenden and Khater (1984) reported an activation energy of 70 kJ/mol for evaporation fouling caused by kerosene autoxidation. The magnitude of the final fouling rate and its weak dependence on surface temperature indicate that it is controlled by conditions in the bulk liquid rather than the surface reaction zone. The differences between the initial fouling regime and the accelerating regime indicate that different processes are involved in the generation of deposit. These processes are likely to be linked to changes in the chemical reaction as autoxidation continues in the bulk liquid. The importance of the bulk chemical reaction prompted the investigation of model solution autoxidation described in Section 4. Figure 5.13 shows how the change in deposition mechanisms is related to the bulk solution chemistry. No fouling is observed prior to generation of gum in the bulk liquid, indicating that the surface reaction zone is not the only region involved in the formation of deposit. The linear, initial fouling regime corresponds 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 heavy fouling 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. 181  5. Initial Fouling EperimenIs  Figure 5.13  Gum Concentration and Fouling Resistance in Thermally Initiated Fouling  6  0.0010  5  0.0008  4 0.0006 3 T’l  t  0.0004  2 0.0002  0  0.0000 1200  Adjusted Time (minutes) Run 028: T 8  225°C; Tbulk  80°C; Re  3050; 0.41M indene in Paraflex  182  P  5. Initial Fouling Experiments  The variation in the initial fouling rate with inclene concentration could not be reliably deduced from the runs in Section 5.3 as these were performed at relatively low temperatures (180°C). The transition in fouling mechanisms was confirmed by examining the deposit formed at the different surface temperatures. All deposits showed the changes in deposition pattern with surface temperature described in Section 5.1. The foulant morphology depended on the surface temperature and the length of a run. Lower surface temperatures, where fouling followed an accelerating profile alone, produced very chunky, soft deposits. Optical microscopy indicated that the surface was covered at random by large blobs of a dark red material in a disordered structure, with greater surface coverage at higher surface temperature. This material lay on top of a hard dark brown deposit which was densely packed and much harder to remove. The chunky material was soluble in acetone but the hard material had to be removed mechanically. The blobs ranged in diameter from 50 to 100 jim and looked as if they had melted in to the deposit. The larger blob dimensions are comparable to the depth of the thermal boundary layer in these experiments and suggest that these larger blobs are agglomerates of insoluble gum formed in the bulk solution. Deposit from the enhanced surface temperature runs was significantly different; the foulant at the hotter end consisted of an ordered array of small gum particulates, 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 in nature down to the dark amorphous solid formed at the PFRU surface. Agglomerated gum particles could be seen to lie on top of this denser matrix in runs where accelerating fouling also occurred. The veneer formed on the less hot PFRU surface suggested that blobs had adhered to the surface and melted in to the veneer. The foulant morphology thus indicated a deposition process involving precursor solubility, as hypothesised by Crittenden et al. (1987b) in their polymerisation fouling studies. Estimates of the deposit thermal  183  5. Initial Fouling Experiments  conductivity using Af = b/Rf ranged from 0.1-0.2 W!m.K, in agreement with earlier results. Table 5.6 summarises the data from elemental analysis of the deposits. The C:H:O , the foulant 2 0 8 H 9 results are compared with the result for indene polyperoxide, C precursor. The deposits from kerosene and Paraflex were similar in composition as well as morphology. The deposit shows reduced oxygen content, particularly in the material exposed to the higher temperatures near the probe wall. FTIR analysis of the deposits indicated no hydroperoxide functionality but a complex mixture of aromatic carbonyl groups generated by the degradation of peroxide linkages. Deposit recovered from the probe wall contained less oxygen and gave very complex FTIR spectra, indicating that the foulant undergoes an ageing process once deposited. The FTIR spectra were very similar to those reported by Lambourn and Durrieu (1983) in their studies of crude oil fouling in the presence of oxygen. The deposits from the chemically initiated runs are discussed in detail in Section 5.5.  The fouling resistance data and the deposit morphologies indicate that fouling in these batch autoxidation experiments involves a transition between two deposition mechanisms; 1.  Linear Fouling Regime  Polyperoxide formation and attachment involving bulk and reaction zone phenomena;  2.  Accelerating Fouling  Bulk precipitation of polyperoxide, dominated by bulk kinetics and mass transfer  Transitions 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 is surpassed by the generation of agglomerates in the bulk when the bulk temperature reaches a critical value for protein denaturation. Fouling from asphaltenes in crude oil has been described as a solubility phenomenon by Eaton and Lux (1984) and by Scarborough et al. 184  5. Initial Fouling Experiments  Elemental Analysis of Fouling Deposits  Table 5.6 Indene Concentration [mol lU  Initiation  Solvent  Surface Temperature [°Cj  0.40  thermal  kerosene  180  9 H C 727 0130  Deposit surface  0.41  thermal  Parallex  180  9 9 C H 67 °0.97  Less hot surface  180  101 01.16 H 9 C  Deposit surface  180  9 C  PFRU Wall  240  9H C 98 01.77  Gum from Solution  240  98 C H 98 00.87  Deposit surface  240  9 Hg C 97 00.75  PFRU Wall  0.41  thermal  Paraflex  Elemental Analysis  133 0056  Origin of Deposit  0.41  chemical  Paraflex  240  9H C 104 Oi  300 minutes  0.41  chemical  Parallex  240  99 C H 24 01.03  400 minutes  0.41  chemical  Parallex  240  9H C 913 °094  480 minutes  9H (C 8 02)nH  Indene polyperoxide  -  -  Table 5.7 Run  ..  Effects of Bulk Chemical Parameters on Fouling and Reaction Behaviour  Tbulk  Re  Initiation  (°C)  Induction Period (hr)  kR ((mol 7L)ihr)  028  80.0°  3050  thermal  14.5  0.00849  031t  80.0°  3050  thermal  24.0  035  100.0°  3295  thermal  038  80.7°  3050  039  80.8°  042  Gum Yield  g*  (g/g)  (giL)  Initial Fouling Rate (x10 ) 7 .KlW.minj 2 [m  Deposit Analysis  5.01  180 H 9 C 0 0 09  -  -  0.00861  0.59  4.8  1.92  183 H 9 C 0 0 03  1.6  0.01574  0.50  9.3  5.6  187 H 9 C 0 0 08  2.5mMABN  1.0  0.01041  0.45  5.5  3.01  H 9 C 1 O 06 20  3050  2.5mM bP  1.0  0.01439  0.41  5.0  2.59  H 9 C 1 O 04 26  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  t  Tgf  200°C, included for diagnostic comparisons  185  -  -  H 9 C 1 0 01 25  5. Initial Fouling Experiments  (1979), who reported deposits generated from small organic particulates. Further work was concentrated on mechanism (1) as this corresponds more closely to conditions in commercial heat exchangers. There are considerable experimental difficulties involved in further study of mechanism (2) owing to the insoluble nature of the precursors.  5.5  Velocity and Surface Temperature Effects in Chemically Initiated Fouling  Chemical free radical initiators were investigated as a means of eliminating the chemical induction periods and scatter in reaction parameters observed in the thermally initiated experiments. This could also be achieved by increasing the bulk liquid temperature, which would in turn increase the operating Reynolds number away from the transition regime threshold (Re  2300). Table 5.7 is a summary of the results from a  series of runs performed to seek an optimum set of operating conditions for further fouling studies. The optimum criteria were chiefly those of experimental feasibility; negligible induction period, reasonable run lengths and fouling behaviour similar to that observed in Section 5.4. The runs were performed using solutions of 0.41 mol!L indene in Paraflex under 79.2 kPa oxygen overpressure in the PFRU operating at a surface temperature of 210°C except where noted and at Re  3000. The SCR experiments had showed that 2.5  mmol/L was a suitable concentration of initiator and this was used for the rest of the initial fouling experiments. Reaction diagnostics are reported as described in Section 4.  Both chemical initiators (benzoyl peroxide, ABN) eliminated the chemical induction period and increased the rate of reaction, as observed in the SCR. The initiated runs at 80°C gave similar initial fouling rates to the thermally initiated run but the experiments were deemed to be long, lasting over 20 hours. The initiated runs at 90°C were repeated because the PFRU gave unusual fouling resistance profiles; Rf increased linearly while the gum 186  5. Initial Fouling Experiments  concentration increased and levelled off, as observed previously, but then decreased or fluctuated rather than accelerating. This was thought to be due to spalling but inspection of the probe showed that a heavy deposit layer had collected; this deposit contained larger agglomerates of insoluble gum than observed previously and presumably yielded a negative fouling resistance via surface roughness effects as described for particulate fouling by Crittenden and Alderman (1988). The accelerating Rf profile was not observed at 90°C despite several attempts. The runs at 100°C showed the same trends observed at 80°C; the chemically initiated runs involved increased reaction rates and slightly higher fouling rates, while the fouling resistance profile remained similar. The initial linear fouling rate operated until after the gum solubility limit, g*, was reached and was followed by the accelerating profile. The increased reaction rate at higher temperatures meant that the initial rate period was shorter than that observed under thermal initiation. These results confirmed that operating at 100°C with chemical initiation did not change the initial fouling rate mechanism whilst yielding shorter runs with advantages in control and Re range. The reaction parameters listed in Table 5.7 and the raw data showed that the autoxidation reaction in the PFRU followed the same trends observed in the SCR in Section 4. Chemically initiated solutions featured enhanced indene reaction rates and slightly lower gum yields compared to thermally initiated runs. Activation energies for thermal 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. This  trend was also seen in the activation energies under benzoyl peroxide initiation; 52±6 kJ/mol for kR and 59±9 kJ/mol for the initial gum formation rate. The SCR studies also reported increased activation energies for initiated solutions. The error estimates are quite large and reflect the difficulty in achieving experimental reproducability in a larger reactor with different indene batches. The presence of the hot fouling probe could also be responsible for the larger activation energies observed in the PFRU Fouling runs. The gum solubility limit, g*, shows reasonable reproducability at each bulk temperature irrespective 187  5. Initial Fouling Experiments  of the mode of initiation. One notable feature was the tendency for the gum level to overshoot g* in later fouling experiments; this irregularity was not seen under thermal initiation and suggested that a lower initiator concentration would be preferable in the subsequent TFU runs. The foulant composition and morphology were similar across the range of conditions used and confirm that the same fouling mechanism is involved. The sequence of fouling events was also similar to that observed in Section 5.4, though the timescale was shortened as reaction rates increased. At higher bulk temperatures, insoluble yellow gum tended to collect on the unheated surface of the PFRU probe; this was soluble in acetone and its FEIR spectrum indicated that it contained indene poiy- and hyciroperoxides. The initial fouling rates in Table 5.7 increase with bulk temperature, which excludes a simple surface reaction fouling model. No further conclusions were drawn as the surface shear stress and residence times also varied in these runs. Panchal and Watkinson (1993) reported similar results for indene in kerosene, although their experiments were performed at constant Urn rather than Re.  Bulk temperatures of 100°C and 2.5 mM benzoyl peroxide initiation were thus used in subsequent fouling studies. Each set of runs used the same batches of indene and Paraflex to reduce variations in the bulk reaction caused by different indene batches observed in the SCR. Table 5.8 sunimarises the reaction diagnostics; the mean initial gum rate was 2.5 ±0.2 g/L.hr, mean kR was 0.0237 ‘./(molIL)!hr and the gum yield varied from 0.52-0.77. Figure 5.14 is an Arrhenius type plot of the initial fouling rates obtained from runs at Re  =  3295 where Tsurf ranged from 210°-255°C. Figure A.2.1 shows the fouling  resistance profiles from the individual runs. The initial fouling rates were calculated as before and increase with surface temperature, fitting an expression of the form dRf!dt  =  977 exp  [- (81 900 ±16 400)/RT(K)] 188  [5.5]  5. Inital Foulin,ç’ Experiments  Table 5.8 Run #  Reaction Diagnostics from Chemically Initiated PFRU Fouling Experiments Tsface  [indene] 0 (molIL)  Re  (CC)  4 k ((mol/L) /hr)  Gum Yield  2 R  Tbulk  (g/g)  (kM)  (°C)  040  207.1  3295  0.331  0.02869  0.613  0.998  100.4  043  224.6  3295  0.340  0.03325  0.772  0.987  100.4  044  239.5  3295  0.361  0.03297  0.524  0.996  100.2  050  247.4  3295  0.301  0.02205  0.535  0.988  100.3  055  252.4  3295  0.295  0.02226  0.736  0.990  100.4  049  254.6  3295  0.365  0.02342  0.682  0.992  100.6  045  222.1  4920  0.372  0.02526  0.527  0.994  100.6  047  218.0  6513  0.368  0.02741  0.603  0.998  99.9  056  246.7  1020  0.365  0.02485  0.700  0.984  100.0  054  247.0  1920  0.302  0.02271  0.593  0.977  100.2  053  246.7  6513  0.3 12  0.02496  0.630  0.979  100.2  Model solution: 5wt% indene in Parallex at 100°C; 79.2 kPa oxygen blanket; 2.5 mM bP initiator  Figure 5.14  Surface Temperature Effects on Initial Fouling Rate in Thermally and Chemically Initiated Fouling Runs -10  • 0  1]  s  -  -14-  Thermal Initiation (1 00°C)  ••  N. &N0 ‘N  -15:z!.  Chemical Initiation (iO0C) Thermal initiation (80°C) Zhang et al. (kerosene)  -,  0-’ -16-17-  —18  I  ..  .  .  ‘_  ..  —‘——‘  0.0018  0.000  0.002  0.0024  0.0026  lIT surface (K) Thermal initiation  -  Tbulk  80°C, 100°C: Chemical initiation  189  -  Tbulk  100°C, 2.5mM benzoyl peroxide  5. Initial Fouling Experiments  The activation energy is similar to the value reported for thermally initiated fouling (84.8 kJ/mol) but is subject to a larger estimate of error due to the scatter in the data. The initial fouling rates from thermally initiated runs at TbUlk  100°C were higher than those reported  at 80°C at the same surface temperature but less than those reported in the chemically initiated runs. The increase in initial fouling rates could be due to the increased rate of reaction in the bulk liquid, the reduced  Urn  used at 100°C or the change in solution  properties with bulk temperature. The flow velocity at 100°C (0.506 m/s) was smaller than that at 80°C (0.687 mIs) but the ratio of chemical/thermal fouling rates is  3.75, which  suggests that other factors are involved. The experimental data indicates that the initial fouling rate is linked to the bulk reaction rate; Panchal and Watkinson (constant Urn) and the runs described above (constant Re) both reported increased initial fouling rates with increased bulk temperature and hence bulk reaction rate. This result would be consistent with a reaction zone fouling model such as those described by Panchal and Watkinson, and Paterson and Fryer. Figure 14 also shows the data reported by Zhang  et  al. (1993) for  fouling from solutions of lOwt% indene in kerosene in the PFRU under similar chemical conditions (TbUlk  86°C, thermal initiation, 86 kPa  oxygen overpressUre)  but at Re  11000. The activation energy in the kerosene solution was reported as 39 kJ/mol, which is significantly smaller than that observed in Paraflex. The deposits obtained from the chemically initiated runs showed the same morphologies and composition trends as observed in the thermally initiated runs. The runs at highest surface temperature involved greater surface coverage, consistent with the development of a thermal boundary layer and a local surface temperature effect in deposition. Photo-micrographs of the deposits confirmed the link between the onset of accelerated fouling and the appearance of large dark red particulates of insoluble gum.  The effect of flow velocity was studied at two different surface temperatures (—222 and 247°C) using different batches of indene in each series. The initial and final fouling 190  5. Initial Fouling Experiments  rates defined in Section 5.4 are summarised in Table 5.9 and plotted on log axes in Figure 5.15. 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 and gum concentration profiles for each series and confirm that the onset of accelerated fouling coincides with the gum level reaching the solubility limit, g*, of 11 gIL. The initial fouling rates decrease with increasing flow velocity while the final fouling rates do not, confirming the 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 figure shows that the rates could be fitted to Red, where -2< n <-1. Zhang et al. (1993) observed similar scatter in their data from indene in kerosene and found that n  -1. The initial  fouling rate was thus proportional to exp (-E/RTsurfe) u ; Paterson and Fryer’s model predicted n  =  -1, whereas Epstein’s model (1993) predicts n  -1.75 when chemical  reaction controls the fouling process. Fouling models are discussed further in Section 7.2. The final fouling rates at different surface temperatures should not be compared directly as they are limited by the maximum surface temperature on the PFRU probe. It is evident, however, that the final fouling rate increases as Re exceeds 3000 and the flow enters the turbulent transition regime. Bulk deposition processes are expected to be faster under such conditions due to the contribution of turbulence to mass transfer rates. The deposit morphology did not change noticeably across the range of flow rates studied and deposit analyses confirmed that the foulant chemistry was consistent with the ageing of polyperoxide gums. The thermal entry length on the PFRU probe was shorter at larger Re, increasing the proportion of the probe surface covered by deposit.  5.6  Stages in Autoxidation Fouling  A series of runs was performed under identical chemical and flow conditions in order to study the sequence of events in the fouling experiments. These were performed 191  5. Inital Fouling Experiments  Log-log plot of Velocity Effects in Chemically Initiated Fouling  Figure 5.15 1000  Final, 222°C 0  Final  240°C  0  0  0  100  A  o  A  Initial, 222 C  •  Initial, 240°C  A  A -  A  .  1000  10000  Reynolds Number  Table 5.9 Run#  Effect of Flow Rate on Initial Fouling Rate in Initiated PFRU Experiments Re  urns  Tsurf  Uref  kR  (mis)  (°C)  .K) 2 (W/m  (rnol/L)/hr  Initial Fouling Rate 7 (x10 .K!W/min) 2 (m  Final Fouling Rate ) 7 (x10 .K/W/rnin 2 (m  043  3295  0.506  224.6 220  790  0.03325  33.3 27.2t  154±20  045  4920  0.756  222.1  1018  0.02526  9.3  200±20  047  6513  1.000  218.0  1245  0.02741  5.6  200±20  050  3295  0.506  247.4  814  0.02205  15.2  90±10  053  6513  1.000  246.7  1262  0.02496  7.4  150±20  054  4920  0.756  247  1053  0.02271  10.3  158±20  055  3295  0.506  252.4 247t  800  0.02226  26.7 21.9t  87± 10  056  4020  0.617  246.7  922  0.02485  12.0  169±15  t estimated using activation energy 83 kJ/mol -  192  5. Initial Fouling Experiments  under severe fouling conditions (Tblk  100°C, Tsurfo.ce  =  240°C, Re  3295) using  thermally initiated solutions of 0.41 mol/L indene in Paraflex. The reaction diagnostics are summarised in Table 5.10 and show similar scatter in reaction diagnostics as those in Section 5.5. This scatter is reflected in the initial fouling rates in Table 5.10 The runs were stopped at the start of the initial fouling rate period; at the end of the initial fouling period, when the gum concentration had reached g*, and during the accelerated fouling period. No fouling was observed until gum was observed as cloudiness in the bulk solution. Figures 5.16.a-c are photomicrographs of the surface deposit in the heavily fouled section and show distinct changes in deposit morphology during the fouling process. The deposit in Figure 5. 16.a appeared as a smooth, even veneer to the naked eye but is clearly composed of striations of gum in the direction of flow. These striations are composed 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 22 jim. The deposit in Figure 5. 16.b appeared as a smooth red/brown material and can be seen to consist of many gum globules (15-20 jim) which have adhered to the surface and melted into the deposit. More large globules of insoluble gum (30 jim) are visible. The deposit in Figure 5. 16.c was a relatively uniform, rough, dark brown solid; the figure shows that this consists of larger globules of red insoluble gum and agglomerates (size 4465 jim) on top of a fine particulate matrix (size  10 jim) which is presumably the aged  polyperoxide gum deposited earlier. Scraping off the earlier veneers often uncovered thin layers of this orange matrix material. Figure 5. 16.d shows the smooth orange veneer found in the thermal entry length of the 500 minute run; melted gum globules are evident in the veneer. The surface beneath shows the darkened (aged) early striations and the scratched surface of the PFRU probe. The scratches in the veneer surface occured during cooling and were observed in the ageing studies in Section 4.7 These interrupted runs confirmed that the change in fouling mechanism is due to the onset of bulk precipitation of insoluble gum agglomerates. The deposit formed in the initial 193  5. Initial Fouling Experiments  Table 5.10  Reaction Diagnostics and Fouling Summary of Interrupted Fouling Runs  Duration  Induction Period  (minutes)  (minutes)  (°C)  v’(mol/ L)lhr)  300  120 ±50  237.7  0.0302  400  120 ±30  237.4  500  90±30  236.6  Tbulk  100°C; Re  kR  3295; P  Gum Yield  Fouling Induction Period (minutes)  Final Rf ) 3 (x10 JJr 2 (m  Initial Fouling Rate ) 7 (x10 K/Wmin) 2 (m  Deposit Analysis  0.464  240 ±18  0.06  7.69  H 9 C 1 0 04 01  0.400  0.0213  0.439  240 ±18  0.10  5.67  903 H 9 C 1 0 24  0.360  0.0337  0.446  180 ±15  0.75  6.66  C9H9130J94  0.401  79.2 kPa oxygen overpressure; thermal initiation  194  [indene]  (moIIL)  5. In i/ia! Fouling  ExpertmenIs  Figure 5.1 6a,b Photographs of PFRU Deposit from Interrupted Fouling Runs using Thermally Initiated Solutions of Indene in Paraflex Fig. 5. 16 a (t  Fig 5. 16 b (t  =  =  300 minutes, 272x)  400 minutes, 272x)  ‘J:  ..  ,  c..  _s  —. . ..  w__’•  L  —  ...‘  ‘  i..:-  ----..-  —  -  ‘-  .—  .-*  -..  -J  3• _4.  --‘.——.  -  -  195  —.  ——  -‘  5. Initial Fouhin2 Experiments  Figure 5.16c,d Photographs of PFRU Deposit from Interrupted Fouling Runs using Thermally Initiated Solutions of Indene in Paraflex Fig 5.16 c (t  Fig 5.16 d (t  =  500 minutes, 272x)  500 minutes, 272x)  5. Initial Fouling Experiments  stages of fouling consisted of smaller gum globules or veneers which were derived from gum melting into the deposit as part of the ageing process. The initial fouling mechanism thus involves gum insolubility and the adhesion of gum globules to the surface. Once adhesion occurs, the polyperoxide gum ages to the darker coloured carbonyl materials discussed earlier. The analysis results in Table 5.10 show that the deposit is low in oxygen compared to its parent polyperoxide gum at even the earliest stage of deposition. This is consistent with an adhesion step involving a chemical reaction, which explains the large activation energy reported for the fouling rate. The estimated values of A 1 ranged around 0.18 W/m.K, which agreed with the values reported earlier and suggested that the ageing process did not alter the deposit thermal conductivity. The globule sizes in the initial fouling phase (<20 jim) can be compared with the estimated momentum and heat transfer boundary layer thicknesses. The former can be estimated fromy = 5, giving Omom  =  153 jim; the latter from 1 O her,1  the thermal conductivity of the liquid, giving ôtIzernwl  =  =  , where A is 0 A/U  148 jim.; the latter figure is  expected to be smaller as Pr> 1. The gum globules could thus be formed in the viscous sublayer or the bulk solution. Several attempts were made to measure the size of gum globules in solution but these were complicated by the tendency of the gum to agglomerate on cooling. Insoluble gum globule sizes were found to lie in three ranges; 20-24 jim, 4570 jim and 110+ jim, the latter including very large gum agglomerates up to 1 mm. The soluble gum size estimates seemed to fit a similar distribution.  5.7  Fouling in the Presence of Antioxidants  The ability of antioxidants to suppress autoxidation in model solutions of indene in Paraflex at 100°C was demonstrated in Section 4.5. The effectiveness of BMP in the harsher operating conditions of a heat exchanger was unknown, although the literature does refer to antioxidant ceiling temperatures above which the antioxidants become relatively 197  5. Initial Fouling Experiments  ineffective. Lloyd and Zimmerman (1965) studied the effect of 2,6-di-butyl-4-cresol on the autoxidation of cumene and found that above the effective ceiling temperature, the induction period varied as the log of the initial cresol concentration. Below the ceiling temperature (at 126.5°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 and found that it was less effective than metal ion inhibitors at temperatures above 100°C. The effect of antioxidants on fouling rates after the induction period was also unknown as the literature usually measures antioxidation effectiveness in terms of the induction period alone. This was of particular interest to the fouling runs performed in kerosene as these featured extended induction periods caused by commercial additives. A series of fouling runs was performed under identical fouling conditions in the PFRU (Re at Tblk  =  =  3300, T ,.f = 240°C) using a model solution of 0.41 mol/L indene in Paraflex 5  100°C and 79.2 kPa oxygen overpressure in air. Thermal initiation was used in  the presence of 0, 50, 100 and 200 ppm BMP. The results are summarised in Table 5.11 and the gum analysis data are plotted in Figure 5.17. This figure and the data in the table show that there is little effect of BMP addition to the reaction diagnostics after the chemical induction period. The scatter in the reaction diagnostics is consistent with that observed in other thermally initiated experiments. GC/PID analysis showed that the chemical induction period ended when the BMP was exhausted. The chemical reaction in the PFRU system thus behaved similarly to the antioxidation experiments in the SCR. Figure 5.18 shows how the fouling resistance profiles behaved differently at higher additive concentrations. The initial fouling rates at lower BMP concentrations did not show any significant change in fouling behaviour. At 200 ppm BMP addition, fouling started before autoxidation in the bulk solution and featured the formation of a thin grey deposit on the hottest part of the heat exchanger. The fouling rates reported in this region are smaller than the initial fouling rate observed in the period when the gum concentration increases towards g*. This grey material was covered by the orange and red/brown indene 198  5. Initial Fouling Experiments  Table 5.11  [BMPJ  t  -  Fouling in the Presence of an Antioxidant (2,6,di-t-butyl 4-methyiphenol)  Fouling Induction Period (hours)  Fouling Regime  (ppm)  Reaction Induction Period (hours)  0  2 ±0.5  3.5 ±0.5  increasing gum  13.0  50  10  11.5  increasing gum  100  17t  18  200  25.5  200  21 ±1  Initial Fouling Rate j 7 {x10 .K!W.min) 2 (rn  kR [R ] 2  Gum Yield  Deposit Analysis  0.03368 0.986  0.464  997 H 9 C 0 0 03  12.3  0.03217 0.993  0.560  908 H 9 C 1 0 35  increasing gum  13.1  >0.0171  0.503  885 H 9 C 0 0 83  12 22.5  pre-gum increasing gum  1.58 13.1  0.03114 0.992  0.509  615 H 9 C 1 0 68  13.3 21.8  pre-gum increasing gum  4.67  0.02549 0.989  0.677  995 H 9 C 0 0 00  estimated value  199  -  (mol/L)/hr  5. Initial Fouling Experiments  Figure 5.17  Autoxidation Fouling in the Presence of Antioxidant; Gum Concentrations  12 0 A  0  10  o  A  0  8 0  El  6 El  bI  7 ‘_-.  .—.  4  0 A El  2  20  25  30  Time (hours) Figure 5.18  Fouling Resistance Profiles in the Presence of an Antioxidant  0.0010 0.0009 0.0008 0.0007 0.0006 0.0005 0.0004 X  0.0003 0.0002  4 ri-  0.0001 0.0000  0  0  0 ppm  300  900 Time [minutes]  600  50 ppm  O  100 ppm  200  1200  200 ppm  1500  1800  200 ppm  5. Initial Foiling Experiments  polyperoxide deposit after the bulk reaction started. Deposit analysis did not show any significant differences, which is expected as the BMP oxidation products would be difficult to determine from the polyperoxide products unless isotopic labelling of the BMP were used. An interrupted run would have yielded further information in a complete study of the phenomenon. Contamination was dismissed following thorough cleaning and a repeated run. This early, linear fouling regime was not observed at lower BMP concentrations and is thought to be caused by reaction of the BMP or its oxidation products at the heat exchanger surface. Figure 5.19 is a comparison of induction periods in the SCR and the fouling runs and shows that the heat exchanger affects antioxidant performance. Chemical induction periods are shorter in the fouling system and are no longer proportional to the initial BMP concentration. The chemical induction periods are expected to be reduced by the enhanced temperatures in the heat exchanger but the change in concentration dependence indicates that effects similar to those reported by Lloyd and Zimmerman are also involved. The figure shows a logarithmic curve fitted to the PFRU induction period data and the fit is reasonable. These results are consistent with the effects of the heat exchanger operating conditions on the chemical induction period observed in Section 5.4. Further experiments would be required to investigate antioxidant ceiling temperatures but this was considered to be beyond the scope of the current study.  The antioxidant experiments confirmed that at low concentrations, phenolic antioxidants such as BMP extend the chemical reaction induction period but do not affect the subsequent autoxidation and fouling rates. The kerosene used in the fouling runs likely contains low levels of phenolic inhibitors to resist atmospheric degradation; the inhibitors are thus unlikely to affect the fouling rates and autoxidation kinetics. The fouling experiments 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 actually 201  5. Initial Fouling Experiments  Figure 5.19  Effect of Antioxidant (BMP) on Reaction and Fouling Induction Periods 50 45  -  40  0 •  PFRU Reaction SCR Reaction PFRU Fouling  -,  35.  0  50  100  150  200  250  [BMPJo (ppm)  0.41 M indene in Paraflex: TbuIk  100°C : 377 kPa air overpressure: Tsurf (PFRU)  202  =  240°C  5. Initial Fouling Experiments  magnify the fouling problem if the BMP dosage were increased in order to suppress autoxidation. The selection of suitable antioxidants has already been described as a commercially sensitive area. The combination of SCR and PFRU experiments does demonstrate, however, the pitfalls in extending simple batch tests to antioxidant performance in a plant. Conversely, the understanding of the mechanisms involved in autoxidation fouling gained in these experiments suggests that the existing fuel stability information can be used to identify streams likely to form gums and foul heat exchangers.  5.8  Summary of Initial Fouling Experiments  Combinations of two dopants and four solvents were tested in order to identify a suitable model solution for use in fouling studies. These model solution experiments were also used to develop suitable experimental procedures and chemical analyses for further fouling experiments. A solution of 5 wt% (0.405 mol/L) indene in Paraflex was identified as the optimum model solution for fouling studies. The chemical analyses confirmed that autoxidation was occurring in the bulk liquid and controlled the fouling behaviour in the batch fouling experiments. The foulant precursors were identified as soluble and insoluble polyperoxides generated by the cooxidation of the alkene, confirming the hypothesis of Asomaning and Watkinson (1992). Parameters which influenced the formation of polyperoxides in the bulk liquid had similar effects on the observed fouling behaviour; this result could be used to rationalise previous fouling 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 fouling resistance behaviour; an initial, pseudo-linear fouling rate region followed by a later, accelerating region where the fouling resistance increased at a markedly larger rate. The transition between the regions coincided with the polyperoxide gum concentration reaching the solubility limit, g*, after which time globules of insoluble gum were found in solution. 203  5. Initial Fouling Experiments  The initial fouling rate increased with surface temperature and decreased with increasing flow velocity, whereas the later fouling rates appeared to increase with flow rate and did not show any systematic variation with surface temperature. The transition in fouling behaviour was shown to correspond to a change in fouling mechanisms, from a surface reaction zone process to the deposition of insoluble gum globules. 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 particulates melted to form veneers on cooler regions of the heat transfer surface. Chemical analysis indicated that the gum aged soon after adhesion occurred. The deposits were depleted in oxygen and contained strong carbonyl group activity consistent with a thermal degradation ageing step. The chemical induction periods observed in thermally initiated solutions of indene in Paraflex and kerosene were affected by the conditions in the fouling probe. Chemical initiators were used to eliminate the chemical induction period and to ensure reaction reproducibility. The chemical initiators increased the reaction rate and also the initial fouling rate.  The fouling behaviour of mixtures of two active alkenes was studied using indene and dicyclopentadiene. The chemical analysis and fouling results showed non-additive effects which were attributed to the competition for available oxygen in solution.  The effect of an antioxidant on fouling behaviour was studied under severe fouling conditions using a commonly used gasoline antioxidant, BMP. The surface temperature conditions were above the ceiling temperature for BMP reported by Lloyd and Zimmerman (1965) and the antioxidant efficiency was reduced markedly. The end of the extended chemical induction period coincided with the exhaustion of the antioxidant. The 204  5. Initial Fouling Experiments  autoxidation kinetics and fouling behaviour following the end of the chemical induction period showed little difference from the thermally initiated runs at low BMP concentrations. Unusual behaviour was observed at larger BMP concentrations.  205  6. TFU Fouling Studies  6.  Fouling Experiments in the Tube Fouling Unit  6.1  Autoxidation of indene in the TFU  The effects of surface temperature and flow velocity on the initial fouling rate and deposit 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  72  kPa. Lower Pair and initiation levels than in Section 5.5 were used in order to extend the initial 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 results from batch reactions performed in the SCR, PFRU and TFU. The air flow rate was maintained at excess oxygen (300 mL/Lsoln/min) in all cases. The figures show similar behaviour in all three reactor configurations. The kinetic parameters are given in Table 6.1 with the results from the fouling runs  02 The high indene P in the TFU and PFRU under the same chemical conditions (except ). concentration in run 065 was caused by an experimental error. The autoxidation of indene proceeds faster in the TFU and PFRU than in the SCR, despite the discrepancy in oxygen partial pressure conditions. The discrepancy in autoxidation rates is thought to be caused by differences in oxygen mass transfer between the units. Runs 501 and 502-12 did not show any significant effect of the heated section on the reaction kinetics; individual fouling run kinetics were subject to some variation, though this was than less than observed in the PFRU runs using 2.5 mM bP initiation. The chemical analysis results confirmed that reasonably 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 indene concentration results from the TFU fouling runs and the kinetic parameters from these runs are summarised in Table A.2.1. The gum profiles show considerable scatter after t,  206  6. TFU Fouling Studies  Figure 6. la-c Comparison of Indene Autoxidation in SCR, TFU and PFRU  50 40-  0  SCR 145  •  SCR 147 TFU5O1  0 0  PFRIJO65  8  30  A  e  20 10  I.,,,  0  0  a.POx  1  2  4  3  5  6  7  8  9  10  20 145 SCR 147 SCR  18 16  501 TFU 065 PFRU  14 12  2  A  A  A  AA A  10  8 6 4  A  0  •O  A.00° A  2 A 0I•G•  0  b.[gurnj  I...I  1  2  3  4  5  6  7  0.45k0400.35  A  A A  8  0  147SCR  •  145 SCR 501 TFU 065 PFRU  A  0.30 fl2c.  A  .  -  0.20-  9  10  0 A  0.15-  0  0.10-  0.05 0.00  c. [indene]  -  0  1  2  3  4  5  Time (hours)  207  6  7  8  9  10  6. TFU Fouling Studies  Table 6.1  Comparison of Indene Autoxidation Kinetics under TFU Conditions  Experiment  Reactor  Yield  g*  i/(mol/L)/hr  Initial Gum Rate (g/L.hr)  (g/g)  (gIL)  Tbulk  02 P  kR  (°C)  (kPa)  145  SCR  100  79  0.02692  2.0 ±0.2  0.65  10.5 ±1  147  SCR  100  79  0.02723  2.1 ±0.1  0.52  11.5±1  501  TFU  100  72  0.03253  3.1 ±0.1  0.57  11 ±0.5  502-12t (avemged)  TFUt  101.7  72  0.03254  2.7 +0.1  0.53  11.5 ±1  PFRUt  100.2  79  0.03996  3.6 ±0.1  0.49  12 +1  065  0.405 M indene in Parallex; air flow rate  300 mLIL.min (ntp);  208  -  fouling experiments;  -  0.43 M  6. 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 and tube processing methods. An electric grinder was used to cut the 501 test section into different lengths but the metal dust caused errors in the deposit mass assay. The tube surface 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 and regularly spaced (separation  1-2 mm). The deposit mass coverage increased rapidly from  zero over the first 15 cm then increased gradually with axial distance. The striations are caused by precipitation of polyperoxide gum at the unheated tube surface, where T ,f < 5 Tb1k due to the small insulation losses. The gum solubility limit, g*, is a strong function of temperature and polyperoxide gum is thus precipitated at the surface. This form of crystallisation fouling was responsible for the loss of cooling capacity observed in the fouling runs, where the cooling utility was mains water at  12-15°C. The striation  phenomena are discussed further in Section 6.2.3  6.2  Initial Fouling Studies  Experiments 503 and 504 were performed under identical chemical and fouling conditions  200°C, W 0.105 kg/s) to study the reproducibility of TFU fouling  data. Run 502 also involved these conditions but was terminated by a power failure after 300 minutes. Good agreement was found in the chemical analysis results and in heat transfer performance. The overall fouling resistance profiles from both runs are plotted in Figure 6.2 and the agreement is well within the estimated error in calculating Rf. The overall fouling resistance, Rf , was calculated using data from those thermocouples which were 0 209  6. TFU Fou1ing Studies  Figure 6.2  Fouling Resistance Profiles Showing TFU Reproducibility  0.00009  I  o  503  •  504  0.00008  0 0  0  0  0.00007  o ,_.  .00  OO  0.00006  o•  9 o’  0•  o  0.00005  o  0 Q  o  0.00004  0  0  o o•  0  •  z  1’•  O%%  0  o 0.00003  O•  o 0  •  0  .  0.00002  •  .. •  •  0  0 0 0  0.00001 0 0  0.00000  0  .  -0.00001 0  100  200  300  Time (minutes)  210  400  500  600  6. TFU Fouling Studies  considered to be downstream of the thermal entry length. The data exhibited more scatter than that observed in the PFRU runs due to magnetic field interference and the effect of averaging over the whole heated section. The runs confirmed that the TFU gave reproducible results. The fouling resistance profiles in Figure 6.2 show a transition from an initially low fouling rate to a significantly larger fouling rate at approximately 300 minutes. The initial rise in fouling resistance was caused by the system adjusting to the change in heat transfer coefficients after indene had been added. The system data were averaged over the first 25 minutes in order to give the initial values. Figure 6.3 shows the fouling resistance and gum concentration profiles for run 503 and confirms that the transition in fouling rates (and presumably mechanisms) coincides with the gum concentration approaching g*. All the TFU fouling runs showed the same relationship between gum concentration and fouling mechanism as reported in the PFRU fouling runs in Section 5.  6.2.1  Tube Pressure Drop  The pressure drop across the tube changed during a fouling experiment as material deposited on the tube wall narrowed the flow channel and affected the roughness of the tube surface. The variation of the pressure drop with time confirmed the transition in fouling mechanisms from a surface reaction zone process to the deposition of insoluble gum. Figure 6.4 shows the tube pressure drop and soluble gum concentration data from run 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.  ) 1 Deposition increased the tube pressure drop by reducing the channel size (increasing u  and by increasing the friction factor via roughness effects. Roughness effects are discussed further in Section 6.2.3.  211  6. TFU Fouling Studies  Figure 6.3  Overall Fouling Resistance and Soluble Gum Concentration in Run 503  0.00009  18.0  •  0  0.00008  °  R f,o  -  16.0  —  •°  •  Gum •°  0.00007  14.0  —  -  —  0  —  •  0.00006  oQ  000 0 • 0  —  •  12.0  008  —  .  0.00005  <‘  —  10.0 00  •  0.00004  0  0  —  0  0 0  —  8.0 ,@0  —  0  0.00003  6’  •  00  —  0  0.00002  0  6,0  °,  & ,p  ‘.  4.0  Oo  °o8°  0 C’  0.00001  2.0 0  0  100  200  I  I  I  300  400  500  Time (minutes)  212  0.0 600  6. TFU Fouling Studies  Tube Pressure Drop and Soluble Gum Concentration in Run 505  Figure 6.4  20.0  14.5  o  APtube  •  Gum  18.0  16.0 —  14.0  —  -  14.0  —  —  .  °—  .00o 0  0  0 0  12.0  00  13.5  -  —  ci) —  0 0  •  10.0  U  • -  8.0  0 00  0  0  -  -o o0 020 0 0 00 000 00 00 0 00 0 00 oo 0000 OOQJ% —00 OOo% 0 9 o 00 0 0 0 00 0 O0 0000 00 0 0 0 —  -  6.0—0 4.0  2.0  —  -  12.5  —  •  0.0  I  0  100  200  Time (minutes)  213  300  ç —  00  •  13.0  -  —  • H  0  00  400  6. TFU Fouling Studies  Table 6.2 is a summary of the initial conditions in the fouling runs and the observed tube pressure drops. AP increases with u,, as expected from Equation [3.33]. The low value of initial AP in run 512 was probably caused by trapped air in the lines. Runs 510 , caused by heavy fouling. Run 508 was 1 and 512 involved very large changes in A.P operated at close to maximum heat flux but cooler fouling caused the bulk temperature to rise steadily after 200 minutes. The deposit morphology was noticeably different from that of runs 503-4 and appeared to be caused by bulk precipitation of insoluble polyperoxides.  6.2.2  Local Fouling Behaviour  The local heat transfer coefficient was not uniform along the heated length owing to the development of the thermal boundary layer in the tube. This produced a non-uniform temperature profile along the heated length, which would be expected to affect the fouling rate as this is a strong function of temperature. The local fouling behaviour varied significantly with axial location x, but this variation could not be explained simply in terms of the surface temperature or the development of the thermal boundary layer. Table 6.3 summarises the local heat transfer and fouling resistance variation observed in run 504. The overall 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 temperature  around 300 mm, which may be caused by variations in tube thickness. The deposit coverage in this region was also larger than average, which indicated that the thermocouples were operating correctly. Most runs showed some deviation from the expected temperature profile along the heated section so a proposed simulation of the TFU was not performed. Table 6.3 shows that negative fouling resistances were observed in the first 150mm of 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. 214  6. TFU Fouling Studies  Table 6.2 Run  Initial Conditions and Tube Pressure Drops in TFU Fouling Runs 0 f 5 T , (°C)  502  -  Tbulk (°C) 101  Uo  W (kg/s)  (W/m . 2 K)  -  -  -  503  199.7  102.8  0.11  1901  504  198.3  101.3  0.11  1873  505  222.6  101.4  0.102  1886  506  187.4  100.2  0.104  1843  507  211.5  100.8  0.104  1883  508  199.4  102.3  0.174  2883  509  212.4  103.6  0.137  2476  510  211.3  101.9  0.056  1153  511  209.1  101.5  0.087  1758  512  211.3  101.5  0.056  1188  0.405 M indene in Paraflex, 1mM bP initiation;  t  q (qt) ) 2 (kWim  -  171.6 (170.5) 166.9 (171.8) 207.7 (200.8) 146.6 (145.2) 189.4 (185.4) 256.0 (256.0) 248.3 (242.7) 113.3 (112.8) 176.1 (171.9) 118.2 (117.5)  Urn 0.01 (mis)  ±  Initial LPtube (kPa)  -  Initial  I  Final AP j 1 (kPa)  -  -  2.18  13.6  0.00767  14.3  2.18  13.2  0.00739  13.5  2.03  13.1  0.00863  13.8  2.07  13.3  0.00845  13.7  2.08  13.1  0.00820  13.9  3.47  32.2  0.00716  33.2  2.73  20.2  0.00726  27.8  1.12  4.8  0.01016  6.4  1.73  9.2  0.00837  9.3  1.12  3.3  0.00704  6.8  average heat flux over Run; f Fanning friction factor  215  -  6. TFU Fouling Studies  Table 6.3  Distance along heated section,  Local Heat Transfer and Fouling Behaviour in Run 504  Tsurf,o  h(x)  (mm)  (°C)  44.4  (W/m . 2 K)  Initial Fouling Rate ) 8 (x10 .K!W.mjn) 2 (m  Final Fouling Rate ) 8 (x10 .K/W.min) 2 (m  Final Fouling Resistance (x10) .K!W) 2 (m  185.91  1950  4.4  32t  0  104.8  195.24  1783  5.7  -136,-33  -8.0  158.8  197.5  1760  10.3  0  2.0  0.48  219.1  195.6  1811  10.2  34  5.5  0.54  323.9  201.5  1747  8.6  42  7.1  0.50  435.0  197.4  1869  8.6  40  7.3  0.65  546.1  194.8  1976  7.4  40  7.0  603.3  199.8  1888  6.6  36  6.5  0.65  657.2  196.1  1995  5.9  35  5.5  0.47  708.0  210.7  1891  6.2  36  6.5  0.50  overall  198.3  1873  7.7  37  7.2  0.60  X  -  initial rate which tends to zero after 30 minutes:  -  216  Deposit Coverage ) 2 (mg/cm  initially rapid rate, which decreased in magnitude  -  -  6. TFU I’ouliizg Studies  Figure 6.5  2  Local Fouling Resistances in Run 504  0.00009 0.00008  -  0,00007  .  a. x  44.4mm  =  0.00006 0.00005 0.00004  0.00003  .1.  -  0.00002  •• .  •  0.00001  4.’.—  .  0.00000  Iv  .  -0.00001  -  -0.00002 0  50  100  150  200  250  300  .$••  ••.i  350  400  450  500  400  450  500  0.00004 N  dPh1M  -  .#  -0.00000 -0.00002 -0.00004  .  -0.00006  b.x=104.8mm  -0.00008  -  C  -0.00010 50  0  100  150  200  300  250  350  0.00006 0.00005  c. x  =  219.1 mm  • .&•  N  -  ••.•••  0.00003 0.00002  S  -  .%  •.1  •.  0.00001  •  .5  •  •  -0.00000 C  -0.00001  -  0  50  100  150  200  250  300  Time (minutes)  217  350  400  450  500  6. TFU Fouling Studies  Negative thermal fouling resistances were not reported in previous chemical reaction fouling studies but have been reported in particulate fouling (Crittenden and Alderman, 1988). The positive mass deposit coverage values indicated that this was not a spalling effect but was caused by enhancement of the local heat transfer coefficient. A negative value of R/x) indicates that h(x) >h(x) , (and at constant heat flux, 0  Tsurj(X) < T(X)surjo),  which occurs when the polyperoxide gum particulates are large enough to disturb the developing thermal boundary layer. The effect on the friction factor is expected to be less significant as the momentum boundary layer thickness, boundary layer thickness,  ôth,  for fluids with Pr  >  ömom,  is larger than the thermal  1.  A negative fouling resistance at constant heat flux invalidates the constant depositlliquid interface temperature assumption which is desired in chemical reaction fouling studies. Equation [5.4] shows that the fouling rate is very sensitive to surface temperature 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 Figure 3.12 to be approximately 180 mm) indicated that this was a localised phenomenon in the thermal entry region and that data from downstream were not affected. This was confirmed by the inspection of the deposit morphology described in Section 6.2.3.  6.2.3  Deposit Distribution and Morphology  The foulant was inspected in situ using optical microscopy, SEM and a simple gravimetric analysis. More information about deposit structure and properties could be obtained using the techniques recommended in Section 9. The deposit morphology varied with position in the fouled heated sections. Figure 6.6 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. The photograph shows that deposition occurred in four different regions which are shown in 218  6. JTU Foniin Sindies  Figure 6.6  Photograph of Fouled TFU Tube from Run 507 after Division into Sections  ‘V  ,(.,  -  ./- ‘t?I  • 501  Figure 6.7  V 5fr  13  TFU Test Section Depositon Regions  Flow  Sec/ion 1]  I  Unheated Entry Length  Figure 6.8  Themial Entry Length  36 Number  26  13-  I  Developed Region  Exit Length  Photograph of Thermal Entry Length Deposition in Run 507  Ill lIi!IIIfIIIIiIlJTTlJTTTjIllIIIijIlilI!Ij1I  ;I!IIIIIllIl1l1lI1l1III1I1(I11T11111[i]J1i1IIIIIIi1I[I9iII1iTf ii  L  MAO IN ENGL  -  Oil  O)I  O6O  O9O  r  CV  L -.  6. TFU Fouling Studies  schematic form in Figure 6.7. Figure 6.8 shows the transition between the unheated entry length (UEL) and the thermal entry length (TEL) in greater detail. The striations observed in 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 is consistent with the striations being generated by the deposition of insoluble gum discussed in Section 6.1. The liquid leaves the heated section with Tb,o  >  Tj so that the gum  concentration is no longer at the solubility limit and the driving force for precipitation is significantly reduced. The striation widths and separation were much larger than the visible grain structure in the drawn tubing, which suggested that the striations were not caused by channels 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 at lower Re. At low Re, smaller striations coexisted with larger ones and often merged with other striations. The striation deposition patterns are thought to be caused by local channelling in the surface turbulence structure, in conjunction with irregularities in the tube surface. Yung et al. (1989) described the locally organised motion in turbulent boundary layers and reported dimensions of the variable speed fluid streaks. The streaks average width was 15-20 v/u* and they were spaced randomly apart, at a mean separation of 80100 v/u. For the TFU, these dimensions were calculated as widths ranging from 251-77 jim and mean separations ranging over 1100-420 jim. The streak widths and separations decrease with u, as observed in the TFU, and the streak separations were comparable to the striation separations observed in the experiments. The variation in deposit coverage with Re may, however, be related to removal and deposition mechanisms. Further study of the striation patterns was recommended. Figure 6.8 shows the fouled thermal entry length from run 507 and the transition between the unheated entry length and the developed fouling regions. The length of the fouling transition region varied with Re in the same manner as the heat transfer entry length. The dimensions of the striations in this region were similar to those in the cool entry 220  6. TFU Fouling Studies  length but the gum itself was red or darker, indicating that the material had been aged by the enhanced surface temperature. The striations then fanned out into a chaotic region where the deposit consisted of randomly sized and distributed globules of aged gum. The deposit patterns bear a striking resemblance to photographs of the onset of turbulence in smoke tracer/wind tunnel studies, where turbulent spots occur and their wakes expand into the chaotic region of developed turbulence. The gum globule sizes in the heated entry length region ranged up to 1 mm, which was significantly larger than those measured downstream in the developed region. The larger globules were thought to originate as entrained agglomerates of insoluble gum which were deposited on contact with the first section of heated surface. The thermal entry length would then act as a trap for such large agglomerates, removing them from solution before they reached the developed region. The thermal entry length observed in the deposit from the PFRU runs was noticeably shorter than in the TFU runs. This suggests that entrained globules of insoluble gum could be deposited in the developed region of the PFRU which would not be deposited in the TFU. This mechanism may explain the differences in fouling resistance profiles reported by Panchal and Watkinson (1993) between a longer, tubular probe and the PFRU 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 fouling resistance profiles. The foulant observed in the developed region consisted of randomly deposited globules of gum which had aged to form a red/brown solid. Figure 6.6 shows a grain structure 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 tube surface; some segments experienced heavier fouling, which was thought to be due to variations in the tube thickness. 221  6. TFU Fouling Studies  The transition from the developed region to the exit length was always very sharply defined, confirming that conduction losses through the terminals were greater than those along 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 in the PFRU experiments; a more even initial deposit formed from smaller gum globules followed 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 showed significantly fewer globules of dark insoluble gum and almost no agglomerates of this material. The deposit coverage observed in run 502 was also smaller due to the lower surface temperature involved (200°C). The deposit coverage profiles from runs 502-4 are plotted in Figure 6.10 and show the deposition patterns observed in Figure 6.6 in a quantitative form. The missing data points refer to sections retained for further analysis. The coverage in the cool entry length 2 irrespective of the run duration increases in the first 300 mm to a mean value of 1 mg/cm (502 5.5 hrs; 503 8.5 hrs; 504 8 hrs). The reduced value at the inlet was probably due -  -  -  to increased turbulence in this region following the tube fittings. The developed region coverage profiles show the local hot spot reported in 503 and the effect of run duration on the final deposit mass. Experiments 503 and 504 differed in duration by 30 minutes but the difference in coverage is greater than that between 504 and 502; this result indicated that the coverage data should not be used to compare fouling rates between runs featuring different durations. The run 502 data confirmed that the initial fouling rate period in Paraflex involved the deposition of relatively small amounts of material under the conditions used.  Globule diameters, d, were estimated from the micrographs and found to belong to four size ranges. The smallest, ranging in from 7-18 Jtm, were the amber/orange particulates involved in the formation of the yellow matrix and the initial, even deposit. The 222  6. 1 FU Foiling Studies  Figure 6.9  Optical and Scanning Electron Micrographs of Deposit fom Run 507  a.  Optical (136x)  b.  SEM  TFU Run 507 Section 16  (bOx)  223  6. TFU Fouling Studies  Variation in Deposit Coverage with Axial Position in TFU Runs 502-504  Figure 6.10 2.c  L8  Heated Zone  -  .503  1.6  A504  1.4 1.2-  *  •  •0  .1)  .  .  •0.t. A •AA  L0  ‘  502  0  I  .  .  00  e  0.8 *0  0.6 I4  c  0.4  A  A 0  Flow  0 00  o0o  -  0  200  0  400  600  .  0  0  0.2 0.0  A  0  • 0.•000  I  I  800  1000  1200  1400  .I  1600  1800  x (mm)  Table 6.4  Comparison of Sand Roughness Criteria with Gum Globule Dimensions in TFU Fouling Run 503  Globule Size Range (Fm) (Tbuak  =  l Re 1 (d/d  Roughness Regime  108°C) (Tfjlm = 154°C)  7-18  0.5-1.3  0.9-2.4  8-19  hydraulically smooth  30-40  2.1-2.8  4.0-5.3  32-43  transitional roughness  66-75  4.6-5.3  8.8-9.9  74-84  transitional roughness  250+  17.6+  33.2  71+  fully rough  224  6. TFU Fouling Studies  smaller dark red/black globules observed in the later stages of the initial fouling rate ranged around 30-40 jim in diameter. The larger black/red globules observed in the bulk precipitation phase ranged from 66-75 p.m and then  —  250 p.m+ in diameter and were  evidently agglomerates of smaller units; these approached millimetre dimensions in the thermal 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 as mom 6  =  5v/u*; the data in Table 6.2 gave u for run 503  /s at Tb,m 2 viscosity of 1.925x 106 m Tfilm  154°C to calculate  V  0.1354 mIs. The kinematic  108°C gave a sublayer thickness of 71 jim; using  6 m /s) gave 2 (1.02 xlO-  mom 3  38 p.m. These estimates  confirm the observations in Section 6.2 where the friction factor was not significantly affected by the formation of deposit until the generation of the larger insoluble gum globules. The larger globules present in the bulk precipitation process were expected to cause measurable roughness effects. White (1991) described the following sand roughness effect criteria based on d: <  4  <  hydraulically smooth wall  4  d  <  >  60  60  (dId) Re  <  (dLjdt) Re  >  10  transitional roughness regime fully rough (no viscosity effect)  1000  The globule dimensions are compared with these criteria in Table 6.4. The largest globule size range belonged to the transitional roughness regime in both cases, whereas the large agglomerates (250 p.m+) are expected to give fully rough behaviour.: The effective heat transfer sublayer thickness can be estimated as gave  thm 6  öth  =  A/U, which  in run 503 as 61 p.m (based on the film temperature). The heat transfer sublayer  thickness is thus comparable with the viscous sublayer thickness  monz 6 (  =  38-7 1 jim),  which is unexpected in turbulent flow where Pr>]. This result indicates that the d sizes are  of the same order of magnitude as  th• 5  Paterson and Fryer’s fouling model describes the  generation of foulant as occurring within the thermal ‘boundary layer’; the relative sizes  225  6. TFU Fouling Studies  confirm that the larger globules must have been formed in the bulk fluid. Fouling models  are discussed further in Section 7.3. The concentration ‘boundary layer’, ômass, can be estimated using the Chilton Colburn mass transfer analogy St Pr 2/3 Assuming St’ ómass  =  2/3  St’ Sc  =  f12  [6.1]  dr/Re Sc ömass gives =  ôth  3 (Pr/Sc)”  At the film temperature in run 503, &  =  [6.2]  17.9; Sc(indene)  =  475 and Sc(oxygen)  =  147,  giving ômass as 20.4 pm (indene) and 30.2 pm (oxygen). The concentration boundary layers thus lie completely within the viscous sublayer. Both ranges of insoluble gum globule sizes are larger than  ömass  and thus indicate that this deposit precursor was  generated in the bulk fluid.  Surface Temperature Effects in TFU Fouling  6.3  The effect of surface temperature was studied using the model solution described in Section 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, as reported 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 the  other 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 observed previously. The fouling data for different surface temperatures and flow velocities are summarised in Table 6.5. The final Rfvalue and the mean deposit coverage both reflect the trends observed in the initial fouling rate. Figure 6.11 is an Arrhenius plot of the averaged 226  6. TFU Fouhin2 Studies  Table 6.5  Surface Temperature and Flow Velocity Effects on Initial Fouling Rate  Duration  Tsurf,o  w  Urn  (hr)  (°C)  (kg/s)  (mis)  503  8.5  199.7  0.105  504  8.0  198.3  505  6.5  506  Run  Initial Fouling Rate (x 108) .K1W.min) 2 (m  Final Fouling Resistance ) 5 (xlO .K/W) 2 (m  Mean Deposit Coverage (Developed Region)  2.05  8.1  8.0  1.15  0.105  2.04  7.8  7.2  0.61  222.6  0.102  2.04  23.4  24  4.06  8.0  187.4  0.104  2.07  5.8  3.6  0.13  507  8.0  211.5  0.104  2.09  13.0  7.7  1.48  508  8.0  199.4  0.174  3.48  2  -O  0.04  509  8.0  212.4  0.137  2.75  6.6t  -8.6  1.85  510  8.0  211.3  0.056  1.12  16.Ot  -15.5  6.03  511  5.8  209.1  0.087  1.73  22.0  19.5  3.54  512  8.0  211.3  0.056  1.12  24.Ot  -14.0  8.37  ) 2 (mg/cm  t final rate becomes negative -  Figure 6.11  Arrhenius Plot of Initial and Final Fouling Rates in TFU Experiments  -12  -13  -14 -15  -16 -17  -18 -19 0.00200  0.00205  0.00210  0.00220  0.00215  0.00225  1/Ts,o(K) o  -  local initial fouling rate;  •  -  local final fouling rate;  227  •  average initial fouling rate  6. TFU Fouling Studies  initial fouling rates and also shows the local initial and final fouling rates. The initial fouling rates were obtained by drawing a line through the fouling resistance data where t<t*; the scatter in the data suggested little benefit from the use of a regression program. The averaged initial fouling rates were fitted to a modified Arrhenius equation dRf/dt  =  23.9 exp (-7M00/RT)  2 (R  =  0.970)  [6.3j  The activation energy, 76.4 ±8.0 kJ/mol, is reasonably close to the values of 81.9 ±16.4 and 84.8 ±13.2 kJ/mol reported in the PFRU studies and indicates that the same fouling mechanism is operating in the TFU runs. The prefactor, 23.9 m .KIW.min, is an order of 2 magnitude smaller than those reported in the PFRU; this is likely linked to the changes in probe geometry and the higher flow rate in the TFU experiments. The final fouling rates are also significantly smaller than those reported in Section 5.4. Comparisons are less useful in this case as the runs were terminated after 8 hours where possible, rather than allowing the precipitation fouling process to accelerate as in the PFRU runs. The 8 hour limit was adopted to avoid excessive precipitation of insoluble gum in the TFU. The similarity in mechanisms was confirmed by the deposit analyses summarised in Table 6.6. The gum recovered from the unheated entry length was similar in composition to the indene polyperoxide gums described previously. Gums recovered from the heated sections showed a marked loss in oxygen content, which was more pronounced at higher initial surface temperatures. The gum from the thermal entry length, where the surface temperatures were not as severe, showed an oxygen content between those of the two other regions. The results are consistent with the gum ageing process described in Section 5. The deposits from runs 509, 510 and 512 show significantly higher oxygen content than reported in the other TFU runs. This is consistent with the negative final fouling resistances reported in these cases; the lower final surface temperature would reduce the rate of ageiñg and thus the loss of oxygen in the deposit.  228  6. TFU Fouling Studies  of the deposit can be estimated from the  The effective thermal conductivity,  deposit coverage and the final fouling resistance when the latter is positive. Assuming constant deposit density p, the final fouling resistance of a mass of deposit (mass  =  area  öf.Pf) is Rf  =  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 fouling resistances where the local surface coverage was also known. The hand drawn line gives a .K; the density of indene is approximately 990 kgm 4 value of pfAf as 184.6 W.kg/m , 3 giving an estimated value of  as 0.19 W/m.K, which is in good agreement with the  values estimated from deposit thicknesses in Section 5. The deviation from the line is largest at low surface coverage (and low surface temperature). These coverage measurements involved the largest percentage errors and the deposit contained more oxygen, which would decrease 2 and thus raise the points above the line. The TFU thus gives a direct measurement of pf? a quantity which is usually estimated in fouling model calculations.  6.4  Velocity Effects in TFU Fouling  The effect of flow velocity was studied using the same model solution of indene in Paraflex and a surface temperature of 210°C. A 770 mm long heated section was used and the bulk fluid velocity varied from 1.12  -  2.75 mIs. The range of velocities and  temperatures involved was limited by the heating and cooling capacity of the system under fouling conditions. Runs 508 and 509 were performed at high heat fluxes and experienced cooling capacity problems due to cooler fouling. The reported initial fouling rates were small and subject to error estimates of ±15%; the rates could have been increased by using higher surface temperatures but this would have required a new transformer or shorter  229  6 TFU Fouling Studies  Elemental Analysis of Deposits from TFU Runs  Table 6.6 Run, Section#  Position  Tsurf,o (‘C)  C  wt% H  0  As peroxide  504.13  TEL  195  74.26  4.74  21  0 6 H 9 C 1 91 9  505. 8 505.15 505.20 505.23  UEL DR DR DR  100 225 225 225  70.0 84.11 81.04 80.2  5.8 7.32 6.57 5.51  24.2 8.57 12.39 14.29  0 9 H 9 C 2 32 0 9 H 9 C 0 0 69 4 8 H 9 C 1 0 03 8 7 H 9 C 1 0 20 4  507. 5 507.15 507.21  UEL DR DR  100 210 210  71.71 78.38 77.07  5.33 4.96 5.25  22.96 16.66 17.68  0 8 H 9 C 2 16 0 6 H 9 C 1 0 43 8 7 H 9 C 1 0 55 4  509.16  DRt  210  72.85  4.70  22.45  0 7 H 9 C 2 08 0  510.16  DRt  210  74.04  4.68  21.28  0 6 H 9 C 1 94 8  511.20  DR  210  80.25  5.35  14.4  0 7 H 9 C 1 21 2  512.20  DRt  210  78.21  4.91  16.88  0 6 H 9 C 1 46 8  t negative final Rf: UEL unheated entry length; TEL thermal entry length; DR developed region -  -  Figure 6.12  -  -  Plot of Final Local Fouling Resistance against TFU Deposit Coverage for Estimation of Deposit Thermal Conductivity  0.00040 0.00035 0.00030 0.00025  I  0.00020 0.00015 0.00010 0.00005 0.00000 0.0  0.5  1.0  1.5  2.0  2.5  3.0  3.5  4.0  Deposit Coverage (mg!cm2)  230  4.5  5.0  5.5  6.0  6. TFU Fouling Studies  heated lengths. The study was interrupted by pump failure; no further runs were undertaken. The results are summarised in Table 6.5 and show that the initial fouling rate decreased with flow velocity, as reported in Section 5.5. The thermal entry lengths were shorter at larger flow velocities, as reported in the literature. Run 511 showed some anomalies in its fouling behaviour; the initial fouling rate was larger than expected and the experiment had to be terminated early as the surface temperature alarm reading was high and larger than the surface temperatures recorded by the PC. The deposit obtained from run 509, after cooling capacity failure, was quite different to that observed in other TFU fouling experiments; the foulant was a shiny, dark mass of randomly deposited, large gum  aggregates of size  >  200 pm. This caused negative fouling resistances in the later part of  run 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 responsible and the onset of decreasing Rf coincided with  t  and the transition to the bulk agglomerate  fouling regime. The deposit consisted of an orange/red veneer composed of dark, amorphous gum agglomerates ranging in size from 150 pm upwards. The agglomerates thus increased the local heat transfer coefficient and caused a cooling effect at the heated surface. The effect of these large agglomerates on the viscous sublayer was confirmed by tube pressure drop, which increased rapidly after  t  as Rf decreased. This can be clearly  seen in Figure 6.13, which shows the fouling resistance and pressure drop profiles from run 510. The increase in pressure drop is caused by both the change in friction factor and the narrowing of the duct by the accumulation of deposition. The increase in friction factor also enhances heat transfer and a rough comparison of these effects can be made using the heat and mass transfer analogy in Equation [6.1]. Substituting for the friction factor in an approximate tube pressure drop expression based on Equation [3.33] gives 231  6. TFU Fouhin2 Studies  Figure 6.13 Tube Pressure Drop and Overall Fouling Resistance in TFU Run 510  0.00010  I  I  -  -  0.00005  o  Rf  •  LP  8&&o0  00  080  0  a,  0 0  0.00000  co  o  ooo%)00 oo 0 a 0 0  o  6.5  I  -  0  0  8  .  0  6.0  00 oO  .• • •  0  0 0  0 0• 0•  ci)  -0.00005  5.5  zC .  -0.00010  .  1•.  1  ..  .  •?*  .1.  .  •  • • ..• ._  H 5.0  0  .  -0.00015  -C  •.  ••  0  •0  00 ‘DO  .  • ...•  -0.00020 0  IIp•  100  •i.  .•  200  .  4.5 300  Time (minutes)  232  400  500  6. TFU Fouling Studies  AP  =  4 f (L/d (Pum )/2 2  =  4 (ff2) Re (L/dt )um 2  i  =4 Nu Pr 2 (L/dt 3 “ ) Um 11  [6.51  Assuming constant fluid properties and flow rate, the ratio of pressure drops and Nusselt numbers is given by /zP(t) 0 AP,  =  fNu(t)] 0 0 [Nu Id, [d(t) ] 4  =  IU(t)] [ 0 0 [U /d, d(t) l 3  [6.6]  Equation 6.6 shows that the pressure drop is more sensitive than the Nusslet number to changes in the duct geometry. Data from run 510 were: U increased from 1153 to 1400 K; zIP increased from 4.8 to 6.45 kPa. Assuming that the pressure drop outside the 2 W/m heated 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 agreement with a visual estimate of the deposit thickness; the material was too brittle to measure mechanically. The negative fouling resistances observed were thus consistent with the deposition of large gum agglomerates which interrupted the viscous sublayer. These aggregates are presumably present at higher flow rates also but do not cause similar effects there; this indicated that the adhesion of the aggregates is controlled by the residence time at the surface (which decreases with u *) or by the shear forces given by ‘r acting at the surface (which are larger at higher flow rates). The fouling mechanism is discussed further in Section 7.  The initial TFU fouling rates decreased with flow velocity, varying as  where -1  Um  <n <-2. The initial rate reported in run 511 was larger than those observed at a lower flow  rate (510,512) and suggested that there was a local maximum in the fouling rate velocity -  curve. Such maxima are a feature of Epstein’s model (Section 2.4). The unusual fouling behaviour in this run rendered the rate value suspect and a repeat experiment was required to verify the result; unfortunately a repeat run could not be performed.  233  6. TFU Fouling Studies  Comparison of TFU and PFRU Fouling Probes  6.5  Negative fouling resistances were reported in the initial PFRU fouling experiments at 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 run was performed using the same model solution in the PFRU and T,j= 210°C. The PFRU had been modified since run 064 and was operated at its maximum flow rate, which gave Urn  0.774 mIs and Re  5040. The flow velocity was less than in runs 510 and 512 (1.1  mis) but the Reynolds number was similar (5360). The heat flux calibration had also been modified and the calculated value, 145.8 kW/m , yielded a higher value of U 2 0 than reported in Section 5. The PFRU was operated under similar chemical conditions as the TFU 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 identical fouling behaviour to that observed in the TFU. The negative fouling resistances therefore arose because of conditions leading to bulk precipitation rather than differences in probe geometry. Figure 6.14 shows the Rfprofiles from runs 512 and 065; the fouling resistance decreased after t and the onset of the bulk precipitation fouling mechanism. The drop in Rf in 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 fouling probes. The initial fouling rate in the PFRU (3.36x10 7m .KIW.min) was larger than in 2 the TFU (1.6 and 2.4 xlO.KIW.min), which was attributed to the lower flow 2 7m velocity. This result suggested that the dimensionless velocity (given by Re) was less important than  Urn  in the fouling mechanism. The fouling resistance data from the PFRU in  Figure 6.14 shows considerably less scatter than the data from the TFU. This is because the PFRU measures a local fouling effect whereas the TFU monitors the overall unit  234  6. TFU Fouling Studies  Figure 6.14  Comparison of Fouling Resistance Profiles in TFU and PFRU  0.00020  o  065(PFRIJ)  •  512(TFU)  • •  •  . • ,..  0.00015 0• • i0’  0.O0010  •  •  •%  0  •  Q 0  •• ••:: ••:.  •  4-e  0.00005  -  00 0  •.  •  •  0 0 0  -  -000000  .•  •0  0  0  9 C  0  •  0  0  ()  •  •  0  o 000  -0.00005  s  0  -  0 0  0  0 0  -000010 0  50  100  150  200  250  300  Time (minutes)  235  350  400  450  500  6. TFU Fouling Studies  performance. The differences in the TFU and PFRU Rj profiles before  t>’  were similar to  those reported by Panchal and Watkinson (1993). The PFRU deposit was remarkably similar to that observed in the corresponding TFU runs. The unheated entry length was coated with yellow gum striations of varying widths, developing swirled patterns at positions further downstream from the probe entrance. The swirl was probably caused by the rotational motion of the fluid imparted by the sharp turn at the probe inlet. The thermal entry length was 25 mm long and contained clearly defined red/orange striations running parallel to the flow. This entry length was significantly shorter than that observed in the TFU runs (100-150 mm) and did not involve the ‘turbulent spot’ zone reported in Section 6.2.3. The heavily fouled region featured very disordered deposition by thick, shiny red/brown agglomerates of insoluble gum which ranged in size up to 1 mm. Occasional light yellow strands of gum were evident in the exit length.  The TFU studies showed the same initial fouling rate behaviour as observed in the thermally and chemically initiated PFRU studies described in Section 5.2. The deposit morphology and composition also confirmed that the same fouling mechanisms occur in both probes. The negative Rf behaviour reported at low velocities in runs 510, 512 and 065 is unusual. A heat exchanger subject to this type of fouling behaviour would suffer increased pressure drops before there was any thermal evidence of deposition. It is likely that the deposit would eventually lose the enhancement in heat transfer coefficient caused by roughness and the fouling resistance would then increase rapidly. Further investigation of the negative fouling resistance phenomenon requires a study of the reaction and agglomeration 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 T U,j= 180°C, 5 236  6. TFU Fouling Studies  Tbik  =  80°C, Re  =  3050 and 79 kPa oxygen saturation. The low surface temperature gave  negligible initial fouling rates so that the fouling behaviour was dictated by bulk precipitation. Figure 6.15 shows how the fouling resistance profiles for these runs deviated from that observed in run 025 (5 wt% indene). Similar deposit morphologies to run 065 were also reported. Experiments 024, 027 and 065 featured larger indene reaction rates than run 025, which may be an important factor in the formation of agglomerates.  6.6  Summary of TFU Fouling Studies  The Tube Fouling Unit (TFU) was shown to give reproducible heat transfer and thermal fouling resistance results, although differences between individual tubes did cause variations in the fouling data. The range of operating conditions was limited by the current design and by the fouling of the cooling system by the polyperoxide gums which also fouled the heated section. This cooling fouling phenomenon is linked to the solubility of the polyperoxide gums and was not noticed in the PFRU. Ten fouling experiments were performed in the TFU to investigate the effects of surface temperature and flow velocity on the initial fouling rate. Chemical analyses confirmed that conditions of reproducible reaction were achieved in the bulk liquid. The initial fouling rate showed the same trends in surface temperature and velocity reported in the PFRU experiments. Anomalous fouling resistance behaviour (negative Rf) was observed at the lowest velocity ( lm/s), where the deposition of large agglomerates of gum enhanced the heat transfer coefficients via roughness effects. This behaviour was also observed in a comparative experiment performed in the PFRU. The experimental evidence confirmed that the same deposition mechanisms were involved in the annular PFRU and tubular TFU fouling probes. The tube pressure drop measurements confirmed the transition in fouling mechanisms observed in the thermal fouling data. The TFU permitted the examination and analysis of deposit in situ and 237  6. TFU Fouhin2 Studies  Figure 6.15  Fouling Resistance Profiles from Initial PFRU Runs at High Indene Concentrations showing Unusual Fouling Behaviour  0.0006  0.0005  0.0004  0.0003  0.0002  0.0001  0.0000 Concentrations of indene in Paraflex in wt%  3500 4000 4500 5000  Time (minutes)  238  6. TFU Fouling Studies  recorded local as well as overall fouling behaviour. Noticeable variations in local fouling behaviour were caused by the development of the thermal boundary layer and precipitation of entrained gum. SEM and optical micrographs confirmed that the foulant consists of particulates of insoluble gum. The profiles of deposit coverage provided a useful comparison for the thermal fouling results and permitted the direct calculation of the factor  PfAf. The unheated regions of the test section were coated with striations of insoluble gum which require further study.  239  7. Models of Aspects of Fouling  7.  Models of Aspects of Autoxidation Fouling  The experimental studies showed that chemical reaction fouling under autoxidative conditions is a complex phenomenon which cannot be explained by a simple, global fouling model. A qualitative model of the fouling process was developed, featuring the following mechanisms.  a.  FOULANT GENERATION The foulant precursor was identified as polymeric peroxides, formed by the reaction  of alkenes with dissolved oxygen. The tendency to form polyperoxides is determined by the alkene structure and by the chemical conditions in solution. Chain transfer to the solvent or other components, free radical scavengers and low oxygen concentrations were all found to interfere with the formation of polyperoxides. Fouling was not detected in the absence of polyperoxide gums, except at high antioxidant levels, so that the fouling process can be described as subject to bulk chemical reaction control. The induction periods observed in thermally initiated fouling experiments were due to the accumulation of hydroperoxide to a concentration large enough to cause the autocatalytic reaction with oxygen which generated polyperoxides. The reaction kinetics in the 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 not adequately described by the kinetic schemes in the literature. Significant solvent effects were observed in the behaviour of the polyperoxide reaction products. The aromatic polyperoxide has a limited solubility in aliphatic solvents; polyperoxide was precipitated as agglomerates of insoluble gum after the polyperoxide concentration exceeded the solubility limit.  240  7. Models of Aspects of Fouling  b.  DEPOSITION Fouling studies were only performed in aliphatic solvents as a suitable aromatic  solvent was not identified. Two deposition regimes were observed, involving different deposition mechanisms; an initial, pseudo linear fouling regime followed by an ‘accelerated fouling’ regime. Fouling rates were much higher in the latter regime and swamped any contribution from the mechanism generating deposit in the initial regime. The mechanism 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 was reached. This deposit was uneven and rough enough to disturb the fluid momentum boundary layer. The initial fouling regime was observed when the dissolved polyperoxide gum concentration 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 be expected if the precursor was formed on the surface alone but these were only observed in regions of low surface temperature where the gum could melt into the surface. Deposits formed 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 fouling rate increased with surface temperature and decreased with flow velocity. It is this regime which is of prime interest for modeling because of its industrial importance. Cases of accelerating fouling could arise in closed heat transfer systems using hydrocarbon oils as the heat transfer medium.  c.  DEPOSIT AGEING The material recovered from the fouled surface was chemically different from the  polyperoxide foulant precursor. The deposit aged to form polyperoxide degradation products on exposure to the enhanced temperatures on the heat exchanger surface. The  241  7. Models of Aspects of Fouling  ageing of deposit followed soon after adhesion to the surface and may be involved in the adhesion process itself. Deposit removal was not observed in the fouling studies.  The current work did not provide a complete understanding of all aspects of the fouling process. A complete mathematical model of the phenomenon requires further research and experimentation and was beyond the scope of this work. Certain aspects of the fouling process were investigated and are described in the following sections.  7.1  Autoxidation Kinetics in the Fouling Apparatus  The chemical induction periods in the thermally initiated fouling runs reported in Section 5.4 and by Zhang et at. (1993) were coupled to the operating conditions in the PFRU. No systematic variation with PFRU conditions was observed in the gum formation rate or indene kinetic constants in these runs; this is consistent with the oxygen mass transfer limitation, but could also be due to the lower activation energy involved in these processes. A simple model of the PFRU system was developed to study the coupling of bulk chemical reaction and fouling conditions.  7.1.1  Model Development  The PFRU system was modelled as a well mixed batch reactor with a pumparound through the fouling probe as shown in Figure 7.1. The system is assumed to be adiabatic apart from the cooling coils which maintain the holding tank at the set temperature, T . 1 Volume 2 represents the pump and the piping to the PFRU, where T 2 was slightly hotter than T 1 due to the dissipation of heat in the pump. The difference was estimated as 0.5C. Volume 3 represented the volume of liquid exposed to the high temperatures at the 6 1 PFRU surface (T ), assumed to be the volume of the thermal boundary layer. It is 3 f= T 1 242  7. Models of Aspects of Fouling  Figure 7.1  Compartmental Kinetic Model of PFRU System  C3 PFRU Volume 3  C2  Table 7.1  PFRU Reaction Coupling Model Parameters  Parameter  Paraflex Simulation  kerosene Simulation  10 L  10 L  Volume 2  0.84 L  0.84 L  Volume 4  0.42 L  0.42 L  80°C  80-84°C  17 L/min  7-15 L/min  Eact  40- 120 kJ/mol  40-74 kJ/mol  f 5 T  180-240°C  165-216°C  880-950 W/m K 2  880-1750 W/m K 2  Total Liquid Volume  Thulk Flow rate  0 U  243  7. Models of Aspects of Fouling  this part of the fouling ioop which operates at higher temperature and thereby gives rise to differences between kinetics in the isothermal and in the fouling experiments. The area available for heat transfer was known and the thermal boundary layer thickness, assumed to be given by öh  , so that V 0 ?JU 3  =  öth.(.rr  =  öth,  was  da,i L), where L is the PFRU  heated length. Thermal boundary layer development was ignored in this analysis. The temperature in the piping back to the reactor (Volume 4) was calculated from 4 T  2 T  =  (3t da,i L) (T 0 U 3 T )!(W.Cp) 2  +  [7.1]  -  Volume 1 was calculated knowing the total volume of liquid and the piping volumes 1 V  Viiq  =  2 V  -  -  3 V  -  4 V  [7.21  The effect of the enhanced temperature zone is modelled by considering the rate of reaction of a general species, concentration C. Assuming constant physical properties and flow rate, a mass balance onV 1 gives dt  =  (t)) 1 C  -  +  1 R  [7.3]  where x 1 is the space time in Volume 1, given by V 1 p1W. R 1 is the volumetric reaction rate in V 1 and has the form R 1 result when C 1  =  =  -  1C k 1 “. Equation [7.3] reduces to the isothermal batch reactor  . C 4 C 4 was calculated by assuming plug flow in the pumparound  volumes; for a first order reaction, where R = 1 1 k, 4 C C is given by -  (t) 4 C  =  (t 1 C  -  -  -  t)  exp -[k t+k 2 t 3  +  ] t 4 k  [74]  where is the space time in Volumej. This gives (t) 1 dC  =  (t) 1 -ç  (1  +  ) 1 kjt  +  3 (t -r- t 1 C -  In the case of zeroth order kinetics, where R 1 (t) 1 dC dt  (t) 1 -C =  +  =  -  -  ) 4 t  exp 3 +r t 2 -[k k  +  1 t 4 k  [7.5]  , [7.3] becomes 1 k  1 ( —(C ) 1 2 3 ) 4 t t-t -u -k k t [7.6]  Ti  Equation [7.61 is a linear o.d.e. with a delay term (t  -  2 + t t 3 + t ) 4  which does not have an  analytical solution. The equations show that the isothermal rate constant is augmented by a  244  7. Models of Aspects of Fouling  contribution from the pumparound which depends on the pumparound volumes and the relative sizes of the kinetic constants. The effective rate constant for the thermal boundary layer is a function of the temperature gradient across the layer. Paterson and Fryer (1988) obtained the following result for the overall rate constant in a boundary layer with a linear temperature gradient across its thickness, kr, where the rate constant obeyed the Arrhenius equation; kr  =  _B_ k(Tsurf) Fact  (  Tf  (Tsurf  -  Tb)1\  i  -  exp  (Tb  -  Tsuris)Eact  R Tf  [7.7]  ,f and Tb are the surface and bulk liquid temperatures respectively. This approximation 5 T was used in calculating k . 3 Equation [7.5] was solved numerically using a FORTRAN 77 program based on a simple Euler method to estimate the derivative. The initial conditions in the pumparound were represented by (t) 4 C  =  1 (0) C  t<t 2 + t 3 + t 4  and short time increments were used (At  =  2 O.2-O.3(r  [7.81 +  3 r  + t)).  The isothermal case was  compared with the analytical solution and showed negligible solution inaccuracy. The effect of the unheated flow loop and the heated section was quantified by comparing the value of C (t) with the isothermal result after a reasonable time period. A 1 first order indene rate constant of 8.5 x10 4 miw 1 was regressed from the data for the thermally 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 of effective rate constants. For first order kinetics, kcoupledfkiso  =  in (Ciso(t)/Ccoupled(t))  [7.91  For zeroth order kinetics, kcoupledlkiso  =  [C(t = 0) Ccoupled(t)lI[C(t = 0) C (t)] 0 -  The model parameters are summarised in Table 7.1.  245  [7.101  7. Models of Aspects of Fouling  7.1.2  Model Performance  A sensitivity analysis of the model showed that the rate constant activation energy was the most important parameter within the range of experimental conditions. The effects on the effective rate constants was similar under zeroth and first order kinetics. Figure 7.3 shows the effect of total volume on the rate constant ratio at activation energies of 120 and 50 kJ/mol at T 1  =  80°C, T 3  =  240°C, U 0  =  .K and W = 0.23 kg/s. The activation 2 950 W/m  energies are those obtained in Section 4 for the initiation and propagation stages in the autoxidation of indene. The effect of changes in liquid volume is larger at the higher activation energy, which indicates that liquid sampling would increase the effective rate constant and hence reduce the observed induction period. The model showed little effect of system parameters such as surface temperature at the lowest activation energy, which agrees with the experimental data in Table 7.2. This small effect of temperature could be due to an oxygen mass transfer effect, however, where the bulk oxygen concentration was negligible and there was no reaction in the bulk liquid. Such mass transfer effects are unlikely 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 runs involving indene in Paraflex in Section 5. The chemical induction period was assumed to be inversely proportional to the initiation rate and the rate enhancement predicted by the model was transformed into induction periods by using the data point at T ,f= 180°C as a 5 reference, 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 in the experimental data. The reaction zone volume had to be increased to unrealistic values in order to yield results close to the experimental data. Zhang et al. (1993) reported the induction periods from their thermally initiated PFRU runs using indene in kerosene and their data showed the same trend as the Paraflex  246  7. Models of Aspects of Fouling  Regression of First Order Indene Rate Constant from Data from SCR Experiment 122  Figure 7.2 0.40  0  0.35 0.30  I  -2  0.25 0.20  ‘S  0.15 -4 0.10  ‘-5  0.05 0.00  0  5  10  15  20  25  35  30  40  -6  Time (hours) 0.41 M indene in Paraflex: 79 kPa oxygen saturation: TbUlk = 80°C: thermal initiation  Figure 7.3  PFRU Reaction Coupling Model Predictions of Volume Effects in Kinetic Parameters in the PFRU System  2.5  •  •  0 0  2.0  0  0 0  1.5  e  1.0  0.5  0.0  4  0  Eact  =  1 20 kJ/mol  •  Eact  =  50 kJ/mol  5  6  7  .-  .-  -.  8  9  10  Volume (L)  247  11  7. Models of Aspects of Fouling  Figure 7.4  Comparison of Reaction Coupling Model Predictions and Experimental Induction Periods in Thermally Initiated Fouling Runs in Paraflex  : 20 15  -  10 0  Experimental  •  Model  I  0  160  180  I  200  •  220  I  240  260  T surface (°C)  Table 7.2  Solvent  Experimental Data from Thermally Initiated Fouling Experiments 0.41M Indene in Paraflex; 79 kPa Oxygen Saturation  Ner  Tsurf  Re  (CC)  kp.  Induction Period (experimental) (hours)  v(mol/L)Ihr  /  Paraflex  025  180  3050  26±2  0.00793  Parallex  031  200  3050  20  0.00861  Paraflex  028  210  3050  7-10  0.00849  Paraflex  032  225  3050  6±1  0.00765  Paraflex  029  240  3050  6±1  0.00723  248  7. Models of’ Aspects of Fouling  data 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 the decrease in  rh. 5 t  The model confirmed that thermally initiated reaction kinetics in the PFRU would be coupled to the heat exchanger performance but gave very poor agreement with the experimental data from the thermally initiated fouling runs. A more detailed analysis was not considered appropriate. The model did show that reproducible chemical reaction is unlikely in thermally initiated batch experiments if the PFRU surface temperature is varied. This conclusion supported the use of chemically initiated model solutions, which yielded more reproducible reaction behaviour.  7.2  Fouling Mechanisms in the Initial Fouling Regime  The mechanisms governing foulant generation and adhesion in the initial fouling regime were not completely understood. Existing fouling models were compared with the experimental data and a fouling model using the Paterson and Fryer approach was proposed.  7.2.1  Particulate Fouling Models  The TFU runs confirmed that the deposit consisted of small globules of insoluble polyperoxide gum which could have been generated in the bulk liquid and deposited on the hot 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 sizes estimated from the TFU micrographs were used to compare the observed fouling rates with estimates of particulate fouling rates. The unagglomerated globule dimensions fell within  249  7. Models of Aspects of Fouling  the range of O-5Ojm reported by Crittenden and Alderman (1988) in their discussion of particulate fouling. The rate of fouling by particulates transported from the bulk liquid to the wall in the absence of boundary layer gum generation is given by dRf/dt where  =  0 C /p ?f K  [7.11]  is the mass concentration of particulates in the bulk liquid and K 0 an overall mass  transfer coefficient including adhesion (ka) and mass transport (km) coefficients similar to those in Equation [2.33] 1/K  =  1/km  +  [7.12]  1/ka  The adhesion term often includes a surface temperature dependency and a sticking probability. Equation [7.12] assumes that the adhesion process is first order in particulate concentration, following the result of Ruckenstein and Prieve (1973). Mass transport of particulates in turbulent flow can occur via diffusion, inertial coasting of particulates through the viscous sublayer or by particle impaction (Melo and Pinhero, 1988). The dominant mechanism is determined by the particle relaxation time, t; t--  2d u 2 f___ p 18v 2  where  [7.131  4 is the particle diameter and Pp the particle density. Melo and Pinhero gave the  following ranges for each mechanism t<O.1  diffusion  0. l<t<l0  inertial coasting km  t> 10  inertial impaction  k  67 u*/11.8ScO.  =  =  (t  u*/5.23) (PIPp) exp (0.48t)  [7.14] [7.151 [7.16]  Turner (1993) discussed other forces (lift, drainage effects) involved in inertial coasting and suggested that the diffusionlcoasting bound (t = 0.]) should be lower in solid/liquid systems than the bounds above, which were obtained from aerosol studies. The Schmidt number in Equation 7.14 was calculated using the Stokes-Einstein equation: =  23 T(K)/(3njd) 1.38x10 250  [7.17]  7. Models of Aspects of Fouling  The particle sizes reported in Section 6 in Paraflex yield Sc> 106, which is outside the original range of Metzner and Friend’s correlation (Equation 7.14); its use at high values of Sc was defended by Muller-Steinhagen et at. (1988). The values of  for the four globule  size ranges were calculated at the conditions used in the fouling runs and are summarised in Table 7.3. The film temperature was used in calculating the fluid properties and the gum . The friction factor in the PFRU runs 3 density was estimated as that of indene, 990 kg/rn was estimated using (Rosenhow and Hartnett, 1973) f  =  [7.18]  ) 25 0.085 (Re-O.  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 the unagglomerated globules is by inertial coasting; the larger agglomerates belong to the particle impaction regime. The fouling rates in the absence of adhesion effects were estimated using [7.11] and the value of p 2 reported in Section 6.3. The maximum initial fouling rates for the group A particles were calculated and are given in Table 7.3 along with the experimental results. Bulk particulate concentrations of 5 g/L and 11 g/L (g*) were used at Tblk = 80 and 100°C respectively. The estimates for mass transfer controlled fouling are orders of magnitude too large and do not show the observed dependence on surface temperature. 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 isothermal turbulent flow regime were adhesion controlled and fitted the data to Epstein’s model with some success. The particles in his study belonged to the diffusion transport regime. The initial fouling rate regime, however, involves certain features which cannot be explained by a particulate fouling model. Firstly, the particles are relatively large and are transported to the surface by inertial coasting, which varies as u . The initial fouling rate 3 decreased with increasing u*; an adhesion term such as that in Epstein’s model would have  251  Table 7.3  Particle Relaxation Times and Mass Transfer Coefficients in Particulate Fouling Calculated at Conditions used in Fouling Runs, Assuming Mass Transfer Control  Conditions  Tsuff  Particle Size  (°C)  Quantity  Thermally Initiated Tbulk 80°C 687 miS um=O.  180  (PFRU)  Chemically Initiated Tbulk 100°C 6 mis um=O.SO  A 7-18j.tm  B 30-401.tm  C 66-75.tm  D 250j.tm+  Fouling Ratet K/Wmin) 2 (m  tp 8 kD (mis) x10 5 kim (mls)x10  0.005-0.03 6.3-4.6 4.025  0.08-0.53 2.4-1.99 66-130  0.41-0.53 1.42-1.3 400540  5.9 0.58 7970  4 4.06x10 [1.0x10 ] 7  240  tp 8 kD (mis) x10 5 kim (mis)x10  0.009-0.064 11-6 7.5-51  0.18-0.32 4.2-3.4 150290  0.86-1.11 2.47-2.3 1010-1480  12.4 0.96 37100  4 8.3x10 [14.4x10 1 7  200  tp 5 kim (mis)x10  0.004-0.029 2.5-16  0.08-0.53 47-86  0.41-0.53 300-360  5.9 4480  4 5.72x10  255  tp kim (mls)x 10  0.008-0.053 4.4-30  0.146-0.26 88-160  0.71-0.914 560-800  10.15 74400  4 10.7x10 [26.7x i0-]  200  tp 4 kim (flL/s)x10  0.014-0.09 1.5-10  0.2620.466 30-60  1.2681.638 242-373  255  t 4 kim (mls)x10  0.026-0.173 2.7-19  0.48-0.856 62-130  2.33-3.01 725-1300  187  tp 4 kim (mis)x10  0.042-0.028 8.5-63  0.784-1.39 220-530  3.79-4.91 459010090  0.0225 [5.8x10 ] 8  3 kim (mls)x10  0.065-0.429 1.3-10.1  1.193-2.12 41-113  5.77-7.46 1780-5160  0.0361 [23.4x10 ] 8  (PFRU) Uml.OO mis  Chemically Initiated Tbulk 100°C .O mis 2 um= 4 f= 0.008  222 (TFU)  -  experimental  kD kim  -  -  =  247°C:  35.8x i0  4 67.9x10 [7.4x10 ]  t calculated for 181.tm particles using Equation [7.11] where lfka 0; -  diffusive mass transfer coefficient inertial coasting mass transfer coefficient  []denotes experimental values  252  7. Models of Aspects of Fouling  1/u* which is difficult to explain in physical terms. The particles in Chen’s , to vary as 4 work involved diffusion mass transport, which varies as  u,  and the adhesion term was  interpreted in terms of the particle residence time at the surface. The fouling rates reported in the bulk precipitation regime in the PERU increased with flow velocity, however, which is consistent with the hypothesised change in fouling mechanism to deposition by agglomerate particulates. Secondly, a particulate fouling model does not account for the observed pseudolinear 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. This discrepancy could not be explained by an increase in the mean particle size as the bulk reaction progressed, as the mass transfer rate (which increases as d increases in the diffusion and inertial coasting regimes) would thus increase over time. A complex adhesion/deposit removal term would have to be invoked in order to predict the psuedo linear Rf behaviour. Thirdly, the particulate flux towards the surface moves along a temperature gradient and would be subject to thermophoresis effects in the thermal boundary layer. The thermophoretic velocity acts against the temperature gradient and would thus prevent particles reaching the surface. The thermophoretic velocity,  Utplj,  can be estimated from  (McNab and Meisen, 1973) —  Utph  —  where A and A  -0.26 ? v VT T (2 +  —  —  q 0.26 + )T (2  [7.19]  are the fluid and particle thermal conductivities respectively. Assuming that  0.12 W/m.K, the thermophoretic velocity at a film temperature of 130°C  /s) and a heat flux 2 m 6 l.37x10  (v  2 is estimated as 2.2x 10 m/s, which is of the 90 kW/m  same order of magnitude as the inertial coasting velocity given in Table 7.3. MUller Steinhagen  et  al. (1988) reported that Equation [7.19] overestimated Utph in their alumina  253  7. Models of Aspects of Fouling  particulate deposition studies and suggested a correction factor of 0.12; this still yields a 1 comparable to km for the group A globules. Thermophoretic effects would value of u thus be important in the initial fouling rate period. The  values are suspect, however, as  Equation [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 surface temperature (2.3x10 4 mis at a film temperature of 160°C) and would thus not contribute to the temperature dependency observed in the experimental data. Particulate fouling analyses are further complicated by the complex and poorly defined particle size distribution in the fouling solutions. The assumption that negligible reaction occurs in the boundary layer region is also highly suspect. Simple particulate fouling models were thus unable to explain the observed fouling behaviour.  7.2.2  Chemical Reaction Fouling Models  Panchal and Watkinson (1993) identified two chemical reaction fouling model cases which showed reasonable agreement with the experimental results from autoxidation fouling 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 to form foulant. The polyperoxide gum was assumed to be soluble so that convective mass transport applied. This case corresponds to Epstein’s model if adhesion effects are included in the foulant generation step and the polyperoxide is assumed to form insoluble globules at the wall. Adhesion effects are likely to control fouling as the convective mass transfer coefficients for the polyperoxide gum are large. Convective mass transfer coefficients km for indene and its monomeric, dimeric and trimeric peroxides under the fouling conditions in run 503 are estimated in Appendix A.3.4 using both Metzner and Friend’s result and the modified form proposed by Epstein (1993b) in his analysis of the fouling data of 5 Crittenden et al. (1987a). The values range from 2.8 to 1.5 xl0 mis and 1.2 to 0.7 x10 254  7. Models of Aspects of Fouling  mis, respectively. A maximum fouling rate can be estimated by assuming that all gum is converted to foulant (i.e. simple stoichiornetry, sticking probability  =  1) and that the  1 = 0.85 x10 soluble gum consisted of peroxide dimers, ROOROOH. Using k 5 mis and a  .s and a maximum 2 gum concentration of lOg/L, this yielded a mass flux of 0.05 1 kg/m 5 rn .KIW.min. 2 fouling rate of 2.78 x10 This rough estimate is two to three orders of magnitude too large and indicates that adhesion effects or the surface reaction rate dominate deposition in this case. This model does not explain the observed linear fouling rate behaviour, but does indicate that surface reaction mechanisms are significant. The rate limiting step in the reaction at the surface could also be the mass transfer of oxygen from the bulk fluid, where the concentration of oxygen is always less than that of indene in these experiments. If the autoxidation reaction in the bulk liquid is controlled by oxygen mass transfer (see Section 4.2), the concentration of oxygen in the bulk will be very small and will control any oxidation process occurring at the heated surface. This concentration is likely to remain constant during an experiment and would thus give rise to a constant fouling rate. A similar maximum rate argument to that above can be applied to the initial fouling rate in experiment 503 (8.1x10 8 rn .K/W.min) to estimate the 2 concentration of dissolved oxygen in the bulk fluid. Equation [7.11] is used with ilka  0  and a stoichiometric assumption that 1 mol of oxygen produces 0.1481 kg foulant (i.e. all products attach; no kinetic resistance; no loss of material on ageing); this is unlikely but yields an order of magnitude estimate. Equation [7.111 gives 2 Co ) 3 (mol/m  !60) (184.6) I [(0.1481)(2.17x10 8 (8.1x10 )] 5  0.08 mol/m 3 [7.20]  .KIW.s) (kg.W/m 2 (m .K) (rnol/kg) (s/rn) 4  The dissolved oxygen concentration under saturation conditions in run 503 was estimated at 4.9 mourn , which is 3  -  60x larger than this estimate of the bulk oxygen concentration.  This value is consistent with the autoxidation kinetic scheme described in Section 4 where reaction in the diffusive boundary layer gave rise to negligible oxygen concentrations in the 255  7. Models of Aspects of Fouling  bulk liquid. The stoichiometric assumption and the mass transfer coefficient are the primary sources of error in this analysis. If all the oxygen transported to the surface does not form foulant, the estimate of C 02 will be increased and could approach the saturation value. The fouling rate would still be controlled by the oxygen concentration and would still give rise to a linear Rtime curve. The case la model does not, however, predict the reduction in fouling rate with increased velocity noted in the fouling experiments. Oxygen mass transfer would be increased at higher flow rates, confirming that surface reaction and/or adhesion steps are involved in the fouling mechanism. The case 2 scenario proposed by Panchal and Watkinson was an attempt to model the complex reaction dynamics in the ‘reaction zone’ in the viscous sublayer next to the heat transfer surface. Convection in and out of the zone as well as enhanced reaction rates in the zone were estimated using a consecutive reaction scheme based on that of Norton and Drayer (1968). This scheme ignored the effect of indene concentration on the formation of insoluble polyperoxide and did not include adhesion effects, but did include oxygen mass transfer. The data from the tubular apparatus showed a quasi-linear initial fouling rate but the Case 2 calculations did not show this behaviour. This discrepancy suggests that the reaction dynamics are more complex than those used and require further investigation. This model assumed that the bulk reaction (and hence the surface reaction) was not oxygen-mass transfer limited. Panchal and Watkinson concluded, however, that the reaction zone analysis did yield the best agreement with the experimental results. There are several problems in modelling the reaction zone dynamics. The kinetic model of indene autoxidation in Section 4.6 indicated that the reaction stoichiometry is complex. The initial fouling rate activation energies observed in the fouling experiments are significantly larger than the activation energies obtained for indene autoxidation in the SCR experiments. This result suggests that the rate determining step in foulant generation differs from the bulk reaction kinetics, particularly as the dissolved oxygen concentration is 256  7. Models of Aspects of Fouling  suspected to be negligible throughout the liquid. The foulant generated, however, contains a 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 the reaction zone dynamics in the current work prevented a quantitative comparison with the case 2 scenario.  Analysis of Fouling Rate Behaviour  7.3  Paterson and Fryer’s ‘reaction engineering’ approach to milk fouling treated the complex reaction zone kinetics as a ‘black box’ and analysed the phenomenon in terms of parameters likely to influence deposition. This approach will be used here, invoking the following assumptions;  (a)  Bulk Chemical Reaction Control of the timing of fouling behaviour. The transition  from the induction period to the initial fouling rate period is determined by the bulk reaction where most free radical formation occurs. The transition from the initial fouling rate period to the bulk precipitation phase is determined by the solubility of the polyperoxide gum products.  (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 formed initially in the reaction zone would be convected out. The formation of deposit is determined by the rate of reaction in the reaction zone and the kinetics of the adhesion mechanism. The rate of reaction is assumed to be proportional to the bulk reaction rate via the 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 by the fluid residence time in the reaction zone, 0.  257  7. Models of Aspects of Fouling  (c)  Rapid Ageing. The adhesion process is assumed to involve part of the ageing  mechanism so that the deposit thermal conductivity does not change significantly after deposition has occurred.  (d)  Negligible Removal. No evidence for a removal mechanism was observed.  7.3.1  Formulation of a Lumped Parameter Fouling Model  A compartmental fouling model such as case 2 above requires a working knowledge of the reaction kinetics in the reaction zone. A lumped parameter approach is used 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 fouling precursor in the reaction zone, which lies in the viscous sublayer adjacent to the heat transfer surface. The fouling rate can be described in terms of a surface renewal model or a film model, shown schematically in Figure 7.5. In the former, elements of fluid remain at the surface for a given residence time during which insoluble gum is generated, which can then react at the wall to form deposit. A surface renewal model involves mean residence times and surface reaction parameters which are not available for this case, so the film model approach will be used.  The film model is similar to that described by Epstein (1993b), where the fouling rate is given by dRf/dt  =  Rf  =  (l/pf) (rate of reaction at surface)(attachment factor)  258  [7.21]  7. Models of Aspects of Fouling  when mass transfer effects are negligible. Epstein assumed that the attachment factor depended 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 Mechanism 0  Bulk Fluid  o  rapid mass transfer  I  elements of fluid in bulk .  i  .  Boundary Layer  0  i  ‘  : :  renewal  /  elements of fluid at surface  -  I  Heat Transfer Surface Surface Renewal Model  Film Model  v/u* which mirrors the periodicity of turbulent bursts in , 0 was taken as proportional to 2 Cleaver and Yates’ model for particle deposition. Assuming that the initial fouling rate, chemical reaction rate and attachment factor are described by Arrhenius-type kinetics, the temperature dependence in equation [7.20] is given by Epj  =  0 EG + Eatt + E  [7.22]  where EG is the activation energy of the reaction generating polyperoxide gum precursors and E 0 is the temperature dependency in the residence time at the surface. The overall fouling rate is then given by Rf.  oc  Rf  =  (l/fpf) (reaction rate) 9 exp (EattfRTsuri’)  [7.23]  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 the concentrations of indene and its oxidation products, {C}.J3 is a constant which has the units <mis> when RG is given in <kglm .s>.. Assuming that 0 3 becomes  259  cc  , Equation [7.24] 2 v/u*  7. Models of Aspects of Fouling  Rf  =  2 exp (Ealt/RTsurf) (C vlu* 0 R ) 13’ (lIX-pf) 1  [7.251  The trends in the experimental data were compared with the behaviour predicted by Equation [7.25].  7.3.2  Analysis of Experimental Fouling Data  inspection of Equation [7.25] shows that the initial fouling rate varies with the rate of chemical reaction, the surface temperature and the flow velocity. Multiple parameter regression analysis was not used owing to the scatter in the data. The approach taken was to determine the dependency on temperature and reaction rate before investigating velocity effects. The calculation results are summarised in Appendix A.4. The effect of surface temperature on the initial fouling rate is shown in the Arrhenius plots in Figure 7.6. Each series of runs was performed with the same indene batch, flow rate and initiation method. The activation energies are given in Table 7.4 and can be seen to lie within one standard deviation of each other. The difference in standard ) for each 2 deviations was not significant for this sample size. The correlation coefficient (R activation energy is not significant at the 1% level, indicating that the rates are related by the Arrhenius equation (for these sample sizes). The initial fouling rate at 187°C in the TFU involves the largest experimental error; in its absence, the TFU activation energy figure would be 88 ±6 kJ/mol, which is closer to the PFRU values. A comparison of flow velocity effects must be performed at a constant surface temperature. The TFU flow variation runs were performed at 210°C so this was used as the reference temperature. Equation [7.22] indicates that the fouling rate activation energy includes a contribution from the mean residence time, mainly via the kinematic viscosity. The fouling rates were scaled to 210°C using Iooc) 2 (Rf7v)(  =  0 (Rf/V)(Tsur  exp [-EactIR (1/483.15 1/T)1 -  260  [7.26]  7. Models of Aspects of Foulin2  Figure 7.6  Arrhenius Plot of Initial Fouling Rates  —11  -12  I  -13  -14  I  -15  -16  -17 0.0018  0.0019  0.0020  0.0021  0.0022  0.0023  1/T(K)  Table 7.4  Comparison of Initial Fouling Rate Activation Energies Statistical Parameters  Fouling Probe  Initiation Mode  Eact  0J3  (kJ/mol)  (kJ/mol)  2 R  t-test ‘t  ti”  (data)  (comoarison)  PFRU  thermal  84.8  13.2  0.953  2.32  0.29  PFRU  chemical (2.5 mM bP)  81.9  16.4  0.976  4.99  0.05  TFU  chemical (1.0 mM bP)  76.4  8  0.968  2.32  0.58  t  calculated using Ernean  =  81 kJ/mol; tt’  =  abs[Eact 81 1/°E -  261  7. Models of Aspects of Fouling  Figure 7. 7  Arrhenius plot of (Rf7V) against Surface Temperature  -2  -3 -4 14  -5 ‘S -6 -7 -8  0.0018  0.0019  0.0020  0.0021  0.0022  0.0023  1JT(K)  Table 7.5  Comparison of Activation Energies of Adjusted Fouling Rate (Rf/v) Statistical  Fouling Probe  Initiation Mode  Eact (kJ/mol)  (kJ/mol)  Parameters t—test t  ‘tt’  (data)  (comparison)  PFRU  thermal  96.6  11  0.963  -8.79  0.23  PFRU  chemical (2.5 mM bP)  97.2  10  0.946  -9.36  0.32  TPU  chemical (1.0mM bP)  88.2  8  0.977  11.4  -0.74  t calculated using Enean = 94 kJ/mol; ‘tt’ = abs[Eact  262  -  7. Models of Aspects of Fouling  The 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 range from 88-97 kJ/mol, which is close to the sum of the energies of the initial gum degradation rate (40 kJ/mol) and the indene consumption rate (ER  48-57 kJ/mol). This result  suggests that the oxygenated reaction occurs in the boundary layer region, which infers that the bulk oxygen concentration is not negligible. A reliable method of measuring the dissolved oxygen concentration is required to verify this hypothesis. The fouling rate data was corrected to 210°C by calculating (Rf/v) and applying Equation [7.261 using the activation energies in Table 7.5. Each series of data in the Arrhenius plots was represented by a single value (taken from the regressed line at Tsurj = 210°C) for subsequent checking of velocity effects. The effect of the chemical reaction rate is demonstrated by the PFRU data in Figure 7.7, where the chemically initiated fouling rate is markedly faster than the thermally initiated fouling rate. The flow velocity conditions were similar (0.51 mIs vs. 0.68 mIs) but the bulk reaction rate constant was markedly larger in the chemically initiated case. This is consistent with a surface renewal model, where the reaction at the wall will be dependent on the concentration of free radicals brought to the surface in a fluid packet. The surface reaction rate in the film model would be expected to be proportionately higher than the bulk reaction rate. The chemical reaction rate in the bulk varied between series of experiments due to differences in indene batch, reactor configuration and bulk liquid temperature. The reaction stoichiometry is also unknown. For the purposes of comparison, the chemical reaction rate was assumed to be proportional to the maximum reaction rate of indene in the bulk liquid. The investigation of indene autoxidation in Section 4 showed that indene reacts to form other products than the gum fouling precursors. Equation [4.34] gives the bulk reaction rate of indene in kg/rn 3 .s as proportional to kR and the concentration of indene. 0 R  a  kR Vtindenej (116.1)/3600 263  [7.27]  7. Models of Aspects of Fouling  where kR is in ‘J(mo1JL)/hr. Equation [7.25] can be arranged to give [(Rf) 2 (pf f )]I(v RG)  J exp [-ERfIRTSUff] 2 B” [1/u*  [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, =  ] exp [-ERfIRTSUff] 2 B” [l/u*  e, gives [7.29]  The PFRU data from Figure 7.7 were replotted as an Arrhenius plot of Equation [7.29] in Figure 7.8; the difference in flow velocities has been ignored in this figure. The value of K. Figure 7.8 shows that the kg.W1m . pf)t.f used was that obtained from Section 6, 184.6 4  transformed fouling rates seem to be in good agreement, so this basis was used in comparing 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 without a working model of the boundary layer kinetics. The maximum difference between transformed values in Figure 7.8 occurred at 225°C and was  50%, which gives an  indication of the size of errors expected in the model. When the fouling rate data is corrected to a common surface temperature, 0 will then represent the velocity dependence in the attachment factor.  The fouling model in Equation [7.25] indicates that the fouling rate is inversely proportional to u , which is given by Um 2 f/2. The friction factor in the PFRU had to be 2 estimated as pressure drop data were not available. An estimation method based on the heat and mass transfer analogy represented in Equation [6.1] is described in detail in Appendix A.4. This analysis gave the following correlations for use in the fouling model calculations; ±0.02  f (PFRU)  =  0344 0.295 (±0.03) Re  f (TFU)  =  0203 0.059 (±0.0 19) Re  [A.4.3]  [A.4.2]  The fouling rate data was used to calculate (Rf/v) at T,’= 2 10°C using Equation 1 [7.26]. This value was used to calculate  O(2100c),  264  which is plotted against the estimated  7. Models of Aspects of Fouling  Figure 7.8  Arrhenius Plot of  e v. Inverse of Surface Temperature for PFRU Runs  12  •  o  thermal initiation  •  2.5 mM bP initiation  •  10  • 0 0  9  • 8  0 0  7 0.0018  I  0.0019  I  0.0020  0.0021  0.0022  1!T(K) Thermal initiation  : 0.41 M indene in Paraflex, Tbujk  =  80°C, Urn  Chemical initiation  : 0.41 M indene in Paraflex, Thulk  =  100°C, Urn  265  =  0.68 rn/s =  0.51 mIs  0.0023  7. Models of Aspects of Fouling  friction velocity in Figure 7.9. The Figure is a log/log plot and shows thatO decreases as friction velocity increases. There is considerable scatter in the experimental data, which is expected given the degree of estimation involved in generating the plot. The data point at the largest friction velocity involved a very small fouling rate and was subject to large experimental errors (-.50%). The figure shows that the TFU data seem to follow the same trend as the PFRU data, suggesting that the same fouling mechanism is involved in both fouling probes. The TFU data are widely scattered and more information is required to confirm this hypothesis. The figure shows the line of best fit obtained by least squares regression of the data, which gave log 0  =  -1.98 (±0.3) log u*  +  0.97 ±0.30 (R 2  =  0.74)  [7.30]  The initial fouling rate can thus be expressed by Rf  =  B”  V  198 RG exp (-11306/Tff) u  [7.31]  where RG  =  and B”  7.4 x10 8 K/W.kg.s° 498 The activation energy in Equation [7.31] is the m . 98  =  0.03225 kR /[indene] 3 (kg/m . s); kR is in (1(mol/L)/hr), [indene] in (molJL)  mean value of the values given in Table 7.5.  Equation [7.29] predicts a friction velocity dependence of -2, which is close to that observed. The average absolute deviation from the regressed result [7.30] was calculated as 49%, which is larger than the experimental error in most runs. The systematic error from the reaction and friction factor estimates was not known. Further regression analysis was not performed given the degree of estimation involved in generating the data. Fortuin et al. (1992) developed an Extended Random Surface Renewal model for f 2 turbulent flow in liquids, which predicted that the mean residence time varied as 4v/u*  This result would also give a velocity dependence similar to that observed in Figure 7.9, but verification requires better quality data.  266  7. Models of Aspects of Fouling  Figure 7.9  Fouling Model Analysis: Variation of 0(210°C) with Friction Velocity  •  •  •  D 0 \  D  3.5  Q D  \  C [  e  3  e  2.5  I  2  -1.6  •  -1.4  I  •  -1.2  I  -1  •  I  -0.8  -0.6  log {u* (mis)]  C  PFRU, 2.5 mM bP  PFRU, 2.5 mM bP, Tseries  o  •PFRU, 1 mM bP  TFU, 1mM bP, Tseries  0  PFRU, thermal, Tseries  TFU, 1mM bP  267  7. Models of Aspects of Fouling  Paterson and Fryer’s analysis (1988) pictured the reaction zone as a differential reactor where the volume was proportional to the thermal boundary layer thickness,  th• 3  Incorporating this factor into the film model gives a fouling model of the form Rf  =  B” (l/ f’pf) RG(Cj) 2  th  2 exp (Eatt/RTsurf) vlu*  [7.32]  wherej3” has units of s. This model can be developed using the same assumptions as above and gives the result Othth  =  ] exp [-ERfIRTSUffJ 2 B” [1/u*  [7.33]  The thermal boundary layer thickness can be estimated from U ,/2i and was used to calculate 1 /öth• 0  The temperature variation in  )L.  was ignored in correcting the data to 210°C. Figure  7.10 is a log-log plot of the experimental data plotted as Equation [7.33] and was regressed using least squares to give log  (e/h) =  5.9  -  1.12 logu*  =0.501) 2 (R  [7.34]  This approach increases the scatter in the data and gave a larger deviation from the predicted velocity dependence.  Further fouling model analysis was not performed as the uncertainties in the reaction dynamics and friction velocity hinder numerical verification. The experimental data involved 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 the observed fouling behaviour. Further modelling studies requires better quality data and further investigation of the reaction at the heat transfer surface.  268  7. Models of Aspects of Fouling  Paterson and Fryer Fouling Model Analysis:  Figure 7.10  Variation of G(2100C)/&h with Friction Velocity  8  •  •  D 7.5  •  o  -  D D D :°  6.5-  6  I  -1.4  •  -1.2  -1  I  -0.8  -0.6  log [u* (mis)]  D  PFRU, 2.5 mM bP  PFRU, 2.5 mM bP, Tseries  0  PFRU, 1mM bP  TFU, 2.5 mM bP, Tseries  o  PFRU, thermal, Tseries  TFU, 1 mM bP  269  8. Conclusions  8.  Conclusions  Chemical reaction fouling caused by the products of the autoxidation of hydrocarbons 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 on the fouling rate in two different fouling probes operating under turbulent flow conditions. The current study represents the first systematic study of autoxidation fouling in heat exchangers operating under conditions of turbulent single phase heat transfer. Chemical monitoring of the process was used successfully to relate the observed fouling behaviour to the known reaction chemistry and to formulate a qualitative model of autoxidation fouling in these systems. This model included deposit ageing, which has been reported in other chemical reaction fouling systems. Chemical initiation was used to decouple the effects of reaction rate and surface temperature in the analysis of fouling rate data. A new fouling probe was commissioned and compared favourably with an existing device. 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 industrially important systems.  1.  MODEL SOLUTIONS The 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 annular PFRU fouling probe in order to select a model solution for further study. The experiments were performed in the batch mode and chemical analyses were developed to follow the progress of the autoxidation reaction. Fouling was caused by deposition of the insoluble polymeric peroxide products of the autoxidation reaction, confirming the hypothesis of Taylor and Wallace (1967). Polyperoxide formation was determined by the reactivity of the 270  8. Conclusions  alkene peroxy radical and was controlled by chemical factors including oxygen concentration, alkene structure, solvent nature and solvent structure. Fouling was thus controlled by the bulk chemical reaction, which explained the fouling behaviour observed by Asomaning and Watkinson (1992) and Zhang et al. (1993). A solution of 5wt% indene in 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 FOULING The fouling behaviour was controlled by the bulk chemical reaction. A chemical  induction period was observed under thermally initiated conditions; this was linked to the accumulation of free radicals in solution, which was affected by the temperature in the fouling probe. A simple model of the annular probe recirculating system (PFRU) did not explain the trends observed in the induction period data of Zhang et at. (1993) and in the current work. The fouling resistance in Paraflex increased at a nearly constant rate until the soluble polyperoxide concentration reached a solubility limit, g*. This initial fouling rate is thought to be controlled by the complex reaction and adhesion processes in the hot boundary layer region adjacent to the heat transfer surface. Once the gum concentration exceeded the solubility limit, fouling was controlled by the deposition of globules and agglomerates of insoluble gum entrained in the bulk fluid. The fouling rate in this phase increased with time and was orders of magnitude larger than the initial fouling rate. Similar fouling resistance behaviour was observed in both the annular PFRU and tubular TFU fouling probes. The pressure drop caused by fouling increased significantly in the later fouling period and the large globules and agglomerates caused roughness effects. Negative fouling resistances were reported at low velocities when large agglomerates deposited, disrupting the heat transfer boundary layer. 271  8. Conclusions  The initial fouling rates were relatively small (106108 m .K/W.min) and 2 increased with surface temperature. Activation energies were estimated assuming an Arrhenius relationship dRf/dt  =  a exp (-Eact/RTsf)  [8.1]  as 84.8 ±13.2 kJ/mol (PFRU, thermal initiation), 81.9  ±  16.4 kJ/mol (PFRU, chemical  initiation) and 76.4±8 U/mo! (TFU, chemical initiation). The initial fouling rate decreased with flow velocity, varying as u where -1<n<-2. The later fouling rate did not vary significantly with surface temperature and appeared to increase with flow velocity, which was consistent with fouling controlled by the deposition of large particulates. The initial fouling rate variation with bulk temperature was complicated by bulk reaction and velocity effects.  3.  AUTOXIDATION KINETICS AND BEHAVIOUR The autoxidation of indene in kerosene and Paraflex was studied in a semi-batch  reactor under the conditions used in the fouling experiments. Solvent effects were observed in 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 the most common mode! solution used (5 wt% indene in Paraflex at Tblk  =  100°C) and  depended on the physical nature of the solvent (temperature, aromaticity). Insoluble polyperoxide was precipitated as gum globules ranging in size from 30 jm+, once the solubility limit was reached. The indene reaction kinetics in the kinetic and fouling experiments were subject to oxygen mass transfer effects after the chemical induction period. The data was successfully fitted to a kinetic scheme based on mass transfer with zeroth order chemical reaction in oxygen, first proposed by Van de Vusse (1961). The activation energies reported for the disappearance of indene (48-58 kJ/mol) did not agree with the literature values for well defined initiation schemes. 272  8. Conclusions  A kinetic model was proposed based on a set of series-parallel reactions which showed the trends observed in the experimental data, but requires experimental verification.  4.  ANTIOXIDATION A 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 extended the chemical induction period but did not affect the autoxidation rate, as reported by Howard and Ingold (1962). GC-PID analysis showed that the end of the induction period corresponded to the exhaustion of BMP. The activation energy of indene initiation was estimated as 120 kJ/mol. The antioxidant was less effective under heat transfer conditions, marked by a reduction in the length of the extended induction period. The fouling and autoxidation mechanisms did not show any noticeable change after the induction period. High concentrations of phenol caused extra fouling in the heat exchanger and indicated that such antioxidants should not be used above their ceiling temperatures.  5.  DEPOSIT MORPHOLOGY AND AGEING The deposit formed during the initial fouling period consisted of coloured gum  veneers on cooler surfaces and small (6-17 pm) globules of insoluble material on hotter surfaces. The deposits formed during the later fouling regime consisted of larger, randomly distributed globules of insoluble gum. The deposit chemical activity was consistent with ageing of the deposit by the degradation of polyperoxides on the hot surface. This thermal degradation mechanism was confirmed by ageing simulation experiments. Ageing was thought to be involved in the adhesion of the polyperoxides to form deposit. The thermal conductivity of the deposit was estimated by a novel method in the  TFU as 0.19 W/m.K, which is in good agreement with the literature values. 273  8. Conclusions  7.  TUBE FOULING UNIT (TFU)  The TFU is a novel device for studying fouling which was design ed and commissioned during the current work. The device allows the inspection of deposits in situ and was found to give reasonably reproducible fouling behaviour. Polyperoxide s were also found to cause cooler fouling in the TFU.  8.  FOULING MODELS  The initial fouling rate behaviour could not be explained by particu late fouling models, or by chemical reaction fouling models which did not accoun t for reaction and adhesion effects in the reaction zone next to the heat transfer surface. The fouling resistance profiles in the initial fouling rate period could not be quantitatively explain ed by the current work. A lumped parameter fouling model was shown to predict the trends observed in the experimental data but involved two parameters which were difficult to quanti fy.  Novel aspects of the study include the investigation of parameters involved in selecting a suitable model solution; the use of chemical initiators to decoup le the effects of reaction rate and surface temperature; the ageing simulation work and many features of the TFU, which represents the application of techniques used in two differe nt cases of chemical reaction fouling to provide extra insight into the fouling processes in these hydrocarbon systems.  274  9. Recommendations  9.  Recommendations for Further Study  The current work explained some of the features observed in autoxidation fouling but also generated questions which require further further investigation. These are  1.  Oxygen in Autoxidation Fouling. A reliable method of determining dissolved  oxygen concentration was not available during the current work. The oxygen mass transfer limitation 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 the development of a reliable fouling model.  2.  Autoxidation Mechanisms. The product analysis methods could be improved to  yield further information required by a kinetic model of indene autoxidation. A method to determine insoluble gum agglomerate sizes in solution would be useful in studying the final fouling rates. The chemical reactions involved in the reaction zone are not completely understood 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 polar  solvent for model solution studies. Fouling behaviour was found to be linked to polyperoxide solubility in aliphatic solvents and it would be useful to investigate the influence of aromatic or polar solvents where the autoxidation products were more soluble.  4.  PFRU Apparatus. Several features of the PFRU could be improved on before  performing further autoxidation studies. The probe alignment, gas sparging system, cooling system and data collection methods need attention. This device was used to a much greater extent than was initially expected.  275  9. Recorntnenda lions  5.  TFU Apparatus. The TFU pump and connections to the heated section should be  replaced. The configuration of the device requires revision if it is to be used to study autoxidation fouling further as the cooling system and transformer imposed limits on the operating range in the current work. Further TFU runs would be useful in verifying the effect of flow velocity in autoxidation fouling. The TFU fouling rates in the current study are small and these runs should 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. These methods 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 UBC  and 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 general model of fouling under autoxidative conditions. The aim of this model would be the extension of the batch fouling studies to continuous systems as found in the petrochemical industry.  276  Abbreviations  Abbreviations  ABN  2-2’-azobis-2-methylpropionitrile  AIBN  Azodiisobutyronitrile  BMP  2,6-di-t-butyl-4-methyl-phenol  bP  Benzoyl peroxide  DCP  Di-cyclopentadiene  RD  Flame lonisation Detector  FTIR  Fourier Transform Infra Red (Spectroscopy)  GCMS  Gas Chromatography Mass Spectrometry  HWP  Hot Wire Probe  MIBK  methyl-isobutyl-ketone  n.m .r.  Nuclear Magnetic Resonance (Spectroscopy)  n.t.p.  Normal Temperature and Pressure (20°C, 101.3 kPa)  PFRU  Portable Fouling Research Unit fouling apparatus  PID  Photo-lonisation Detector  POx  Peroxide Number  SCR  Stirred Cell Reactor  TBA  Tri-t-butylamine  TCD  Thermal Conductivity Detector  TFU  Tube Fouling Unit  TFU  Tube Fouling Unit  TGA  Thermal Gravimetric Analysis  THF  Tetrahydrofuran  XPS  X-ray Photoionisation Spectroscopy  -  277  Nomenclature  Nomenclature  A  surface area  a  generic constant  1 a  stoichiometric constant, species i  AM b  surface area, mass transfer  2 m  time constant (Kern and Seaton model)  1/s  Bif  fouling Biot number  1 C  concentration, speciesj orifice discharge coefficient  3 moIIm  concentration of precursor or particulates  , g/m 3 mourn 3  Cp  heat capacity  J/kg.K  D  diffusion coefficient  /s 2 m  dh , 0 d  hydraulic diameter  m  diameter; outer,inner  m  e  kinetic model, propagation efficiency  Eatt  activation energy of attachment process  J/mol  activation energy of gum formation process  J/mol  activation energy of stepj  J/mol  Cor C  2 m  -  -  -  -  0 Ej,  overall activation energy J/mol activation energy/temperature dependence in residence time J/mol  Epj F  overall activation energy of fouling process  f  friction factor (Fanning, unless specified)  G g*  mass flux (Epstein, 1993b)  .s 2 kg/rn  soluble gum solubility limit  g/L  H h  enthalpy local heat transfer coefficient  J/mol  I  current  amps  general function  J/mol -  -  .K 2 W!m  Souder viscosity constant  -  J  mass flux  K  kinetic model ratio of kinetic constants  K, K  pressure drop coefficients; entry, exit  kciep  mass transfer factor  kD  rate constant, Equation [5.2]  .s 2 kg/m -  278  /mol.min 3 m -  Nomenclature  1 k  rate constant, initiation  1 k  overall rate constant, initiation  molJL.min  klmG  mass transfer coefficient; liquidlgeneral, gas phase lumped kinetic constant, equation [4.**]  mis, mol/N.s  kM,po 14 k 3 , 1  kinetic constant, mass transfer, peroxide  ./(mol/L)/hr  kj  first order kinetic rate constant (Norton and Drayer model)  0 lcd, k’ 0 K  zeroth order reaction rate constants  1/hr 1I(L/mol.hr)  overall mass transfer coefficient  rn/s  kpD  kinetic constant, peroxide decomposition  kR  kinetic constant inc. mass transfer, Equation [4.34]  kr KVa,b,c  general reaction rate constant  L  dimension, length  Km  kinetic rate constant, step n i.e. Equation ,  [341  /(mo1/L)Ihr  constants in kinematic viscosity correlation, Equation [3.5] m  Ostwald coefficient, oxygen  -  M  Molecular weight  m  mass  kg  1 N  total number of mols, species i  mol  Nr Nu  foulant generation rate, Paterson and Fryer model  kg/s  P  pressure  Pa  pair  saturating air pressure  kPa  POx  Peroxide Number  meq/L  Pr  Prandtl number  q  heat flux  2 kW/m  Qe  heat supplied as electrical power  W  Qi  heat losses to ambient  W  Q  heat transferred to liquid  W  R  universal gas constant  J/mol.K  -  Nusselt number  -  -  Regression coefficient of variation  -  [RH]  concentration of species RH  3 molIm  Rf  fouling resistance  .K/W 2 m  Rf, 0 Rf*  overall fouling resistance, TFU  .KJW 2 m  asymptotic fouling resistance  .K/W 2 m  R  linearfouling rate  .K/W.min 2 m  RG  volumetric reaction rate, foul ant generation  .s 3 moIIm  279  Nomenclature  R  free radical initiation rate  .s 3 mourn  R  volumetric reaction rate, species j  .s 3 mourn  0 , 1 r  radius; inner, outer  m  0 R  volumetric disappearance rate, zeroth order  .s 3 mourn  Re  Reynolds number  S Sc  sticking probability  Sh  Sherwood number  St, St’  Stanton, modified Stanton number  T  temperature  °C, K  t  time  s, mm, hr  t  time taken to reach g*  hr  tD  diffusional contact time  s  tres  dimensionless particle relaxation time residence time  Ttr U 0 U  Schmidt number  -  -  s  triple point  K  overall heat transfer coefficient  .K 2 W/m  clean overall heat transfer coefficient  W/m K . 2 rn/s  friction velocity bulk mean velocity thermophoretic velocity  m/s  V v  volume voltage  m 3 Volts  1 V  molar volume, speciesj  rnlJmol  W w  mass flow rate  kg/s  Utph  x  rn/s  weight fraction, species j  -  XE  dimension dimension oxygen exhaustion  m  Xf  deposit thickness  m  x  PFRU thermocouple depth  m  -  Subscripts a,att  adhesion,attachment  a,i/o  annulus, inner, outer  act  activation energy  280  m  Nomenclature  ageing  ageing processes  amb  ambient  AOx  antioxidant  b,bulk  bulk liquid  back  back diffusion (Crittenden et at. 1987b)  cony  convective heat transfer  coupled  coupled kinetics  D, diff  diffusion  dep  deposition, deposit  eff  effective  f  foulant  f.pt.  freezing point gas phase  g in  inlet  md  induction  unit  initiation  ins  insulation  mt iso  gas/liquid interface isothermal  I  liquid phase  m  mean  M,mass  mass transfer  met  metal  mix  mixture  or out  orifice outlet  p POx  particle or precursor  prop  propagation  rem  removal reaction  rxn RM  peroxide  Russell’s kinetics, including mass transfer effects  RP  Russell’s kinetics, excluding mass transfer effects  s, surf  surface  t,  i, o  term  tube, inner, outer termination 281  Nomenclature  w, i, o  tube wall, inner, outer  Greek 1 etc. B,B  general constant Hildebrand solubility parameter  05 (J/cm3)  AHf  enthalpy change, solvation  i/mo!  AP  pressure drop  Pa  temperature difference  K  o  time increments filmldeposit thickness  s m  E  Hayduk & Minhas factor  0  fouling flux mol fraction, species i  AL  y  -  .KJJ 2 m -  activity coefficient  -  thermal conductivity  W/m.K  v  dynamic viscosity kinematic viscosity  Pa.s /s 2 m  e  ratio of fouling rate to reaction rate  (s/rn)  0  residence time  s  p  density  3 kg/rn  a  standard deviation  WD  induction period, space time  hr  wall shear stress  Pa  deposit strength factor  -  282  References  References Alexander, J.M. (1990). Gas Phase Ignition in Liquid Phase Air Oxidation Processes, Trans.I.Chem.E. 68B, p17-23. ,  Asomaning, S. (1990). The Role of Olefins in Fouling of Heat Exchangers, M.A.Sc thesis, University of British Columbia. Asomaning, S. & Watkinson, A.P. (1992). Heat Exchanger Fouling by Olefin-Kerosene Mixtures, 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, Amsterdam Atkins, 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 Whey Protein Solutions. Proc. 3rd UK Natl. Heat Tfr. Conf., Birmingham, UK, I.Chem.E., Rugby, UK Bohnet, 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 Chemical Engineer #255, . p 3 91 95 ,  Bott, 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, UK Bowen, B.D. (1978). Fine Particle Deposition in Laminar Flow. Ph.D thesis, University of British Columbia. Bradley, S.E., Fryer, P.J., Griffin, T.R. & Wilson, D.I. (1989).The Mechanisms of Food 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 Heat Exchangers, Biofouling, 5, 295 14 p 3 Braun, R. & Hausler, R.H. (1976). Contribution to Understanding Fouling Phenomena in the Petroleum Industry, Paper 76-CSME/CSChE-23, 16th National Heat Transfer Conference, Aug. 8-11, St. Louis, MO, Brill, W.F. (1965). Selectivity in the Liquid Phase Autoxidation of Olefins. Adv. Chem. Ser. 51, . 70 p 8 -  283  References  Butler, R.C. & McCurdy, W.N.J. (1949). Fouling Rates and Cleaning Methods in Refinery Heat Exchangers, Trans. A.S.M.E. 71, p843-849. ,  Canapary, R.C. (1961). How to Control Refinery Fouling, Oil & Gas J. 9, p 1 14-118. ,  Carisson, D.J. & Robb, J.C. (1966). Liquid Phase Oxidation of Hydrocarbons, Trans. Faraday Soc., 62, . p 3 403 415 Chantry, W.A. & Church, D.M. (1958). Design of High Velocity Forced Circulation Reboilers for Fouling Service, Chem. Eng. Frog. 54, . p 6 4 7 ,  Charlesworth, D.H. (1963) Atomic Energy of Canada Ltd. Report, Report #AECL-1 761 Chen, 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 and Technology. Kiuwer Academic Publishers. p49-66. Chenoweth, J.M. (1990). Final Report of the HTRI/TEMA Joint Committee to Review the Fouling Section of the TEMA Standards, Heat Tfr. Eng., 11,(1), p73-107 Chhabra, R.P. (1992). Prediction of Viscosity of Liquid Hydrocarbon Mixtures, A.I.Ch.E. Journal 38, . p 1 657 661 ,  Cleaver, J.W. & Yates, B. (1975). A Sublayer Model for the Deposition of Particles from Turbulent Flow, Chem. Eng. Sci. 30, . p 9 83 92 ,  Cleaver, J.W. & Yates, B. (1976). The Effect of Re-entrainment on Particle Deposition, Chem. Eng. Sci. 31, . p 1 47 51 ,  Coulson, J.M., Richardson, J.F. & Sinnott, R.K. (1983). Chemical Engineering Vol.6. Pergamon, Oxford Cowan, J.C. & Weintritt, D.J. (1978). in Water Formed Scale Deposits. pubi. Gulf, Houston, TX Crawford, J.E. & Miller, J.M. (1963). Refinery Process Fouling Effect of Trace Metals and 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 of Chemical Reaction Fouling, Chem. Eng. Res. Dev. 65, p165-170. ,  Crittenden, B.D., Kolaczkowski, S.T. & Hout, S.A. (1987b). Modelling Hydrocarbon Fouling, Chem. Eng. Res. Des. 65, . p 1 79 71 ,  Crittenden, B.D. (1988a). Basic Science and Models of Reaction Fouling, in Fouling Science and Technology, pubi. Kiuwer Academic Publishers, 293 14. p 3 Crittenden, B.D. (1988b). Chemical Reaction Fouling of Heat Exchangers. in Fouling Science and Technology, publ. Kiuwer Academic Publishers. p315-332. 284  References  Crittenden, B.D. & Alderman, N.J. (1988). Negative Fouling Resistances: The Effect of Surface Roughness, Chem. Eng. Sci. 43, . 829 p 3 8 ,  Crittenden, B.D., Kolaczkowski, S.T. & Downey, I.L. (1992). Fouling of Crude Oil Preheat Exchangers, Chem. Eng. Res. Des. 70, . p 5 47 57 ,  Crittenden, B.D., Kolaczkowski, S.T. & Takemoto, T. (1993). Use of In-Tube Inserts to Reduce Fouling from Crude Oils. Proc. Heat Transfer, Atlanta, GA, A.LCh.E. Symp. Ser. 295, p 3 00 07 Danckwerts, P.V. (1970). Gas-Liquid Reactions. McGraw-Hill, London Daubert, T.E. & Danner, T.E. (1989). Physical and Thermodynamic Properties of Pure Chemical Compounds. publ A.I.Ch.E., NY Davies, A.G. (1961). Organic Peroxides. publ. Butterworths, London Dickakian, G. & Seay, S. (1988). Asphaltene Precipitation Primary Crude Exchanger Fouling 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, MD Dugan, C.F., Van Nostrand Jr., W.L. & Haluska, J.L. (1978). How Antifoulants Reduce the Energy Consumption of Refineries, Chem. Eng. Frog., 74, . p 5 3 7 Eaton, P. & Lux, R. (1984). Laboratory Fouling Test Apparatus For Hydrocarbon Feedstocks, ASME HTD 35, p33-42. Epstein, N. (1981) Fouling: Technical Aspects, in Fouling of Heat Transfer Equipment, eds. Somerscales, E.F.C. & Knudsen, J.G., publ. Hemisphere, p31-53  Epstein, N. (1983). Thinking About Heat Transfer Fouling : A 5x5 Matrix, Heat Transfer Engng., 4, . 46 p 5 3 Epstein, N. (1988). General Thermal Fouling Models. in Fouling Science and Technology., publ. Kiuwer Academic Publishers. p15-30. Epstein, N. (1993a). The Velocity Effect on Initial Fouling Rates, Israel Chemical Engineering, 22, . p 3 2 7 Epstein, N. (1993b). A Model of the Initial Chemical Reaction Fouling Rate For Flow Within a Heated Tube and its Verification, accepted for presentation at the 10th Intnl. Ht. Tfr. Conf., Brighton, UK (1994),  Fernandez-Baujin, J.M. & Solomon, S.M. (1976). in Industrial and Laboratory Pyrolyses. eds. Aibright, L.F. & Crynes, B.L. Washington, A.C.S. Symp. Ser. 32. . p 3 45 72  285  References  Fetissoff, P.E., Watkinson, A.P. & Epstein, N. (1982). Comparison of Two Heat Transfer Fouling Probes. Proc. 7th Int’l Heat Transfer Conference, Munich Fetissoff, P.E. (1982). Comparison of Two Fouling Probes. M.A.Sc. thesis, University of British Columbia. Fisher (1959). Fisher Density/Temperature Tables. Fisher/TAG Manualfor Inspectors of Petroleum. 29th edn., publ. Fisher Scientific Fortuin, J.M.H., Musschenga, E.E. & Hamersma, P.J. (1992). Transfer Processes in Turbulent Pipe Flow Described by the ERSR Model, A.I.Ch.E.J. 38, . p 3 43