Mathematical Modeling of Complex Reaction Systems in The Oil and Gas Industry

Downloaded from https://onlinelibrary.wiley.com/doi/ by ibrahim ragab - Universita Di Firenze Sistema , Wiley Online Library on [29/07/2024].
See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Mathematical Modeling of Complex
Reaction Systems in the Oil and
Gas Industry
Downloaded from https://onlinelibrary.wiley.com/doi/ by ibrahim ragab - Universita Di Firenze Sistema , Wiley Online Library on [29/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Mathematical Modeling of Complex Reaction
Systems in the Oil and Gas Industry
Instituto Mexicano del Petróleo
Boreskov Institute of Catalysis
University of Tyumen
Andrey Zagoruiko
Mexico City, Mexico
Novosibirsk, Russia
Andrey Elyshev
Jorge Ancheyta
Tyumen, Russia
Edited by
This edition first published 2024
© 2024 John Wiley & Sons Ltd
All rights reserved, including rights for text and data mining and training of artificial intelligence technologies
or similar technologies. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except
as permitted by law. Advice on how to obtain permission to reuse material from this title is available at
http://www.wiley.com/go/permissions.
The right of Andrey Elyshev, Andrey Zagoruiko, and Jorge Ancheyta to be identified as the authors of the editorial
material in this work has been asserted in accordance with law.
Registered Offices
John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK
For details of our global editorial offices, customer services, and more information about Wiley products visit us at
www.wiley.com.
Wiley also publishes its books in a variety of electronic formats and by print-on-demand. Some content that appears in
standard print versions of this book may not be available in other formats.
Trademarks: Wiley and the Wiley logo are trademarks or registered trademarks of John Wiley & Sons, Inc. and/or
its affiliates in the United States and other countries and may not be used without written permission. All other
trademarks are the property of their respective owners. John Wiley & Sons, Inc. is not associated with any product or
vendor mentioned in this book.
Limit of Liability/Disclaimer of Warranty

While the publisher and authors have used their best efforts in preparing this work, they make no representations or
warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all
warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose.
No warranty may be created or extended by sales representatives, written sales materials or promotional statements for
this work. This work is sold with the understanding that the publisher is not engaged in rendering professional
services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a
specialist where appropriate. The fact that an organization, website, or product is referred to in this work as a citation
and/or potential source of further information does not mean that the publisher and authors endorse the information
or services the organization, website, or product may provide or recommendations it may make. Further, readers
should be aware that websites listed in this work may have changed or disappeared between when this work was
written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other
commercial damages, including but not limited to special, incidental, consequential, or other damages.
Library of Congress Cataloging-in-Publication Data is Applied for

Hardback ISBN: 9781394220021
Cover Design: Wiley

Cover Image: © Created and Designed by Darya Obanina together with the Shedevrum neural network
Set in 9.5/12.5pt STIXTwoText by Straive, Pondicherry, India

v
Contents
List of Contributors xiii

Preface xv
1 Modeling the Kinetics of Hydrocracking of Heavy Oil with Mineral Catalyst 1

Guillermo Félix, Fernando Trejo, and Jorge Ancheyta
1.1 Introduction 1
1.1.1 Reserves and Production of Heavy Crude Oils 1
1.1.2 Heavy Crude Oil Upgrading Processes 2
1.1.3 Reactions During Slurry Phase Hydrocracking 6
1.1.4 Catalysts for Hydrocracking of Heavy Crude Oils in Slurry Phase 6
1.2 Kinetic Models 7
1.2.1 General Types of Kinetic Models 8
1.2.1.1 Lumping Kinetic Models 8
1.2.1.2 Continuous Lumping Kinetic Models 8
1.2.1.3 Single-Event Kinetic Models 10
1.2.2 Kinetic Models Reported in the Literature for Hydrocracking of Heavy
Crude Oils Using Dispersed Catalysts 10
1.2.2.1 Kinetic Models Based on Distillation Curves 10
1.2.2.2 Kinetic Models Based on SARA Fractions 18
1.2.3 Kinetic Models Based on Continuous Lumping 21
1.2.4 Thermodynamic Model to Predict the Asphaltenes Flocculation and Sediments
Formation 22
1.3 Kinetic Parameters Estimation 24
1.3.1 Assumptions 26
1.3.2 Initialization of Parameters 27
1.3.3 Nonlinear Optimization 28
1.3.4 Objective Function 28
1.3.5 Sensitivity and Statistical Analyses 29
1.3.5.1 Perturbations 29
1.3.5.2 Parity Plots 29
1.3.5.3 Residuals 29
1.3.5.4 AIC and BIC 30
1.4 Results and Discussion 30
1.4.1 Kinetic Parameters 30
vi Contents
1.4.1.1 Assumptions 30
1.4.1.2 Reaction Rate Coefficients 32
1.4.1.3 Activation Energies 38
1.4.2 Accuracy of the Kinetic Models 38
1.4.2.1 SARA-Based Models 38
1.4.2.2 Distillation Curves-Based Models 41
1.4.3 Reactions in Parallel and in Series 44
1.4.4 Thermodynamic Model 45
1.4.5 General Comments 48
1.5 Conclusion 50
References 50
2 Modeling Catalyst Deactivation of Hydrotreating of Heavy Oils 56

Javier Jurado, Vicente Samano, and Jorge Ancheyta
2.1 Introduction 56
2.2 Mechanisms of Deactivation 57
2.2.1 Coking Deposition (Fouling) 59
2.2.2 Metal Deposition (Poisoning) 59
2.3 Deactivation Models 60
2.3.1 Deactivation Models by Coke Deposition 60
2.3.2 Deactivation Models by Metal Deposition 65
2.3.3 Deactivation Models by Coke and Metal Deposition 70
2.4 Development of Models for HDT Catalyst Deactivation 78
2.4.1 Important Issues 78
2.4.2 Final Remarks 82
2.5 Development of a Reactor Model for Heavy Oil Hydrotreating with Catalyst
Deactivation Based on Vanadium and Coke Deposition 83
2.5.1 The Model 84
2.5.1.1 Description 84
2.5.1.2 Solution of the Model 86
2.5.1.3 Advantages of the Model 86
2.5.1.4 Procedure for Parameter Estimation 88
2.5.2 Results and Discussion 89
2.5.2.1 Profiles of Sulfur and Vanadium Concentration in Products 89
2.5.2.2 Comparison of Predictions with Literature and Proposed Model 90
2.5.2.3 Profiles of Coke and Vanadium on Catalyst 91
2.5.2.4 Final Remarks 93
2.5.3 Usefulness of the Model 95
2.5.4 Conclusion 96
2.6 Application of the Deactivation Model for Hydrotreating of Heavy Crude
Oil in Bench-Scale Reactor 96
2.6.1 Properties of Heavy Oil 96
2.6.2 Properties of the Catalyst 96
2.6.3 Bench-Scale Reactor 98
2.6.4 Catalyst Activation 98
2.6.5 Operating Conditions 99
2.6.6 Characterization Methods 99
Contents vii
2.6.7 Parameter Estimation 100

2.6.8.1 Evolution of Sulfur and Metals Concentration in Products 101
2.6.8.2 Coke and Metals on Catalyst 102
2.6.9 Conclusion 105
Nomenclature 105
References 111
3 Simulation of the Oxidative Regeneration of Coked Catalysts: Kinetics,

Catalyst Pellet, and Bed Levels 116
Sergey Zazhigalov , Osman Abdulla, and Andrey Zagoruiko
3.1 Introduction 116
3.2 Process Chemistry and Laboratory Experiments 117
3.2.1 Catalyst and Proposed Reactions 117
3.2.2 Reaction Kinetics 119
3.2.3 Experimental Setup 121
3.2.4 Experiments 124
3.3 Mathematical Model 126
3.4 Model Solution Method 132
3.5 Modeling Results 133
3.6 Conclusion 134
3.7 Notation 136
Abbreviations 136
Acknowledgment 137
References 137
4 Modeling of Unsteady-State Catalytic and Adsorption–Catalytic

Processes: Novel Reactor Designs 138
Sergey Zazhigalov , Andrey Elyshev, and Andrey Zagoruiko
4.2 Novel Reactor Designs for Catalytic Reverse-Flow and Adsorption–Catalytic
Processes 141
4.2.1 Unsteady-State Catalytic Reverse-Flow Process 141
4.2.2 Adsorption–Catalytic Process 142
4.3 Mathematical Models of the Processes 145
4.4 Results 148
4.4.2.1 Reactor with Truncated Cone Entrance 153
4.4.2.2 Multisectional Reactor 156
4.5 Conclusion 164
4.6 Notation 165
Abbreviations 165
Acknowledgments 165
References 166
viii Contents
5 Molecular Reconstruction of Complex Hydrocarbon Mixtures for Modeling

of Heavy Oil Processing 168
Nikita Glazov and Andrey Zagoruiko
5.2 The Problem 168
5.3 Illustration 169
5.4 Reconstruction by Entropy Maximization (REM) 169
5.5 Stochastic Reconstruction (SR) 174
5.6 SR-EM 179
5.7 Structure-Oriented Lumping (SOL) Method 181
5.8 State Space Representation Method 182
5.9 Molecular Type-Homologous Series Matrix 183
5.10 Conclusion 184
Acknowledgment 184
References 184
6 Modeling of Catalytic Hydrotreating Reactor for Production of Green Diesel 187

Alexis Tirado, Fernando Trejo, and Jorge Ancheyta
6.2 Conversion of Vegetable Oils into Renewable Fuels 187
6.2.1 Commercial Production of Renewable Diesel 189
6.3 Hydrotreating Kinetic Models and Reaction Pathways 190
6.3.1 Model Compounds 190
6.3.2 Vegetable Oils 197
6.4 Models for Catalytic Deactivation 204
6.5 Reactor Modeling for Vegetable Oil Hydrotreating 205
6.5.1 Deviation from Ideal Flow Pattern 208
6.6 The Importance of Modelling Reactors for Vegetable Oil Hydrotreating 210
6.7 Study Case for the Development of Dynamic Reactor Model 210
6.7.1 Equations and Assumptions for Hydrotreating Reactor Modeling 210
6.7.2 Kinetic Model for Hydrotreating of Vegetable Oil 213
6.7.3 Hydrogen Consumption and Gas Generation 213
6.7.4 Solution of Reactor Models 215
6.8 Analysis and Discussion of Results 217
6.8.1 Criteria to Ensure Ideal Behaviors in Trickle-Bed Reactor 217
6.8.2 Dynamic Profiles of Feedstock and Products of a Bench-Scale Reactor
for Catalytic Hydrotreating of Vegetable Oil 219
6.8.3 Validation of Hydrotreating Reactor Model with Pilot Plant Data 222
6.8.4 Dynamic Simulation of a Non-isothermal Reactor 225
6.8.4.1 Comparison of Non-isothermal Model with Experimental Results
in Isothermal Reactor 225
6.8.4.2 Comparison of Bench-Scale and Pilot-Scale Reactor Under Non-isothermal
Operating Condition 227
6.8.5 Dynamic Simulation of an Adiabatic Commercial Reactor 229
6.8.5.1 Configuration of Hydrogen Quenching 232
6.8.5.2 Liquid-Phase Yields and Gas Composition 232
6.9 Conclusions 235
References 236
Contents ix
7 Modeling of Slurry-Phase Hydrocracking Reactor 242

Cristian Calderón and Jorge Ancheyta
7.1.1 Characteristics of Slurry-Phase Reactors for Hydrocracking 242
7.1.1.1 Type of Reactors 242
7.1.1.2 Catalyst Properties 245
7.1.2 SPR Modeling 246
7.1.2.1 Classification 246
7.1.2.2 Model Complexity 249
7.1.2.3 Models for Slurry Reactors 249
7.2 Proposed Generalized Model 253
7.2.1 Equations for the Generalized Model 253
7.2.2 Solids Concentration 257
7.2.3 Initial and Boundary Conditions 257
7.2.4 Estimation of Model Parameters 260
7.2.5 Gas Holdup 260
7.2.6 Gas–Liquid Mass Transfer Coefficients 262
7.2.7 Gas–Liquid Equilibrium 264
7.2.8 Liquid–Solid and Gas–Solid Mass Transfer Coefficients 264
7.2.9 Dispersion Coefficients 265
7.2.10 Heat Transfer Coefficients 267
7.2.11 Example of Simplification of the Generalized Model 267
7.3 Simplified Models 268
7.3.1 SPR 1D Model 268
7.3.2 SPR 2D Model 269
7.3.3 Continous Stirred Tank Reactor Model 270
7.3.4 Parameters 270
7.3.5 Reaction Kinetics 273
7.3.6 Solution Method 274
7.4 Numerical Simulations 275
7.4.1 Experimental Reactors 275
7.4.1.1 Dynamic Simulations of CSTR and SPR 275
7.4.1.2 Steady-State Simulations of a SPR 278
7.4.2 Industrial-Scale Reactor 280
7.4.2.1 Dynamic Simulations of the Industrial Slurry-Phase Reactor 283
7.4.2.2 Sensitivity Analysis for the Industrial Slurry-Phase Reactor 287
7.5 Conclusions 291
Nomenclature 294
References 297
8 Modeling of Fischer–Tropsch Synthesis Reactor 303

César I. Méndez and Jorge Ancheyta
8.1 Fundamentals of the Fischer–Tropsch Synthesis to Produce Clean Fuels 303
8.1.1 Fischer–Tropsch Synthesis Technology 304
8.1.2 Fischer–Tropsch Synthesis Catalysts 307
8.1.2.1 Cobalt-Based Catalysts 307
8.1.2.2 Iron-Based Catalysts 308
8.1.2.3 Catalyst Support 309
x Contents
8.1.3 Fischer–Tropsch Synthesis Kinetic Models 309

8.1.3.1 Kinetic Models Developed with Iron Catalyst 310
8.1.3.2 Kinetic Models Developed with Cobalt Catalyst 310
8.1.4 General Aspects of Fischer–Tropsch Catalytic Mechanisms 315
8.1.5 The Fischer–Tropsch Synthesis Product Distribution Models 321
8.2 Modeling of Catalytic Fixed-Bed Reactors for Fuels Production
by Fischer–Tropsch Synthesis 324
8.2.1 Introduction 324
8.2.2 Modeling of Fixed-Bed Fischer–Tropsch Reactors 324
8.2.2.1 Classification of Fixed-Bed Fischer–Tropsch Reactor Models 325
8.2.2.2 One- and Two-Dimensional Pseudohomogeneous Model 325
8.2.2.3 One- and Two-Dimensional Heterogeneous Model 326
8.2.3 Development of a Generalized Fixed-Bed Fischer–Tropsch Reactor Model 326
8.2.3.1 General Equations of the Model 326
8.2.3.2 Boundary Conditions of the Proposed Generalized Model 334
8.2.3.3 Pressure Drop 337
8.2.4 Model Parameters 340
8.2.4.1 Mass Transfer Parameters 340
8.2.4.2 Heat Transfer Parameters 341
8.2.4.3 Phase Equilibrium 343
8.2.4.4 Catalyst Particles Parameters 345
8.2.4.5 Catalytic Bed Parameters 352
8.3 Importance of Proper Hydrodynamics Modeling in Fixed-Bed Fischer–Tropsch
Synthesis Reactor 354
8.3.2 Mathematical Modeling of the Fixed-Bed Fischer–Tropsch Synthesis Reactor 354
8.3.2.1 Reactor Model 355
8.3.2.2 Kinetics 356
8.3.2.3 Other Parameters and Correlations 357
8.3.2.4 Numerical Method 357
8.3.3.1 Simulations for the One-Stage Reactor 358
8.3.3.2 Simulations for the Two-Stage Reactor 364
8.4 Dynamic One-Dimensional Pseudohomogeneous Model for Fischer–Tropsch
Reactors 371
8.4.2 Formulation of the Model 371
8.4.2.1 Model Equations and Solution 371
8.4.2.2 Model Parameters, Correlations, and Kinetics 372
8.4.3.1 Experimental Data 373
8.4.3.2 Conversion of CO and H2 373
8.4.3.3 Temperature Profiles 378
8.4.4 Product’s selectivity 381
Contents xi

8.5 Modeling and Control of a Fischer–Tropsch Synthesis Reactor with
a Novel Mechanistic Kinetic Approach 390
8.5.2 Formulation of the Model 392
8.5.2.1 Model Equations and Solution 392
8.5.2.2 Model Parameters and Correlations 394
8.5.2.3 The Mechanistic FTS Kinetic Model 395
8.5.3 Implementation of the PI Controller 396
8.5.4.2 Simulations of the Syngas Conversion, Light Gases, and Heavy
Liquid Selectivity 397
8.5.4.3 Simulations of the Fischer–Tropsch Fixed-Bed Reactor and the Cooling
Jacket Thermal Behavior 404
8.5.4.4 Surfaces of the Syngas Conversion and the Heavy Liquids Selectivity
as a Function of the FTS Reactor Temperature 405
8.6 On the use of Steady-State Optimal Initial Operating Conditions for the
Control Scheme Implementation of a Fixed-Bed Fischer–Tropsch Reactor 408
8.6.2 Methodology 408
8.6.2.1 Model Equations and Numerical Solution 408
8.6.2.2 Model Parameters, Correlations, and Kinetics 409
8.6.2.3 Steady-State Nonlinear Constrained Optimization Problem 409
8.6.2.4 Implementation of the Control Scheme 413
8.6.3 Results and discussion 414
8.6.3.2 Simulations of the Steady-State Nonlinear Constrained Optimization Problem: CO
Conversion, SC5+ Selectivity, and Temperature Profiles 415
8.6.3.3 Simulations of the Control Scheme Implementation: CO Conversion,
SC5+ Selectivity, and Temperature Profiles 416
References 421
9 Computational Fluid Dynamics Modeling of Mass Transfer Processes in

Structured Beds of Microfibrous Catalysts 434
Sergey Lopatin, Andrey Elyshev, and Andrey Zagoruiko
9.2.1 Model Description 437
9.2.2 Computing Domain 437
9.2.3 Simulation Object Geometry 437
9.2.4 Reaction 439
9.2.5 Model Parameters 440
9.3 Simulation Results 440
9.3.1 Cartridge Channel with Corrugated Structuring Mesh 440
443
Cartridge Channel Without Corrugated Structuring Mesh
The General Description of Mass Transfer in GFC 451

Convective Flow Inside the GFC Thread 449
Two-Sided Washing of GFC Textiles 447
Influence of GFC Textile Shape 443
Acknowledgement 453
Abbreviations 453
Conclusion 452
References 453
456
Index
Contents
9.3.2
9.3.3
9.3.4
9.3.5
9.3.6
9.4
xii
xiii
List of Contributors
Osman Abdulla Sergey Lopatin

Tyumen State University, Tyumen, Russia Boreskov Institute of Catalysis, Novosibirsk
Russia and Tyumen State University
Jorge Ancheyta Tyumen, Russia
Department of Petroleum Engineering
Kazan Federal University, Kazan, Russia; César I. Méndez
Instituto Mexicano del Petróleo Centro de Investigación en Ciencia Aplicada
Mexico City, Mexico; and Escuela Superior de y Tecnología Avanzada del Instituto Politécnico
Ingeniería Química e Industrias Extractivas Nacional, Mexico City, Mexico
Instituto Politécnico Nacional, UPALM
Zacatenco, Mexico City Mexico
Alexis Tirado
Cristian Calderón Department of Petroleum Engineering
Tecnológico de Monterrey, Mexico State Kazan Federal University, Kazan, Russia
Mexico
Fernando Trejo
Andrew Elyshev Instituto Politécnico Nacional
Tyumen State University, Tyumen, Russia CICATA–Legaria, Ciudad de México, Mexico
Guillermo Félix Andrey Zagoruiko

Department of Petroleum Engineering Boreskov Institute of Catalysis, Novosibirsk
Kazan Federal University, Kazan Russia and Tyumen State University
Russia and Tecnológico Nacional de Tyumen, Russia
México/IT de Los Mochis, Los Mochis
Sinaloa, Mexico
Sergey Zazhigalov
Boreskov Institute of Catalysis, Novosibirsk
Nikita Glazov
Russia and Tyumen State University
Boreskov Institute of Catalysis SB RAS
Tyumen, Russia
Prospekt Akademika Lavrent’yeva
5, Novosibirsk, Novosibirsk Oblast
Russian Federation Vicente Samano
Instituto Politécnico Nacional
Javier Jurado Centro de Investigación en Ciencia
Instituto Politécnico Nacional, Centro de Aplicada y Tecnología Avanzada
Investigación en Ciencia Aplicada y Tecnología Unidad Legaria, Legaria 694
Avanzada, Unidad Legaria, Legaria 694 Col. Irrigación
Col. Irrigación, Ciudad de México, Mexico Ciudad de México, Mexico
xv
Preface
After more than 30 years of experience working in the kinetic and reactor modeling of different
processes for the petroleum and gas industry as well as for environmental protection applications,
we have identified the need to have a collection of the most relevant cases that would have great
impact on the industry. The general areas that are described in this book correspond to reaction
kinetics, catalyst deactivation, reactor model, computational fluid dynamics, and steady-state
and dynamics simulations using own experimental data or information reported in the literature.
Mathematical Modeling of Complex Reaction Systems in the Oil and Gas Industry is focused on the
step-by-step description of kinetics and reactor modeling of various important processes used in the
oil and gas industry. The topics covered in this book are hydrocracking of heavy oils, catalyst deac-
tivation, oxidative regeneration of catalyst, adsorption–desorption catalytic processes, molecular
reconstruction, production of green diesel via catalytic hydrotreating, slurry-phase hydrocracking
reactor, Fischer–Tropsch synthesis and structured beds of micro-fibrous catalyst. Nowadays, all
these topics are relevant to the oil and gas industry, particularly in this era of energy transition
and decarbonization, which makes the book original and unique among previous reactor modeling
books.
Each chapter describes the development of kinetic and reactor models for steady-state and
dynamic simulations. To develop the kinetic model for each reaction or set of reaction for further
reactor modeling, exhaustive experimental data either reported in the literature or generated in
own laboratories are used. The developed models are validated with laboratory experimental data
and simulation are reported to predict the commercial performance of the involved reactors. In
addition, such a modeling in some cases allows for a deeper understanding of the qualitative
essence of the phenomena observed in experiments, that is, to apply the augmented reality
approach and make visible what cannot be directly observed in experiments, or at least to reproduce
the most plausible picture of the physical events taking place.
This book is expected to be a reference guide for researchers, PhD students, postdoctoral
researchers, catalyst manufacturers, process designers, and professors, to help them develop kinetic
and reactor models at different reaction scales. It is also anticipated that this textbook can be used to
cover part of the content of courses of different carriers at undergraduate and postgraduate levels.
Jorge Ancheyta, National Polytechnic Institute and Mexican Institute of Petroleum, Mexico
Andrey Zagoruiko, Boreskov Institute of Catalysis, Novosibirsk, Russia
Andrey Elyshev, Tyumen State University, Russia
1
Modeling the Kinetics of Hydrocracking of Heavy Oil with

Mineral Catalyst
Guillermo Félix1,2, Fernando Trejo3, and Jorge Ancheyta4,5
1
Department of Petroleum Engineering, Kazan Federal University, Kazan, Russia
2
Tecnológico Nacional de México/IT de Los Mochis, Los Mochis, Sinaloa, Mexico
3
Instituto Politécnico Nacional, CICATA-Legaria. Legaria 694, Col. Irrigación, Ciudad de México, Mexico
4
Instituto Mexicano del Petróleo, Eje Central Lázaro Cárdenas Norte 152, San Bartolo Atepehuacan, Mexico City, Mexico
5
Escuela Superior de Ingeniería Química e Industrias Extractivas, Instituto Politécnico Nacional, UPALM, Zacatenco, Mexico City, Mexico
1.1 Introduction
The continuously increasing global demand for energy and the decrease of light crude oil reserves
have caused researchers to look toward heavy crude oil reserves. Despite the importance of renew-
able sources is growing, fossil energy sources will continue to meet most of the global energy
requirements (Bellussi et al. 2013; Félix and Ancheyta 2019a; Martínez-Grimaldo et al. 2014; Rana
et al. 2007; Santos et al. 2014). The driving forces of the oil industries to improve heavy crude oils
and vacuum residues (VRs) are as follows: (i) obtain high-quality transportation fuels, (ii) middle
distillates, (iii) reduction of conventional crude oil supply, and (iv) price increase. Heavy crude oils
or VRs are considered suitable alternative sources for transportation fuels, energy, and petrochem-
icals to meet the demands of industries. However, because of the low quality and difficulties in
transporting heavy crude oils, it is necessary to develop technologies to upgrade their properties
(Castañeda et al. 2014; Martínez-Grimaldo et al. 2014; Sahu et al. 2015).
In recent decades, researchers have prioritized refining techniques that use low-cost feedstock,
such as coal, low-grade oil or wax, heavy crude oil, VRs, and natural gases, which are aimed to
improve the cost-effectiveness of the refining process. The goal of these techniques is to transform
such feedstock into transportation fuels and other low-boiling point liquid products. Refining heavy
crude oils and VRs decreases their viscosity, initial boiling point, metal contents, sulfur, and other
impurities, while increasing the H/C ratio to meet the commercial standards (Rodríguez et al. 2018;
Sahu et al. 2015; Speight 2004).
1.1.1 Reserves and Production of Heavy Crude Oils

The assessment of production capacity of oil field reveals that conventional light and medium oil
reserves peaked in the 1960s. Since then, these reserves have steadily declined. An important focus
of the global oil industry is to develop new technologies to be feasible: the use of unconventional
resources as heavy and extra heavy crude oils (Heidary et al. 2017; Martínez et al. 2010; Santos et al.
2014). At the end of 2021, more than 1730 billion barrels of proven reserves are reported worldwide,
Mathematical Modeling of Complex Reaction Systems in the Oil and Gas Industry, First Edition.
Edited by Jorge Ancheyta, Andrey Zagoruiko, and Andrey Elyshev.
© 2024 John Wiley & Sons Ltd. Published 2024 by John Wiley & Sons Ltd.
2 1 Modeling the Kinetics of Hydrocracking of Heavy Oil with Mineral Catalyst
Crude oil reserves (billion bbl)

0 50 100 150 200 250 300 350
Venezuela
Saudi Arabia
Iran
Canada
Iraq
Kuwait
United Arab Emirates
Russia
Libya
United States
Nigeria
Kazakhstan
China
Qatar
Brazil
Figure 1.1 Proven reserves of crude oil by country.
as shown in Figure 1.1 (International—U.S. Energy Information Administration (EIA) 2021; Oil Gas
Journal 2022; Organization of the Petroleum Exporting Countries (OPEC) 2020). Venezuela, Saudi
Arabia, and Iran possess the largest crude oil reserves; however, the quality of their crudes varies
significantly. Currently, these reserves contain 434.3 billion barrels of heavy and extra-heavy crude
oil, while 650.7 billion barrels correspond to bitumen, representing over 60% of the total volume
(Bata et al. 2019; Pratama and Babadagli 2022).
1.1.2 Heavy Crude Oil Upgrading Processes

The processing of the heavy crude oil presents difficulties due to the complex nature of its heavier
fractions, which commonly possess low API gravity (high density), high viscosity, elevated initial
boiling point, higher contents of heteroatoms (sulfur, nitrogen, and oxygen), metals (Ni and V), and
asphaltenes. In general, most of sulfur and nitrogen species present in crude oils are found in
heavier fractions, while metals are found in porphyrins and as concentrated, exclusively in residue
fractions (Ancheyta 2011; Memon et al. 2010; Quitian and Ancheyta 2016b). Therefore, the addition
of hydrogen or rejection of carbon is the main process focused to increase the hydrogen-to-carbon
(H/C) ratio in heavy crudes. These two routes can be catalytic and noncatalytic (Ancheyta 2013;
Félix et al. 2017).
Over the years, several technologies based on carbon rejection, the addition of hydrogen, and the
combination of both have been used to upgrade heavy crude oils (Figure 1.2). The noncatalytic
(deasphalting, gasification, coking, and visbreaking) and catalytic (catalytic cracking of VR) carbon
rejection technologies account for 56.6% of the global processing technology due to its cost-effective
operation. The noncatalytic (hydrovisbreaking) and catalytic (hydrotreating and hydrocracking)
hydrogen addition technologies display advantages and disadvantages when applied to heavy crude
oil processing as well as carbon rejection technologies (Castañeda et al. 2014; Rana et al. 2007; Sahu
et al. 2015; Speight 2004).
1.1 Introduction 3
0.03% 3.62%
3.92%
Slurry-phase
18.25% Deasphalting
40.29%
Ebullating bed
Fixed bed
Cracking/Visbreaking
Coking
33.89%
Figure 1.2 Processing capacity for worldwide heavy crude oil upgrading technologies.
For heavy and extra-heavy crude oils, the choice between carbon rejection and hydrogen addition
technologies depends on technical and economic assessments, as well as the quality of products
from downstream refining processes. Therefore, deciding on which upgrading technology is appro-
priate for certain heavy crude oil is not an easy task and several factors must be taken into consid-
eration, such as the following (Ancheyta 2013; Ancheyta and Speight 2007; Quitian and
Ancheyta 2016b):
✓ Price of heavy crude oil

✓ Amount of impurities in feedstock
✓ Level of quality of the upgraded oil
✓ Process outline of the refinery
In the past, the carbon rejection technologies were preferred for feed with high metal content,
such as heavy crude oils. Nonetheless, these technologies have the following disadvantages
(Ancheyta 2013; Sahu et al. 2015):
✓ Low yield of upgraded oil

✓ Enhanced formation of coke
✓ High aromatic content in the upgraded oil
✓ Lower yield and quality of gasoline and diesel fuels
✓ Low value of upgraded oil
Technologies based on the hydrogen addition produce crude oils that possess superior quality and
greater added value than carbon rejection methods do; hence, these approaches are becoming
increasingly attractive. Highlights of the hydrogen addition technologies are the following:
✓ Minimizes the coke formation

✓ Enhances the yield of liquid products obtained by hydrocracking or hydrogenolysis mechanisms
✓ Increases the H/C ratio of the products
✓ Reduction of impurities, such as sulfur, nitrogen, metals, and asphaltenes
✓ Increases hydrogen consumption, giving higher liquid yields
These technologies also present the following disadvantages (Ancheyta 2013; Quitian and
Ancheyta 2016b; Rana et al. 2007):
✓ Large volumes of hydrogen are necessary to carry out the hydrogenation of carbon-rich crude oil
✓ Difficulties arise from the carbon and metal deposition on the catalyst surface
✓ Requires well-designed catalysts that are able to process high levels of metals and asphaltenes
within the feed
There are also some emerging technologies at different levels of development or even close to be
commercialized, which have been developed due to the growing production of heavy and extra-
heavy crude oils (Castañeda et al. 2014; Speight 2004). Hydrocracking is a well-known refining
process where heavy crude oil in hydrogen atmosphere is transformed into light products with bet-
ter quality. There are different types of reactors to carry out this process and the choice depends
mainly on the level of conversion required and the number of impurities (metals and asphaltenes)
in the feed. Three-phase catalytic reactors are highly valued due to their broad applications in var-
ious reaction systems and being notably used in the petroleum industry, including hydrocracking
processes. Fixed-bed reactors have stable and reliable performance, but strong limitations are feed
properties due to fast catalyst deactivation by coking and metal depositions. Ebullated-bed reactors
perform better than fixed-bed reactors since they can process feed containing higher quantities of
metals and coke at the expense to be conversion-limited and reaching values up to 80%. This is
because of the rapid formation of sediments, by which commonly aromatic solvents are used to
keep the soluble in the reacting phases. Slurry phase reactors achieve higher conversions perform-
ing superior levels of upgradation of feed rich in sulfur, metals, and asphaltenes (Ancheyta 2011;
Jarullah et al. 2011; Martínez-Grimaldo et al. 2014; Sahu et al. 2015).
The degree of conversion is affected by the properties and nature of the feed, thus lighter crude oil
possesses lower number of impurities, while heavy fractions contain higher content of asphaltenes
(Alonso et al. 2019; Marafi et al. 2005; Ortega-García et al. 2012; Stanislaus et al. 2005; Stratiev et al.
2014, 2019; Tirado and Ancheyta 2018). Asphaltene is the most polar fraction found in heavy crudes
and residues that is soluble in aromatic solvents but not in alkanes. This fraction comprises of both
aromatic and aliphatic carbons, which contribute to its high molecular weight. The complex struc-
ture (Figure 1.3) makes asphaltenes one of the most studied fractions as they are coke precursors.
The asphaltene fraction is formed by large aromatic sheets, side alkyl chains, heteroatoms, and
metalloporphyrins. These latter components exhibit a remarkable influence of p-electronic inter-
actions. The alteration in solubility during the processing of heavy crudes or residues is linked to
solid formation. This disruption in solubility equilibrium relies on the chemical characteristics of
the feed and the processing conditions. Based on the colloidal model (Figure 1.4), asphaltenes are
surrounded by resin-building micelles keeping the asphaltenes dispersed in crude oil that avoids
the contact with saturated compounds, since their nonpolar nature cause the instability of asphal-
tene micelles. Sediment formation is the result of a disturbance in the balance between asphaltenes
and those compounds that keep them dispersed in solution during hydrotreating/hydrocracking; in
other words, resins lose their peptization properties causing aggregation and precipitation of
asphaltenes (Ashoori et al. 2017; Félix and Ancheyta 2019a; Ortega-García et al. 2012; Rana
et al. 2007; Rogel et al. 2013; Stratiev et al. 2014, 2019; Tirado and Ancheyta 2018).
In addition to the loss of solubility of asphaltenes by oil and resin fractions, hydrocracking leads
to an increase in the aromaticity and condensation of unconverted asphaltene cores. Reports indi-
cate that the asphaltenes found in upgraded products have higher aromaticity and a lower H/C ratio
in comparison with original asphaltenes in the feed. The results of structural analysis of the hexane-
insoluble fraction indicate a higher degree of polarity due to the presence of a significant number of
5
1.1 Introduction
Asphaltenes
Aromatics
Saturates
Resins
s
Metalloporphyrin
Saturates
Figure 1.4 Heavy crude oil composition based on SARA fractions.

N
Aromatic closter
N
Figure 1.3 Hypothetical asphaltene molecule.

N
s
s
N
N
s
o
N
Heteroatoms
heteroatoms with shorter alkyl chains because of the dealkylation reactions, which causes asphal-
tenes to become more aromatic and precipitate. The precipitation of sediment has also been attrib-
uted to the possible liquid−liquid phase separation of the material at high temperatures of
processing (Mochida et al. 1989; Rogel et al. 2013; Tirado and Ancheyta 2018).
Sediment formation typically has a maximum limit of 0.8–1.0 wt.% in refineries. Higher values
may cause commercial plants to shut down due to plugging issues. For this reason, sediment for-
mation is a critical process parameter during hydrotreating and hydrocracking of heavy oils and
thus dictates the maximum allowable residue conversion. Furthermore, a more extensive compre-
hension of sediment formation during upgrading processes is essential for enhancing both perfor-
mance and control strategies. Improving the predictability of sediment deposition is highly
desirable as it would enable the preliminary screening of feed and the optimization of processing
(Alonso et al. 2019; Rogel et al. 2013; Tirado and Ancheyta 2018).
1.1.3 Reactions During Slurry Phase Hydrocracking

For hydrocracking reactions in slurry phase, the operating conditions can be classified as follows
(Quitian and Ancheyta 2016a; Speight 2013):
Mild conditions can be attained if the pressure and temperature do not exceed 80−100 kg/cm2
and 410 C, respectively, while the hydrocracking of heavy crude oil or residua leads to a VR of
less than 50%.
Severe conditions are reached at pressure and temperature exceeding 80−100 kg/cm2 and 410 C,
respectively, and the VR conversion is greater than 50% during the hydrocracking of heavy crudes
or residua.
Hydrocracking and hydrogenolysis are exothermic reactions and may occur at different extents
depending on the operating conditions. At mild temperatures, hydrogenation is the dominant reac-
tion unlike hydrocracking and thermal cracking reactions that take place at high temperatures
(Quitian and Ancheyta 2016a; Rașeev 2003). The polyaromatic compounds break the weak bonds,
such as C–S, C–O, and so on, and transform into free radicals. Meanwhile, the H2 molecule dissoci-
ates to form Mo–H or S–H bonds, which in turn release hydrogen radicals to cap unstable com-
pounds. Consequently, the formation and saturation of free radicals is a cyclic process. It has
been reported that thermal and catalytic hydrocracking occur in series, the residue is converted
to vacuum gas oil (VGO), medium distillate, naphtha, and gases, and the combination of thermal
cracking and hydrogenation promotes the formation of the requested products (Martínez-Grimaldo
et al. 2014; Nguyen et al. 2016; Quitian et al. 2015).
1.1.4 Catalysts for Hydrocracking of Heavy Crude Oils in Slurry Phase

The types of catalysts used in hydrocracking process are the solid heterogeneous powders and dis-
persed catalysts. The first type contains at least one metal (cobalt, molybdenum, nickel, iron) sup-
ported on alumina, silica–alumina, etc., by which the catalytic system is heterogeneous, containing
the feed in liquid phase, the hydrogen in gaseous phase, and the catalyst in solid phase. As the reac-
tion takes place, heavier molecules block the pore mouths and reduce the catalytic activity, thereby,
reducing the reaction rate (Nguyen et al. 2016; Rodríguez et al. 2018). To solve this issue, unsup-
ported catalysts were developed. The dispersed catalysts are particulate materials that enhance the
catalytic hydrotreating at high reaction rates due to their large specific surface area, which makes
the kinetics in the slurry phase faster than in fixed-, moving-, or ebullated-bed reactors. The faster
the mass and energy transfer, the lower the over-hydrocracking activity, and thereby, the lower the
coke and gas formation (Liu et al. 2009; Quitian and Ancheyta 2016a).
Dispersed catalysts are highly stable by which optimized contact between oil and hydrogen is pos-
sible. As a result of high metal contents in dispersed catalysts, the heavy hydrocarbons reach the
active sites very quickly and avoid plugging the pores as in supported catalysts. For these reasons,
dispersed catalysts can be used directly in presence of heavier feedstock making slurry processes more
suitable than other hydroconversion technologies when processing heavy feedstock (Angeles et al.
2014; Bellussi et al. 2013). Dispersed catalysts consist in different types: (i) soluble in water,
(ii) soluble in oil (in this case heavy crude oil), (iii) mineral, and (iv) ionic liquid. The dispersed
oil-soluble catalysts are generally more expensive than the water-soluble catalysts, but the water
can evaporate during the reaction and reduce the catalytic activity, resulting in particle agglomera-
tion. Meanwhile, the oil-soluble catalyst can be uniformly dispersed in the feed enhancing hydrogen
uptake and suppressing the coke formation. The conversion as well as the quality of liquid product
increases (Galarraga et al. 2012; Kim et al. 2017; Liu et al. 2009; Nguyen et al. 2016).
The mineral catalysts are natural ores with a high content of one or more transition metals, usu-
ally the Fe-based and Mo-based minerals are the most employed for slurry-phase hydrocracking.
Different works have been employed magnetite, limonite, hematite, ferrite, laterite, pyrite, and
molybdenite, among others. This type of catalyst does not require any special preparation, instead
drying, grinding, and sieving are the usual operations required before using, making them cheaper
than other catalysts (Al-Attas et al. 2019; Fukuyama and Terai 2007; Quitian and Ancheyta 2017).
The ionic liquid catalysts are a relatively new alternative to dispersed catalysts. These compounds
are organic salts that can be liquefied at near room temperature. Their catalytic activity depends
primarily on the cations and anions used during synthesis and usually contain molybdenum,
cobalt, iron, or nickel (Cai et al. 2022; Ma et al. 2022). It is preferred to use molybdenum compounds
as dispersed catalysts due to their greater hydrogenation activity (Calderon and Ancheyta 2016; Du
et al. 2015; Panariti et al. 2000). It has been shown that metal sulfide is the active species in hydro-
cracking process, and it is formed by reacting the metal precursors with sulfur compounds in the
feed or additional sources of sulfur (Nguyen et al. 2016; Panariti et al. 2000). The sulfidation (acti-
vation of catalysts) is done before the hydrocracking reaction and starts by two routes: (i) ex situ and
(ii) in situ sulfidation. In the case of ex situ sulfidation, the sulfur sources are often found in car-
bonaceous compounds (as heavy crude oils) or using hydrogen sulfide (Nguyen et al. 2016; Quitian
and Ancheyta 2016a). Small concentration of low-cost, finely dispersed catalyst makes recovery
impractical after reaction. In addition, slurry catalysts are deactivated after reaction, which occurs
by coke and metal deposition hindering its reuse (Quitian and Ancheyta 2016a; Rezaei 2013).
1.2 Kinetic Models
For adequate modeling of a new hydrocracking process is imperative to take into account the feed
properties, design, performance, process control, and cost to meet the specification of the products.
The main issue when modeling hydrocracking process is the complex nature of feedstock compared
with model compounds by which it is necessary to understand the changes in the rate constants,
conversion, and API gravity at different reaction scales. This will improve the methodologies for
upscaling reaction parameters from the laboratory to industrial scale (Ancheyta et al. 2005; Becker
et al. 2016; Celse et al. 2015; Elizalde et al. 2009; Hassanzadeh and Abedi 2010; Marafi et al. 2010;
Rashidzadeh et al. 2011).
1.2.1 General Types of Kinetic Models

Discrete and continuous lumping models have been largely used to represent the reaction kinetics.
Recently, the use of more complex and fundamental microkinetic models to simulate hydrocracking
has increased the interest of the research community (Becker et al. 2016; Martínez and Ancheyta
2012; Sánchez et al. 2005a). The yield of gases, naphtha, middle distillates, VGO, and residue based
on boiling point intervals is adequately predicted by kinetic models. In addition, the so-called SARA
fractions, that is, saturates, aromatics, resins, and asphaltenes, are also included in the modeling.
Kinetic studies for the transformation of asphaltenes into maltenes fractions (saturates, aromatics,
and resins) have received less attention than those involving the pseudocomponents based on boiling
point ranges (Asaee et al. 2014; Félix and Ancheyta 2019a; Moustafa and Froment 2003).
1.2.1.1 Lumping Kinetic Models

Traditional models by pseudocomponents take macroscopic properties (for example, measurable
properties in the product) and derive the equation system to be solved. These models can be spe-
cifically grouped into compounds in the feed or products in a few cuts, often characterized by ranges
of boiling points that undergo different reactions in series or parallel (Ancheyta 2013; Ancheyta
et al. 2005; Balasubramanian and Pushpavanam 2008; Elizalde et al. 2009; Galarraga et al. 2012;
Hassanzadeh and Abedi 2010; Moghadassi et al. 2011).
Models based on pseudocomponents (up to 10) are commonly reported in the literature to sim-
plify the large number of hydrocarbons present in heavy crude oil fractions for lumping purposes
during hydrocracking. These models are used to simulate the yield of fractions of interest for refi-
ners or process designers in spite of the reduced amount of information obtained from pseudocom-
ponent-based models. When a distillation curve is required, a small number of pseudocomponents
must be interpolated, while the number of kinetic parameters to be calculated increases exponen-
tially, which clearly is not feasible when modeling (Asaee et al. 2014; Becker et al. 2016; Quitian and
Ancheyta 2016a; Sadighi et al. 2010).
1.2.1.2 Continuous Lumping Kinetic Models

Continuous models overcome limitations in lumping models by considering that petroleum is
formed by an infinite number of components, allowing crude to be considered as a continuous mix-
ture instead of pseudocomponents. Continuous mixing takes into account a large number of unde-
fined species contained in the oil fraction, and the difference between adjacent species is relatively
smaller (Ancheyta 2013; Elizalde et al. 2009).
A continuous function describes the reactivity, while the distillation curve is considered to solve
the model keeping the number of parameters constant. In addition, properties, such as the carbon
number or the true boiling points in distillation, are used to characterize the feedstock and products
in the continuous kinetic modeling. The combination of continuous models and pseudocompo-
nents allows more than one continuous distribution to be obtained. However, this model is highly
depending on feed properties, and its advantages are as follows (Ancheyta 2013; Becker et al. 2016):
✓ Prediction of the distillation curve with few parameters to be estimated

✓ Track the chemistry of the process accurately
✓ Individual reaction order has physical meaning
✓ The properties of fractions between the initial and final boiling points are easily obtained
✓ The heteroatom distribution curve is well predicted
✓ Easy to adapt to heavy, extra-heavy crudes and residue kinetics
The model involves the dimensionless temperature (θ) and weight fraction, which is calculated as
follows:
TBP − TBP l
θ= 11
TBP h − TBP l
where TBP is the true boiling point temperature (in kelvin) of any component, TBP(h) and TBP(l)
are the highest and lowest boiling points in the mixture, respectively. Because the TBP of the dis-
tillation curve is a function of the reactivity (λ), a coordinate transformation between discrete and
continuous forms (from the TBP to reactivity for any mixture) is expressed as follows:
λ
=θα
1
12
λmax
where λmax and α are positive constants and model parameters; λmax is the reactivity for species with
the highest dimensionless temperature. The distribution function of species type (D(λ)) is necessary
to transform the coordinates adequately from θ to λ, and it is defined as follows:
ns α
D λ = α λα − 1 13
λmax
where ns is the total number of species in the mixture. The equation of mass balance as a function of
the reactivity and residence time (t or LHSV−1) after changing the coordinates, becomes
λmax
dc λ, t
= − λc λ, t + p λ, Λ Λc Λ, t D Λ dΛ 14
dt λ
where c(λ, t) is the concentration of the component with reactivity λ, c(Λ, t) is the concentration of
the component with reactivity Λ, p(λ, Λ) is the probability density function of species with reactivity
Λ that can form species with reactivity λ, Λ is the maximum reactivity of the same component with
reactivity λ, and D(Λ) is the Jacobian that transforms the discrete coordinate i into a continuous
coordinate Λ. The form of the function p(λ, Λ) and its parameters are given by:
2
λ a0 −0 5
− Λ
1
p λ,Λ = e a1
− A1 + B1 15
S0 2π
2
A1 = e −
05
a1
16
λ
B1 = σ 1 − 17
Λ
where a0, a1, and σ are parameters of the model, and S0 is obtained from the following mass balance
criteria:
Λ
p λ, Λ D λ dλ = 1 18
0
where D(λ) is the Jacobian that transforms the discrete coordinate i into the continuous coordinate
λ. Using this equation and solving Eq. (1.5) to S0, the following solution is reached:
2
λ a0 − 0 5
Λ − Λ
1
S0 = e a1
−A1 + B1 D λ dλ 19
0 2π
The hydrocracking phenomena occurring within the reactor are described by the differential
equation, which illustrates the formation of compounds resulting from the cracking of components
with higher boiling points (second term on the right side of the equation) and the cracking of lower
boiling products (first term of the equation). This equation is solved for any residence time in terms of
c(λ, t), and the desired cuts in mass fraction depending on its dimensionless temperature as follows:
λ2
Cθ1 ,θ2 t = c λ, t D λ dλ 1 10
λ1
where Cθ1 ,θ2 t is the mass fraction with dimensionless temperature ranging from θ1 to θ2 with its
corresponding reactivity λ1 and λ2 as shown in Eq. (1.2) (Ancheyta 2013; Becker et al. 2016; Elizalde
et al. 2009, 2016; Elizalde and Ancheyta 2014; Laxminarasimhan et al. 1996; Martínez-Grimaldo
et al. 2011).
1.2.1.3 Single-Event Kinetic Models

Microkinetic models are developed at a molecular scale considering the chemistry of the reaction
system by deriving the kinetic equations for individual molecules. The solution adequately con-
verges in reacting systems that contain a limited number of chemical species. However, the set
of equations suggested for intricate systems, such as those represented by heavy crude fractions,
presents a challenging task to accomplish. Even modern computational tools are unsatisfactory
because the size of the reaction network increases exponentially with the number of carbons
(Ancheyta et al. 2005; Becker et al. 2016).
The advancement of these types of models for hydrocracking of heavy crude oils is propelled by
the rising availability of experimental data gathered using sophisticated analytical techniques and
increasing computational capacities. The microkinetic modeling method (single events) was ini-
tially proposed by Froment et al. (Baltanas et al. 1989; Baltanas and Froment 1985; Froment
1987; Hillewaert et al. 1988), and the reaction coefficients were calculated for each individual reac-
tion. Computational regrouping algorithms reduce the size of reaction networks without losing
information. These can be applied in various reactions, such as hydrocracking, catalytic reforma-
tion, isomerization and alkylation; however, the original regrouping method remains as unsuitable
for implementation in huge network of reactions, notwithstanding the current computational cap-
abilities. To find an approach, two alternative methods have been proposed (Becker et al. 2016;
Guillaume et al. 2011; Mitsios et al. 2009; Valéry et al. 2007): (i) a method based on decomposition
of side chains that was developed by Valéry et al. (2007) and (ii) a method based on structural
classes, which was developed by Martens and Marin (2001).
1.2.2 Kinetic Models Reported in the Literature for Hydrocracking of Heavy

Crude Oils Using Dispersed Catalysts
1.2.2.1 Kinetic Models Based on Distillation Curves
Hassanzadeh and Abedi (2010) as well as Galarraga et al. (2012) carried out experiments for
the catalytic heavy crude oil upgrading with dispersed NiWMo catalyst and Athabasca bitumen
(9.5 API) as feed in a batch reactor. The tests were performed at temperature ranges of
320–380 C, 3–70 h of reaction time, total hydrogen pressure of 3.45 MPa, and stirring rate of
500 rpm. Both works used the same kinetic model (Figure 1.5a) that was first reported by Sánchez
et al. (2005a) based on boiling point intervals. The reaction scheme takes into account five lumps:
Residue (R, 538 C+), VGO (343–538 C), distillates (D, 204–343 C), naphtha (N, IBP-204 C), and
(a) (b)
R R
k1 k2
k7
VGO VGO
k5 k1 k4
k6
k2 k9
D D
k8
k3
k3
N N
k10 k5
k4 k6
G G
(c) (d)
k5
L+ UCO G
k3
k1 k1 k4
k7
k4
k3
L–
k6 VGO k5
k9 k2
k8 k2
k6 k7
C G C D
(e)
VR
kL1
kL5
VGO
kL4
kL2
D
kL6
kL3
N
Figure 1.5 Four- and five-lump kinetic models based on distillation curves reported in the literature for heavy
oil hydrocracking using dispersed catalyst.
gases (G). Nonetheless, some pseudocomponents proposed by Galarraga et al. (2012) have different
boiling point ranges: residue (545 C+), VGO (343–545 C), D (216–343 C), N (IBP–216 C). The
mass balance equations are the same in both cases, and a first-order reaction was considered giving
the following expressions:
r R = − k 1 + k2 + k 3 + k4 yR 1 11
r VGO = k 1 yR − k 5 + k6 + k7 yVGO 1 12
r D = k 2 yR + k5 yVGO − k 8 + k9 yD 1 13
r N = k3 yR + k6 yVGO + k8 yD − k 10 yN 1 14
r G = k4 yR + k7 yVGO + k9 yD + k10 yN 1 15
where ri is the reaction rate of the component i, yi the mass fractions of component i, kj the reaction
rate coefficients for each reaction pathway j. The only difference is that Hassanzadeh and Abedi
(2010) used the weight fraction, while Galarraga et al. (2012) employed a dimensionless concentra-
tion (the concentration at any time divided by the initial concentration). Furthermore, the latter
authors performed a general reaction for the residue conversion into light products assuming
first-order reaction with k0 as the only reaction rate coefficient as follows:
k0
Residue Products 1 16
After the catalytic hydrocracking of Athabasca bitumen with ultradispersed catalyst at bench-
scale, kinetic studies on a large-scale reactor were developed by Loria et al. (2011) using the same
model as Galarraga et al. (2012). The experimental results were obtained in a pilot plant tubular
reactor at reaction temperatures of 320–380 C, residence times (τ, for continuous reactors is
assumed to be the ratio of the reactor volume to the volumetric flow rate) of 9–51 h, keeping
the total pressure constant (2.76 MPa), and a hydrogen-to-oil ratio of 3509 (scf/bbl). Although
the reaction scheme is the same, some boiling point intervals are different: VGO (343–550 C)
and R (550 C+). Because an ultradispersed catalyst is employed in the study, the internal and exter-
nal diffusion and the axial dispersion were negligible in a continuous gas-oil hydrocracking reactor,
which allowed to set the reaction equations as follows:
y τf
τf τf τf τf
dyR = − k 1 yR dτ − k 2 yR dτ − k 3 yR dτ − k 4 yR dτ 1 17
τ0 τ0 τ0 τ0
y τ0
y τf
τf τf τf τf
dyVGO = k1 yR dτ − k 5 ydτ − k 6 yVGO dτ − k 7 yVGO dτ 1 18
τ0 τ0 τ0 τ0
y τ0
y τf
τf τf τf τf
dyD = k2 yR dτ + k 5 yVGO dτ − k 8 yD dτ − k 9 yD dτ 1 19
τ0 τ0 τ0 τ0
y τ0
y τf
τf τf τf τf
dyN = k3 yR dτ + k 6 yVGO dτ + k 8 yD dτ − k 10 yN dτ 1 20
τ0 τ0 τ0 τ0
y τ0
y τf
τf τf τf τf
dyG = k 4 yR dτ + k 7 yVGO dτ + k 9 yD dτ + k 10 yN dτ 1 21
τ0 τ0 τ0 τ0
y τ0
where τ0 and τf are the initial and final residence times, respectively, and y(τ0) and y(τf) are the
initial and final weight percentages, respectively, as a function of the residence time.
Nguyen et al. (2013) carried out a kinetic study for the hydrocracking of the Arabian light atmos-
pheric residue in a batch reactor using Mo-dispersed catalyst (molybdenum naphthenate). The
experiments were carried out at temperatures of 420–430 C, residence times of 0.5–1 h, and hydro-
gen pressure of 15.2 MPa. The proposed reaction scheme (Figure 1.5b) considered the following
boiling point intervals: R (510 C+), VGO (350–510 C), D (180–350 C), and N (IBP–180 C).
A first order was assumed for all the reactions and the reaction from R to N was neglected;
furthermore, the stoichiometric coefficients of the reactions were considered, giving the following
equations:
r R = v1R k 1 CLR C LH2 + v2R k 2 CLR CLH2 1 22
r VGO = v1VGO k 1 CLR C LH2 + v3VGO k 3 C LVGO CLH2 + v4VGO k 4 CLVGO CLH2 1 23
r D = v2D k 2 CLR C LH2 + v3D k 3 C LVGO C LH2 + v5D k 5 C LD CLH2 1 24
r N = v4N k 4 C LVGO C LH2 + v5N k 5 C LD C LH2 + v6N k 6 CLN CLH2 1 25
r G = v1G k 1 C LR CLH2 + v2G k 2 CLR CLH2 + v3G k 3 CLVGO C LH2 + v4G k 4 C LVGO C LH2 + v5G k 5 CLD CLH2
1 26
+ v6G k 6 C LN C LH2
r H2 =v1H2 k 1 C LR CLH2 +v2H2 k 2 C LR C LH2 +v3H2 k 3 CLVGO CLH2 +v4H2 k 4 C LVGO C LH2 +v5H2 k 5 CLD CLH2
1 27
+v6H2 k 6 CLN CLH2
where vji is the stoichiometric coefficients of species i for reaction pathway j and C Li the molar con-
centration of species i in the liquid.
A kinetic model of four pseudocomponents (Figure 1.5c) for VR hydrocracking with Maya crude
and powder NiMo/Al2O3 catalyst was developed by Puron et al. (2014). The experimental data were
obtained from a batch reactor at three temperatures (400, 425, and 450 C), 18.5 MPa of H2 pressure
and four reaction times (10, 30, 60, and 90 min). The model was divided on four fractions: (i) boiling
point products >450 C (L+), which are formed by asphaltenes and heavy maltenes, (ii) boiling
point products <450 C (L−) that include light maltenes, (iii) gases (G), and (iv) coke (C). The reac-
tion rate equations were set as follows:
Boiling point products (L+):
rL + = − k1 + k3 + k5 yL + + k4 yC + k7 yL− 1 28
Boiling point products (L−):
rL − = k1 yL + + k9 yC − k2 + k7 + k8 yL− 1 29
Gases (G):
rG = k2 yL− + k5 yL + + k6 yC 1 30
Coke (C):
rC = k3 yL + + k8 yL − − k4 + k6 + k9 yC 1 31
da Silva De Andrade (2014) established a kinetic model for in situ upgrading of Athabasca bitumen
VR and pitch (asphaltene-rich phase) with ultradispersed catalyst. All of the tests were conducted in
a downflow reactor at temperatures between 320 and 395 C, residence times of 20–50 h, hydrogen
pressure of 3.45 MPa, and hydrogen-to-oil ratio of 90 standard cm3/cm3. The same lumped kinetic
model reported by Hassanzadeh and Abedi (2010) was used by the authors.
A six-lump kinetic model was developed by Asaee et al. (2014) using the experimental information
reported by Song et al. (2004) for the Chinese Gudao residue upgrading with ammonium phospho-
molybdate catalyst. The reaction scheme (Figure 1.6a) includes R (480 C+), VGO (350–480 C),
(a) R
k1 k5
k2 k3 k4
k6
VGO AGO N G C
k10
k7
k8
k9
(b)
kHDS k1
RS R
k3
H2S k2 VGO
D
k4
N
G
(c)
HRi Sediment RS + H2 R + H2S
LR9 to 11
LR8
LR7
LR6
VGO5
VGOi
VGO1
D
N
G
Figure 1.6 Detailed-lumping kinetic models based on distillation curves reported in the literature for heavy
oil hydrocracking using dispersed catalyst.
atmospheric gas oil (AGO, 180–350 C), N (IBP-180 C), G, and C. The heavier fraction cracks into
lower molecular weight fractions, and G and C are generated only from heavy fractions (R and VGO),
obtaining the following equations:
r R = − k1 + k 2 + k3 + k4 + k 5 yR 1 32
r VGO = k 1 yR − k 6 + k7 + k8 yVGO 1 33
r AGO = k2 yR + k6 yVGO − k 9 + k10 yAGO 1 34
r N = k3 yR + k 7 yVGO + k 9 yAGO 1 35
r G = k4 yR + k8 yVGO + k10 yAGO 1 36
r C = k5 yR 1 37
The kinetic studies for in situ upgrading process of VR and heavy oil at Mexican carbonate res-
ervoir conditions were developed by Orozco Castillo (2016) and were performed in a plug flow reac-
tor employing a Ni–W–Mo catalyst (the same as in the aforementioned studies) at temperatures of
320–360 C, residence times of 24–96 h (heavy oil) and 48–120 h (VR), and hydrogen pressure of
10.34 MPa. The kinetic model is the same as used by Loria et al. (2011).
Kim et al. (2017) used an oil-soluble (MoS2) catalyst for the hydrocracking of a residue fraction at
temperatures of 380–400 C, reaction times of 1–16 h, and hydrogen pressure of 9.5 MPa in a batch
reactor. The five-lump reaction scheme (Figure 1.5d) includes unconverted oil (UCO, 560 C+),
VGO (320–560 C), D (IBP-320 C), G, and C. The cracking of heavier fractions produces lighter
fractions, except when UCO forms C by condensation reactions. A second-order of reaction was
considered for UCO, a first-order of reaction for VGO, and zero-order of reaction for D and G,
obtaining the following equations:
r UCO = − k1 + k3 + k 6 + k7 y2UCO 1 38
r VGO = k 1 y2UCO − k 2 + k4 yVGO 1 39
r D = k 2 yVGO + k 7 y2UCO − k 5 1 40
r G = k3 y2UCO + k4 yVGO + k5 1 41
r C = k6 y2UCO 1 42
A kinetic model for the hydrocracking of Iranian heavy crude oil with an oil-soluble dispersed
catalyst in a batch reactor was developed by Huang et al. (2017). The experimental data were carried
out at temperatures of 405–435 C, reaction times of 1–10 h, and hydrogen pressure of 9 MPa. The
kinetic model was the same as reported by Hassanzadeh and Abedi (2010), but the boiling point
ranges were similar to those reported by Nguyen et al. (2013), with different R (524 C+) and
VGO (350–524 C) boiling points.
Ortega-García et al. (2017) conducted a kinetic study for heavy oil (12 API) upgrading in a contin-
uous stirred tank reactor (CSTR) reactor using a liquid-acid catalyst at moderate conditions: tempera-
tures of 350–370 C, residence times of 12–180 h, hydrogen pressure of 9.8 MPa, and hydrogen-to-oil
ratio of 62.9 standard cm3/cm3. The five-lump kinetic model is similar to that reported by Hassanza-
deh and Abedi (2010) (with different names, VGO is named HGO and D is termed LGO) based on the
following boiling point ranges: R (540 C+), heavy gas oil (HGO, 343–540 C), light gas oil (LGO,
221–343 C), N (IBP-221 C).
Elahi et al. (2019) performed a kinetic study for in situ upgrading inside a carbonate rock (porous
media) by injecting VR from Mexican heavy oil, hydrogen, and NiMo-ultradispersed nanocatalyst.
The experimental part was carried out in a continuous reactor where a mixture of VR and nano-
catalysts (0.22 wt.% of the carbonate cores) is submitted at reservoir conditions (temperatures of
320–360 C, residence times of 12–72 h, hydrogen pressure of 10 MPa, and hydrogen-to-oil ratio
of 150 standard cm3/cm3). The kinetic model scheme is the same as employed by Hassanzadeh
and Abedi (2010).
Félix and Ancheyta (2019b) proposed a four-lump kinetic model (Figure 1.5e) for the hydrocrack-
ing of heavy crude oil using molybdenite ore catalyst in a batch reactor. The tests were conducted at
VR
kL1
kL5
VGO
kL4
Liquid
kL2 phase
D
kL6
kL3
N
Coke As k4 Gases
k13 k1 k7
k11 k9
Re
k5 k10
k2 k12
Ar
k8
k3 k6
Sa
Figure 1.7 Global kinetic model to predict liquid, gas, and coke yields combining kinetic models based on
SARA fraction and distillation curves.
mild conditions (temperatures of 360–400 C, reaction times of 2–5 h, and hydrogen pressure of 3.9
MPa). The boiling point ranges for R, VGO, D, and N were similar to those reported by Sánchez
et al. (2005a, b): R, VGO, D, and N. The reaction scheme is based on the conversion from heavier
fractions to lighter fractions, but the G and solid (Coke, C) fractions were calculated (Figure 1.7)
following a complementary kinetic model (Félix and Ancheyta 2019a), which is presented later,
giving the following equations:
r R = − k1 + k 2 + k3 yR 1 43
r VGO = k 1 yR − k 4 + k5 yVGO 1 44
r D = k2 yR + k4 yVGO − k 6 yD 1 45
r N = k3 yR + k 5 yVGO + k 6 yD 1 46
Complementary reactions rate equations:
r G = k 4 yAs + k 7 yRe + k9 yAr + k 10 ySa 1 47
r C = k13 yAs 1 48
A detailed (seven-lump) kinetic model was developed by Álvarez et al. (2019) based on the
residua upgrading (Safaniya residue and Arabian light atmospheric residue) in a semibatch reactor
with an oil-soluble Mo precursor catalyst (molybdenum octoate). The boiling point range is similar
to those reported by Nguyen et al. (2013), and the pseudocomponents are involved in one hydro-
desulfurization (HDS) and four hydrocracking reactions (Figure 1.6b). All reactions are performed
in series, except for the parallel cracking of R into D. The cracking of R and VGO to produce lighter
fractions also forms G, while D, which comes from heavy fractions (R and VGO), only produces N.
The stoichiometric coefficients are taken into account in hydrocracking reaction rates, and the reac-
tion orders for each lump are estimated, leading to the resulting kinetic equation:
4 EA,j
ni
ri = vji A0,j e − RT CLi 1 49
j=1
where A0,j is the pre-exponential or collision factor for the reaction j, EA,j is the activation energy of
the reaction j, ni the order of reaction for component i, R is the universal gas constant, and T is the
reaction temperature. Considering the sulfur removal from the residue fraction (RS) to produce
H2S, the stoichiometry of this reaction is based on the addition of hydrogen to 4,6-DMDBT to obtain
the following rate equation:
EA,HDS
nRS
r HDS = vHDS A0,HDS e − RT CLRS CLH2 1 50
Another detailed kinetic model (Figure 1.6c) was reported by Browning et al. (2019) for the
hydrocracking of VR in a semibatch reactor using the same catalyst as reported by Álvarez et al.
(2019) at similar operating conditions. The main fractions were divided in different boiling point
ranges: R (525 C+), VGO (350–525 C), D (160–350 C), N (IBP-160 C), and G. Additionally, the
VGO and R fractions were separated into intervals of 35 C in detailed lumps: five lumps for VGO,
six lumps for light residue (LR, 525–770 C), and seven lumps for heavy residue (HR, 770–980 C).
One HDS reaction and four hydrocracking reactions are assumed as reported by Álvarez et al.
(2019), and the conversion from LR sublumps to VGO sublumps produced G. All reactions are
believed to consume hydrogen and the cracking of LR and VGO sublumps to produce D also forms
N. Furthermore, the HR lump yields D and sediments, giving the following general equation:
r i = vji k j Ci ni 1 51
where Ci is the concentration of the species i. Only the reaction in series from LR sublumps to VGO
sublumps to D is assessed without the formation of any by-products. The first-order reaction for
hydrogen is assumed in the latter and HDS reactions obtaining the following overall equation:
r i = vji k j Ci ni CH2 1 52
A kinetic model for in situ upgrading of residue fraction from Aguacate heavy oil using ultradis-
persed nanocatalyst (Ni–Mo) in four dolomite cores was developed by Duran Armas (2021) at near-
reservoir conditions: temperatures of 335–365 C, residence times of 12–72 h, and hydrogen pres-
sure of 10 MPa. The five-lump kinetic model is similar to that used by Elahi et al. (2019) and Has-
sanzadeh and Abedi (2010).
Pham et al. (2021) reported a kinetic model for the slurry-phase hydrocracking of VR using an oil-
soluble catalyst (Mo-octoate) in a CSTR. Severe operating conditions were employed to obtain the
experimental data: temperatures of 400–450 C, residence times of 1–4 h, hydrogen pressure of 16
MPa, and hydrogen-to-oil ratio of 1500 standard cm3/cm3. The study utilized a five-lump kinetic
model resembling that of Galarraga et al. (2012), encompassing the following cuts: R (524 C+),
VGO (343–524 C), D (177–343 C), N (IBP-177 C), and G. A second-order of reaction for residue
is supposed, giving the following equations:
r R = − k 1 + k2 + k 3 + k4 y2R 1 53
r VGO = k 1 y2R − k 5 + k6 + k7 yVGO 1 54
r D = k 2 y2R + k5 yVGO − k 8 + k9 yD 1 55
r N = k3 y2R + k 6 yVGO + k 8 yD − k 10 yN 1 56
rG = k 4 y2R + k7 yVGO + k9 yD + k10 yN 1 57
A fine-dispersed catalyst (nickel nanoparticles in polyethylene glycol, PEG300) was employed by

Coronel-García et al. (2021) for the heavy oil upgrading in a batch reactor at mild conditions (tem-
peratures of 310–370 C, reaction times of 4–72, and hydrogen pressure of 7 MPa). The kinetic
model is the same as reported by Hassanzadeh and Abedi (2010), with a reaction order for residue
conversion of 2.5 used to establish the subsequent system of differential equations:
r R = − k1 + k 2 + k3 + k4 yR2 5 1 58
r VGO = k 1 y2R 5 − k 5 + k6 + k7 yVGO 1 59
rD = k2 y2R 5 + k5 yVGO − k 8 + k9 yD 1 60
rN = k3 yR2 5 + k6 yVGO + k8 yD − k 10 yN 1 61
r G = k 4 y2R 5 + k7 yVGO + k9 yD + k10 yN 1 62
Cai et al. (2022) reported a kinetic model for the hydrocracking of Karamay atmospheric residue
using Mo-based ionic liquid (1-alkyl-3-methylimidazolium cations with molybdate anion,
[Cnmim]2[MoO4]) as dispersed catalyst in a batch reactor at temperatures of 390–430 C, reaction
times of 1–8 h, and hydrogen pressure of 12 MPa. A similar kinetic model to that reported by Kim
et al. (2017) was employed with different boiling points: R (524 C+) instead of UCO, VGO
(350–524 C), and D (<350 C). The set of differential equations is analogous in both cases because
the UCO and R followed a second-order kinetics.
The trialkylmethylammonium molybdate ionic liquid was used for the slurry-phase hydrocrack-
ing of Karamay atmospheric residue in a batch reactor by Ma et al. (2022). The experimental data
were obtained at temperatures of 380–430 C, reaction times of 0.5–6 h, and hydrogen pressure of
12.3 MPa, and the kinetic model utilized was in line with that of Cai et al. (2022).
1.2.2.2 Kinetic Models Based on SARA Fractions

A combined kinetic model (based on SARA fractions and distillation curves) was reported by
Fukuyama and Terai (2007) for the hydrocracking of VR using two different catalysts (pyrite
and pyrite-active carbon) in a semibatch reactor at temperatures of 415–445 C, residence times
of 1–4 h, and a hydrogen pressure of 10 MPa. The products were classified as gas (G), naphtha
(N, IBP-171 C), kerosene (K, 171–232 C), gas oil (GO, 232–343 C), VGO (343–525 C), vacuum
residue (VR, 525 C+), and coke (C). Additionally, the VR was further separated before and after
reaction into SARA fractions: Asphaltenes (As), resins (Re), aromatics (Ar), and saturates (Sa). The
authors supposed a reaction scheme where As in VR is the main reactant, which only produces
C and VGO, meanwhile VGO generates G, N, K, and GO lumped into light oil (LO), which is also
produced by Sa. Concerning the SARA fractions, Ar cracks into Sa, LO, and VGO, while Re yields
VGO. Furthermore, other condensation reactions are performed from Ar to Re and Re to As. All of
these hydrocracking reactions (Figure 1.8) are assumed to be first-order of reaction, resulting in the
subsequent equations:
dyAs
= − k 8 + k 9 yAs + k3 yRe 1 63
dt
dyRe
= − k 3 + k7 yRe + k2 yAr 1 64
dt
Figure 1.8 SARA- and distillation k3 k2 k1

curves-based kinetic model for heavy As Re Ar Sa
oil hydrocracking using mineral
catalysts.
k9 k8 k7 k6 k5 k4
C VGO LO
k10
dyAr
= − k 1 + k2 + k 5 + k6 yAr 1 65
dt
dySa
= k 1 yAr − k 4 ySa 1 66
dt
dyLO
= k4 ySa + k5 yAr + k10 yVGO 1 67
dt
dyVGO
= − k 10 yVGO + k6 yAr + k7 yRe + k8 yAs 1 68
dt
dyC
= k 9 yAs 1 69
dt
Sheng et al. (2017) developed a kinetic model for asphaltenes liquefaction using a small volume of
industrial distillate as hydrogen donor and dialkyldithiocarbamate molybdenum as a dispersed oil-
soluble catalyst. A four-lump model (asphaltenes, maltenes, gas, and coke) describes the reaction
scheme (Figure 1.9a). The reactions were carried out in a 250-mL batch reactor at 7 MPa of
Figure 1.9 Four-lump kinetic models based on (a)

SARA fractions reported in the literature for heavy oil k2 k3
hydrocracking using dispersed catalyst. Asphaltenes
Gases k1 Coke
k4
Maltenes
(b)
k2
As
k1
k7
k5
Re
k4
k8
Ar
k6
k3
Sa
hydrogen pressure, temperature of 380–400 C, reaction time of 3–9 h, 5:1 asphaltenes-to-hydrogen

donor mass ratio, and 500 μg Mo/g asphaltenes. A first-order reaction was assumed and the math-
ematical equations of the mentioned kinetic model are as follows:
Asphaltene (y1):
dy1
= − k12 + k13 + k14 y1 1 70
dt
Maltene (y2):
dy2
= k12 y1 − k 23 y2 1 71
dt
Gases (y3):
dy3
= k13 y1 + k23 y2 1 72
dt
Coke (y4):
dy4
= k14 y1 1 73
dt
Based on the data reported by Félix and Ancheyta (2019b), these authors developed four kinetic
models for the hydrocracking of heavy crude oil using dispersed Mo-based catalyst based on SARA
fractions (Félix and Ancheyta 2019a), where the heavy fractions produce lighter fractions by par-
allel and in series reactions. The first model (Figure 1.9b) is a four-lump kinetic model that takes
into account only SARA fractions in the liquid phase; furthermore, Ar and Re fractions can undergo
condensation reactions to generate Re and As fractions, respectively. A first-order reaction is
assumed to obtaining the following equations:
r As = − k1 + k2 + k3 yAs + k7 yRe 1 74
r Re = k1 yAs + k8 yAr − k 4 + k5 + k 7 yRe 1 75
r Ar = k2 yAs + k4 yRe − k 6 + k8 yAr 1 76
r Sa = k3 yAs + k5 yRe + k 6 yAr 1 77
The second kinetic model (Figure 1.10a) considers the same reactions between SARA fractions
carried out in the four-lump model. Additionally, this reaction mechanism (five-lump) also consid-
ers the production of G from all SARA fractions, giving the following equations:
r As = − k1 + k2 + k3 + k 4 yAs + k11 yRe 1 78
r Re = k1 yAs + k12 yAr − k 5 + k6 + k7 + k11 yRe 1 79
r Ar = k2 yAs + k5 yRe − k 8 + k9 + k 12 yAr 1 80
r Sa = k3 yAs + k6 yRe + k 8 yAr − k 10 ySa 1 81
r G = k 4 yAs + k 7 yRe + k9 yAr + k 10 ySa 1 82
Another five-lump kinetic model (Figure 1.10b) was reported by Félix and Ancheyta (2019b), which
considers the same reaction mechanism as in the four-lump kinetic model. Furthermore, the conden-
sation reaction of As to generate C is included in this model. Therefore, in Eq. (1.74), the term −k9yAs
is added as well as in the reaction rate equation for C:
r C = k9 yAs 1 83
Figure 1.10 Five- and six-lump kinetic models (a)

based on SARA fractions reported in the literature k9 k2
for heavy oil hydrocracking using dispersed catalyst. C As
k1
k7
k5
Re
k4
k8
Ar
k6
k3
Sa
(b)
As k4 G
k1
k11 k7
Re k9
k5 k10
k12
k2
Ar
k8
k6
k3
Sa
(c)
C As G
k13 k4
k1
k11 k7
Re k9
k5 k10
k12
k2
Ar
k8
k6
k3
Sa
A six-lump kinetic model (Figure 1.10c) was also reported by Félix and Ancheyta (2019b) con-
sidering the five-lump reaction mechanism mentioned before that includes the production of G. In
addition, the condensation reaction of As to produce C is considered and the term −k13yAs is
included in Eq. (1.78), likewise the reaction rate equation for the C fraction. This kinetic model
is employed to calculate the G and C fractions in the four-lump model based on distillation curves
that were previously discussed.
r C = k13 yAs 1 84
1.2.3 Kinetic Models Based on Continuous Lumping

A continuous kinetic model was developed by Martínez-Grimaldo et al. (2011) for the Maya heavy
oil upgrading in a batch reactor using a dispersed catalyst (powder ammonium heptamolybdate) at
400 C, reaction times of 1–24 h, and hydrogen pressure of 5.5 MPa. The continuous kinetic model
forecasted the composition based on distillation curves of liquid phase during the hydrocracking
reaction using Eqs. (1.1)–(1.10). Since the continuous kinetic model solely predicts the composition
of the liquid phase, the quantities of gas (G), liquid (L), and solid (S) are computed as follows:
dL − k L L − LL
= γ 1 85
dt 1 + K LL + K SS + K GG
dS k L L − LL
= PS γ 1 86
dG k L L − LL
= PG γ 1 87
PS + PG = 1 1 88
where kL, KL, LL, KS, PS, KG, PG, and γ involves the liquid, solid, and gas phases to be used as para-
meters in the continuous lumping equations.
1.2.4 Thermodynamic Model to Predict the Asphaltenes Flocculation

and Sediments Formation
Asphaltene fraction is in equilibrium surrounded by resins, aromatics, and saturates in crude oil,
with the resins acting as peptizing agents. Therefore, a regular solution solubility model can be
applied (Akbarzadeh et al. 2004), contemplating the liquid–solid equilibrium. The equilibrium
ratios of each component are used to determine the quantity of precipitated asphaltene. The activity
of the solid phase for asphaltenes approaches 1 when this fraction is in solid–liquid equilibrium.
Furthermore, by disregarding the heat of fusion (ΔH/RT), the equilibrium ratio (Ki) can be obtained
according to the following:
x Si VL V Li V Li 2
Ki = L
= exp 1 − i + ln + δi − δmix 1 89
xi V mix V mix RT
where x Si and x Li are the mass fractions of the component i in the solid and liquid phase, respectively,
V Li is the liquid molar volume of the component i, Vmix is the average molar volume, δi is the
solubility parameters of the component i, and δmix represents the solubility parameters of the
mixture. The solubility parameter of the liquid (δL) correlates with the internal energy of vapori-
zation (ΔULV) and the molar volume of the liquid phase (VL) according to the equation:
1 2
ΔU LV
δL = 1 90
VL
The energy of vaporization and the molar volume are calculated using a cubic equation, which
enables the determination of the solubility parameter. It has been reported that the SRK EOS with a
Peneloux correction (c) gives the best approaches of liquid volumes whereby accurate values of the
solubility parameter are obtained (Akbarzadeh et al. 2004). The SRK EOS has the following form:
RT aT
P= − SRK SRK 1 91
VSRK
−b V V +b
where P is the pressure, VSRK is the molar volume calculated with the SRK EOS, and a(T) and b are
EOS parameters. The corrected liquid molar volume is obtained as
V L = V SRK − c 1 92
Substituting this term in Eq. (1.90) gives

RT aT
P= − 1 93
VL − b − c VL + c VL + c + b
The Peneloux correction term (c) can be estimated by

0 40768T c
c= 0 29441 − Z RA 1 94
Pc
where Tc is the critical temperature, Pc is the critical pressure, and ZRA is the Racket compressibility
factor. If the value of the ZRA is not available, it is calculated using the acentric factor (ω) with the
following equation:
Z RA = 0 29056 − 0 08775ω 1 95
The internal energy of vaporization is related with the EOS parameters (a and b) by
a T da b
ΔU LV = − ln 1 + SRK 1 96
b b dT V
Therefore, at any temperature and pressure, Eq. (1.92) can be solved for the molar volume, and
the solubility parameter is calculated by
1 2
da
a−T b
δL = dT ln 1 + 1 97
L L
bV V +c
The SRK EOS parameters (a and b) for a given component are calculated from the critical proper-
ties and the acentric factor as follows:
1 2 2
0 42748R2 T 2c T
aT = 1 + 0 48 + 1 574ω − 0 176ω2 1− 1 98
Pc Tc
0 08664RT c
b= 1 99
Pc
The critical properties and the acentric factor of each SARA fraction to calculate the EOS para-
meters can be estimated with some correlations reported in literature (Akbarzadeh et al. 2004; Riazi
2005). For the saturate fraction, the correlation depends on some constants shown in Table 1.1 and
molar mass as can be seen in the following equation:
ε = ε∞ − exp A − EM C 1 100
Table 1.1 Constants and their correlations for Eq. (1.100).
ε ε∞ A E C
Tb 1070 6.98291 0.02013 0.67

Tb/Tc 1.15 −0.41966 0.02436 0.58
−Pc 0 4.65757 0.13423 0.50
−w 0.3 −3.06826 −1.04987 0.20
where ε, ε∞, A, E, and C are constants, and M is the molar mass. For aromatics, resins, and asphal-
tenes, the correlations are solely based on the molar mass. Due to the fact that these fractions con-
sist of polynuclear aromatics with slight structural variations, a correction factor (cf, Table 1.6) was
implemented for adjusting the acentric factor and account for such variations, even in the saturate
fraction. These correlations are presented as follows:
T c = 77 856M 0 4708 1 101

− 0 7975
Pc = 1891 4M 1 102
w = cf 0 5837 ln M − 2 5389 1 103
The application of mixing rules allows for the computation of crude oil properties, based on the
SARA fractions:
M mix = xiMi 1 104

i = SARA
V mix = xi V i 1 105
i = SARA
x i V i δi
δmix = 1 106
i = SARA
V mix
where Mmix is the molar mass of crude oil, xi is the mass fraction of component i, and Mi is the molar
mass of component i. After estimating the equilibrium ratio, the subsequent task is to solve the
phase equilibrium for asphaltene precipitation. When flocculation starts, the fraction of asphaltene
that is dissolved is equivalent to the highest fraction that can dissolve in crude oil. The solid phase
fraction is 1 when asphaltenes are precipitated, enabling the maximum soluble fraction to be
obtained as described below:
1
x Lmax = 1 107
K
where x Lmax is the maximum fraction soluble in the crude oil. Once the maximum amount of soluble
asphaltenes within crude oil has been determined, this value is compared with the calculated
asphaltene amount through the SARA-based kinetic model to establish the sediment fraction as
follows:
As − x max
Sediment fraction = ycal L
1 108
where ycal
As is the calculated composition of asphaltenes. Finally, the sediment formation is deter-
mined through asphaltene precipitation, and this methodology is depicted in Figure 1.11.
1.3 Kinetic Parameters Estimation
Once experimental data have been collected, it is essential to devise a reaction scheme capable of
accurately predicting the product composition. The reaction rate equations are obtained by deriving
them from the proposed reaction scheme, while the mass balance equations are obtained by
making some assumptions. For estimating parameters in kinetic modeling of complex reaction net-
works, a nonlinear optimization process is typically employed. However, when the number of para-
meters to be estimated is significant, convergence issues may arise. Additionally, the initialization
Critical properties
Solve modified SRK
(Tc, Pc and w)
SARA molar mass EOS to obtain molar
estimation
volume (Eq. 1.91)
(Eqs. 1.101–1.103)
Calculate the
solubility parameter
(Eq. 1.97)
Calculate SARA
Mixing rules
fractions and
Experimental data estimation
upgraded oil
(Eqs. 1.104–1.106)
composition
Calculate equilibrium
ratio with the regular
solution model
(Eq. 1.89)
Compare calculated
Solve phase
fraction with
Fraction of sediments equilibrium
maximum soluble
(Eq. 1.107)
fraction (Eq. 1.108)
Figure 1.11 Methodology to calculate the fraction of precipitated asphaltenes to form sediments.
of the kinetic parameter values is another problem frequently found in nonlinear estimation that
may converge to a local minimum of the objective function instead the global minimum (Ancheyta
and Sotelo-Boyás 2000; Quitian and Ancheyta 2016a).
After optimizing the reaction rate coefficients, the activation energies and collision factors are
calculated using the Arrhenius equation.
− EA,j
k j = A0,j e RT 1 109
MATLAB software is widely used for estimating kinetic parameters (Álvarez et al. 2019; Brown-
ing et al. 2019; Félix and Ancheyta 2019a, b; Huang et al. 2017; Loria et al. 2011; Ortega-García et al.
2017; Pham et al. 2021; Puron et al. 2014; Sheng et al. 2017); however, other commercial packages,
such as Microsoft Excel (da Silva De Andrade 2014; Duran Armas 2021; Elahi et al. 2019; Galarraga
et al. 2012), R (Coronel-García et al. 2021), 1st opt (Cai et al. 2022; Ma et al. 2022), and PROII, have
been used to calculate kinetic parameters (Nguyen et al. 2013). Regarding the optimization algo-
rithms, the majority of them use the Levenberg–Marquardt methodology (Coronel-García et al.
2021; Félix and Ancheyta 2019a, b; Huang et al. 2017; Kim et al. 2017; Nguyen et al. 2013;
Ortega-García et al. 2017; Pham et al. 2021). In addition, the trust-region-reflective (Álvarez
et al. 2019; Browning et al. 2019) and the Universal Global Optimization are also reported (Cai
et al. 2022; Ma et al. 2022). The mass balance equations are preferred to be solved numerically
by integral factor technique (Hassanzadeh and Abedi 2010), the trapezoidal rule (Galarraga
et al. 2012), the Petzold–Gear backward differentiation formulae (Nguyen et al. 2013), Runge–Kutta
(Félix and Ancheyta 2019a, b; Huang et al. 2017; Ortega-García et al. 2017; Pham et al. 2021; Puron
et al. 2014; Sheng et al. 2017), and the Runge–Kutta–Fehlberg method (Cai et al. 2022; Ma et al.
2022) instead of obtaining analytical solutions (Asaee et al. 2014; Coronel-García et al. 2021).
Regarding the objective function, the most used is the SSE (sum of the squared error) (Álvarez
et al. 2019; Browning et al. 2019; Cai et al. 2022; Félix and Ancheyta 2019a, b; Galarraga et al.
2012; Huang et al. 2017; Kim et al. 2017; Loria et al. 2011; Ma et al. 2022; Ortega-García et al.
2017; Pham et al. 2021; Puron et al. 2014). Nonetheless, the SSECC (correlation coefficient) (da Silva
De Andrade 2014; Duran Armas 2021; Elahi et al. 2019) and the SSEWF (weighting factor) (Asaee
et al. 2014; Coronel-García et al. 2021) are also employed.
1.3.1 Assumptions
The mass balance equations for kinetic models during the hydrocracking reactions carried out in
batch reactors using slurry phase catalysts assume the following considerations (Matsumura et al.
2005; Quitian and Ancheyta 2016a; Rezaei et al. 2010):
✓ Liquid and gas phases are solely present when using dispersed catalysts
✓ Each phase is considered as pseudohomogeneous
✓ Reactions are carried out a constant temperature
✓ The volume is kept as a constant
✓ The mass transfer between phases is invariant with the time
✓ Reactions take place in liquid phase
Kinetic studies can be conducted under various experimental conditions, depending on whether
kinetics is intrinsic or not. However, data from reaction kinetics acquired under conditions of mass
transfer limitations cannot be directly applied to develop kinetic models, as these limitations can
conceal results and cause misinterpretations. There are specific prerequisites for conducting kinetic
experiments in circumstances where transport resistances are insignificant. If this is possible, the
intrinsic kinetics and its parameters are obtained (rate constant, activation energies, and collision
factors). Therefore, it is important to perform kinetic experiments under conditions where transport
resistances are negligible (Bej 2002; Mederos et al. 2009; Perego and Peratello 1999).
Gradients of concentration and temperature may arise at the phase boundary within the reactor.
In the case of solid catalysts, these gradients can occur between the internal solid surface and the
bulk fluid (interphase), as well as within the catalyst particles themselves (intraphase). These gra-
dients are also referred to as external or internal, respectively, and their presence complicates data
analysis (Fogler 1992; Perego and Peratello 1999). For batch reactors, the agitation or external dif-
fusion test (Figure 1.12a) is useful to avoid interphase gradients. In these experiments, the influence
of changing the stirring rate on conversion at constant operating conditions is analyzed. The exter-
nal transport-free operation regime allows the conversion to remain constant as the stirring speed is
increased, whereas the internal diffusion test (Figure 1.12b) is employed to find the particle size of
the dispersed catalyst at which kinetic control is carried out. Intraphase gradients are excluded dur-
ing reaction if catalyst is finely crushed into progressively smaller particles until the reactant con-
version is not changed (Perego and Peratello 1999).
A significant number of researchers determine the reaction order from the experimental
findings or adopt values documented in the literature. The most inferred reaction order in the
kinetic model for the hydrocracking of heavy oil using dispersed catalyst is the first order. Only
few works corroborate the assumed order by linearizing the reaction rate equation with the
supposed reaction order and plotting the yield (yi) or conversion of the main reactant (Xi) term
(a)
Stirring motor
Conversion
Limitations by
external diffusion
No limitations by
external diffusion
RPM
(b)
dp
dp/2
Conversion
Limited by
internal diffusion
dp/3
Kinetic control
dp/4
1 2 3 4
1/dp
Figure 1.12 External diffusion test (a) and internal test (b) to evaluate the influence of catalyst particle size on
the conversion at constant operating conditions.
against the reaction time. For reaction order different than 1 (n 1), “y1i − n − y1i,0− n” or “(1 − Xi)1 − n
− 1” is plotted against t, being the slope “(n − 1)k” or “ n − 1 yni,0− 1 k”, respectively, while for n=1,
“ln(yi,0/yi)” or “ln(1/(1 − Xi))” is plotted against t and the slope is k in both cases.
1.3.2 Initialization of Parameters

The parameter initialization is the foremost step of the optimization process as the estimation of
kinetic parameters heavily relies on the initial guesses. Commonly, the initial guess is obtained
from literature or calculated using the Monte Carlo algorithm, as documented earlier with satisfac-
tory outcomes (Asaee et al. 2014; Félix et al. 2019; Félix and Ancheyta 2019a, b; Hassanzadeh and
Abedi 2010; Huang et al. 2017; Ortega-García et al. 2017; Pham et al. 2021). The Monte Carlo
method assigns random numbers to the kinetic parameters and the objective function is calculated.
This calculation is repeated, and a plot with different kinetic parameters and their corresponding
objective function is made. The kinetic parameters that minimize the objective function are selected
as initial values for nonlinear optimization.
1.3.3 Nonlinear Optimization

Some models may contain numerous parameters requiring a precise estimation due to their high
nonlinearity. In such cases, when calculating kinetic parameter values, multiple solutions of the
objective function can arise during the optimization process by which optimal values could not be
reached. The optimal solution depends mostly on initial guesses of kinetic parameters
(Alcázar and Ancheyta 2007; Félix et al. 2019). A general method for estimating kinetic para-
meters (Figure 1.13) that considers various steps to calculate and validate the best parameter
set minimizing the objective function has been found to yield favorable results in the literature
(Alcázar and Ancheyta 2007).
Once the mass balance of reactants and products is obtained, the kinetic parameters are calcu-
lated by minimizing the difference between the experimental and calculated values. The set of ordi-
nary differential equations (ODEs) is solved analytically (integrating factor for example) or using
numerical methods (fourth-order Runge–Kutta method). The nonlinear optimization problem can
be solved using optimization algorithms, such as Levenberg–Marquardt, interior-point, trust-
region-reflective, among others, to find the minimum value of the objective function (Alcázar
and Ancheyta 2007; Angeles et al. 2014; Ortiz Moreno et al. 2014; Quitian and Ancheyta 2016a).
1.3.4 Objective Function

The most used objective function for the kinetic modeling during the hydrocracking of heavy oil is
the SSE. Nevertheless, other objective functions, such as the SSEWF and the SSECC, also can be
used together with SSE. Similar objective functions, such as the average absolute difference (AAD),
and more recently, the average absolute error (AAE) have been utilized to compare calculated and
experimental data (Alcázar and Ancheyta 2007; Ancheyta and Sotelo-Boyás 2000; Asaee et al. 2014;
da Silva De Andrade 2014; Stratiev et al. 2021). The AAE objective function has recently gained
attention (Félix et al. 2022a,b; Stratiev et al. 2021; Tirado et al. 2022) due to its ability to provide
more balanced weightings to all yields, even when their differences in order of magnitude are high,
which contributes to have an improved level of accuracy.
nc
2
SSE = yexp
i − ycal
i 1 110
i=1
Optimal kinetic
parameters
Perturbations
Experimental data, objective
Graphical analysis Yes
Monte Carlo function, and reaction equations
Reported values
Sensitivity
analysis Global
minimum
Initialization of Nonlinear
of the
parameters optimization
objective
Statistical function?
Interior point algorithm analysis
Constraints
Residual No
Parity plot
Figure 1.13 Methodology for optimal estimation of kinetic parameters.

nc
exp 2
SSEWF = wi yi − ycal
i 1 111
i=1
nr
SSECC = wi SSE + W R nr − ϕ2j 1 112
j=1
nc
AAD = yexp
i − ycal
i 1 113
i=1
nc
exp
yi − ycal
i
exp
yi × 100
i=1
AAE = 1 114
N
where wi is the weighting factor, WR is the Arrhenius fitting index weight factor, and ϕ2j is the cor-
relation coefficient of the reaction j. The most used weighting factors are 1 y2exp and the ratio of the
arithmetic mean yield of the main reactant to the arithmetic mean yield of each pseudocomponent
(Alcázar and Ancheyta 2007; Asaee et al. 2014; Félix et al. 2022a,b; Gauthier et al. 2006; Rana et al.
2021; Saleh 2018; Stratiev et al. 2021).
1.3.5 Sensitivity and Statistical Analyses

1.3.5.1 Perturbations
After estimating the kinetic parameters, the sensitivity analysis is applied through perturbations by
changing one parameter at a time while keeping the others constant. This allows to re-evaluate the
objective function for each perturbation and plot the corresponding percentage of perturbation
against the value of the objective function. If the original parameters (0% perturbation) give the
minimum value for the objective function after applying all perturbations, it can be concluded that
the global minimum has been achieved. Conversely, if one or more parameters fail to achieve the
global minimum at 0% perturbation, it suggests that the nonlinear optimization was inadequate
and the kinetic parameters should be recalculated.
1.3.5.2 Parity Plots

The parity plot is another tool to evaluate the adjustment between experimental and calculated
values. A satisfactory model fit is observed when values align with a 45 line (Alcázar and Ancheyta
2007; Félix et al. 2019).
1.3.5.3 Residuals
It is common to use a plot of residual values against the number of experimental observations to
verify the adequacy of the proposed model and calculated parameters. The plot should demonstrate
a regular distribution of residuals with no prediction bias. In the case of a pattern or tendency, it can
be concluded that the parameters were estimated incorrectly. Another useful method to observe the
fit quality is the parity plot, in which experimental values are plotted against calculated values
(Alcázar and Ancheyta 2007; Félix et al. 2019; Félix et al. 2022a,b; Sámano et al. 2020).
Residual = yexp
i − ycal
i 1 115
1.3.5.4 AIC and BIC

To compare the kinetic models based on SARA fractions, both the Akaike and Bayesian informa-
tion criteria (AIC and BIC, respectively) were used (Sánchez et al. 2007). The AIC takes into account
the intricacy of a model and how well it fits the data. BIC is more rigorous in penalizing free para-
meters in comparison with AIC. Both approaches strive to choose the model with the best fit and
minimum number of free parameters, as illustrated below:
SSE
AIC = 2h + Nln 1 116
N
SSE
BIC = hln N + Nln 1 117
N
A penalization term (2h) is incorporated in the AIC that increases with the number of parameters
(h), enabling effective comparison of models with different parameters. In both cases, the preferred
model is that with the lowest value of AIC and BIC.
1.4 Results and Discussion
Among all the reported kinetic models for the hydrocracking of heavy oil using dispersed catalyst,
only the kinetic models developed by Félix and Ancheyta (2019a, b), and Fukuyama and Terai
(2007) employed mineral catalyst (molybdenite and pyrite, respectively), thus this section is focused
to their results.
1.4.1 Kinetic Parameters

1.4.1.1 Assumptions
In the assumptions of both kinetic models, the fist-order of reaction for asphaltenes and residue
conversion is supposed, as well as external and internal diffusion phenomena are neglected.
The plot to corroborate the first-order of reaction for asphaltenes conversion reported by Félix
and Ancheyta (2019a, b) during the hydrocracking of heavy oil using mineral catalysts is shown
in Figure 1.14. Here, t versus ln(1/(1 − XAs)) is plotted at the different studied temperatures, where
the experimental data follow a straight-line relationship, which suggests that the reaction order is
indeed one. In addition to the hydrocracking of asphaltenes, Figure 1.14 shows a plot of t versus
ln(1/(1 − XR)) to confirm first-order reaction in hydrocracking of residue at different temperatures,
as reported by the same authors. The experimental data follow a straight line, which confirms the
first-order of reaction. These results are in accordance with most of the reported kinetic models for
residua (Asaee et al. 2014; da Silva De Andrade 2014; Elahi et al. 2019; Galarraga et al. 2012; Has-
sanzadeh and Abedi 2010; Huang et al. 2017; Loria et al. 2011; Nguyen et al. 2013; Orozco Castillo
2016; Ortega-García et al. 2017; Puron et al. 2014) and asphaltenes (Sheng et al. 2017) conversion
during the hydrocracking of heavy oils using dispersed catalysts. Nonetheless, this supposition not
always is corroborated as in the case of Fukuyama and Terai (2007), and only few studies have
corroborated the supposed first-order reaction (Duran Armas 2021; Elahi et al. 2019; Orozco
Castillo 2016).
Additional experiments were carried out by Félix and Ancheyta (2019a, b) to find the conditions
at which the external and internal diffusions are disregarded and the kinetic control predominates.
The external diffusion experiments, where different stirred rates versus the residue conversion are
plotted, can be noticed in Figure 1.15. The higher the stirring rate, the higher the residue
0.18 0.15
360°C 360°C
In (1/(1–XAs))
In (1/(1–XR))
0.12 0.10
0.06 0.05
0.00 0.00
0.9 0.9
380°C 380°C
In (1/(1–XAs))
In (1/(1–XR))
0.6 0.6
0.3 0.3
0.0 0.0
1.2 1.8
390°C 390°C
In (1/(1–XR))
In (1/(1–XAs))
0.8 1.2
0.4 0.6
0.0 0.0
1.8 400°C 2.7 400°C
In (1/(1–XAs))
In (1/(1–XR))
1.2 1.8
0.6 0.9
0.0 0.0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Time (hours) Time (hours)
Figure 1.14 First-order reaction corroboration for R and As conversion during the hydrocracking of heavy oil
using molybdenite.
conversion. This trend is observed up to 800 rpm, beyond which the conversion remains constant
and unaffected by diffusion. Figure 1.15 also shows the internal diffusion experiments to find the
catalyst particle size at which the intrinsic kinetic is performed. The residue conversion against the
inverse of the particle size was plotted, and it can be noticed that the conversion is increased as the
particle size is reduced (the inverse of size is increased). Nevertheless, when the inverse of particle
size achieves the value 0.01571/μm (63.5 μm), the residue conversion is constant. Hence, with 800
rpm of stirring rate and 63.5 μm of molybdenite particle size, the intrinsic kinetics can be obtained.
Inverse of particle size (1/μm)

0.00 0.01 0.02 0.03 0.04 0.05
50
48
Residue conversion (%)
46
44
42
40
38
36
400 600 800 1000
Stirring rate (rpm)
Figure 1.15 Experiments for internal and external diffusion tests for the hydrocracking of heavy oil using
molybdenite.
1.4.1.2 Reaction Rate Coefficients

SARA-Based Model The reaction rate coefficients for the kinetic models based on SARA fractions
reported by Félix and Ancheyta (2019a, b) during the hydrocracking heavy oils using mineral cat-
alyst are summarized in Table 1.2. The kinetic parameter values of the four-lump kinetic model
(Figure 1.9b, Model 1) suggest that specific reactions do not occur at particular temperatures.
The reaction pathways from asphaltenes to resins (k1) and resins to saturates (k5) are not carried
out at 360 C. Furthermore, the agglomeration of aromatics to produce resins (k8) is only carried out
at 400 C because the thermal cracking is enough to break small molecules from heavy fractions
(asphaltenes and resins) but unable to produce heavy fractions with lower molecular weight, such
as resins. In addition, the thermal cracking produces free radicals from resins that form asphaltenes
by agglomeration. However, aromatics do not form resins since thermal cracking requires higher
temperature. Some of these kinetic parameters are not affected by temperature changes (k2, k3, k7)
under the mild conditions used. Accordingly, changes are noticeable only at high temperature
(400 C).
The reaction rate coefficients for the five-lump kinetic model considering the SARA fractions and
coke (Figure 1.10a, Model 2) are observed in this table. The conversion of resins to saturates cannot
be performed, as the constant k5, obtained by optimizing the reaction rate coefficients, is zero at all
studied temperatures. Moreover, the reactions that involve the coefficients k1 and k8 are not carried
out at 360 C as in the previous model, as well as the same reaction pathways (k2, k3, k7) are not
influenced by the temperature increase.
The kinetic parameters obtained for five-lump SARA-based Model 3 (considering all SARA and
gas fractions, Figure 1.10b) are observed in Table 1.2. The reaction from aromatics to gas (k9) is not
performed in this model. The reaction rate coefficients k1 (asphaltenes to resins), k6 (resins to satu-
rates), and k10 (saturates to gases) are not carried out at 360 C, while k4 (asphaltenes to gases) and
k12 (aromatics to resins) are performed at higher temperatures (380 and 400 C, respectively). As
stated earlier, breaking these molecules at low temperatures is difficult, and thermal cracking
requires temperatures exceeding 400 C to occur. Additionally, most of the reaction rate coefficients
are not sensitive to the temperature change until high temperatures are reached.
The kinetic parameters obtained for the 6-lump kinetic model based on SARA fractions
(Figure 1.10c, Model 4) are shown in the same table. In this model as in Model 3, the reaction from
Table 1.2 Kinetic parameters reported for kinetic models based on SARA fractions during the heavy oil hydrocracking using molybdenite catalyst.
Temperature AAE
( C) As Re As Ar As Sa As G As C Re Ar Re Sa Re G Re As Ar Sa Ar G Ar Re Sa G SSE (%)
Model 1 (4 lumps, 8 constants)

360 0 0.09499 0.02088 0.01666 0 0.07782 0.01936 0 4.87 × 10−4 2.07
380 0.11916 0.09499 0.02088 0.01666 0.00321 0.07782 0.05004 0 2.07 × 10−3 4.09
390 0.35913 0.09499 0.02088 0.08568 0.00815 0.07782 0.05004 0 2.03 × 10−3 6.20
400 0.35913 0.23637 0.05205 1.58695 0.00815 0.07782 0.05004 1.17453 4.03 × 10−3 9.28
EA (kJ/mol) 202.64 63.34 63.47 367.34 170.97 — 87.15 —
A0 (h−1) 2.29 × 015 1.33 × 1004 3.01 × 1003 1.49 × 1028 1.76 × 1011 — 3.49 × 1005 —
360 0 0.09305 0.02058 0.00248 0.01737 0.07760 0.01932 0 4.86 × 10−4 3.18
380 0.10049 0.09305 0.02058 0.03636 0.01737 0.08391 0.04959 0.00668 2.50 × 10−3 25.70
390 0.26486 0.09305 0.02058 0.12208 0.01737 0.08391 0.04959 0.00668 3.74 × 10−3 18.38
400 0.26497 0.13053 0.03286 0.25659 2.74437 0.08391 0.04959 2.03092 4.39 × 10−3 11.75
EA (kJ/mol) 178.10 23.52 32.50 419.72 396.63 7.18 86.52 1039.55
A0 (h−1) 2.06 × 1013 7.59 x1000 9.01 × 1000 1.17 × 1032 3.64 × 1030 3.07 × 10−01 3.08 × 1005 3.52 × 1080
360 0 0.09562 0.01892 0 0.01365 0 0.00752 0.07514 0.01967 0 0 4.79 × 10−4 2.71
380 0.12056 0.09562 0.01892 0 0.01365 0.00703 0.00752 0.07514 0.05295 0 0.01484 2.03 × 10−3 5.50
390 0.33659 0.09562 0.01892 0.02374 0.07866 0.00942 0.00752 0.07514 0.05295 0 0.01484 1.99 × 10−3 6.36
400 0.33659 0.21383 0.07895 0.02374 1.43887 0.00942 0.00752 0.07514 0.05295 1.06501 0.03190 3.40 × 10−3 9.67
EA (kJ/mol) 188.59 55.91 99.27 — 377.86 53.66 — — 90.87 — 139.18
A0 (h−1) 1.72 ×1014 3.35 × 1003 2.21 × 1006 — 8.90 × 1028 1.44 × 1002 — — 7.24 × 1005 — 1.76x1009
360 0 0.09381 0.01858 0.00248 0.01428 0.00753 0.07499 0.01965 0 0 4.77 × 10−4 3.46
380 0.09647 0.09381 0.01858 0.03586 0.01428 0.00753 0.07772 0.05547 0.00361 0.01489 2.41 × 10−3 23.26
390 0.25858 0.09381 0.01858 0.11950 0.06722 0.01393 0.07772 0.05547 0.00361 0.01991 3.49 × 10−3 17.33
400 0.25858 0.11424 0.05680 0.24924 1.58861 0.01393 0.07772 0.05547 1.17341 0.03459 3.77 × 10−3 11.53
EA (kJ/mol) 181.11 13.69 77.62 417.01 375.34 61.83 3.28 95.26 1051.84 153.85
A0 (h−1) 3.45 × 1013 1.22 × 1000 3.78 × 1004 7.00 × 1031 5.65 × 1028 8.68 × 1.41 × 10−01 1.68 × 1006 1.81 × 1081 2.86 ×1010
1002
aromatics to gas is not carried out at the temperatures studied, furthermore, k4 and k6 correspond-
ing to the reactions from asphaltenes to gas and resins to saturates, respectively, are not accom-
plished. This is due to the fact that, under these conditions, the reactions occur primarily in
series rather than in parallel. Furthermore, parameters k1 (asphaltenes to resins), k10 (saturates
to gases), and k13 (aromatics to resins) exhibit a value of zero at 360 C. Some reactions are not
affected by temperature as the reaction rate coefficients (k2, k3, and k11) remain invariant despite
the increase in temperature.
The sensitivity analysis performed to the obtained reaction rate coefficients for the SARA-based
kinetic models at 400 C is depicted in Figure 1.16. Almost all the obtained reaction rate coefficients
achieved the minimum value of the objective function (SSE) at 0% of perturbation. Thus, these param-
eter values are optimal and have been accurately calculated. For Model 1, some of the obtained reac-
tion rate coefficients (k2, k5, k6, and k7) at 400 C give a lower value of the objective function when the
perturbation is negative (lower than 0%). For Model 2, the acquired kinetic parameters k2, k3, k6, and
k7 at 390 and 400 C have a lower value of the SSE in a perturbation lower than the original value,
whereas for Model 3, some of the obtained reaction rate coefficients (k2, k3, k4, k6, k7, k10 and k11) have
a lower value of the objective function when the perturbation is negative at high temperatures (390
and 400 C). Concerning Model 4, the optimized kinetic parameters for the reactions from asphal-
tenes to aromatics, aromatics to saturates, resins to gases, aromatics to gases, and resins to asphaltenes
at all temperatures except for 360 C exhibit a lower objective function value with a perturbation of
less than 0%. This suggests that the parameter value ought to be lower, but it cannot be the case, since
to satisfy the Arrhenius equation, the reaction rate coefficient has to rise with temperature.
Distillation Curves-Based Models The reaction rate coefficients for combined kinetic model obtained
by Fukuyama and Terai (2007) are depicted in Table 1.3. All the reaction pathways are carried out at
the conditions studied. Nevertheless, the reaction from Ar to Re for both catalysts, and the produc-
tion of VGO from Ar and As for the pyrite-activated carbon catalyst are not sensitive to the tem-
perature because the condensation reactions of Ar to produce heavier fraction, such as Re and VGO,
require higher temperature to be performed, meanwhile for As being a high polar and polycon-
densed fraction is easier to agglomerate to produce C than crack into VGO. The highest reaction
rate coefficient was k4 (Sa to LO) for pyrite and k5 (Ar to LO) for pyrite-activated carbon catalysts,
due to the similarity of these light fractions, such as Sa with N, and Ar with K, which enhance
the reaction. Therefore, the lowest values are for k2 (Ar to Re) and k1 (Ar to Sa) for the pyrite
and pyrite-activated carbon catalysts, respectively.
The estimated kinetic parameters for the distillation curves-based kinetic model reported by Félix
and Ancheyta (2019a, b) are presented in Table 1.4. Some reaction pathways (VR to D, VR to N, and
D to N) do not occur at 360 C because their kinetic parameters (kL2, kL3 and kL6) were zero. This
trend is also shown at 380 C for the reaction from D to N (kL6) owing to the mild reaction condi-
tions used in the experiments, which cannot achieve severe hydrocracking to produce lighter dis-
tillates. Among the three reaction pathways of VR hydrocracking (kL1, kL2, and kL3), kL1 (VR to
VGO) gave the lowest value, while kL2 (VR to D) exhibited higher values, indicating that VR hydro-
cracking is more selective toward light distillates. kL4, kL5, and kL6 had even lower values indicating
that over cracking of VGO and D proceeded at low rates.
As expected, all the obtained reaction rate coefficients are in accordance with Arrhenius equa-
tion, contrary to some reaction rate coefficients reported in the literature (Coronel-García et al.
2021; Galarraga et al. 2012; Hassanzadeh and Abedi 2010; Kim et al. 2017; Loria et al. 2011; Orozco
Castillo 2016) that displayed higher values at lower temperatures. The sensitivity analysis on the
obtained reaction rate coefficients for distillation curves-based kinetic model is shown in
Figure 1.17. With the original value of reaction rate coefficients (0% of perturbation), the global
4.5E-03 0.015 5.3E-03 1.4E-02
Model 1 Model 2
5.1E-03 1.2E-02
4.3E-03 0.011
4.9E-03 1.0E-02
SSE
SSE
4.7E-03 8.0E-03
4.1E-03 0.007
4.5E-03 6.0E-03
3.9E-03 0.003 4.3E-03 4.0E-03
3.8E-03 0.017 4.20E-03 0.013

Model 3 Model 4
3.7E-03 0.014 4.10E-03 0.011
3.6E-03 0.011 4.00E-03 0.009

SSE
SSE
3.5E-03 0.008 3.90E-03 0.007
3.4E-03 0.005 3.80E-03 0.005
3.3E-03 0.002 3.70E-03 0.003

–20 –15 –10 –5 0 5 10 15 20 –20 –15 –10 –5 0 5 10 15 20
% perturbation % perturbation
Figure 1.16 Sensitivity analysis of the obtained reaction rate coefficients (kj) for the SARA-based kinetic models at 400 C. k1 (○), k2 (□), k3 (△), k4 (⋄), k5 (×), k6 (+), k7
(▽), k8 (∞), k9 (γ), k10 (ε), k11 (ω), k12 (η), and k13 (Π).
Table 1.3 Kinetic parameters obtained for the kinetic model based on combined fractions (SARA and
distillation curves) in the hydrocracking of heavy oil using mineral catalysts.
Temperature ( C)
Parameter 415 425 435 EA (kJ/mol) A0 (h−1)
Pyrite
k1 4.76 × 10−3 1.66 × 10−2 3.03 × 10−2 253.00 5.17 × 1018
k2 1.35 × 10−5 1.41 × 10−5 1.46 × 10−5 10.00 4.94 × 10−3
−6 −4 −3
k3 3.54 × 10 6.34 × 10 7.62 × 10 1050.00 1.29 × 1076
k4 1.82 × 10−2 4.65 × 10−2 7.27 × 10−2 189.00 2.65 × 1014
−4 −3 −2
k5 7.96 × 10 5.67 × 10 1.45 × 10 397.00 7.45 × 1028
k6 2.44 × 10−5 3.45 × 10−3 3.71 × 10−3 1002.00 2.14 × 1073
−3 −2 −2
k7 5.42 × 10 1.66 × 10 2.83 × 10 226.00 4.90 × 1016
k8 4.84 × 10−6 1.30 × 10−4 6.32 × 10−4 667.00 1.28 × 1047
−5 −3 −2
k9 3.56 × 10 4.23 × 10 4.18 × 10 967.00 6.17 × 1070
k10 7.09 × 10−4 7.11 × 10−3 2.15 × 10−2 467.00 1.19 × 1034
Pyrite-activated carbon
k1 8.84 × 10−6 1.18 × 10−5 1.25 × 10−5 47.00 2.18 × 100
−5 −5 −5
k2 1.49 × 10 1.54 × 10 1.57 × 10 7.00 3.02 × 10−3
k3 1.22 × 10−5 1.37 × 10−5 1.45 × 10−5 24.00 4.74 × 10−2
−5 −3 −2
k4 1.41 × 10 1.27 × 10 1.09 × 10 910.00 1.14 × 1066
k5 4.68 × 10−3 1.95 × 10−2 3.87 × 10−2 289.00 2.65 × 1021
−5 −5 −5
k6 1.40 × 10 1.49 × 10 1.54 × 10 13.00 8.07 × 10−3
k7 4.65 × 10−3 1.36 × 10−2 2.28 × 10−2 217.00 9.31 × 1015
−5 −5 −5
k8 1.24 × 10 1.29 × 10 1.31 × 10 7.00 2.66 × 10−3
k9 1.57 × 10−5 9.12 × 10−4 6.39 × 10−3 822.00 2.61 × 1059
−3 −3 −3
k10 1.29 × 10 4.28 × 10 7.61 × 10 243.00 2.38 × 1017
Table 1.4 Kinetic parameters obtained for the kinetic model based on distillation curves during the heavy oil
hydrocracking using molybdenite catalyst.
Temperature ( C)
Parameter 360 380 390 400 EA (kJ/mol) A0 (h−1)
kL1 0.03203 0.03443 0.03615 0.07771 64.69 6.01 × 103

kL2 0 0.06499 0.09512 0.12896 125.33 6.92 × 108
kL3 0 0.04349 0.06515 0.11827 182.66 1.70 × 1013
kL4 0.00219 0.00230 0.00543 0.06503 263.40 6.42 × 1018
kL5 0.00339 0.00355 0.01119 0.01175 123.08 4.08 × 107
kL6 0 0 0.01088 0.02711 338.81 5.29 × 1024
SSE 3.96E−4 4.97E−4 1.61E−3 7.04E−4
AAE (%) 2.25 2.43 3.24 2.84
(a) (b)
4.05E-04 1.1E-03 1.9E-03 5.08E-04
1.4E-03 5.03E-04
SSE
SSE
4.00E-04 7.0E-04
9.0E-04 4.98E-04
3.95E-04 3.0E-04 4.0E-04 4.93E-04
(c) (d)
3.5E-03 1.66E-03 2.6E-03 7.4E-04
SSE
SSE
2.5E-03 1.63E-03 1.6E-03 7.2E-04
1.5E-03 1.60E-03 6.0E-04 7.0E-04

–20 –15 –10 –5 0 5 10 15 20 –20 –15 –10 –5 0 5 10 15 20
% perturbation % perturbation
Figure 1.17 Sensitivity analysis of the reaction rate coefficients (kj) for the kinetic model based on distillation curves at 360 (a), 380 (b), 390 (c), and 400 C (d). k1 (□),
k2 (△), k3 (○), k4 (⋄), k5 (+), and k6 (×).
minimum of the objective function (SSE) is achieved confirming that they are the optimal values.
The same results were observed at different temperatures demonstrating accurate estimation of
optimal values even at other temperatures.
1.4.1.3 Activation Energies

SARA-Based Models As can be seen in Table 1.2, certain activation energies were not calculated
because several reactions did not proceed as Re to Sa (Models 2 and 4.), As to G, and Ar to
G (Models 3 and 4) at any temperature, and Ar to Re (Models 1 and 3) is only carried out at
400 C. Additionally, other reactions were unchanged with variations of the temperature, such
as Re to As (all models), and Re to G (Model 3), whereas the range of the activation energies values
is lower for Models 1 (63.34–367.34 kJ/mol) and 3 (53.66–377.86 kJ/mol) than Models 2
(7.18–1039.55 kJ/mol) and 4 (3.28–1051.84 kJ/mol). This is owed to the reaction from Ar to Re is
carried out only at 400 C for the former models, while for the latter ones the reaction is performed
at 380–400 C with low-rate values at low temperature (380 and 390 C) and considerably higher at
400 C, resulting in a high activation energy. Based on the activation energies in all models, the
cracking of the heavy fractions (As and Re) into lighter fractions (Ar and Sa) through parallel reac-
tions are more feasible under the studied conditions, while higher values of activation energies were
calculated for in-series reactions (As to Re to Ar to Sa).
Distillation Curves-Based Models The activation energies for the kinetic model based on distillation
curves using pyrite and pyrite-activated carbon (pyrite-AC) are noticed in Table 1.3. Some of the
activation energies for both catalysts showed low values because of the reactions are insensitive
to temperature, such as Ar to Re. The values of the activation energies are in the range of
10–1050 kJ/mol (pyrite) and 7–910 kJ/mol (pyrite-AC). All the activation energies present lower
values using pyrite–AC catalyst, indicating that this catalyst enhances all the reactions. The greater
activation energies observed when using pyrite as a catalyst pertain to condensation reactions (Re to
As, Ar to VGO, and As to C), while employing pyrite–AC as catalyst the largest values corresponded
to the cracking of Sa and condensation of As. The lower activation energies using pyrite corre-
sponded to the cracking of light fractions (Ar and Sa), meanwhile utilizing pyrite–AC, lower values
were observed for condensation reactions (Ar to Re, Ar to VGO, and Re to As) because these reac-
tions were not influenced by temperature. Thus, the addition of activated carbon to pyrite restricts
the secondary cracking of light fractions, such as Sa.
The activation energies (64.69–338.81 kJ/mol) reported by Félix and Ancheyta (2019a) for the
distillation curves-based kinetic model are depicted in Table 1.4. The lower values of activation
energies pertained to the cracking of VR, and the activation energy increased as the molecular
weight of products decreased (VR to VGO < VR to D < VR to N). This behavior suggests that
the conversion of heavy fractions was performed with ease, while the conversion of D demands
high activation energy. Similar results were obtained by other authors (da Silva De Andrade
2014; Galarraga et al. 2012; Loria et al. 2011) using dispersed catalysts since the lighter the products,
the higher the activation energy for the cracking of heavy fractions.
1.4.2 Accuracy of the Kinetic Models

1.4.2.1 SARA-Based Models
The calculated and experimental mass fractions varying the reaction times at 390 and 400 C for the
kinetic models based on SARA fractions are observed in Figure 1.18. In all models, the mass fraction
of As decreases while Sa and G (only considered in Models 3 and 4) fractions increases as reaction
390 °C 400 °C
0.36 0.46 0.36 0.46
Model 1 Model 1
Mass fraction
Mass fraction
0.24 0.41 0.24 0.40
0.12 0.36 0.12 0.36
0.00 0.31 0.00 0.31

0.32 0.46 0.24 0.46
Model 2
Model 2
0.24 0.43 0.18 0.36

Mass fraction
Mass fraction
0.16 0.40 0.12 0.32
0.08 0.37 0.06 0.28
0.00 0.34 0.00 0.24

0.45 0.27 0.42
Model 3 Model 3
Mass fraction
Mass fraction
0.30 0.18 0.37
0.15 0.09 0.32
0.00 0.00 0.27

0.24 0.41 0.32 0.40
Model 4 Model 4
0.18 0.36 0.24 0.36

Mass fraction
Mass fraction
0.12 0.31 0.16 0.32
0.06 0.26 0.08 0.28
0.00 0.21 0.00 0.24

0 1 2 3 4 5 6 0 1 2 3 4 5 6
Reaction time (h) Reaction time (h)
Figure 1.18 Calculated (lines) and experimental (symbols) mass fractions for the kinetic models based
on SARA fractions at 390 and 400 C. As (○), Re (□), Ar (+), Sa (△), G (×), and C (⋄).
time and temperature raise. This is due to the thermal cracking that produces lighter fractions (Sa
and G), and these products do not undergo further cracking. In addition, as temperature and reac-
tion time increase, C (only considered in Models 2 and 4) fraction increases by the formation of free
radicals in thermal cracking. All the experimental data are accurately predicted in the four SARA-
based models, with the C fraction being the most difficult to forecast.
The residual plots of the kinetic models based on SARA fractions are shown in Figure 1.19. The
residual plot confirms the correct estimation of kinetic parameters, as they neither overestimate nor
underestimate the experimental data. This is evidenced by the number of positive and negative resi-
duals in the statistical analysis (Table 1.5). The residual values showed that the prediction is better
0.04 0.5
Model 1
Model 1
Calculated mass fraction

0.4
0.02
0.3
Residual
0.00
0.2
–0.02
0.1
–0.04 0.0
0.04 0.5
Model 2
Model 1

0.4
0.02
0.3
Residual
0.00
0.2
–0.02
0.1
–0.04 0.0
0.03 0.5
Model 3
0.02 Model 3
0.4
0.01
0.3
Residual
0.00
0.2
–0.01
0.1
–0.02
–0.03 0.0
0.04 0.5
Model 4 Model 4
0.4
0.02
0.3
Residual
0.00
0.2
–0.02
0.1
–0.04 0.0
0 3 6 9 12 15 18 0.0 0.1 0.2 0.3 0.4 0.5
Experiments Experimental mass fraction
Figure 1.19 Residual and parity plots for the kinetic models based on SARA fractions. As (○), Re (□), Ar (+),
Sa (△), G (×), and C (⋄).
at low temperatures, especially for coke fraction because the values increase as temperature rises
due to the constraint that the reaction rate coefficients must increase with temperature restraining
their values. The lowest residuals corresponded to the gas fraction since its mass fraction is smaller,
thus the residuals were closer to zero. In Figure 1.19, the parity plots of the kinetic models based on
Table 1.5 Statistical analysis for the kinetic models during the heavy oil hydrocracking using molybdenite
catalyst.
Model Residuals (+) Residuals (−) Residual range Intercept Slope AIC BIC
SARA-based
1 31 33 0.0575 0.0023 0.9909 −23.94 −27.07
2 40 40 0.0613 0.0031 0.9887 −20.84 −24.35
3 42 38 0.0537 0.0009 0.9960 −16.23 −20.92
4 49 47 0.0525 0.0028 0.9945 −13.19 −18.27
Distillation curves-based
— 26 38 0.0428 0.0021 1.0084 — —
SARA fractions is also shown. The experimental and calculated data follow a straight line present-
ing good fit. Furthermore, in the statistical analysis (Table 1.5), the slope and the intercept to “y”
axis are close to 1 and 0, respectively, confirming the good prediction.
The AIC and BIC values for the kinetic models based on SARA fractions are shown in Table 1.5.
Model 1 is deemed the most appropriate model due to its lower AIC and BIC values. Nevertheless,
all models presented similar values for both criteria and accurately forecasted the experimental
data, regardless of the number of parameters present.
1.4.2.2 Distillation Curves-Based Models

The experimental and calculated yields at different reaction times and temperatures using a kinetic
model combining distillation curves and SARA fractions is noticed in Figure 1.20. The yields of VR,
Re, and Ar are diminished as the reaction time and temperature are increased, indicating that the
main reactants to produce light and middle fractions in VR are Re and Ar using both catalysts,
whereas the yields of LO and C rise as the reaction time and temperature increase, due to the crack-
ing of Ar and Re, as well as the in-series condensation reactions of the latter to produce As and then
C. The yield of VGO increases with the reaction time at 415 C; however, at 425 C, it increased and
then decreased, because the production of this fraction is mainly from the cracking of Re, while at
higher temperature (425 C) the rate of secondary cracking is higher and results in a slight
decrement.
The yield of As does not show any noticeable change except a slight decrement at 425 C owed to
the higher rate of free radicals production to generate C, which agrees with the increment of this
latter fraction. It can be noticed that the addition of activated carbon to pyrite restricts the conden-
sation reactions, enhancing the generation of lighter fractions, such as LO. A reasonably good fit of
the kinetic model to the experimental data is observed. The sensitivity or statistical analyses are not
provided for this model. Hence, some of the experimental fractions are underestimated or overes-
timated, due to the lack of optimization procedure.
The calculated and experimental mass fractions at different reaction times and temperatures of
390 and 400 C for the kinetic model based on distillation curves are observed in Figure 1.21. The
mass fraction of VR is decreased as the reaction time and temperature are raised, while the opposite
behavior is observed for the mass fractions of D, N, G, and C. This is because the cracking and
45 100 60 100
Pyrite-AC(415 °C) 90 Pyrite-AC(425 °C) 90
40
50
35 80 80
70 70
30 40
VR yield (wt.%)
VR yield (wt.%)
Yield (wt.%)
Yield (wt.%)
60 60
25
50 30 50
20
40 40
15 20
30 30
10 20 20
10
5 10 10
0 0 0 0
45 100 70 100
Pyrite (415 °C) 90 Pyrite (425 °C) 90
40
60
35 80 80
70 50 70
30
VR yield (wt.%)
VR yield (wt.%)
Yield (wt.%)
Yield (wt.%)
60 60
25 40
50 50
20 30
40 40
15
30 20 30
10 20 20
10
5 10 10
0 0 0 0
0 50 100 150 200 250 0 30 60 90 120 150 180
Residence time (min) Residence time (min)
Figure 1.20 Calculated (lines) and experimental (symbols) yields for the combined kinetic model based on SARA fractions and distillation curves. VR ( ), VGO (▪), •
LO (▲), C (⧫), As (○), Re (□), Ar (△), and Sa (⋄).
0.45 0.08
390 °C 390 °C
0.35 0.06
Mass fraction
Mass fraction
0.25 0.04
0.15 0.02
0.05 0.00
0.45 0.12
400 °C
400 °C
0.35 0.09
Mass fraction
Mass fraction
0.25 0.06
0.15 0.03
0.05 0.00
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Reaction time (h) Time (hours)
Figure 1.21 Calculated (lines) and experimental (symbols) mass fractions for the kinetic model based on distillation curves at 390 and 400 C. VR (○), VGO (□), D (+),
N (△), G (×), and C (⋄).
0.03
0.02
0.01
Residual
0.00
–0.01
–0.02
–0.03
0 3 6 9 12 15 18
Experiments
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5
Experimental mass fraction
Figure 1.22 Residual and parity plots for the kinetic model based on distillation curves. VR (○), VGO (□), D (+),
and N (△).
condensation reactions of VR generates lighter fractions, such as D, N, and G, as well as highly

polycondensed molecules, such as C. Concerning the VGO mass fraction, at 390 C, it slightly
increases with the reaction time because of its small production from VR and secondary cracking
into lighter fractions is presented. Nevertheless, at 400 C the secondary cracking rate is higher,
causing a reduction in the mass fraction of VGO as reaction time increases. Almost all mass frac-
tions are well predicted, being the C fraction the most difficult to forecast since the free radical pro-
duction that leads to C is faster than the hydrogenation reactions at higher temperatures, hindering
to predict its behavior at high temperatures.
Figure 1.22 shows the residual and parity plots for the kinetic model based on distillation curves.
The small residual values confirm that the kinetic parameters were assessed appropriately. Addition-
ally, the parity plot demonstrates that the experimental and calculated mass fractions exhibit similarity,
as evidenced by a 45 -straight line in the data, thereby, validating the fitting of the kinetic model.
1.4.3 Reactions in Parallel and in Series

The results obtained from the kinetic models based on SARA fractions indicate that some in-series
(As to Re to Ar) and in-parallel reactions (As to Ar) are benefited using molybdenite as catalyst.
Regarding the production of G, the reactions of As to G and Ar to G are hard to be performed at the
conditions used, where the formation of G fraction is mainly carried out from Sa fraction. Further-
more, the condensation reaction of middle (Ar to Re) and heavy (Re to As and As to C) fractions are
intricate to be done as higher temperatures are required.
In the combined kinetic model, the kinetic parameters suggest that the in-series reactions from
Ar to Sa to LO and the parallel reaction from Re to VGO are easily carried out using pyrite as cat-
alyst. Meanwhile, the addition of activated carbon to pyrite induces in-series reaction from Re to
VGO to LO. Furthermore, both catalysts (in higher degree for pyrite–AC) hinder the condensation
in-series reactions from Ar to Re to As to C. The results of distillation curves-based kinetic model
infer that the cracking of heavy fraction (VR and VGO) by parallel reactions (VR to D, VR to N, and
VGO to N) are favored. Additionally, the cracking of D is hindered at these conditions.
The experimental and calculated As and VR conversions of kinetic models based on distillation
curves and SARA fractions for hydrocracking of heavy crude oil using molybdenite as catalyst is
observed in Figure 1.23. The similarity between the experimental As and VR conversions suggests
that a majority of the As fraction is present in VR. Furthermore, the similarity between the calcu-
lated and experimental As and VR conversions confirms the reported kinetic models have a good fit.
1.4.4 Thermodynamic Model

Figure 1.24 illustrates the impact of temperature on the molar mass of SARA fractions in the heavy
oil hydrocracking. It is evident that the molar mass of all fractions (As, Re, Ar, and Sa) decreases
with an increase in temperature. This trend can be attributed to the hydrogenation of aromatic rings
and dealkylation reactions of side aliphatic chains in the polyaromatic fractions. Linear correlations
are observed among all molar masses, and the equations showed below are obtained accordingly:
M As = − 16 517T + 8816 521 1 118
M Re = − 8 326T + 4446 031 1 119
M Ar = − 11 304T + 5282 505 1 120
M Sa = − 10 686T + 4768 261 1 121
Figure 1.23 Experimental (full 100

symbols) and calculated (empty
symbols) residue and asphaltenes
conversions at 360 (circle), 380
Asphaltenes conversion (%)
80
(square), 390 (triangle), and 400 C
(diamond).
60
40
20
0
0 20 40 60 80 100
3000
2500
Molar mass (g/mol)
2000
1500
1000
500
350 360 370 380 390 400 410
Temperature (°C)
Figure 1.24 Effect of temperature on the molar mass of As ( ), Re ( ), Ar ( ), and Sa ( ).
The sediment content resulting from the experiments at varying reaction times and temperatures
is plotted in Figure 1.25a. As depicted, increasing both temperature and reaction time results in
elevated sediment content. When conducting reactions at low temperature (360 C), sediment for-
mation is insignificant for all reaction durations, whereas at high temperature, sediment formation
occurs even during brief reaction intervals. Based on the experimental findings, it was observed that
the Sa fraction increases with an increase in reaction time and temperature. As the solubility of As
in crude oil decreases, the agglomeration and flocculation of this fraction intensify and form
sediments.
The kinetic model based on SARA fractions (6-lump) was employed as well as the regular solu-
tion model to calculate the molar volumes and solubility parameters of SARA fractions, and then
estimate the content of sediments during the hydrocracking of heavy oil using molybdenite as cat-
alyst. Molar volumes and solubility parameters for SARA fractions at various temperatures are dis-
played in Table 1.6. The molar volume increases with increasing molecular weight of the fractions.
Furthermore, the molar volume of asphaltene, resin, and aromatic fractions decreases, while the
saturate fraction increases as the reaction temperature rises. At low temperatures, the solubility
parameter of the four fractions is similar and diminishes as the temperature increases. This behav-
ior was more noticeable in the Sa fraction as it is responsible for altering the solubility of asphal-
tenes in crude oil.
The experimental and calculated sediment contents using the regular solution model for dif-
ferent residue conversions and temperatures are shown in Figure 1.25b. The fit of the proposed
thermodynamic model is moderate, since the molar masses of SARA fractions are calculated at
different temperatures using the obtained correlations. Nevertheless, these molar masses also
vary as the reaction time increases, but in this case, they are assumed to be constant because there
are no such correlations reported. The formation of sediments is minimal at temperatures below
380 C, corresponding to a residue conversion lower than 40–50%. However, when residue con-
version is higher than this value, the sediment formation increases substantially. This behavior is
also reported in literature due to insolubility of heavy fractions as aliphatic chains are converted
(Marroquín 2007). Similar results are attained for the experimental and calculated sediments con-
tent obtained for different asphaltene conversions and temperatures (Figure 1.25c), since the for-
mation of sediment increases rapidly when asphaltene conversion is near 50%. The higher the
temperature, the higher the conversion of asphaltene and residue. Furthermore, the formation
of sediments is observed. The calculated and experimental sediment contents using the thermo-
dynamic model for different Sa/As ratios and temperatures are depicted in Figure 1.25d. It is
(a)
1.2
Sediments content (wt. %)
1.0
0.8
0.6
0.4
0.2
0.0
2 3 4 5
Time of reaction (h)
(b)
1.6
1.2
0.8
0.4
0.0
0 20 40 60 80 100
(c)
1.6
1.2
0.8
0.4
0.0
0 20 40 60 80 100
Asphaltenes conversion (%)
(d)
1.6
1.2
0.8
0.4
0.0
0 2 4 6 8 10
Sa/As ratio
Figure 1.25 Experimental (symbols) and calculated (lines) sediment content for different reaction times
(a), residue conversion (b), asphaltenes conversion (c), and Sa/As ratio (d) at 360 C ( ), 380 C ( ), 390 C ( ),
and 400 C ( ).
Table 1.6 Correction factor for acentric factor, molar volume, and solubility parameter obtained for the
different fractions during the heavy oil hydrocracking using molybdenite catalyst.
Correction Molar volume Solubility Temperature

Fraction factor (cf) (cm3/mol) parameter (MPa0.5) ( C)
Saturates 0.9024 447.98 19.36 360

Aromatics 0.8098 1131.76 18.78
Resins 0.7910 1368.47 18.54
Asphaltenes 0.7940 2188.98 19.58
Saturates 0.9024 833.32 12.30 380
Aromatics 0.8098 984.85 18.45
Resins 0.7910 1210.03 18.28
Asphaltenes 0.7940 2014.36 19.22
Saturates 0.9024 826.31 11.40 390
Aromatics 0.8098 898.99 18.25
Resins 0.7910 1148.58 18.16
Asphaltenes 0.7940 1925.81 19.05
Saturates 0.9024 787.71 10.87 400
Aromatics 0.8098 762.82 17.92
Resins 0.7910 1129.45 18.07
Asphaltenes 0.7940 1866.94 18.93
noticed that the regular solution model predicts the behavior since the sediment content
increases as Sa/As ratio increases. This is because when the Sa fraction increases, the insolubility
of As also increases, leading to the agglomeration and flocculation of this fraction, resulting in
sediment formation.
1.4.5 General Comments

It was corroborated that the hydrocracking reactions at mild conditions using molybdenite indeed
follow a first-order reaction. Furthermore, the optimal particle size and stirring rate were deter-
mined to obtain the intrinsic kinetics for hydrocracking reactions using mineral catalysts.
According to the AIC and BIC criteria, Model 1 is the most accurate among all the kinetic models
based on SARA fractions. Nonetheless, the analysis solely focuses on the liquid phase, which
involves the smallest number of reaction pathways. Additionally, the AAE is marginally elevated.
Model 2 considers the liquid and solid phases, but it displays the highest AAE due to the difficulty to
predict C fraction. Model 3 exhibit similar AAE values to Model 1, even considering two phases
(liquid and gas) because of the good prediction of gas phase. However, it was not possible to cal-
culate some activation energies because most of the kinetic parameters do not change with tem-
perature. Finally, Model 4 considers the greatest quantity of pseudocomponents, reaction
pathways, reaction rate coefficients, and phases (liquid, solid, and gas). Although it presents high
AAE values, it enables the calculation of all activation energies. All these features turn into the most
complete model and explain better As hydrocracking under the reaction conditions studied. The
results for these SARA-based kinetic models indicate that some hydrocracking in-series (As to
Re to Ar) and in-parallel (As to Ar) reactions are promoted, differing from the hydrocracking of
lighter fractions (Ar and Sa). However, the formation of G fraction is caused by the breaking of Sa
molecules. The activation energies of condensation reactions (Ar to Re to As) are not calculated or
their values are significantly high due to their insensitivity to temperature. These reactions require
more energy to be carried out.
Concerning the distillation curves- and SARA fractions-based kinetic models, the results indicate
that some condensation reactions need higher temperatures to be performed. Moreover, the hydro-
cracking of middle fractions, such as Ar and Re, to produce LO are preferred at the conditions stud-
ied. Both catalysts promote the in-series reactions following different pathways: with pyrite as
catalyst, the reactions from Ar to Sa to LO are enhanced as well as from Re to VGO, while for
pyrite–AC catalyst, the reactions from Re to VGO to LO are promoted. The activation energies
for this kinetic model exhibit higher values using pyrite as catalyst than the pyrite–AC catalyst.
In addition, a synergic effect between pyrite and activated carbon hinders the condensation reac-
tions (Ar to Re to As to C), increasing the yield of LO. The results for the kinetic model based on
distillation curves imply that the hydrocracking of heavy fractions (VR and VGO) is promoted by
parallel reactions (VR to D, VR to N, and VGO to N), while the secondary cracking of D is difficult to
be performed at the studied conditions.
According to the statistical analysis, the kinetic parameters for the models based on SARA frac-
tions and distillation curves were accurately estimated. Furthermore, the sensitivity analysis for all
these kinetic parameters denotes that all these values are optimized for the experimental data. Nev-
ertheless, some of the reaction rate coefficients did not reach the minimum value of the objective
function (especially at high temperatures), since the established constraints (the reaction rate coef-
ficient must increase with temperature) limited the range of permissible values. These analyses are
not performed to the estimated parameters for the combined (distillation curves and SARA frac-
tions) kinetic model. Therefore, these kinetic parameters might not be the optimal set as certain
fractions are not accurately predicted.
The molar masses of SARA fractions display a linear relationship with the temperature since they
are reduced as the temperature increased. Hence, a correlation for each SARA fraction was
obtained. The solubility parameters for SARA fractions exhibit a slightly reduction with the incre-
ment of temperature, excluding those for Sa fractions since the drop is considerable. With these
data, a regular solution model and a kinetic model based on SARA fractions were applied to predict
the As flocculation to form sediments, which increases as the reaction time and temperature rise.
The experimental and calculated sediment amounts with the thermodynamic model were com-
pared with As and VR conversion, obtaining a remarkable increase in the sediment contents after
reaching a value of 50% and 40–50%, respectively. Additionally, it was confirmed that increasing the
amount of Sa in heavy oil rises the flocculation of As to produce sediments, because more sediments
are obtained when the Sa/As ratio is incremented. Likewise, it was proved that almost all As are
present in the VR fraction because similar As and VR conversions are acquired at the different con-
ditions studied.
Comparing the results obtained for the hydrocracking of heavy crude oil using mineral catalyst
with other processes (supported and dispersed catalysts), it is noticed that different levels of VR or
As conversion are achieved. Higher conversions are obtained employing commercial or synthesized
catalysts at high reaction temperatures (above 420 C) or at high hydrogen pressures than using
mineral catalysts. The conversion of VR or As using mineral catalysts reaches 80% at high temper-
ature (400 C) with formation of sediments. One advantage of utilizing this kind of catalysts is that
the mineral catalysts are cheaper compared with commercial catalyst used in other technologies. In
addition, if mild reaction conditions are used in these works, the technical and operational costs are
reduced.
1.5 Conclusion
From the results obtained from the developed kinetic models for the catalytic hydrocracking of
heavy oil using mineral catalysts, the following can be concluded:
– The reaction conditions to estimate the intrinsic kinetics using molybdenite as catalysts were
obtained
– The first-order of reaction for the As and VR hydrocracking was confirmed
– The 6-lump model proved to be the most accurate one (largest number of parameters and pseu-
docomponents) in comparison with other SARA-based kinetic models
– The reactions from Re to Sa, Ar to G, and As to G did not occur at the reaction conditions studied
using molybdenite as a catalyst
– The reactions in-series (As to Re to Ar) and in-parallel (As to Ar) of heavy fractions are promoted
using molybdenite as a catalyst
– For the combined-based (distillation curves and SARA fractions) kinetic model, the pyrite (Re to
VGO and Ar to Sa to LO) and the pyrite–AC (Re to VGO to LO) catalysts promoted in-series
reactions
– The addition of activated carbon to pyrite suppresses the condensation reactions (Ar to Re to As
to C) and enhances the yield of LO
– For the kinetic model based on distillation curves, the use of molybdenite enhances the hydro-
cracking of VR and VGO by parallel reactions (VR to D, VR to N, and VGO to N)
– Based on the statistical and sensitivity analyses, the kinetic parameters for the models based on
SARA fractions and distillation curves were accurately estimated and corresponded to optimal
values
– All the kinetic models fit the experimental data well, although predicting the C fraction proved to
be the most difficult to forecast
– Linear correlations were developed for each SARA fraction because their molar masses reduced
as the temperature increased
– The sediment content enlarged with the increase of the reaction time, temperature, and Sa/As
ratio during hydrocracking of heavy oils using molybdenite
– The sediment formation increases promptly when As and VR conversions reached 40–50% and
50%, respectively
– Most of As fraction in heavy oil is included on VR fraction
References
Akbarzadeh, K., Ayatollahi, S., Moshfeghian, M. et al. (2004). Estimation of SARA fraction properties
with the SRK EOS. Journal of Canadian Petroleum Technology 43: 31–39.
Al-Attas, T.A., Ali, S.A., Zahir, M.H. et al. (2019). Recent advances in heavy oil upgrading using dispersed
catalysts. Energy Fuels 33: 7917–7949.
Alcázar, L.A. and Ancheyta, J. (2007). Sensitivity analysis based methodology to estimate the best set of
parameters for heterogeneous kinetic models. Chemical Engineering Journal 128: 85–93.
Alonso, F., Ancheyta, J., Centeno, G. et al. (2019). Effect of reactor configuration on the hydrotreating of
atmospheric residue. Energy Fuels 33: 1649–1658.
Álvarez, P., Browning, B., Jansen, T. et al. (2019). Modeling of atmospheric and vacuum petroleum
residue hydroconversion in a slurry semi-batch reactor: study of hydrogen consumption. Fuel
Processing Technology 185: 68–78.
References 51
Ancheyta, J. (2011). Modelling and Simulation of Catalytic Reactors for Petroleum Refining, 1ste. Wiley.
Ancheyta, J. (2013). Modeling of Processes and Reactors for Upgrading of Heavy Petroleum, 1ste. CRC Press.
Ancheyta, J. and Sotelo-Boyás, R. (2000). Estimation of kinetic constants of a five-lump model for fluid
catalytic cracking process using simpler sub-models. Energy Fuels 14: 1226–1231.
Ancheyta, J. and Speight, J.G. (2007). Hydroprocessing of Heavy Oils and Residual, 1ste. CRC Press.
Ancheyta, J., Sánchez, S., and Rodríguez, M.A. (2005). Kinetic modeling of hydrocracking of heavy oil
fractions: a review. Catalysis Today 109: 76–92.
Angeles, M.J., Leyva, C., Ancheyta, J., and Ramírez, S. (2014). A review of experimental procedures for
heavy oil hydrocracking with dispersed catalyst. Catalysis Today 220–222: 274–294.
Asaee, S.D.S., Vafajoo, L., and Khorasheh, F. (2014). A new approach to estimate parameters of a lumped
kinetic model for hydroconversion of heavy residue. Fuel 134: 343–353.
Ashoori, S., Sharifi, M., Masoumi, M., and Mohammad Salehi, M. (2017). The relationship between
SARA fractions and crude oil stability. Egyptian Journal of Petroleum 26: 209–213.
Balasubramanian, P. and Pushpavanam, S. (2008). Model discrimination in hydrocracking of vacuum gas
oil using discrete lumped kinetics. Fuel 87: 1660–1672.
Baltanas, M.A. and Froment, G.F. (1985). Computer generation of reaction networks and calculation of
product distributions in the hydroisomerization and hydrocracking of paraffins on Pt-containing
bifunctional catalysts. Computers Chemical Engineering 9: 71–81.
Baltanas, M.A., Van Raemdonck, K.K., Froment, G.F., and Mohedas, S.R. (1989). Fundamental kinetic
modeling of hydroisomerization and hydrocracking on noble metal-loaded faujasites. 1. Rate
parameters for hydroisomerization. Industrial Engineering Chemistry Research 28: 899–910.
Bata, T., Schamel, S., Fustic, M., and Ibatulin, R. (2019). AAPG Energy Minerals Division Bitumen and Heavy
Oil Committee Annual Commodity Report. In American Association of Petroleum Geologists (AAPG).
Becker, P.J., Serrand, N., Celse, B. et al. (2016). Comparing hydrocracking models: continuous lumping vs
single events. Fuel 165: 306–315.
Bej, S.K. (2002). Performance evaluation of hydroprocessing catalysts. A review of experimental
techniques. Energy Fuels 16: 774–784.
Bellussi, G., Rispoli, G., Landoni, A. et al. (2013). Hydroconversion of heavy residues in slurry reactors:
developments and perspectives. Journal of Catalysis 308: 189–200.
Browning, B., Pitault, I., Couenne, F. et al. (2019). Distributed lump kinetic modeling for slurry phase
vacuum residue hydroconversion. Chemical Engineering Journal 377: 119811.
Cai, Z., Ma, Y., Zhang, J. et al. (2022). Tunable ionic liquids as oil-soluble precursors of dispersed catalysts
for suspended-bed hydrocracking of heavy residues. Fuel 313: 122664.
Calderon, C.J. and Ancheyta, J. (2016). Modeling of slurry-phase reactors for hydrocracking of heavy oils.
Energy Fuels 30: 2525–2543.
Castañeda, L.C., Muñoz Arroyo, J.A.D., and Ancheyta, J. (2014). Current situation of emerging
technologies for upgrading of heavy oils. Catalysis Today 220–222: 248–273.
Celse, B., Costa, V., Wahl, F., and Verstraete, J.J. (2015). Dealing with uncertainties: sensitivity analysis of
vacuum gas oil hydrotreatment. Chemical Engineering Journal 278: 469–478.
Coronel-García, M.A., Reyes de la Torre, A.I., Domínguez-Esquivel, J.M. et al. (2021). Heavy oil
hydrocracking kinetics with nano-nickel dispersed in PEG300 as slurry phase catalyst using batch
reactor. Fuel 283: 118930.
Du, H., Liu, D., Li, M. et al. (2015). Effects of the temperature and initial hydrogen pressure on the
isomerization reaction in heavy oil slurry-phase hydrocracking. Energy Fuels 29: 626–633.
Duran Armas, J. L. (2021). Application of In-situ Upgrading in Naturally Fractured Reservoirs [PhD.
Thesis, University of Calgary].
Elahi, S.M., Scott, C.E., Chen, Z., and Pereira-Almao, P. (2019). In-situ upgrading and enhanced recovery
of heavy oil from carbonate reservoirs using nano-catalysts: upgrading reactions analysis. Fuel 252:
262–271.
Elizalde, I. and Ancheyta, J. (2014). Modeling the deactivation by metal deposition of heavy oil
hydrotreating catalyst. Catalysis Today 220–222: 221–227.
Elizalde, I., Rodríguez, M.A., and Ancheyta, J. (2009). Application of continuous kinetic lumping
modeling to moderate hydrocracking of heavy oil. Applied Catalysis A: General 365: 237–242.
Elizalde, I., Trejo, F., Muñoz, J.A.D. et al. (2016). Dynamic modeling and simulation of a bench-scale
reactor for the hydrocracking of heavy oil by using the continuous kinetic lumping approach. Reaction
Kinetics, Mechanisms and Catalysis 118: 299–311.
Félix, G. and Ancheyta, J. (2019a). Comparison of hydrocracking kinetic models based on SARA fractions
obtained in slurry-phase reactor. Fuel 241: 495–505.
Félix, G. and Ancheyta, J. (2019b). Using separate kinetic models to predict liquid, gas, and coke yields in
heavy oil hydrocracking. Industrial Engineering Chemistry Research 58: 7973–7979.
Félix, G., Quitian, A., Rodríguez, E. et al. (2017). Methods to calculate hydrogen consumption during
hydrocracking experiments in batch reactors. Energy Fuels 31: 11690–11697.
Félix, G., Ancheyta, J., and Trejo, F. (2019). Sensitivity analysis of kinetic parameters for heavy oil
hydrocracking. Fuel 241: 836–844.
Félix, G., Ríos, J.J., Tirado, A. et al. (2022a). Monte Carlo and sensitivity analysis methods for kinetic
parameters optimization: application to heavy oil slurry-phase hydrocracking. Energy Fuels 36:
9251–9260.
Félix, G., Tirado, A., Al-Muntaser, A. et al. (2022b). SARA-based kinetic model for non-catalytic
aquathermolysis of heavy crude oil. Journal of Petroleum Science and Engineering 216: 110845.
Fogler, H.S. (1992). Elements of Chemical Reaction Engineering, 5the. Prentice-Hall.
Froment, G.F. (1987). Kinetics of the hydroisomerization and hydrocracking of paraffins on a platinum
containing bifunctional Y-zeolite. Catalysis Today 1: 455–473.
Fukuyama, H. and Terai, S. (2007). Kinetic study on the hydrocracking reaction of vacuum residue using
a lumping model. Petroleum Science and Technology 25: 277–287.
Galarraga, C.E., Scott, C., Loria, H., and Pereira-Almao, P. (2012). Kinetic models for upgrading
athabasca bitumen using unsupported NiWMo catalysts at low severity conditions. Industrial and
Engineering Chemistry Research 51: 140–146.
Gauthier, S., Keane, C.J., Niemela, J.J. et al. (2006). Optimization algorithms and weighting factors for
analysis of dynamic PET studies. Physics in Medicine Biology 51: 4217.
Guillaume, D., Valéry, E., Verstraete, J. et al. (2011). Single event kinetic modelling without explicit
generation of large networks: application to hydrocracking of long paraffins. Oil Gas Science and
Technology – Revue d’IFP Energies Nouvelles 66: 399–422.
Hassanzadeh, H. and Abedi, J. (2010). Modelling and parameter estimation of ultra-dispersed in situ
catalytic upgrading experiments in a batch reactor. Fuel 89: 2822–2828.
Heidary, S., Dehghan, A.A., and Mahdavi, S. (2017). Feasibility study on application of the recent
enhanced heavy oil recovery methods (VAPEX, SAGD, CAGD and THAI) in an Iranian heavy oil
reservoir. Petroleum Science and Technology 35: 2059–2065.
Hillewaert, L.P., Dierickx, J.L., and Froment, G.F. (1988). Computer generation of reaction schemes and
rate equations for thermal cracking. AIChE Journal 34: 17–24.
Huang, T., Liu, B., Wang, Z., and Guo, X. (2017). Kinetic model for hydrocracking of Iranian heavy crude
with dispersed catalysts in slurry-phase. Petroleum Science and Technology 35: 1846–1851.
International – U.S. Energy Information Administration (EIA). (2021). https://www.eia.gov/
Jarullah, A.T., Mujtaba, I.M., and Wood, A.S. (2011). Kinetic parameter estimation and simulation of
trickle-bed reactor for hydrodesulfurization of crude oil. Chemical Engineering Science 66: 859–871.
Kim, S.-H., Kim, K.-D., and Lee, Y.-K. (2017). Effects of dispersed MoS2 catalysts and reaction conditions
on slurry phase hydrocracking of vacuum residue. Journal of Catalysis 347: 127–137.
References 53
Laxminarasimhan, C.S., Verma, R.P., and Ramachandran, P.A. (1996). Continuous lumping model for
simulation of hydrocracking. AIChE Journal 42: 2645–2653.
Liu, D., Kong, X., Li, M., and Que, G. (2009). Study on a water-soluble catalyst for slurry-phase
hydrocracking of an atmospheric residue. Energy Fuels 23: 958–961.
Loria, H., Trujillo-Ferrer, G., Sosa-Stull, C., and Pereira-Almao, P. (2011). Kinetic modeling of bitumen
hydroprocessing at in-reservoir conditions employing ultradispersed catalysts. Energy Fuels 25:
1364–1372.
Ma, Y., Zhang, J., Wu, W. et al. (2022). Trialkylmethylammonium molybdate ionic liquids as novel oil-
soluble precursors of dispersed metal catalysts for slurry-phase hydrocracking of heavy oils. Chemical
Engineering Science 253: 117516.
Marafi, M., Al-Barood, A., and Stanislaus, A. (2005). Effect of diluents in controlling sediment formation
during catalytic hydrocracking of Kuwait vacuum residue. Petroleum Science and Technology 23: 899–908.
Marafi, A., Stanislaus, A., and Furimsky, E. (2010). Kinetics and modeling of petroleum residues
hydroprocessing. Catalysis Reviews – Science and Engineering 52: 204–324.
Marroquín, G. (2007). Sediment formation during the catalytic hydrotreatment of heavy crude oils [PhD.
Thesis, Instituto Politécnico Nacional].
Martens, G.G. and Marin, G.B. (2001). Kinetics for hydrocracking based on structural classes: model
development and application. AIChE Journal 47: 1607–1622.
Martínez, J. and Ancheyta, J. (2012). Kinetic model for hydrocracking of heavy oil in a CSTR involving
short term catalyst deactivation. Fuel 100: 193–199.
Martínez, J., Sánchez, J.L., Ancheyta, J., and Ruiz, R.S. (2010). A review of process aspects and modeling
of ebullated bed reactors for hydrocracking of heavy oils. Catalysis Reviews 52: 60–105.
Martínez-Grimaldo, H.J., Chavarria-Hernandez, J.C., Ramírez, J. et al. (2011). Prediction of sulfur
content, API gravity, and viscosity using a continuous mixture kinetic model for maya crude oil
hydrocracking in a slurry-phase reactor. Energy Fuels 25: 3605–3614.
Martínez-Grimaldo, H.J., Ortiz Moreno, H., Sánchez Minero, F. et al. (2014). Hydrocracking of Maya
crude oil in a slurry-phase reactor. I. Effect of reaction temperature. Catalysis Today 220–222: 295–300.
Matsumura, A., Sato, S., Kondo, T. et al. (2005). Hydrocracking Marlim vacuum residue with natural
limonite. Part 2: Experimental cracking in a slurry-type continuous reactor. Fuel 84: 417–421.
Mederos, F.S., Ancheyta, J., and Chen, J. (2009). Review on criteria to ensure ideal behaviors in trickle-
bed reactors. Applied Catalysis A: General 355: 1–19.
Memon, A. I., Gao, J., Taylor, S. D., Engel, T. L., and Jia, N. (2010). A systematic workflow process for
heavy oil characterization: experimental techniques and challenges. Canadian Unconventional
Resources and International Petroleum Conference.
Mitsios, M., Guillaume, D., Galtier, P., and Schweich, D. (2009). Single-event microkinetic model for
long-chain paraffin hydrocracking and hydroisomerization on an amorphous Pt/SiO2 Al2O3 catalyst.
Industrial Engineering Chemistry Research 48: 3284–3292.
Mochida, I., Zhao, X., Sakanishi, K. et al. (1989). Structure and properties of sludges produced in
the catalytic hydrocracking of vacuum residue. Industrial Engineering Chemistry Research 28: 418–421.
Moghadassi, A., Amini, N., Fadavi, O., and Bahmani, M. (2011). Hydrocracking lumped kinetic model
with catalyst deactivation in Arak refinery hydrocracker unit. Journal of Petroleum Science and
Technology 1: 31–37.
Moustafa, T.M. and Froment, G.F. (2003). Kinetic modeling of coke formation and deactivation in the
catalytic cracking of vacuum gas oil. Industrial Engineering Chemistry Research 42: 14–25.
Nguyen, T.S., Tayakout-Fayolle, M., Ropars, M., and Geantet, C. (2013). Hydroconversion of an
atmospheric residue with a dispersed catalyst in a batch reactor: kinetic modeling including vapor-
liquid equilibrium. Chemical Engineering Science 94: 214–223.
Nguyen, M.T., Nguyen, N.T., Cho, J. et al. (2016). A review on the oil-soluble dispersed catalyst for slurry-
phase hydrocracking of heavy oil. Journal of Industrial and Engineering Chemistry 43: 1–12.
Oil Gas Journal. (2022). https://www.ogj.com/
Organization of the Petroleum Exporting Countries (OPEC). (2020). https://www.opec.org/opec_web/
en/index.htm
Orozco Castillo, C. (2016). In situ heavy oil upgrading through ultra-disperse nano-catalyst injection in
naturally fracture reservoirs [PhD. Thesis, University of Calgary].
Ortega-García, F.J., Mar Juárez, E., and Schacht Hernández, P. (2012). Controlling sediments in the
ebullated bed hydrocracking process. Energy Fuels 26: 2948–2952.
Ortega-García, F.J., Muñoz Arroyo, J.A.D., Flores Sánchez, P. et al. (2017). Hydrocracking kinetics of a
heavy crude oil on a liquid catalyst. Energy Fuels 31: 6794–6799.
Ortiz Moreno, H., Ramírez, J., Sanchez-Minero, F. et al. (2014). Hydrocracking of Maya crude oil in a
slurry-phase batch reactor. II. Effect of catalyst load. Fuel 130: 263–272.
Panariti, N., Del Bianco, A., and Marchionna, M. (2000). Petroleum residue upgrading with dispersed
catalysts Part 1. Catalysts activity and selectivity. Applied Catalysis A: General 204: 203–213.
Perego, C. and Peratello, S. (1999). Experimental methods in catalytic kinetics. Catalysis Today 52:
133–145.
Pham, H.H., Kim, K.H., Go, K.S. et al. (2021). Hydrocracking and hydrotreating reaction kinetics of heavy
oil in CSTR using a dispersed catalyst. Journal of Petroleum Science and Engineering 107997.
Pratama, R.A. and Babadagli, T. (2022). A review of the mechanics of heavy-oil recovery by steam
injection with chemical additives. Journal of Petroleum Science and Engineering 208: 109717.
Puron, H., Chin, K.K., Pinilla, J.L. et al. (2014). Kinetic analysis of vacuum residue hydrocracking in early
reaction stages. Fuel 117: 408–414.
Quitian, A. and Ancheyta, J. (2016a). Experimental methods for developing kinetic models for
hydrocracking reactions with slurry-phase catalyst using batch reactors. Energy Fuels 30: 4419–4437.
Quitian, A. and Ancheyta, J. (2016b). Partial upgrading of heavy crude oil by slurry-phase hydrocracking
with analytical grade and ore catalysts. Energy Fuels 30: 10117–10125.
Quitian, A. and Ancheyta, J. (2017). Characterization of upgraded oil fractions obtained by slurry-phase
hydrocracking at low-severity conditions using analytical and ore catalysts. Energy Fuels 31: 9162–9178.
Quitian, A., Leyva, C., Ramírez, S., and Ancheyta, J. (2015). Exploratory study for the upgrading of
transport properties of heavy oil by slurry-phase hydrocracking. Energy Fuels 29: 9–15.
Rana, M.S., Sámano, V., Ancheyta, J., and Diaz, J.A.I. (2007). A review of recent advances on process
technologies for upgrading of heavy oils and residua. Fuel 86: 1216–1231.
Rana, A., Khan, I., and Saleh, T.A. (2021). Advances in carbon nanostructures and nanocellulose as
additives for efficient drilling fluids: trends and future perspective – a review. Energy Fuels 35:
7319–7339.
Rașeev, S.D. (2003). Thermal and Catalytic Processes in Petroleum Refining, 1ste. Marcel Dekker.
Rashidzadeh, M., Ahmad, A., and Sadighi, S. (2011). Studying of catalyst deactivation in a commercial
hydrocracking process (ISOMAX). Petroleum Science and Technologyetroleum 1: 46–54.
Rezaei, H. (2013). Catalyst Recycle in Slurry-phase Residue Upgrading [PhD. Thesis, The University of
British Columbia].
Rezaei, H., Liu, X., Ardakani, S.J. et al. (2010). A study of Cold Lake vacuum residue hydroconversion in
batch and semi-batch reactors using unsupported MoS2 catalysts. Catalysis Today 150: 244–254.
Riazi, M.R. (2005). Characterization and Properties of Petroleum, 1ste, vol. 407. ASTM International.
Rodríguez, E., Félix, G., Ancheyta, J., and Trejo, F. (2018). Modeling of hydrotreating catalyst
deactivation for heavy oil hydrocarbons. Fuel 225: 118–133.
References 55
Rogel, E., Ovalles, C., Pradhan, A. et al. (2013). Sediment formation in residue hydroconversion processes
and its correlation to asphaltene behavior. Energy Fuels 27: 6587–6593.
Sadighi, S., Ahmad, A., and Seif Mohaddecy, S.R. (2010). 6-Lump kinetic model for a commercial vacuum
gas oil hydrocracker. International Journal of Chemical Reactor Engineering 8.
Sahu, R., Song, B.J., Im, J.S. et al. (2015). A review of recent advances in catalytic hydrocracking of heavy
residues. Journal of Industrial and Engineering Chemistry 27: 12–24.
Saleh, T.A. (2018). Nanotechnology in Oil and Gas Industries, 1ste. Springer.
Sámano, V., Tirado, A., Félix, G., and Ancheyta, J. (2020). Revisiting the importance of appropriate
parameter estimation based on sensitivity analysis for developing kinetic models. Fuel 267: 117113.
Sánchez, S., Rodríguez, M.A., and Ancheyta, J. (2005a). Kinetic model for moderate hydrocracking of
heavy oils. Industrial Engineering Chemistry Research 44: 9409–9413.
Sánchez, S., Ancheyta, J., and McCaffrey, W.C. (2007). Comparison of probability distribution functions
for fitting distillation curves of petroleum. Energy Fuels 21: 2955–2963.
Santos, R.G., Loh, W., Bannwart, A.C., and Trevisan, O.v. (2014). An overview of heavy oil properties and
its recovery and transportation methods. Brazilian Journal of Chemical Engineering 31: 571–590.
Sheng, Q., Wang, G., Zhang, Q. et al. (2017). Kinetic model for liquid-phase liquefaction of asphaltene by
hydrogenation with industrial distillate narrow fraction as hydrogen donor. Fuel 209: 54–61.
da Silva De Andrade, F. J. (2014). Kinetic modeling of catalytic in situ upgrading for Athabasca bitumen,
deasphalting pitch and vacuum residue [PhD. Thesis, University of Calgary].
Song, F.M., Liu, C.G., Zhou, G.S., and Yang, G.J. (2004). Thermal hydrocracking kinetics of Chinese
gudao vacuum residue. Petroleum Science and Technology 22: 689–708.
Speight, J.G. (2004). New approaches to hydroprocessing. Catalysis Today 98: 55–60.
Speight, J.G. (2013). Heavy and Extra-Heavy Oil Upgrading Technologies, 1ste. Elsevier Science.
Stanislaus, A., Hauser, A., and Marafi, M. (2005). Investigation of the mechanism of sediment formation
in residual oil hydrocracking process through characterization of sediment deposits. Catalysis Today
109: 167–177.
Stratiev, D.S., Russell, C.A., Sharpe, R. et al. (2014). Investigation on sediment formation in residue
thermal conversion based processes. Fuel Processing Technology 128: 509–518.
Stratiev, D.S., Shishkova, I.K., Nikolaychuk, E. et al. (2019). Effect of catalyst condition on sedimentation
and conversion in the ebullated bed vacuum residue H-Oil hydrocracking. Petroleum Science and
Stratiev, D., Nenov, S., Nedanovski, D. et al. (2021). Different nonlinear regression techniques and
sensitivity analysis as tools to optimize oil viscosity modeling. Resources 10: 99.
Tirado, A. and Ancheyta, J. (2018). A batch reactor study of the effect of aromatic diluents to reduce
sediment formation during hydrotreating of heavy oil. Energy Fuels 32: 60–66.
Tirado, A., Félix, G., Kwofie, M. et al. (2022). Kinetics of heavy oil non-catalytic aquathermolysis with and
without stoichiometric coefficients. Fuel 323: 124365.
Valéry, E., Guillaume, D., Surla, K. et al. (2007). Kinetic modeling of acid catalyzed hydrocracking of
heavy molecules: application to squalane. Industrial Engineering Chemistry Research 46: 4755–4763.
56
Modeling Catalyst Deactivation of Hydrotreating of Heavy Oils

Javier Jurado1, Vicente Samano1, and Jorge Ancheyta2
1
Instituto Politécnico Nacional, Centro de Investigación en Ciencia Aplicada y Tecnología Avanzada, Unidad Legaria, Legaria 694,
Col. Irrigación, Ciudad de México, Mexico
2
Instituto Mexicano del Petróleo, Eje Central Lázaro Cárdenas Norte 152, San Bartolo Atepehuacan, Ciudad de México, Mexico
2.1 Introduction
The demand for high-value petroleum products, such as middle distillate, gasoline, and lube oil, is
increasing, while the demand for low-value products, such as fuel oil and residua-based products, is
decreasing. Therefore, maximizing liquid products’ yield from various processes and valorization of
residues are of immediate attention to refiners. Also, the changing prices of crude oils as well as
increasing production of heavy and extra-heavy crude oils has motivated more research and devel-
opment aiming at upgrading such heavy materials. These trends have emphasized the importance
of processes that convert the heavier oil fractions into lighter and more valuable clean products,
such as hydrotreating (HDT) and hydrocracking (Castañeda et al. 2014; Rana et al. 2007).
Catalytic HDT is a process to hydrogenate and cleave the C─C bonds of hydrocarbons using
hydrogen gas and a catalyst, and thereby converting them into compounds of lower molecular
weight (Miki et al. 1983). Although it is often mentioned that HDT is a mature technology and there
is not too much to investigate, the increasing availability of heavy crude oils and the need to
upgrade them have motivated the researchers to continue to study the conversion of these heavy
hydrocarbons (Rana et al. 2007). The nonmetallic (sulfur, oxygen, and nitrogen) and metallic
(nickel and vanadium) compounds mainly present in heavy crude oils are undesirable even before
refinery processing because they are poisons for many catalysts. Thus, reducing the content of het-
eroatoms and asphaltenes of heavy crude oils improves their processability considerably in down-
stream conversion processes (Gary et al. 2007).
In industrial HDT, trickle-bed reactor (TBR) is the most used for heavy fractions because of the
advantages it offers compared with other types of reactors (Mederos et al. 2009), such as low catalyst
loss, capability to operate at high pressures and temperatures, lower investment and operating
costs, among others, as compared with other technologies, for example, ebullated-bed reactor,
slurry-phase reactor (SPR) (Gianetto and Specchia 1992). SPR offers several advantages over other
types of reactors, such as low pressure drop, nearly isothermal operations, capacity to achieve high
conversion rates, operational flexibility, low catalyst concentration and online catalyst addition.
However, reduction of conversion due to high back-mixing, separation of solid particles, among
others make SPR to operate at high temperature and pressure with the consequent high investment
and operating costs. That is the main reason why SPR-based technologies for conversion of
residues have shown poor commercial experience (Calderón and Ancheyta 2017). However,
the main disadvantage during HDT of heavy crude oils is catalyst deactivation due to the
metals and coke deposition, which is more dramatic at long time-on-stream (TOS) and high
reaction severity (Mohammed et al. 2006). Coke is formed very rapidly during the first hours
of stream, meanwhile deposition of metals takes place at longer period. During HDT reactions
at high temperature, asphaltene molecules are mostly converted into coke and together with
metals deposited on the catalyst in due course of the reaction (Ancheyta et al. 2003; Elizalde
and Ancheyta 2014a).
The study of catalyst deactivation during HDT of heavy oil fractions is one of the most important
aspects to improve the catalytic performance in petroleum refining processes (Ancheyta et al. 2002).
A good commercial catalyst is known through three characteristics, namely activity, selectivity, and
stability (Moghadassi et al. 2011; Ward 1993).
The loss of catalytic activity and/or selectivity over time is a problem of great and continuing
concern in the practice of industrial catalytic processes. The degree of catalyst deactivation depends
mainly on the properties of the feed and operating conditions (Kohli et al. 2016). The performance
of catalysts decreases with time due to the already commented reasons, thus to maintain constant
product yields and/or quality, the loss of catalytic activity must be compensated by periodic
increases of reaction temperature (Ancheyta et al. 2003; Moghadassi et al. 2011). Costs to industry
for catalyst replacement and process shutdown are of billions of dollars per year. However, it is
inevitable that all catalysts activity will decay (Bartholomew 2001).
Kinetic studies provide valuable information that is part of the database used for comparison and
selection of catalysts as well as matching the type of catalyst with reactor and feed. The parameters
determined from such studies form the basis for the development of kinetic and deactivation mod-
els, which can be used for assessing the reactivity of heavy feeds, selection of catalysts and optimiz-
ing the operation of catalytic systems. Kinetic and deactivation parameters are part of the models
used for predicting catalyst life. In fact, the determination of such parameters for a particular cat-
alyst may be viewed as the modeling of a catalyst activity scale (Marafi et al. 2010a,b).
2.2 Mechanisms of Deactivation
Catalyst deactivation during HDT of heavy oils comprises three stages: (i) a rapid loss of activity due
to coke deposition on the catalyst surface, usually within the first hundred hours on stream, (ii) a
slower stage where metal covers the catalyst surface, and finally (iii) a rapid loss of activity due to
metal and coke deposited on the catalyst (Callejas et al. 2001; Furimsky and Massoth 1999). Deac-
tivation of HDT catalysts occurs by single or multiple causes, such as active site blockage, active site
coverage by deposits of coke and/or metals, pore mouth blockage, and sintering of the active phase
(Dautzenberg et al. 1978). The degree and type of deactivation will depend on the feed properties
and reaction conditions and will be evidenced by an S-shape curve of temperature vs TOS
(Ancheyta 2016) as shown in Figure 2.1. Stage 1 is characterized by rapid formation of initial coke
with significant loss in catalyst activity; Stage 2 continues with gradual buildup or accumulation of
metal deposit, causing further loss in the activity of the catalyst; and Stage 3 finishes with the con-
striction or blockage of pore by continuous deposition of metals with complete loss of catalytic
activity due to diffusion limitation (Gray et al. 1999). Initially, coke deposition causes a rapid deac-
tivation, which in several hours attains a pseudo-steady state (Stage 1). Then, deactivation by metal
58 2 Modeling Catalyst Deactivation of Hydrotreating of Heavy Oils
Stage 1 Stage 2 Stage 3

405
400
Operating temperature (°C)
395
390
385
380
375
370
0 0.2 0.4 0.6 0.8 1
Normalized TOS
Figure 2.1 Typical profile of catalyst deactivation. Stage 1: deactivation by coke, Stage 2: deactivation by
metals, and Stage 3: pore blockage (Ancheyta 2016 / John Wiley & Sons).
deposits is observed during a longer period of time until a drastic deactivation caused by pore
restriction and blockage is observed (Ancheyta 2016).
In catalytic HDT of heavy oil, asphaltenes are the main cause for coke formation via different
reaction mechanisms. They also contribute to metal deposition as nickel and vanadium are mostly
concentrated in asphaltenes, and when they are hydrocracked, metals are released and deposited
on the catalyst surface. Asphaltenes are widely known to be extremely large and complex mole-
cules, and their diffusion into catalyst pores is difficult. Therefore, when designing catalysts for
HDT of heavy oil, it is important to study the properties of such large molecules (not only bulk
properties but also structural and crystalline properties) and textural properties of the catalyst (pore
size distribution, average pore size). For instance, various authors have reported that the activity of
the catalyst is strictly related to the properties of the catalyst, such as size and volume of the pores
(Galiasso et al. 1983; Oyekunle et al. 2005; Shimura et al. 1986; Song et al. 1991).
In the case of heavy feeds, resins play the role of stabilizer. If there is not a proper ratio of asphal-
tenes, oil, and resins, the colloid stability of the feed cannot be maintained provoking asphaltene
precipitation (Mochida et al. 1989). During HDT, it is important to remove the resins at about the
same rate as asphaltenes, otherwise, precipitation of asphaltenes from the feed can be a major cause
of coke formation (Beuther et al. 1980a). Hydrotreated asphaltenes become more aromatic than the
original asphaltenes. Condensation of asphaltenes on the catalysts surface may produce coke con-
taining small anisotropic regions consistent with mesophase development (Eser et al. 1986;
Nagaishi et al. 1997).
Coke deposition increases with the boiling range of the feed. However, for feeds with similar boil-
ing range, the one with higher concentration of coke precursors, such as aromatics and heterocyclic
compound, will cause more severe poisoning. About one half of metals (Ni, V) present in heavy
feeds are present as porphyrins, the other part of metals are in less-characterized forms that include
bonds with nitrogen, oxygen, and sulfur in the defect centers of asphaltene sheets. Due to the bulk-
iness of the porphyrins and asphaltene molecules, the demetallization reaction is affected by pore
diffusion limitations. Consequently, metals are generally deposited on the outer zone of the catalyst
particle (Ancheyta 2016; Beaton and Bertolacini 1991; Mitchell 1990).
2.2.1 Coking Deposition (Fouling)

Fouling is the physical deposition of species from the fluid phase onto the catalyst surface, which
results in activity loss due to blockage of site and/or pores mainly due to deposits of coke in porous
catalysts. The deactivating influence of coke on catalyst depends on the nature of coke, its structure
and morphology, and the exact location of its deposition on the catalyst surface. The chemical
structures of cokes formed in catalytic processes vary with reaction type, catalyst type, and reaction
condition. When coke deposition occurs mainly on the carrier surface and not directly on the active
sites, the catalyst can tolerate even 6–10 wt.% coke without much loss in activity (Bartholomew
1982; Menon 1990; Nielsen and Trimm 1977; Trimm 1977, 1983).
The coke content of the catalyst typically reaches its maximum steady-state value at about 15 to
25% in a very short period (Thakur and Thomas 1985). About one-third of the total pore volume of
fresh catalyst is filled by coke in the first hours of operation of the process. The formation of coke is
due to polymerization and polycondensation of high molecular weight compounds of the feedstock
(Furimsky and Massoth 1999). Analyses made by Tamm et al. (1981) have shown that the coke
deposition starts on the external surface of the catalyst and its concentration in the interior of
the catalyst increases with process time. In residue HDT, the rapid decline in catalyst activity during
the early hours-on-stream is due to buildup of strongly adsorbed surface coke (Hauser et al. 2005),
which forms a monolayer on the catalyst pellet (Richardson et al. 1996).
Coke deposition increases with the boiling range of the feed. However, for feeds with similar boil-
ing range, the one with higher concentration of coke precursors, such as aromatics and heterocyclic
compounds, will cause a more severe deactivation (Ancheyta 2016). Beuther et al. (1980b) classified
coke deposits formed during hydrodesulfurization (HDS) of residue into three types: Type I deposits
are the normal aromatic background of non-asphaltenic structures deposited during the first part of
the cycle at low temperature. Type II deposits are structures of asphaltenes deposited early in the
coking process. Type III deposits are materials laid after long reaction time and high temperature,
resulting from condensation of aromatic concentrates into clusters and then crystals which
constitute a mesophase.
2.2.2 Metal Deposition (Poisoning)

Poisoning is the strong chemisorption of reactants, products, and impurities, which may affect cat-
alytic activity, changing the geometry of the catalytic surface. Poisoning has an operational mean-
ing: whether a species acts as a poison, its effect on catalyst activity depends on the adsorption
strength relative to the other species competing for the catalytic sites (Bartholomew 2001; Hegedus
and McCabe 1994; Maxted 1951).
In heavy feeds, porphyrin forms of V and Ni are the main organometallic compounds, they are
the main cause of the metal deposits formation on catalyst surface, and most of these metals are
associated with the asphaltene entities. In this type of feed, the amount of V-porphyrins is greater
than that of Ni-porphyrins. Therefore, the metal deposition patterns are influenced by the former to
a much greater extent (Marafi et al. 2010a,b).
The most commonly used catalytic materials for removing both vanadium and nickel present in
heavy feeds are Ni/Mo and Co/Mo sulfide catalysts supported on alumina (Speight 1999). The spe-
cific surface area of the catalyst is usually large (200–300 m2/g), but almost the entire surface is con-
tained within the pore space of the alumina. The Co, Mo, and Ni metals are dispersed in a thin layer
within the pore system of the alumina to form active sites. These types of catalysts are frequently
deactivated and the main cause of deactivation is the metal sulfide deposits (Oyekunle et al. 2005).
Heavy oil
Entrance
a b
Hydrogen
Catalytic bed length

Catalyst
Hydrotreated Exit
Porcentage of deposit
product
Figure 2.2 Schematic representation of position-dependent foulant deposition profiles in a fixed-bed reactor:
(a) coke and (b) metals.
The deposits tend to grow with TOS blocking the pore mouths, reducing the pore radius to the same
magnitude as the size of the organometallic molecules, and thus the effect of restricted diffusivities
is presented at the pore mouth of the catalyst (Kam et al. 2005; Rajagopalan 1979). In a fixed-bed
reactor coke content increases from the entrance to the exit, while metal content decreases from the
entrance to the exit of the reactor as shown in Figure 2.2 (Thakur and Thomas 1985).
2.3 Deactivation Models
From technical point of view, studying the deactivation phenomena due to HDT reactions is a very
complex task, because it requires high costs of experiments and time-consuming, for estimating the
performance and life of the catalysts (Elizalde and Ancheyta 2014a). Mathematical modeling is a
powerful tool to minimize the costs and time required in experimental work to achieve such
research targets (Kam et al. 2005). For this reason, it is desirable to account for a model that simu-
lates the deactivation patterns during catalytic HDT reactions. There are various reactor modeling
works reported in the literature (Alvarez and Ancheyta 2008; Ferreira et al. 2010; Korsten and
Hoffmann 1996; Kumar et al. 1997; Mederos et al. 2006, 2012; Sau et al. 1997; Verstraete et al.
2007) that do not take into consideration the catalytic deactivation to describe long time operations,
that is, middle-of-run (MOR) and end-of-run (EOR) periods. Various models have been reported in
the literature to explain the catalysts deactivation based on different mechanisms of coke and metal
deposition. The most relevant deactivation models are summarized in the following sections.
2.3.1 Deactivation Models by Coke Deposition

It is known that coke is deposited on HDT catalyst surfaces very quickly in the initial stage of run
(Kam et al. 2005). Newson (1975) observed that this behavior is due to a period of uncontrolled
hydrocracking in a short time, about 40 h; subsequently, coke content on catalyst remains constant
with increasing process time. Gualda and Kasztelan (1996) reported that carbon deposits
accumulated on the catalyst surface at a level of 11 wt.% after 6 h and reach 14 wt.% after 140 h on
stream. This initial deposition of coke was termed “young coke” and had a relative high hydrogen-
to-carbon ratio (H/C), ranging from 0.8 to 1.2 (Kam et al. 2005).
One of the most used coking model to describe coke deposition was proposed by Froment and
Bischoff (1962), which involves a deactivation function with coke content on the catalyst through-
out the reactor, relating the carbon content to the reaction rate coefficient as follows:
k = kη φ c 21
where kη is the reaction rate coefficient when no carbon has been deposited, k is the reaction
rate coefficient when carbon has been deposited, and φ is an activity function. An exponential form
of φ could be used as follows:
φ = exp − C0 22
where is the deactivation constant and C0 the carbon content on catalyst. This is based on
the observation that after a strong initial decrease of conversion, the product of interest’s
yield decreases slowly tending asymptotically to a limiting value.
Another form of the activity function was also suggested (Eq. 2.3), which may arise from
Langmuir–Hinshelwood concepts. For mathematical simplicity, it neglects the adsorption of the
components of the main reaction, so that the constant α was then a type of adsorption equilibrium
constant that expresses the proportionality between the amount of deposited carbonaceous
compound, measured as carbon, and the number of active sites fouled.
1
φ= β
23
1 + α C0
Most of the authors (Jacob et al. 1976; Lee et al. 1989; Martínez and Ancheyta 2012; Van et al.
1996) represent the catalytic activity decline by the exponential law or the hyperbolic function given
by Eqs. (2.2) and (2.3), respectively, to simplify the overall kinetic model and parameter estimation.
Voorhies (1945) proposed a model to describe coke deposition on a cracking catalyst in a fixed-
bed reactor, considering only the catalytic carbon produced in the reaction and not including the
removable carbon formed by the absorbed and interleaved gaseous and liquid hydrocarbon
remaining in the spent catalyst after the cracking. This model considers that the weight percent
of carbon formed on the catalyst is approximately a logarithmic function of the length of the time
elapsed since the catalyst is freshly regenerated, by means of the following equation:
V
C 0 = A τ 60 24
for 0.5 ≤ V ≤ 1
where C0 is the amount of coke on the catalyst at time τ, and the correlation coefficients A and V are
obtained experimentally. This equation is not recommended for hydroprocessing catalyst where the
effect of coke on the activity loss is more complicated than that of cracking catalysts.
Rashidzadeh (2010) proposed a 5-lump kinetic model for hydrocracking of vacuum gas oil (VGO)
in a commercial plant. The model considers VGO, diesel, kerosene, naphtha, and gas as products
and that deactivation of catalyst is caused by its surface coke. Catalyst decay function is based on its
life, an exponential form formulated as follows:
− αij Life
φij = e 25
where ij shows the deactivation constant for converting lump i to lump j and life value can be
expressed in days on stream.
0.5
0.4
Yield (mass)
0.3
0.2
0.1
0
0 100 200 300 400 500
Life (day)
Figure 2.3 Comparison of predicted yield of model with measured data. Diesel, predicted (---) and measured
(■). Kerosene, predicted (___) and measured (●). Naphtha, predicted (…) and measured (▲) (Adapted from
Rashidzadeh et al. 2010).
The catalyst used was an amorphous type with the following properties: specific surface area of
176 m2/g, pore volume of 0.51 cm3/g, and bulk density of 0.75 g/cm3. The feed was a mixture of
fresh VGO and unconverted oil. The reactor had 4 beds with a total amount of loaded catalyst
in the reactor of about 69,000 kg. Feed flow rate and reactor operating condition are the following:
fresh feed and recycle feed of 48,545.5 kg/h and 32,634.3 kg/h, respectively, temperature of 400 C,
and total pressure of 153 kg/cm2.
Figure 2.3 shows comparisons between measured product yields from the industrial hydrocrack-
ing reactor and the model predictions. The prediction of the model for 1.5 years is in good agree-
ment with the actual commercial data. The authors reported deviations of predictions for naphtha,
kerosene, and diesel yields of about 1.78%, 1.98%, and 1.97%, respectively.
Richardson et al. (1996) reported a model to describe the initial coke buildup and to predict coke
formation during processing of heavy residues in the first hours of operation, using bitumen and
commercial NiMo/γ-Al2 O3 catalyst in batch and continuous stirred tank reactor (CSTR) reactors,
represented by the following equation:
C0 = Cmax 1 − exp − Kw 26
where w is the cumulative feed-to-catalyst ratio, K an adsorption constant, and Cmax is the
maximum carbon deposition that can be accumulated on the surface equal to monolayer coverage.
If the substance adsorbed on the surface is similar in carbon content, molecular weight (MW), and
density of the residue fraction of the oil (ρ), the radius for a typical residue molecule (rM) would be
given by the following equation:
1 3
3M W
rM = 27
Nρ4π
where N is the Avogadro’s number. To obtain the maximum carbon deposition, Cmax is given as
follows:
MW X c
Cmax = 28
Nπ r 2M
where Xc is the mass fraction of carbon in the residue.
Ancheyta and Lopez (2000) analyzed five deactivation models based on TOS theory for studying
the deactivation of catalyst for catalytic cracking of VGO, in which the main cause of catalyst deac-
tivation is coke deposition. The models were derived from a general m-order expression: Model 1,
exponential law; Model 2, a second-order model; Model 3, power law; Model 4, hyperbolic function;
and Model 5, an empirical model. The following deactivation rate was considered:
dφ
− = k d φm 29
dt
Integration of Eq. (2.9) for m 1 gives,
− 1m − 1
φ = 1 + kd m − 1 t 2 10
Using Eqs. (2.9) and (2.10), the following reported models can be deduced:
Model 1: for m = 1, integrating Eq. (2.9) gives the exponential law,
φ = e − kd t 2 11
Model 2: for m = 2, a second-order model is obtained from Eq. (2.10):
1
φ= 2 12
1 + kd t
Model 3: for m > 2, a power law model is obtained from Eq. (2.10):
φ = A t−B 2 13
where
−1 m−1
A = kd m − 1 2 14
B = 1 m−1 2 15
Model 4: the hyperbolic function is obtained from Eq. (2.10):
−B
φ = 1 + Gt 2 16
where
G = kd m − 1 2 17
Model 5: Empirical deactivation model, reported elsewhere (Jacob et al. 1976; Oliveira and
Biscaia 1989), g = 1 and d = kd
a
φ= 2 18
pq 1 + d tg
To test the models, experiments were carried out with industrial catalyst and feedstock using
the 3-lump kinetic scheme proposed by Weekman (1968) in a micro-activity unit with a
fixed-bed reactor at a reaction temperature of 500 C and constant cat-to-oil ratio of 5.
Figure 2.4 shows the calculated and experimental conversion, gasoline yield, and gas plus coke
yield for each deactivation model. The authors found that Models 4 and 5 were the best for repre-
senting the deactivation of catalyst by coke, and Models 1 and 3 were slightly inferior to Models 4
and 5; however, they represent the catalyst deactivation sufficiently well. Model 2 was highly
inaccurate.
(a)
0.57
0.56
Calculated gasoline yield, wt%
0.55
0.54
0.53
0.52
0.51
0.50
0.49
0.49 0.50 0.51 0.52 0.53 0.54 0.55 0.56 0.57
Experimental gasoline yield, wt%
(b)
0.22
0.21
Calculated gas+coke yield, wt%
0.20
0.19
0.18
0.17
0.16
0.15
0.15 0.16 0.17 0.18 0.19 0.20 0.21 0.22
Experimental gas+coke yield, wt%
(c)
0.78
0.76
Calculated conversion, wt%
0.74
0.72
0.70
0.68
0.66
0.64
0.64 0.66 0.68 0.70 0.72 0.74 0.76 0.78
Experimental conversion, wt%
Figure 2.4 Experimental vs calculated values: (a) Gasoline yield, wt.%; (b) Gas + coke yield, wt.%;
(c) Conversion, wt.%. Model 1 (€), Model 2 (▲), Model 3 (×), Model 4 (o), and Model 5 (+) (Ancheyta and Lopez
2000 / Scientific Electronic Library Online (scielo)).
2.3.2 Deactivation Models by Metal Deposition

Oyekunle and Ikpekri (2004) and Oyekunle et al. (2005) developed a model to describe catalyst
deactivation by metal deposition during residuum HDS. They used three types of catalysts with
macro-, micro-, and random pore structures. It is assumed that the HDS reaction occurs in macro-
and micropores, and there are interconnections between them. An idealized pore model of Wheeler
(1951) was applied making use of an average value for tortuosity factor for interconnecting pores.
The model considers activity decays with time both linearly and nonlinearly. To determine the
catalyst activity decline with TOS, the following equations had to be solved:
∂ 2 αD ∂αD
2 = Φ αD + ∂τ
2 2
2 19
∂βD D
d ϵkSA C F M MS γ D 2
= 2 20
dt ρm r
Oyekunle and Ikpekri (2004) tested the model using data of hydroprocessing of Maya heavy crude
oil published by Ancheyta et al. (2002, 2003). The catalyst life was estimated, and the total activity
loss was predicted to occur after 462, 316, and 150 days for macro-, random, and micropore systems.
Bourseau et al. (1985) and Toulhoat et al. (1990) presented a deactivation model for hydrodeme-
tallization (HDM) catalysts, which considers the catalyst’s textural properties, its initial activity, the
catalyst size, the feed reactivity, and includes the concept of ultimate storage capacity (USC), which
is defined as the mass of metal accumulated in the catalyst particle, relative to the unit mass of fresh
catalyst, until the catalyst activity became zero. The USC can be expressed as follows:
Ωρm ϵ0 C m
USC = 2 21
ρcat
where ϵ0 is the the catalyst’s initial porosity, ρcat is its initial grain density, ρm is the deposited metal
sulfides’ true density, Cm is the metal content on the catalyst, and Ω is the fraction of initial porosity
occupied by the solid deposited on catalyst.
Pereira (1990) proposed a model that describes the influence of feed and catalyst properties on the
demetallation performance of a HDT catalyst. This model includes the metal distribution factor, θM
(Tamm et al. 1981).
1 1
θM = C m ξ dξ Cmax
m ξ dξ 2 22
0 0
where Cm is the local concentration of metal deposit, Cmax

m is the maximum concentration of metals,
and ξ is the fractional distance from the center of the pellet. For demetalation reactions, θM is small
when the intraparticle metal deposit profile is not uniform, and approaches unity when metals are
uniformly deposited throughout the catalyst particle. A set of mathematical equations was
developed, and their solution was obtained for first-order HDM kinetics. A series of correlations
was developed which involves the change of parameters, such as distribution factor, pore radius,
the Thiele modulus, reaction rate, and so on, with TOS (Pereira 1990).
Dautzenberg et al. (1978) developed a model for describing the deactivation behavior of residue
desulfurization catalysts. The deposition of metal compounds was the principal parameter to
describe the deactivation phenomena. They assumed that in residue hydrodesulfurization (RDS)
two reactions take place in parallel: one-and-a-half order HDS; and pseudo-first-order vanadium
removal. Sulfur is removed as H2S and ends up in the gas stream, while the metal removed is depos-
ited on the walls of the catalyst pores as sulfide, diminishing the original desulfurization activity,
due to the reduction of the pore radius (rA) with time (t) according to the following:
∂r A
= k v Cv V dep 2 23
∂t
where kv, Cv, and Vdep are the reaction rate constant for vanadium removal, molar concentration of
vanadium, and deposit volume per mole of vanadium, respectively.
The degree of deactivation (θ) is related to the volume deposited per unit surface area per unit
time represented by the following equation:
t
k v V dep
θ= cv0 dt 2 24
r p0 0
where rp0 and cv0 are the pore radius of fresh catalyst and molar vanadium concentration outside
the catalyst particle, respectively. This model does not consider configurational diffusion effects and
the corresponding changes in diffusivity of metal-bearing molecules with pore diameter.
The experiments were performed at fixed standard temperature and pressure in bench-scale
equipment. Two feedstock were used: Caribbean long residue and Middle-East long residue.
Catalysts of Ni/Mo and Co/Mo type with different particle size, shape bulk density, and support
were tested. The decline in metal-removal activity of catalyst used in stirred-tank reactor experi-
ments was evaluated using a vanadium weight balance. Figure 2.5 shows the metal content of
the product (vanadium) as a function of TOS.
Arbabi and Sahimi (1991a) proposed a percolation model catalyst deactivation, where the porous
catalyst is represented by a network of interconnected pores. Because of continuous deposition of
solid products on the surface of the pores, the morphology of the catalyst’s pore space changes with
time. The effective radius of the pores is distributed according to a probability density function
represented by the following expression:
2
r p0 − r min − 1 r p0 − r min
f r = exp 2 25
r a − r min 2 r a − r min
3
(Vf/Vp)2
0
0 200 400 600 800 1000
Time (hours)
Figure 2.5 Vanadium removal during hydrodesulfurization of Caribbean residue with Co/Mo/Al2O3 (●)
and Middle-East long residue with Ni/Mo/Al2O3 (▲) (Adapted from Dautzenberg et al. 1978).
where rmin is the minimum pore radius and ra is the average pore radius. The model was applied to
hydrodemetalation reaction. The results for the deposition profile were found to be in good agree-
ment with the experimental data (Arbabi and Sahimi 1991b).
Rajagopalan (1979) developed a model for predicting the total life of catalyst during first-order
demetallation reaction considering only the period following the initial fast coking. In this model,
the demetallation activity decays with time is an almost linear function, due to the influence of the
restricted diffusion of the large reactant species in the pores. The model considers the demetallation
in a single cylindrical pore, where the rate of metal deposit per unit length of the pore is as follows:
dCm
= ζ2πr p0 ck fm M MS 2 26
dt
They found that the catalyst becomes inactive when the pore month is reduced to the size of the
organometallic molecule, this behavior can be represented by the following equation:
ρmd r m 1 − λ0
tm = 2 27
ζk fm CF M MS λ0
where ρmd and rm are the metal deposit density and the radius of molecule, respectively, MMS is the
molecular weight of metal deposit, kfm is the demetallation reaction rate constant, ζ is the number
of metal atoms per reactant molecule, CF is the concentration of reactant, rp0 is the pore radius of
fresh catalyst, tm is the maximum lifetime of the catalyst, t is TOS (>50 hours), Cm is the concen-
tration of metal, and λ0 is the ratio of molecular to pore radius.
The model was used for predicting pore sizes, which yield the optimal initial activity and optimal
lifetime activity. Figure 2.6 describes the demetallation rate per unit volume of a slab catalyst as a
function of pore radius for several ages of the catalyst. The graph indicates that the pore radius cor-
responding to the optimum initial activity is 3.1 × 10−9 m. For pores smaller than the optimal
radius, increasing the pore radius improves the catalyst performance as this increases the initial
activity; however, for pores larger than the optimal size, an increase in pore size reduces the initial
activity but increases the total life.
16
Dmetallation rate [10–7] (moles/m3s)
14
12
10
0
0 2 4 6 8 10 12
Pore Radius [10–9] (m)
Figure 2.6 Dependence of demetallation rate on catalyst pore size and time-on-stream: 0 days (__),
8.22 days (---), 18.5 days (…), and 26.72 days (- -) (Adapted from Rajagopalan 1979).
Khang and Mosby (1986) proposed a model to describe deactivation of hydroprocessing catalyst
due to slow and uniform deposition of reaction products in their microporous, mostly metal
compounds. The model does not include mass-transfer restriction in the main channels of the
pores and considers the creation of active sites by metal compounds in the deposited layers. The
remaining pore diameter (Rpd) can be represented using the fraction of the volume occupied by
metal deposited,
1 2
Rpd = 1 − M d M 0 2 28
where Md is the actual amount of metal deposited and M0 is the maximum amount of metal depos-
ited when the pore is filled completely. From Eq. (2.28), the term M d M 0 as a function of time is
2
M d M = 1 − 1 − td h 2 29
0
2
hz = kz Dz r p0 2 2 30
2
h0 = k0 Dz r p0 2 2 31
h = h20 hz 2 32
where hz, h0, and h are dimensionless factors, Dz is the diffusivity of metal, Cm is the metal concen-
tration, rp0 is the pore radius of fresh catalyst, k0 and kz is the initial demetallation constant and the
demetallation rate constant for the additional active sites created by metal deposits, and td is the
dimensionless time defined as follows:
k0 Cm t
td = 2 33
2ρL
The model is appropriate for predicting deactivation curves of first-order HDM and second-order
HDS reactions for less than 50% of pores filled with metals.
Ahn and Smith (1984) formulated a model, which considers two causes of deactivation in HDS
catalysts, partial poisoning of the interior pore surface for HDS and pore mouth plugging for HDM
first-order reactions; these two mechanisms of deposition are described in terms of two parameters,
the Thiele modulus Øm for metal deposition and a partial poisoning parameter, Θ. These para-
meters are defined as follows:
1 2
ρcat k ms
Ø m = r0 2 34
Dm0
Θ = Q0 M p ρcat 2ρmd ϵp 2 35
where Dm0 is the effective diffusivity of the organometallic compounds at zero time, r0 is the catalyst
particle radius, km is the rate constant for demetallizaion, ρcat is the apparent density of catalyst
particle, Q0 is the concentration of deposited metals corresponding to a monolayer, ϵp is the porosity
of catalyst particle, and Mp and ρmd are the molecular weight and density of metal deposited,
respectively.
The rates of demetallization (ηm) and desulfurization (ηs) are represented by effectiveness factor η,
which is related to the rate of organometallic entering to the pore with intraparticle diffusion resist-
ance and without intraparticle diffusion resistance.
2
rA dCmp
r 20 Dm0
r p0 dr p
ηm t = 2 36
1 3
r ρ k ms Cm0
3 0 cat
rp r0 2
rA
r p 2 ρ bk ss Cs dr p + ρr p 2 k ss Cs dr p
0 ri r p0
ηs t = 2 37
1 3
r ρ k ss C s0
3 0 cat
A restricted effective diffusivity was employed for metal compounds on spherical pellet by the
next equation:
2
εD rm
Deff = 1− 2 38
τ rA
Haynes and Leung (1983) formulated a catalyst deactivation model for HDT processes of heavy
fractions. The model considers the deposition of poison on the catalyst and the effects of this dep-
osition on catalyst activity; furthermore, it incorporates the combined effects of pore plugging and
active side poisoning. Catalyst deactivation is represented by effectiveness factor, which is defined
as the ratio of the observed rate at any TOS to the rate at time zero in the absence of significant
resistance to diffusion:
∂C
Dp X 0,t X 0, t
η= ∂X 2 39
X 0 ρcat kC AS
where the effective diffusivity is represented by the relation used by Spry and Sawyer (1975):
DP = DA np πr A 2 τf 2 40
The effect of pore plugging in catalytic deactivation is represented by the following equation:
αp = 2W s ρp r 0 2 41
where the value of Ws corresponds to the mass of adsorbed poison at complete site coverage, ρp is
the adsorbed poison density, and r0 is the initial pore radius. This approach may represent a
diffusion limited coke type of deactivation.
Galiasso (2007) developed a model based on the effect of recycling the unconverted bottom on
catalyst deactivation as a way to improve the hydrocracking conversion. The author assumed that
the six-month test represents the catalyst deactivation by metal loading and neglects the coke
contribution in the inner part of the particle because its content was approximately constant after
10 days on stream. The model considered that the recycle introduces an additional deactivation
through the insoluble material deposition on catalyst in each pass that reduces the incoming
molecules’ diffusivity. Two sites are considered for catalyst deactivation. A simplified type of the
Levenspiel model is used (HCK on Site 1, HDS and HDM reactions on Sites 2):
da1
= k d1 Cm a1 2 42
dtc
da2
= k d2 Cm a2 2 43
dtc
70
60
50
Conversion, wt%
40
30
20
10
0
0 30 60
Days on stream
Figure 2.7 Activity as a function of time–on-stream (Galiasso 2007). For ( ) HCK experimental and
( ) HCK model, ( ) HDS experimental and ( ) HDS model, ( ) HDM experimental, and ( ) HDM model.
where kd1 and kd2 are the deactivation constants for Sites 1 and 2, Cm is the metal loading, and tc is
the solid residence time.
The experiments were performed at ebullated-bed reactor with a NiMo/Al2O3 catalyst at the fol-
lowing operating conditions: temperature of 420 C, pressure of 140 kg/cm2, and liquid-hourly
space-velocity (LHSV) of 0.5 h−1 during two-month of TOS. The feedstock was a 500 C+ residue from
extra-heavy crude oil. The residue has 5.02, 0.93, 22, and 21 wt.% of sulfur, nitrogen, Conradson
carbon, and C7 asphaltene content, respectively, and 714 ppm of vanadium. Figure 2.7 shows the
results of the model prediction for the aforementioned operating conditions. The model was able
to accurately predict the initial activities, with approximately 5% deviation of the 60-day run lengths.
Shah et al. (1976) developed empirical relationships for studying the life cycle of catalyst in a
residual oil trickle-bed HDS reactor. Catalyst activity functions for desulfurization and demetalli-
zation were developed from pilot plant data. The catalyst activity for sulfur and metal removal reac-
tions decreases with time due to both coke and metal deposition (Newson 1975), however, the
authors found that coke deposition is important only after a short TOS, so that most of the activity
loss is a result of continued metals deposition, for this reason activity functions, φsf and φm, are
dependent only on the metals concentration (cm) of the catalyst represented by the following
equations:
ym
φsf = 1 − ya Cm 2 44
yn
φm = 1 − y b Cm 2 45
where ya, yb, ym, and yn are fitting parameters.
2.3.3 Deactivation Models by Coke and Metal Deposition

Kodama et al. (1980) proposed a deactivation model in which the deposits of coke and metals
(mainly V) cause both active site poisoning and pore plugging on catalyst. Second-order kinetics
was used to describe both HDM and HDS reaction rates. The amounts of coke and vanadium depos-
ited were expressed as follows:
t
Qv = 100 dv dt ρcat 2 46
0
t
Qc = 100 dc dt ρcat 2 47
0
In dimensionless forms, these two equations can be rearranged as follows:

φv = Qv 100 Pv ρv 2 48
φc = Qc 100 Pv ρc 2 49
The results showed that coke deposition on the catalyst never increased, but if decreased
abruptly, then coke deposition may be reversible (Eq. 2.50), indicating that reducing the area of
the active surface (A0C) gives a negative value of coking rate, causing a decrease in the content
of coke on the catalyst. Thus vanadium deposition increases, ejecting the coke deposited previously
(Eq. 2.51).
r c = ρcat A0c k c1 − ρcat A0 k c2 Ch Qc 2 50
rV = ρcat A0V ηV k V Ch C 2V 2 51
Isothermal HDS test was made under constant reaction conditions in an up-flow reactor with
the following characteristics: inner diameter of 25.4 mm, thermowell diameter in the center of
the reactor of 6.35 mm, and catalyst bed volume of 400 cm3. An Iranian heavy vacuum residue
was used as feedstock at the following operating conditions: pressure of 140 kg/cm2, LHSV of
0.46 h−1, H2/oil ratio of 5614 scf/bbl, and temperature of 385 C and 400 C. The catalyst used
was of Co-Mo type with 0.601 cm3/g and 155 m2/g of pore volume and specific surface area,
respectively. Figures 2.8 and 2.9 show the calculated results of HDS and HDM respectively,
compared with the experimental data.
Skala et al. (1991) suggested the following model of catalyst activity for sulfur and metal removals,
which decreases over time due to both coke and metal deposition. Coke deposition was important
only during the first part of the run, so that most of the activity loss is the result of continued metal
deposition. The model is based on the three reactions (HDS, HDO, and HDM), each of them being
pseudo-first-order. The catalyst activity is dependent upon both coke and metals depositions on the
100
80
60
% HDS
40
20
0
0 2000 4000 6000 8000 10000
TOS (hours)
Figure 2.8 HDS results reported by Kodama et al. (1980), experimental data: (●) 385 C and (▲) 400 C,
simulated data (__) 385 C and ( ) 400 C. Kam et al. (2005) experimental (––) and simulated (♦) data.
100
80
60
% HDM
40
20
0
0 400 800 1200 1600 2000 2400
TOS (hours)
Figure 2.9 HDM results reported by Toulhoat et al. (2005), Catalyst A: experimental (●) and simulated (-) data.
Catalyst B: experimental (■) and simulated (---) data. Kodama et al. (1980), experimental data: (♦) 385 C and
(▲) 400 C, simulated data (---) 385 C and ( ) 400 C.
catalyst surface. The catalyst activity decline was calculated with the equation proposed by Shah
et al. (1976).
06
φ = 1 − 1 58 Q0 5 2 52
The concentration of coke and metal (Q) was calculated with the following equation:
Q = 0 185 C coke,0 + C ash,0 Lt w 2 53
where Cash,0 is the initial concentration of metals determined as ash of the used oil.
Since the catalyst activity decline is dependent on the concentration of deposits, it was considered
obvious that catalyst deactivation is related to the decrease in bed porosity. Skala et al. (1991)
assumed that the change in catalyst bed porosity (ϵ) is related to the catalyst activity by an equation
similar to Eq. (2.52):
05 06
φ = 1 − 1 58 ϵ0 − ϵt 2 54
for the calculation of the decrease in the bed porosity for time different to zero,
2
ϵt = ϵ0 − 1 − φ1 67 1 58 2 55
Newson (1975) proposed a pore-plugging model for residuum hydroprocessing in an axial-flow

TBR to describe catalyst deactivation. The model has three stages for determining the loss in
porosity of the catalyst: The first stage (θC1) by fast coke deposition, which takes up one-third of
the total porosity; second stage (θC2) by slow coke deposition; and third stage (θMS) by metal sulfides
plugging. These three stages are represented by the next equation:
θT = θC1 + θC2 + θMS 2 56
The rate of metal sulfides deposition, which depends on the properties of the feed and catalyst,
and the operating conditions used are as follows:
L ρ Δl
C F e − k1 τ 1 − e − k1 τ
l
r MS = MW MS LHSV L L 2 57
Δl ρcat
Coke contribution to the pore plugging rate is represented by Eq. (2.58), which represents the
difference between fast and slow coke, and neglects metal deposited in the first 50 h of operation.
The slow coke is assumed to remain approximately constant after the first hours-on-stream:
Csc + W f Rc2 t − 50
C0 = 2 58
1 + W f Rc2 t − 50 + W f RMS t − 50
The rate of deposition of coke plus metal sulfides is estimated from the kinetics of demetallation
and desulfurization first-order reactions.
Alvarez et al. (2011) developed the following relationship for describing the catalyst activity
decline during the hydroprocessing of heavy oils, caused by coking reaction at start-of-run
(SOR) conditions and metal deposition during the whole cycle, the contribution of these two
processes to the overall catalyst deactivation was expressed as follows:
Ψ z, t = Ψ coke t + Ψ metal z, t 2 59
where Ψ coke and Ψ metal stand for the deactivation functions for coking at time t ≤ 100 h and local
metals buildup (Ni + V) at time t, respectively. To represent the effect of these two processes for the
overall catalyst deactivation, the authors proposed the following empirical expression:
1 γ
φ z, t = β
− X MOC z, t 2 60
1 + δt
The initial activity decline Ψ coke during the first 100 h is represented by the first term of Eq. (2.60),
as function of time. At t = 100 h, it is considered that the initial period is terminated and the value of
Ψ coke is kept constant for the rest of the operation cycle. The author proposes that the best way to
describe this process is to link this function to a coking agent instead of time; unfortunately, there
was no experimental information available that provided insight into this mechanism. For this
reason, the application of such functionality is restricted to the experimental operating conditions.
Nevertheless, SOR data obtained with different heavy crudes at different operating conditions
(Centeno 2008) indicated that the HDS and HDM activity decay in this period interestingly follows
the same pattern even in such a wide range of conditions. This behavior is represented in
Figure 2.10 as the change in the deactivation functions of HDM and HDS with respect to the change
in time (ΔφHDM/Δt, ΔφHDS/Δt).
0.025 0.015
0.02
0.01
Δφ HDM/Δt
Δφ HDS/Δt
0.015
0.01
0.005
0.005
0 0
0 20 40 60 80 100
TOS (hours)
Figure 2.10 HDS and HDM deactivation curves at SOR, experimental data for HDS at (○) 385 C and (Δ)
400 C, and simulated data (---). HDM experimental data at (●) 385 C and (▲) 400 C, and simulated
data (-) (Alvarez et al. 2011).
The second term represents the deactivation function due to metals deposit. The amount of metal
deposited is expressed in terms of XMOC, defined as the ratio between the local concentration of
metal-on-catalyst (MOC) at time t and the specific maximum metals uptake of each catalyst.
MOC is expressed in wt.% and obtained from the following metals mass balance equation:
dMOC z, t
= U MOC r HDM z, t 2 61
dt
Where rHDM is the local rate of HDM at time t, and UMOC is a unit conversion factor. This equation
allows for determining a deactivation profile with respect to time t and axial position z. Experimen-
tal data reported elsewhere (Centeno 2008) were used to represent the deactivation of the catalyst
caused by these two mechanisms. Kinetics and catalyst aging data were obtained from bench-scale
tests using a heavy crude oil (13 API) at the following operating conditions: LHSV of 0.5 h−1, pres-
sure of 70 kg/cm2, H2/oil ratio of 5000 scf/bbl, and temperature of 380 and 400 C.
Elizalde and Ancheyta (2014a) applied a deactivation model that considers the three stages of
deactivation, SOR, MOR, and EOR during HDT of heavy oil in an isothermal bench-scale plant.
The model is based on that reported by Idei et al. (1995, 1998, 2003) and Elizalde and
Ancheyta (2014b).
φ = φ A + φD 2 62
The first term of Eq. (2.62) represents the catalyst decay during SOR with absence of diffusional
resistances due to deposits, in which it is considered with different active sites in the catalyst, sites
Type I and sites Type II. The former correspond to the sites which are deactivated at fast rate, and
the latter correspond to those that are deactivated slower than sites Type I.
After a series of deductions and substituting initial conditions, Elizalde and Ancheyta (2014a)
represented the first term of Eq. (2.62) as follows. Detailed deduction can be found elsewhere
(Ancheyta 2016).
k1 α1 − α2 t k0
ln φA = ln 1 + e − ln + α1 t 2 63
k2 k1
The loss of catalyst pore volume and pore diameter reduction due to the metal sulfide and the
aging coke at MOR provokes diffusional resistance, this fact is represented by the second term
of Eq. (2.62) with the following expression:
1 1
ln φD = ln 2 64
2 t
n20 + 1 − n20 1−
t∞
Eqs. (2.63) and (2.64) are substituted in Eq. (2.62) and the following final expression is obtained:
k1 α1 − α2 t k0 1 1
ln φ = ln 1 + e − ln + α1 t + ln 2 65
k2 k1 2 t
n20 + 1 − n20 1−
t∞
Experimental data were obtained from a fixed-bed catalytic reactor with internal diameter of
2.54 cm. Details of the bench-scale plant were given elsewhere (Ancheyta et al. 2001). Commercial
NiMo/Al2O3 catalyst was used, its properties are the following: specific surface area of 175 m2/g,
pore volume of 0.56 cm3/g, and average pore diameter of 127 Å. The heavy oil was hydrotreated
under the following reaction conditions: temperature of 400 C, pressure of 70 kg/cm2, LHSV of
1 h−1, and H2/oil ratio of 5000 scf/bbl. Figure 2.11 shows a comparison of experimental data
of deactivation for hydrodevanadization (HDV) together with simulated profiles.
–0.1
–0.2
In (Φ)
–0.3
–0.4
–0.5
–0.6
0 100 200 300 400 500 600
TOS (hours)
Figure 2.11 Simulation profiles of deactivation for HDV for the range of experimental TOS: (symbols)
experimental, (line) simulation (Elizalde and Ancheyta 2014a).
Elizalde and Ancheyta (2014b) used a continuous lumping kinetic model to represent both the
reaction kinetics and catalyst deactivation of hydrocracking of heavy oil. The model involves site
coverage and pore blockage as the main causes of catalyst deactivation, these functions were repre-
sented by power law, considering first-order dependence on activity (Corella et al. 1985). According
to Monzón et al. (2003), taking into account the considerations mentioned earlier, modeling the
deactivation as TOS function can be expressed as follows:
φ = φs + 1 − φs e − αt 2 66
where φ is the catalyst activity, α is the parameter decay, and φs is the steady-state catalyst activity.
In previous experiments (Ancheyta et al. 2002, 2003), a change in activity has been observed in
two intervals of TOS. In the first interval, the change in activity is during the first hours of TOS by
site coverage due to coke precursors. After certain time, simultaneous sites coverage and pore
blockage provoke loss of activity. According to this, the following function defined by intervals
is proposed:
φ1 0 ≤ t ≤ tb
φ= 2 67
φ1 φ2 tb < t
where φ1 was considered to be equal to Eq. (2.66) and φ2 was defined as follows:
φ2 = φ s + 1 − φs e − ω t − t b 2 68
In Eq. (2.68), tb is the time that marks the change in activity during TOS. Experimental informa-
tion was obtained in a CSTR operated under the following conditions: total initial pressure of 100
kg/cm2, LHSV of 0.75 hours−1, hydrogen-to-oil ratio of 5000 scf/bbl, temperature of 380 C and
400 C, and 1000 rpm of stirring rate. An extruded catalyst was employed and its properties were
the following: nominal size of 1/18 in., Ni content of 0.58 wt.%, Mo content of 2.18 wt.%, specific
surface area of 197.2 m2/g, pore volume of 0.85 cm3/g, and average pore diameter of 172.6 Å.
The first part of curve in Figure 2.12 represents the loss of catalyst activity by site coverage, and
the sharp changes on reactivity are attributed to some catalyst pore constrictions together with
deactivation by coke precursors.
2.5
1.5
kmaxΦ
0.5
0
0 50 100 150 200 250
TOS (hours)
Figure 2.12 Effect of time-on-stream on kmax model parameter at (●) 380 C and (▲) 400 C (Elizalde
and Ancheyta 2014b).
Toulhoat et al. (2005) and Kressmann et al. (1998) developed a model for residue upgrading using
several fixed-bed reactors in series. In this model, catalyst deactivation is represented by two
mechanisms: coke and metal depositions. They represented the local concentration of metals
(Ni, V) in solid phase with the following:
dCm r, z, t Mm
= ρm r, z, t 2 69
dt ρcat j
The local coke deposition caused by the reversible dehydrogenation of the asphaltenic material
and its first-order reaction rate with respect to a driving force equal to the difference between actual
and equilibrium local coke concentrations on catalyst surface are represented by the following
equation:
dC0 r, z, t E±
= k coke,k r, z, t Sk r, z, t exp coke,k × C coke r, z, t − C 0 r, z, t 2 70
dt RT
To express the rate of hydroprocessing reactions, a pseudo-Langmuir–Hinshelwood law was
used, the next equation is a simplified form of this equation for HDS:
± C 1asph r, z, t
MO r, z, t = k r, z, t S r, z, t exp E RT 2 71
1 + K ads T z, t C ∗asph r, z, t
Catalyst tests were run on a bench-scale unit, which has an up-flow trickle-bed tubular reactor.
Two catalysts with various textural parameters were used: A is a NiMo/Al2O3 HDM catalyst, which
is characterized by a continuous pore distribution between macropores and mesopores, and B is a
Co/Mo/Al2O3 HDS catalyst that is characterized by a mesoporous monomodal pore distribution.
A feedstock (Ni and V content of 196 and 1150 ppm, respectively, and 5.28, 0.67, 13, 12 wt.% of
sulfur, nitrogen, asphaltenes, and CCR contents, respectively) was used to evaluate the evolution
of HDM and HDS performances vs TOS. This feed was injected until complete deactivation of the
catalyst for evaluating metal retention capacity at the following operating conditions: total pressure
of 153 kg/cm2, catalyst volume of 40 cm3, LHSV of 1 hour−1; H2/oil ratio of 6737 scf/bbl,
temperature of 390 C, and the TOS was variable.
Figure 2.9 shows the comparisons of simulated and experimental deactivation curves for HDM
for catalysts A and B.
Chao et al. (1991) proposed a mathematical model that describes catalyst deactivation in a com-
mercial RDS unit. The model takes into account metal and coke deposition, and the decrease of
effective diffusivities during HDS and HDM reactions. The reaction rate coefficients of desulfuri-
zation and demetallization reactions, varying according to deposited metal amounts, can be repre-
sented as follows:
k fs = ks Λ when Δr pm ≥ Δr pc 2 72
1 Δr pm
k fs = ks 1 + −1 when Δr pm < Δr pc 2 73
α Δr pc
km
k fm = when Δr pm ≥ Δr pc 2 74
Γ
1 Δr pm
k fm = km 1 + −1 when Δr pm < Δr pc 2 75
α Δr pc
Δr pm = r p0 − r pm 2 76
where Λ and Γ are the relative reaction rate coefficients of fresh catalyst and deposited metal for
desulfurization and demetallization, rp0 is the pore radius of fresh catalyst, rpm is catalyst pore
radius which neglects coke thickness, and Δrpc is the thickness of monolayer deposited metal.
The coke deposited covers the active sites causing them to become inactive, and this fraction of
surface is assumed to be proportional to the relative volumes of carbonaceous compounds and cat-
alytic pore volume. In this model, catalyst decay by coke deposition is represented as follows:
1 − x s = α c Qc 2 77
2 2
Δr pm rp
Qc = − 2 78
r p0 r p0
where Qc is the volume of carbonaceous compounds, xs is the fraction of specific surface area that
is left uncovered by coke, and αc is a proportional constant. The decrease of catalyst effective
diffusivity is represented by a second-order equation with respect to pore radius:
2
rp
Deff = D0 2 79
r p0
Kam et al. (2005) developed a kinetic model of rapid initial catalyst deactivation for HDS, HDM
(by nickel and vanadium), and hydrodeasphaltenization reactions. An exponential decay function
was more appropriate, and moreover, as the HDS catalyst is protected by the HDM catalyst, the
deactivation is only by coke. Therefore, they proposed an empirical decay function (Ψ i) where
the changes in the reaction rate coefficient of the main reactions can be represented as follows:
k Ai = k i + 0 5Δk i Ψ i 2 80
n
C p,i Target
Ψi = − e C p,i Product from previous reactor

2 81
Δk 0,i = k i t = 12 h − k i 2 82
dc
cni
k 0,i t = 12 h = 2 83
dt
For i = 1 to 4
10
8
Asphaltenes (wt %)
0
0 2000 4000 6000 8000 10000
TOS (hours)
Figure 2.13 Comparison of model predictions (-) and experimental (●) data of asphaltenes with
processing time (Adapted from Kam et al. 2005).
where kAi and ki are apparent reaction rate coefficients in the initial rapid deactivation period for
species i estimated by Eq. (2.80) and determined experimentally, respectively; i is the species index
(1 for sulfur, 2 for vanadium, 3 for nickel, and 4 for asphaltenes); Cp and Cf are concentrations of the
product and the feed, respectively; and n is the reaction order. The initial rapid deactivation period
is completed when the decay function for sulfur becomes Ψ i ≤ 0.01. They found that the amount of
coke deposited on the catalysts increased rapidly and reached about 15 wt.% within 12 hours. After
this period, the rate of coke formation slowed down reaching around 20 wt.% after 240 h.
The model was used to simulate the pilot plant performance and predict the catalyst deactivation
beyond the 4000 hours of pilot plant test at the following operating conditions: temperature of
375 C, pressure of 122 kg/cm2, LHSV of 1 hours−1, H2/oil ratio of 3200 scf/bbl, and TOS of 9500
hours. The predicted sulfur and asphaltene concentrations in the product oil beyond the 4100 hours
are illustrated in Figures 2.8 and 2.13, respectively. The pilot-plant sulfur and asphaltenes data clearly
follow the model predictions.
2.4 Development of Models for HDT Catalyst Deactivation
2.4.1 Important Issues

It has been found from the literature review that deactivation models for HDT catalysts have been
developed mainly for heavy oil applications, which is certainly the case that exhibits the most severe
loss of catalyst activity. The deactivation models take into consideration the activity decay as a
function of coke deposition, metals deposition, and both, which are the main agents (coke, metals)
that deactivate the catalyst. Other models use catalyst activity as a function of TOS as the
experiments are easier and cheaper to carry out.
Conducting experiments for developing models based on coke-on-catalyst (COC) or MOC
requires a number of tests to characterize the spent catalyst, which is developing one test at certain
value of TOS, another test at longer TOS, and so on. In each test, the different stages of evaluation
must be performed, that is, catalyst loading, catalyst activation, operation at defined conditions, and
catalyst unloading. It means that for obtaining one point (content of coke and metals on spent
2.4 Development of Models for HDT Catalyst Deactivation 79
catalyst), a complete experimental run must be done. The following run must be conducted at
longer time, which for having enough experimental information for developing a robust deactiva-
tion model can be as long as several weeks. All these experiments will be costly and time-consum-
ing, that is why researchers think twice when intending to do them and decide either to develop
models based on TOS or to conduct experiments at shorter time to develop models based on COC
and early stage of MOC.
The deactivation models reported in the literature have different forms, terms, and parameters
for representing activity loss. Usually, they include parameters that are calculated by fitting
experimental data; however the applicability of these models is only valid for the range of oper-
ating conditions at which they were derived. Tables 2.1–2.3 summarize different deactivation
Table 2.1 Deactivation parameters for deactivation models based on coke deposition and TOS.
Authors Model Reactor Feedstock Parameter
Froment and k = kηφ(C0)

Bischoff (1962)
φ = exp(− C0)
1
φ β
1 α C0
Voorhies (1945) C0 = A(τ/60)V Fixed-bed East Texas gas oil A = 0.52

V = 0.38
Fluid catalyst Tinsley gas oil A = 0.24

cracking V = 0.53
Rashidzadeh φij e − αij Life Trickle-bed Vacuum gas oil αDK = 4.13E−05
et al. (2010) αFG = 8.43E−05
αFN = 8.28E−05
αNG = 9.12E−05
αKN = 1.59E−04
Richardson et al. C0 = Cmax[1 − exp (−Kw)] Batch Athabasca bitumen feed
(1996) CSTR
Ancheyta and Model 1 Fixed-bed Vacuum gas oil Model 1

Lopez-Izunza φ e − kd t kd = 0.0875
(2000)
Model 2 Model 2
1 kd = 1.2864
φ
1 kd t
Model 3
Model 3 B = 0.7888
φ = A t−B
Model 4
Model 4 B = 1.5741
φ = (1 + Gt)−B G = 0.5889
Model 5 Model 5
a a = 0.9335
φ
pq 1 d tg g = 1.3576
Table 2.2 Deactivation parameters for deactivation models based on metal deposition.
Model Reaction Reactor Feedstock Parameter
Oyekunle ∂ 2 αD ∂αD HDM Fixed-bed Heavy Maya η = 0.2

et al. (2004, = Φ2 αD 2 + crude
∂βD 2 ∂τD 0 < βD < 0.6
2005)
km = 4.5 × 10−12
kg/s m
Toulhoalt Ωρm ϵ0 Cm First-order CSTR Deasphalted Ω = 4.26
USC =
et al. (1990) ρcat HDM Boscan crude
oil
ym
Shah and φsf = 1 − ya First-order Trickle- Petroleum ya = 1.58
Cm
Mhaskar yn
HDM-HDS bed residue yb = 1.58
(1976) φm = 1 − yb Cm ym = 0.8
yn = 0.6
Dautzenberg k v V dep t First-order Fixed-bed Caribbean long ηm= 0.5
et al. (1978) θ= cv0 dt HDM residue
r p0 0
Middle East
long residue
Rajagopalan dm First-order Øm= 3

= ζ2πr p0 k fm M MS
(1979) dt HDM ϵp = 0.5
km= 5 ×
10−13 m s−1
Khang and 1 2 First-order Trickle- Vacuum
Rpd = 1 − M d M0
Mosby (1986) HDM bed residues
Ahn and ρcat k ms 1 2 First-order Øm = 3–9

Smith (1984) Ø m = r0 HDM Θ = 0.025–0.075
Dm0
Θ = Q0Mpρcat/2ρmdϵp First-order
HDS Øs = 1
b = 1, 4, 9
Haynes and αp = 2Ws/ρp r0 τf = 3
Leung (1983) η = 0.7,0.2, 0.1,
0.05
Galiasso da1 Ebullated- 500 C+ residue η = 0.83–0.91
= k d1 C m a1
(2007) dtc bed kd1 = 1E + 15
da2 kd2 = 1E + 17
= k d2 C m a2
dtc
models based on coke deposition, metal deposition, and coke/metals deposition, respectively.
Information about the values of parameters, type of reactor, type of reaction, and type of feedstock
obtained by different authors is also included. More details about experimental reaction condi-
tions and additional values of parameters can be found in the corresponding references, as well as
shown in the tables.
Metals deposition is present preferentially near the entrance of the pores of the catalyst (Kodama
et al. 1980) resulting in the blocking of the access to the active sites, the level of blocking depends, of
course, on several factors among them type of support, type of active metals, size of the pores, pore
size distribution, amount and size of asphaltenes, amount of asphaltenic and non-asphaltenic
metals, etc. For these reasons, deactivation models based on metal deposits take into consideration
some properties of the catalyst pore (volume, structure, and size distribution) and intraparticle
Table 2.3 Deactivation parameters for deactivation models based on metals and coke depositions.
Authors Model Reaction Reactor Feedstock Parameter
Kodama et al. (1980) rc = ρcat A0c kc1 Second-order HDM- Up-flow Iranian heavy vacuum residue k s = 1 22 × 1010 e − 31 3 × 10
3
RT
− ρcat A0 kc2 Ch Qc HDS
k v = 1 28 × 1010 e − 24 2 × 10
3
RT
r V = ρcat A0V Ef V k V C h C 2V
k c1 = 3 × 10 − 2 e − 20 × 10
3
RT
103
k c2 = 5 35 × 10 − 5 e − 10 × RT
Skala et al. (1991) φ = [1 − 1.58( ϵ0 − ϵt) ] 0.5 0.6
First-order HDM-HDS Trickle-bed yn = 0.6
Newson (1975) θT = θC1 + θC2 + θMS First-order HDS Trickle-bed Residuum Em = 30, Es = 40
ηm = 0.2, ηs = 0.6
m = 1, 2
Alvarez et al. (2011) Ψ (z, t) = Ψ coke(t) + Ψ metal(z, t) HDM Fixed-bed Atmospheric residue δHDS = 3.81
HDS βHDS = 0.06
γ HDS = 0.34
η0 = 0.7
Elizalde and Ancheyta (2014a) φ = φA + φD HDM Fixed-bed Heavy crude oil (13 API) α1 = 9.0 × 10−3
HDS α2 = 4.0 × 10−4
η0 = 0.73
Elizalde and Ancheyta (2014b) φ2 = φs + 1 − φs e − α t − tb Kinetic lumping Stirred Residue α380 C = 0.070
tank φs = 0.660
tb = 80
Toulhoat et al. (2005) dC m r, z, t Mm HDM Fixed-bed Boscan crude petroleum ks = 16.5 × 103
= ρm r, z, t
dt ρcat j
HDS km = 3.9 × 104
Em = 40, Es = 40
Chao et al. (1991) 1 − xs = αcQc HDM Fixed-bed Iranian heavy A.R. k c1 = 3 × 10 − 2 e − 20 × 10
3
RT
HDS 10 − 24 2 × 103 RT
k m = 1 28 × 10 e
3
k s = 1 22 × 1010 e31 3 × 10 RT
αc = 1.8
n
Kam et al. (2005) Cp,i Target HDM Fixed-bed Atmospheric residue Es = 26.1
Ψi = − e C p,i Product from previous reactor
HDS ks = 0.492
kv = 0.021
nHDS = 2
nHDV = 2
mass-transfer phenomena, however, they do not consider the effect of coke deposition on activity
loss as coke buildup occurs only during the first hour-on-stream.
Most of the authors use first-order kinetics for HDM and HDS reactions for evaluating catalysts
performance. Others utilized different values of global effectiveness factor for determining catalyst
life (Dautzenberg et al. 1978; Weekman 1968; Wheeler 1951). Idei et al. (1998), Skala et al. (1991),
and Arbabi and Sahimi (1991a) calculated deactivation parameters from experimental data and
proposed two different deactivation mechanisms: one involves coke and metals deposits and
another one considers only metals deposits.
Table 2.3 shows the parameters reported by various authors for deactivation models considering
both mechanisms of deactivation. Chao et al. (1991) and Kodama et al. (1980) proposed models
based on different properties of the catalyst and obtained good agreement with experimental data.
Toulhoat et al. (2005) and Newson (1975) obtained similar activation energies for HDM and HDS
reactions. Despite on using different catalysts, these two models were tested in a commercial plant,
the model proposed by Newson (1975) showed better agreement with experimental data.
For the development of deactivation models, it is then important to consider several factors, such
as chemical properties of the feed (particularly metals and asphaltenes contents) and physical and
chemical properties of the catalyst. If the experiments are conducted in tubular reactor, the changes
in catalyst properties should be determined as a function of its position along the reactor length
(Isaza et al. 2000); of course, this depends on the length of the catalytic bed. For small length,
the behavior of the bed may be assumed to behave as integral reactor and the changes in reactor
position can be neglected.
Additionally, it is necessary to take into account the different stages of catalyst deactivation since
it is well-known that at long TOS, various mechanisms are involved, such as coke deposition during
SOR causing loss of catalytic surface area and decrease in mean pore diameter and volume; at MOR,
deactivation of catalyst is mainly due to loss of sites by poisoning and pore plugging by metal sul-
fides deposits; and finally at EOR, catalyst presents severe diffusional resistance due to almost total
pore plugging.
Some models reported in the literature do not take into consideration the effective diffusivity,
which is affected by coke and metal deposits. Deposits cause decrease in pore diameter and loss
of catalytic specific surface area preventing the access to the active sites of the catalyst; consequently
in some cases, a wrong value of effectiveness factor is utilized, resulting in falsified data by apparent
kinetics.
Most models based on coke deposits are a simple function of time, nevertheless, to develop a bet-
ter deactivation model, it is important to consider the concentration of reacting species, which pro-
voke coke deposition on catalyst. Some deactivation models are based on TOS and do not consider
the concentration of heteroatoms in the feedstock. It should be highlighted that the reason for cat-
alyst deactivation is not TOS, but metals and coke deposition. For heavy oils with high content of
impurities (metal and asphaltenes) at long TOS, metal buildup is more relevant; then each feed will
exhibit different deactivation patterns, and for this reason, the proper way for describing the cat-
alyst deactivation is with a model that includes the agents present in the feedstock, which provoke
activity decay of the catalyst, instead of a function based on TOS.
2.4.2 Final Remarks

The formation of coke and metals deposits during HDT reactions of heavy oil is a real and complex
problem that affects the performance of the process and the quality of products due to catalyst deac-
tivation. The main reasons for catalyst deactivation are attributed to loss of active sites due to poi-
soning and pore mouth plugging. It has been found that most of the coke deposition on the catalyst
occurs during the first stage of the process and after that deposited coke remains constant, and
2.5 Development of a Reactor Model for Heavy Oil Hydrotreating with Catalyst Deactivation Based on Vanadium 83
catalyst decay continues but mostly affected by metals deposition. For this reason, the models based
on both stages of deactivation fit better to experimental data.
The models that consider the deactivation of the catalyst due to metals accumulation,
generally take into account the size of the pores, since it is necessary to have a suitable pore
size that allows the metal molecules to enter into the pores. Some authors have developed
models considering both metal and coke deposits. These types of models are more appropriate
for predicting catalyst life because they cover the three stages of deactivation in the HDT
process.
It should be recognized that modeling catalyst deactivation during HDT of heavy oils is not an
easy task. More efforts need to be put to develop robust models that consider the different stages
(SOR, MOR, and EOR) of deactivation, and the effect of the agents that are indeed affecting the
activity of the catalyst (i.e., coke and metals).
2.5 Development of a Reactor Model for Heavy Oil Hydrotreating

with Catalyst Deactivation Based on Vanadium and Coke Deposition
A quasi-dynamic one-dimensional pseudohomogeneous TBR model for HDT of heavy vacuum res-
idue was developed. The model includes catalyst deactivation based on the concentration of impu-
rities in heavy oil, as well as coke and vanadium deposition on catalyst surface during TOS.
Literature experimental data of heavy vacuum residue HDT were used for model validation.
The proposed model well-predicted the typical “S” shape profiles during TOS due to catalyst deac-
tivation for heavy oil HDT.
The proposed model was developed considering a previous work reported in the literature
(Kodama et al. 1980), which was used to describe experimental data obtained for HDT of a
vacuum residue with sulfur and vanadium contents of 3.67 wt.% and 270 ppm, respectively.
Experimental data include the concentrations of vanadium and sulfur in products along
TOS at 140 kg/cm2, 1100 m3/m3 H2/oil ratio, and 0.46 hours−1 LHSV at three temperatures
of 385 C, 400 C, and 415 C. At 1360 hours for the temperature of 400 C, the amount of vana-
dium deposits on catalyst along reactor axial length was measured. From the profiles of impu-
rities concentration of products along TOS, the first stage, second stage, and the beginning of
the third stage of catalyst deactivation can be clearly distinguished. According to the classifi-
cation of Froment et al. (2010), the proposed model would be a dynamic one-dimensional pseu-
dohomogeneous model. However, since there is not an accumulation term, the model was
classified as a quasi-dynamic. The model may be considered of medium complexity. It can sim-
ulate the operation of a TBR at isothermal conditions. By mass balance, the reactor model can
predict the concentration profiles of impurities with respect to TOS. Effective diffusivities have
the strongest influence on catalytic deactivation, so that the limitation of the access of heter-
oatom compounds to the catalytic active sites due to pore plugging is included. This consider-
ation is based on literature reports. For instance, Kobayashi et al. (1987) conducted a study of
heavy oil HDT considering the catalyst pore size. They found that for catalysts with 3% molyb-
denum on alumina and on silica alumina, the optimum pore size was from 100 to 150 Å at
400 C and the optimum pore diameter increases as the temperature increases. Vanadium
and nickel impurities deposited on catalyst surface are high during HDT reactions, while coke
is deposited in lower amount. Active surface area is affected as pore plugging progresses,
reducing conversion of impurities. The base model only includes vanadium deposition; how-
ever, it can be easily adapted to consider nickel content separately from vanadium or as a
sum of them.
2.5.1 The Model

2.5.1.1 Description
The mass balance equations for the quasi-dynamic one-dimensional pseudohomogeneous reactor
model are as follows:
dCS
ul + εS r S = 0 2 84
dz
dCV
ul + εS r V = 0 2 85
dz
where ul is the mass flux velocity of liquid phase, CS is the sulfur concentration in liquid phase, CV is
the vanadium concentration in liquid phase, εs is the catalyst fraction in the reactor, and z is the
reactor length.
Eqs. (2.84) and (2.85) were integrated by the Runge–Kutta method. The exit values of the reactor
at certain TOS can be compared with the respective experimental value. Catalyst fraction (εs) was
calculated by the relationship between void and solid fractions:
εs = 1 − B 2 86
Bed void fraction was calculated by a correlation of particle size and diameter of the reactor pro-
posed by Froment et al. (2010).
2
dR dpe − 2
B = 0 38 + 0 073 1 + 2 2 87
dR dpe
Reaction rates (ri) for vanadium or sulfur depend on the remanent catalyst surface area,
effectiveness factor, and concentration of heteroatoms in the liquid phase. All these parameters
change along the length of the reactor and during TOS. The equations to obtain the reaction rates
for S and V are as follows:
r i = ρcat Ari ηi k i C h C2i 2 88
where ρcat is the packed bed density of catalyst, Ari is the remanent surface area, ηi is the effective-
ness factor, ki is the reaction rate coefficient, Ch is the concentration of hydrogen in the liquid phase,
Ci is the concentration of heteroatom i in process stream, and i represents sulfur and vanadium. It is
well-known that kinetics follows first-order of reaction when model compounds are used, however,
when mixture of model compounds or more complex petroleum fractions are studied, the reactor
order tends to increase. For the case of HDS and HDV reactions when residues are hydrotreated, the
most common reaction order reported in the literature is two. Second-order of reaction was also
used in the literature model (Kodama et al. 1980). The concentration of hydrogen in the liquid
phase was calculated with
Ch = 8 91 × 10 − 6 P + 4 16 × 10 − 6 T − 273 − 1 40 × 10 − 3 2 89
where P is the reactor pressure and T the temperature.

Rate coefficient for each HDT reaction is based on the Arrhenius equation:
E
− RgAT
k i = A0 e 2 90
where A0 is pre-exponential factor, EA the activation energy, and Rg the ideal gas constant.
Effectiveness factor (ηi) is calculated from the Thiele modulus (ϕi):
ρcat Ari k i C h C i
ϕi = r 2 91
ρoil Di
3 1 1
ηi = − 2 92
ϕi tan hϕi ϕi
where r is the catalytic particle equivalent radius, ρoil is the density of the oil, and Di is the diffusivity
of component i inside the pores of the catalyst. Reaction rate equations include catalytic activity in
the term of remanent surface area of catalyst active sites. Remanent surface area of catalyst is
reduced as the amount of impurities deposited on catalyst pores increases and can be calculated
as follows:
Ars = Ao 1 − αV ψ V − αc ψ c 2 93
ArV = Ao 1 − βV ψ V − βc ψ c 2 94
Arc = Ao 1 − γ V ψ V 2 95
where Ars, ArV, and Arc are the remanent surface areas for HDS, HDV, and for the deposition of
coke, respectively; Ao is the fresh surface area; αV, βV, and γ V are the parameters for the effect of
deposition of vanadium on HDS, HDV, and coking active sites including the effect of nickel; αc
and βc are the parameters for the effect of deposition of coke on HDS and HDV active sites.
Remanent surface area of active sites for HDS and HDV are affected by both vanadium and coke
deposition. Meanwhile remanent surface area for coke deposition is affected by vanadium deposits.
ψ is the volumetric amount of impurities deposited on catalyst calculated according to the
following:
Qi
ψi = 2 96
100PV ρi
where Qi is the dimensionless mass of impurity i deposited on catalyst, PV is the catalyst pore vol-
ume, and ρi is the impurity density. The mass amounts of heteroatom deposits on the catalyst in a
time interval generated by HDV reaction and coke deposition are obtained by as follows:
t
100
Qi = r i dt 2 97
ρcat 0
where t is TOS.
The diffusivity of heteroatoms in the liquid phase is limited by the deposited metals and coke. The
parameter ψ max refers to the maximum retention capacity. From this parameter, the diffusivity is
affected as the deactivation of the catalyst progresses.
ψV
Ds = Dso 1 − δV 2 98
ψ max
ψV
DV = DVo 1 − 2 99
ψ max
Effective diffusivities depend on the volumetric amount of impurity deposit on catalyst. δV is the
parameter of the effect of vanadium deposits on sulfur diffusivity. Coke deposition reaction is
reversible, and the rate equation has a different expression than those of the vanadium or sulfur:
r C = ρcat Arc k C1 − ρcat Ao k C2 C h qc 2 100
where kC1 is the reaction rate coefficient of deposition of coke, kC2 is the reaction rate coefficient of
removal of coke, and qc is the mass of impurity deposit per volumetric unity of catalyst.
Coke and vanadium deposits generated affect reaction rates by covering surface area and diffu-
sivity of sulfur and vanadium by plugging catalyst pores.
2.5.1.2 Solution of the Model

The sequence of calculus starts with integration of mass balance equation for vanadium and sulfur
with input data about properties of feedstock and catalyst and characteristics of the reactor. The
deposits of vanadium and coke for fresh catalyst are zero, and the surface area of active sites is
the initial value in the fresh catalyst. The deposition rate of sulfur, vanadium, and coke along
the catalyst bed determines the amount of deposits at each coordinate according to TOS and axial
length steps. Then, diffusivity of sulfur and vanadium, remanent surface areas, and volumetric
quantity of impurities on catalyst are determined. The sequence is repeated for different TOS,
and in each step calculated and experimental values are used in the objective function for optimi-
zation of parameters.
To obtain a more exact simulation, intervals of experimental TOS were divided into forty steps for
the proposed model. For literature model (Kodama et al. 1980), it was reported that intervals were
divided into 20 steps.
Operating conditions of HDT, and feedstock and catalyst properties are essential input data for
better application of the proposed model. It requires at least both concentration of impurities in
products and deposits on catalysts to perform a reliable parameter estimation. Pore volume, rem-
anent surface area, and other heteroatoms concentration values are considered additional informa-
tion that can contribute to more precise modeling.
The structure of the algorithm was coded in Python, whereby the concentration of sulfur and
metals in process stream profiles were obtained with respect to TOS. The calculation of model para-
meters was done by minimizing an objective function based on the squared differences between
calculated and experimental values including their covariances. Levenberg–Marquardt’s method
of optimization was applied at last to rectify the best set of parameter values by assuring the global
minimum of the objective function.
2.5.1.3 Advantages of the Model

The following are the main advantages of the proposed model compared with the literature model
(Kodama et al. 1980). Some of them are related to unreported data and inconsistencies of the lit-
erature model and others more important for simplification of equations, optimization of methods,
and verification of model parameters.
• Maximum capacity of vanadium deposit. An inconsistency in deposited vanadium profile with

respect to reactor length was observed in the literature model (Kodama et al. 1980). While in a
plot a maximum capacity of vanadium deposition of 34.1% was reported, in the text the corre-
sponding value for this parameter is 31.4%. This difference causes a mathematical error because
the value of effective diffusivity (Eq. 2.99) becomes negative and then the value of the Thiele mod-
ulus (Eq. 2.91) results in erroneous values. In any case, using 31.4% or 34.1% of maximum capac-
ity of vanadium deposition gives erroneous values for the prediction of concentration profiles,
which differ from the values reported by the literature model (Kodama et al. 1980). This discrep-
ancy was corrected in the proposed model.
• Calculation of effectiveness factor. The literature model (Kodama et al. 1980) determines the
Thiele modulus and effectiveness factor by solving the intraparticle dimensionless concentration
differential equations by numerical methods. In the proposed model, general Eqs. (2.91)
and (2.92) were used to calculate effectiveness factor from the Thiele modulus. This change
simplifies the calculation and improves optimization of model parameters without losing
accuracy. From the results discussed later on, a reduction of effectiveness factor value
for HDS and HDV reactions was observed along TOS. The rate and distribution of coke
and vanadium deposition on catalyst active sites in residue HDT has a particular behavior
compared with HDT of light fractions and catalyst poisoning in other processes. The shape
of the catalyst pores in residue HDT changes to a bottle ink as metals and coke depositions
occur (Ancheyta et al. 2005). Moreover, residue conversion is reduced as the covering of
active sites progresses, being the reduction of effective diffusivity due to pore plugging
what contributes considerably to conversion decay. Effective diffusivity in conventional
equations is a function of tortuosity, porosity, and constriction factor. In the proposed
model, Eqs. (2.16) and (2.98) describe the effect of vanadium deposits on effective diffusiv-
ity of sulfur and vanadium compounds, which was accurate due to the impossibility to
model the specific effect of bottle ink shape of the catalyst pores for the calculation of
effective diffusivity.
• Density of vanadium deposits. The density of vanadium deposited on catalyst was calculated with
Eq. (2.96) with maximum vanadium volume and mass capacities of 0.261 m3/m3 and 0.314 kg/kg,
respectively. The calculated density of vanadium deposits was 2001.77 kg/m3.
• Density of coke deposits. Coke density in reforming catalyst has been reported to be 966 kg/m3
(Baghalha et al. 2010). The density of bitumen sample with 82% carbon was calculated to be 1050
kg/m3 (Richardson et al. 1996). Density ranges for different types of coals were from 1290 kg/m3
to 1470 kg/m3 (Flores 2014). Based on these references, for this work, coke density was assumed
to be 990 kg/m3. In the literature model (Kodama et al. 1980), neither vanadium density deposits
nor coke density was reported.
• Solid phase fraction volume. In the proposed model, it was calculated through Eqs. (2.86) and
(2.87), while literature model (Kodama et al. 1980) does not report this value.
• Method of optimization. In the literature model (Kodama et al. 1980), complex method was
used to estimate reaction rate coefficients for HDS and HDV, least squares to estimate the
Arrhenius parameters, while coke deposition rate coefficients were estimated by trial and error
with αV, αc, βV, βc, and γ V. For the proposed model, sequential quadratic programming (SQP)
and Powell methods were used to determine reaction rate coefficient by minimization of the
following objective function. The use of more accurate optimization methods and different sta-
tistical approaches for parameter estimation is a new aspect added to the literature model
(Kodama et al. 1980).
N
1 2
SSQQ = y − yi,jcalc 2 101
j=1
covi,j i,j exp
where i is sulfur or vanadium, j is the number of experimental data at different TOS, and y is the
concentration of heteroatom in products or on catalyst surface.
The optimization methods used in the proposed model are much more accurate than those
used in the literature model (Kodama et al. 1980), this latter being developed more than
40 years ago.
A sensitivity analysis on the estimated parameters was carried out to ensure their optimal
values, and that the objective function corresponds to the global minimum and not local minima
(Félix et al. 2019). This type of analysis was not performed in the literature model (Kodama
et al. 1980).
2.5.1.4 Procedure for Parameter Estimation

Parameter estimation for the proposed model was done in three steps to obtain better agreement
with experimental data. In the first step, the reaction rate coefficients for HDS and HDV are esti-
mated for the middle operation temperature. Initial values were taken from the literature (Kodama
et al. 1980). Then, preexponential factors and activation energies are calculated for the SOR con-
dition (300 hours TOS). The other parameters are kept constant in value of 1 and for αV, αc, βV, βc,
γ V, and ψ max the value was 0.4. The reaction rate coefficients for coke deposition and desorption are
estimated based on the content of coke on catalyst at the middle temperature. A simulated curve
approaching to experimental data can be observed after this step. The optimization method used to
estimate preexponential factor and activation energy for coke deposition and desorption is SQP and
Powell for reaction rate coefficients.
The second step of parameter estimation consists of re-estimating αV, αc, βV, βc, γ V, δV, and ψ max
by SQP. This is a key step for having a good agreement with experimental data. So that, the typical
“S” shape of the curve can be reproduced. The value of the objective function is drastically reduced
since these parameters describe the effect of the deposition of impurities on catalytic deactivation
during TOS. For this step, all the experimental values were used.
The last step of parameter estimation consists of carrying out the minimization of the objective
function by SQP method changing all parameter values. Parameters in this point considerably
changed from the initial guesses, indicating that SQP method only approaches to the optimum
set of parameter values. Then, Powell method is used to continue the optimization parameter
values. The curve does not change substantially in this step, but minimization by Powell method
yields a better set of parameter values. A rectification by Levenberg–Marquardt is done at the end of
the parameter estimation. Table 2.4 summarizes the values of parameters in each stage of
Table 2.4 Values of parameters (developed model).
Values in each step of estimation
Symbol Initial First Second Third Unit
kS 1.512 1.212 1.212 1.156 kg/m3h

EA 20,000 23,679 23,679 23,127.45 cal/mol
kV 238.0 225.42 225.42 220.98 kg/m3h
EA 25,000 29,784.2 29,784.2 30,989.38 cal/mol
kC 1 × 10−9 0.98 × 10−7 0.98 × 10−7 1.04 × 10−8 kg/m3h
EA 10,000 11,058.2 11,058.2 11,083.01 cal/mol
kDC 1 × 10−9 3.89 × 10−8 3.89 × 10−8 3.97 × 10−8 kg/m3h
EA 10,000 7723.8 7723.8 7612.83 cal/mol
αV 1.0 1.0 1.8405 1.884 Dimensionless
αc 1.0 1.0 1.7632 1.7759 Dimensionless
βV 1.0 1.0 1.7379 1.819 Dimensionless
βc 1.0 1.0 1.7753 1.85 Dimensionless
δV 1.0 1.0 0.1870 1.0697 Dimensionless
γV 1.0 1.0 0.6038 0.5768 Dimensionless
ψ max 0.4 0.4 0.3424 0.3248 m3V m3cat
DS0 2.10 × 10−6 2.10 × 10−6 2.10 × 10−6 8.22 × 10−7 m2/h
−6 −6 −6 −7
DV0 1.57 × 10 1.57 × 10 1.57 × 10 7.22 × 10 m2/h
parameter estimation. To be sure about the optimal parameter values, a sensitivity analysis method
reported in literature (Félix et al. 2019) is used.
2.5.2 Results and Discussion

2.5.2.1 Profiles of Sulfur and Vanadium Concentration in Products
The concentration profiles of sulfur and vanadium in products along TOS at 385 C, 400 C, and
415 C obtained with the proposed model are shown in Figure 2.14. The expected trend of profiles
with respect to operation temperature is observed, that is the higher the temperature, the higher the
removal of vanadium and sulfur. The first stage (SOR) of catalyst deactivation can be observed until
approximately 500 hours of TOS, the second stage (MOR) until 1100 hours of TOS, and the begin-
ning of the third stage (EOR) in the few available last points after MOR. The proposed model is able
to reproduce the “S” shape profile of sulfur and vanadium concentration in products at constant
temperature, which means that the prediction of the first, second, and the start of the third stages
of catalyst deactivation is accurate with good agreement with experimental data.
A comparison of predictions with both proposed and literature models (Kodama et al. 1980)
is shown in Figure 2.15 for the concentration of sulfur and vanadium profiles at 400 C and
31500
S concentration in products (ppm)
28000
24500
21000
17500
14000
10500
7000
3500
0
180
V concentration in products (ppm)
150
120
90
60
30
0
0 500 1000 1500 2000 2500
Time on stream (h)
Figure 2.14 Profiles of vanadium and sulfur content in products vs TOS. Calculated with the proposed model:
(——) 385 C, ( ) 400 C, and (---) 415 C. Experimental data (Kodama et al. 1980): (■) 385 C, (▲) 400 C, and (♦)
415 C.
20000
17500
S concentration in products (ppm)
15000
12500
10000
7500
5000
2500
120
105
V concentration in products (ppm)
90
75
60
45
30
15
0
0 250 500 750 1000 1250 1500 1750
Time on stream (h)
Figure 2.15 Comparison of profiles of vanadium and sulfur content in products vs TOS. Calculated with the
proposed model: ( ) 400 C, and (——) 415 C. Calculated with the literature model (Kodama et al. 1980): (– –)
400 C, and (---) 415 C. Experimental data (Adapted from Kodama et al. 1980): (▲) 400 C, and (⧫) 415 C.
415 C. The curve corresponding to the first stage of catalyst deactivation for sulfur profiles is the
same for both models till 500 hours of TOS. Calculated and experimental vanadium concentrations
at short TOS (500 hours) show higher deviations at 415 C as compared with 385 C and 400 C,
although the deviations are lower than the literature model (Kodama et al. 1980). For TOS longer
than 500 hours, the literature model (Kodama et al. 1980) tends to underpredict the experimental
data, which is more notorious for sulfur than for vanadium concentration. The proposed model
corrects all these deviations and matches with more accuracy with experimental data.
2.5.2.2 Comparison of Predictions with Literature and Proposed Model

The differences of predictions with the literature (Kodama et al. 1980) and proposed models can be
clearly seen in Figure 2.15 for sulfur and vanadium concentrations, respectively. Both models over-
predict the smallest values of concentrations corresponding to short TOS (SOR conditions), partic-
ularly for vanadium, although the overprediction is more notorious for the literature model
(Kodama et al. 1980). For longer TOS, that is, MOR conditions, both models perform similarly.
However, for EOR conditions, the literature model (Kodama et al. 1980) drastically fails in the pre-
dictions giving very low concentrations values of sulfur, while the proposed model reproduced the
experimental values with high accuracy. The calculated coefficient of determination for proposed
model was equal to 0.9119, which confirms a good description of experimental data. The coefficient
of determination was also determined for the literature model (Kodama et al. 1980) resulting in a
value of 0.879. In other words, the proposed model gives a better description of the experimental
data especially at higher temperature of operation and offers the advantage of simplicity. Statistical
analysis was carried out for the proposed and literature models (Kodama et al. 1980). Residual bal-
ance for the proposed model was 10; meanwhile for the literature model (Kodama et al. 1980), it was
−36, which indicates that the proposed model profiles have better distribution of predictions with-
out tendency to over- or sub-estimation of experimental data, which is not the case for the literature
model (Kodama et al. 1980). Relative standard error for the proposed model was 0.4471 and for the
literature model was 0.4634 (Kodama et al. 1980).
2.5.2.3 Profiles of Coke and Vanadium on Catalyst

Figure 2.16 shows a comparison of predicted and experimental profiles of vanadium and coke
deposits along the reactor length. It is observed that vanadium deposits on catalyst decreases from
the inlet to the outlet of the reactor. The same behavior is observed for the vanadium concentration
profile in liquid phase. In other words, the higher the concentration of vanadium in liquid phase (at
reactor inlet), the more abundant the vanadium deposits on catalyst. Even the predictions of exper-
imental data with the proposed model have some deviations, and the literature model (Kodama
et al. 1980) also presented high deviations. In spite of these differences, predictions with the pro-
posed model follow the tendencies of coke and vanadium contents along reactor axial coordinate.
One aspect that should be taken into consideration is that an inconsistency was found in the
40
Impurities content on catalyst (kg/kgcat)
35
30
25
20
15
10
0.0 0.2 0.4 0.6 0.8 1.0

Dimensionless reactor length
Figure 2.16 Deposits along the axial length of catalyst bed at 400 C and 1360 h. (---)
Calculated content of V, ( ) calculated content of coke, (■) experimental content of vanadium,
(▼) experimental content of coke (Adapted from Kodama et al. 1980).
40
Impurities content on catalyst (kg/kgcat)
35
30
25
1360 h
20 1050 h
600 h
600 h
1050 h 1360 h
15
10 100 h
60 h
5 60 h
100 h
0
0.0 0.2 0.4 0.6 0.8 1.0
Figure 2.17 Evolution of deposit profiles along TOS. Calculated content of V (——), calculated content of coke
(---) at different TOS values.
calculated profiles by authors (Kodama et al. (1980), regarding the profile of vanadium deposits on
catalyst, since the maximum quantity of vanadium deposits on catalyst does not agree with the ini-
tial calculated profile. As mentioned in Section 5.3, 34.1% is the highest value of the vanadium
deposits on catalyst profile, but the reported value is 31.4%. When using this value, an error is
obtained in the Thiele modulus calculation (Eq. 2.91), causing the vanadium diffusivity to be neg-
ative at the beginning of the profile of vanadium deposits on catalyst.
As shown in Figure 2.17, the concentration of coke on catalyst at SOR depicts a linear profile due
to the more uniform distribution of catalyst active sites along reactor axial length. As vanadium
deposits occupy catalyst active sites for long TOS, the coke deposits profile tends to increase. This
is because near the reactor input catalyst active sites are occupied by more vanadium deposits and
more catalyst active sites are free near the exit of the reactor. As remanent surface area of catalyst
active sites is reduced as the quantity of deposits increases on catalyst surface, catalyst deactivation
is dependent on the poisoning of active sites. Meanwhile, pore plugging metals deposits reduce dif-
fusivity of components to internal remanent active sites. External hindrances are not included in
the equations of the proposed model. The increased coke content on catalyst may be also due to the
low hydrogen availability at the bottom of the reactor caused by hydrogen consumption. In com-
mercial reactors, high temperatures at the exit of reactor are another reason for high coke content of
catalyst; however, in this case simulated reactor is isothermal. The earliest profile of coke deposits
on catalyst is uniform since vanadium deposition at SOR is minimal along the catalyst bed. Coke
content on catalyst increases until approximately 600 hours of TOS. This TOS value is near the end
of the first stage of catalyst deactivation as shown in Figure 2.17. It would be expected that coke
content profile on catalyst at 1360 hours had higher values with respect to those at shorter TOS,
which is not the case. This is because the coke deposit reaches a maximum at the end of SOR. Then,
coke deposition reversibility uncovers active sites surface area and metals are deposited instead. On
the contrary, vanadium deposits increase along the catalyst bed and desorption is not occurring.
150
10 h
140
130
HDV remanent surface area (m2/gcat)
120
110
70 h
100
90
80
70 200 h
60
400 h
50
600 h
40
800 h 1000 h 1200 h 1360 h
30
20
10
0
0.0 0.2 0.4 0.6 0.8 1.0

Figure 2.18 Evolution of surface area of HDV active sites along TOS.
The covering of active sites by vanadium deposits reduces the surface area where HDS, HDV, and
coking reactions take place. Reduction of surface area by coking is diminished as the content of
vanadium deposits increases. The quantity of coke deposits not only is reduced due to its revers-
ibility, but vanadium deposition on catalyst surface reduces the quantity of coke on catalyst until
the end of operation. Something important to clarify is that vanadium compounds do not remove
coke from catalyst. Figure 2.18 shows the evolution of remanent surface area of HDV active sites
along the catalyst bed. As expected, vanadium deposition is higher near the inlet of the reactor and
remanent surface area of HDV active sites is lower at the inlet of the reactor. The loss of active sur-
face area during TOS has significant effect on the concentration of sulfur and vanadium in products
at SOR and part of MOR. The last stage of catalyst deactivation (EOR) regarding the concentration
of sulfur and vanadium in products along TOS is dominated by Eqs. (2.98) and (2.99) due to the
considerable fulfilling of catalyst pores, which provokes the nearly zero value of diffusivity and
the consequent low value of effectiveness factor.
The proposed model can describe the behavior of catalytic deactivation in heavy oil HDT. It
reproduces well the first, second, and the start of the third stages of catalyst deactivation in vacuum
residue HDT, that is, SOR, MOR, and EOR conditions.
2.5.2.4 Final Remarks

The proposed model could be described as of intermediate complexity. As shown in Table 2.5, it
considers oil and catalyst properties in the model equations, which give the opportunity to simulate
atmospheric or vacuum residues. As seen in Table 2.4, final values of HDS and HDV reaction rate
coefficients are similar to their initial values, since it is convenient to maintain the intercept near
the first experimental data points. Coking formation and desorption reaction rate coefficients
change according to the experimental quantity of coke deposits on catalyst. The values of para-
meters, αV, αc, βV, βc, γ V, δV, and ψ max, are determined by SQP method according to the performance
of a specific HDT catalyst for a given feedstock, that is, these parameter values will be unique for
Table 2.5 Values of input data (developed model).
Symbol Value Unit
ρcat 660 kg/m3

ρcoke 990 kg/m3
ρvanadium 2001.77 kg/m3
ρoil T avg 766.3 kg/m3
εs 0.511 m3cat m3R
CS0 0.0367 kgS/kgT
CV0 0.00027 kgS/kgT
rcat 1.34 m
PV 0.601 m3pore gcat
QV 0 kgV/kgcat%
Qc 0 kgC/kgcat%
Ars 155 m2/gcat
ArC 155 m2/gcat
ArV 155 m2/gcat
each combination of catalyst and feedstock. The proposed model can represent more accurately the
experimental data at intermediate and high temperatures, which are more common when proces-
sing heavy residues. Figures 2.15 and 2.16 help demonstrate the capability of the proposed model to
predict the TOS in which coking formation reaches equilibrium and coke desorption begins, as well
as dynamic coke and vanadium deposits on the catalyst. Figure 2.18 shows the evolution of rem-
anent surface area for HDV at 10 hours of TOS, which depicts a horizontal line that indicates that
coke deposition is predominant and vanadium deposition is small. However, for longer TOS (MOR
and EOR), the profile of remanent surface area achieves a hybrid shape, which is a result of both
coke and vanadium covering. In general, the proposed model offers a simplified manner to describe
residue HDT and allows for determining the evolution of the main parameters, that is, heteroatom
concentrations in products, impurity contents on catalyst, and catalyst deactivation. From the
results given in Table 2.4, it is observed that at SOR, coke deposition has the highest contribution
to the catalyst deactivation. From MOR to EOR, vanadium deposition dominates catalyst deacti-
vation, thus decreasing significantly the rate of HDS and HDV reactions.
The experimental data show a rapid loss of catalyst activity, that is, HDS and HDV reduce with
TOS. This behavior is typical in laboratory experiments that are conducted at constant reaction tem-
perature, which is necessary to develop kinetic and deactivation models. In commercial applica-
tion, this temperature is constantly increased to compensate for catalyst deactivation, so that
the quality of the product is maintained constant during TOS. The most important aspect for com-
mercial application of TBR for HDT of petroleum residues is the metal retention capacity of the
catalyst, and its life to guarantee operation cycles that can give a positive cost benefit.
Even the number of parameters is large; we used a three-step parameter estimation procedure
and verification of their optimal values was done by sensitivity analysis that assures proper calcula-
tions. Then the set of estimated parameter values that fits the model to experimental data is unique.
The proposed model gives a better description of experimental data than literature model
(Kodama et al. 1980). One important characteristic of the proposed model compared with others
is the simpler manner to calculate the accumulation of impurities on catalyst to describe the catalyst
deactivation.
According to the results, it is deduced that concentrations of impurities in products and on cat-
alyst at temperatures from 385 C to 415 C can be predicted for the same reactor, hydrotreated
feedstock, and used catalyst. This can be applied to find the conditions in which vanadium and coke
deposition are lower to minimize catalyst deactivation. The proposed model can be adapted to
describe experimental data of other residue HDT reactors and reaction conditions. The proposed
model can predict the performance of residue HDT reactor with the same feedstock and catalyst at
operation temperatures between 385 C and 415 C. It should be commented that experimental
data of the three stages of catalyst deactivation are necessary to tune the parameters to be able
to accurately predict the whole run length, typically observed in commercial operation.
Using the information of short TOS (SOR) for the prediction of EOR conditions is risky. It is nec-
essary to have some experimental data corresponding to EOR for model validation. Although the
available experimental data used to develop the proposed model was obtained at SOR, MOR, and
the beginning of EOR conditions, and not during the complete EOR, in which catalyst has been
totally deactivated, the predicted values are in good agreement with the typical “S” shape profiles
of impurities content during TOS. This means that the proposed model is validated at these con-
ditions, and it requires additional experimental values to evaluate its application for long-term
operations.
2.5.3 Usefulness of the Model

The proposed model represents a useful tool to make predictions for the HDT of petroleum resi-
dues. The model parameters were estimated using a sequential method that assure their optimal
values and accurately represent the typical concentration profiles during TOS. It means that the
model can predict the effect of temperature and TOS on catalyst deactivation by proper calculation
of metals and coke deposition. Changes in feed properties (sulfur and vanadium content) can also
be evaluated so that different feed’s nature can be correlated with catalyst deactivation. In sum-
mary, with some properly designed experiments for new catalyst and feed, the proposed model
can be useful for the following:
• Screening of HDT catalysts for heavy oil feeds. Typical experiments for HDT catalyst screening
only consider data at SOR conditions, that is, short TOS. The performance of an HDT catalyst can
be better than others at SOR, but deficient at long TOS. This situation cannot be anticipated at
SOR operation. The model parameters can be tuned with experimental data for new catalyst and
feed, so that it can predict the long-term operation of the catalyst.
• Deactivation of catalyst. With short-term experiments, only early catalyst deactivation can be
observed. Appropriate evaluation of loss of catalyst activity needs to be carried out at long
TOS. The model can predict long-term catalyst deactivation and the extent of metals and coke
deposition, as well as losses in specific surface area. This is of great importance to determine
the catalyst life.
• Selection of operating conditions. Having the concentration profiles of impurities in the products
as function of TOS is useful to define suitable operating conditions to achieve the desired quality
of the products. Increases in reactor temperature to mitigate catalyst deactivation can be properly
predicted by the model. In addition, proper selection of SOR conditions can be done to optimize
the catalyst utilization and life.
• Design of new reactors. Although at present some reactor designs are still done based on experi-
ences collected during decades, and correlations derived from huge number of experimental data,
the use of reactor model with catalyst deactivation is also important for design, simulation, and
optimization of new reactors. For HDT of residue, it is well-known that more than one catalyst is
needed, for example, HDM, hydrocracking, and HDS catalysts. The amount of each catalyst and
operating conditions in a series of HDT reactor depends on the feed properties and desired prod-
uct quality. The proposed model can be used to help define the reactor configuration, since it
properly describes the concentration profiles along the reactor length.
• Economy. The use of reactor models during the development of new processes or optimization of
actual processes is a common task for technology developers. The main advantage of a model
reactor for HDT of residue is that it substantially reduces the experimental work (pilot plant test-
ing, product characterization, analysis of catalyst) and time; therefore, important savings of time
and money are obtained.
2.5.4 Conclusion
The following conclusions can be pointed out from the results obtained with the proposed model:
Catalyst deactivation is modeled through vanadium and coke deposition as a function of TOS.
The proposed model assumes that catalyst deactivation is caused by poisoning mechanism through
deposition of vanadium and coke on catalyst active sites affecting HDS and HDV conversions. Dif-
fusivity of sulfur and vanadium to internal catalyst active sites is also limited by vanadium and coke
deposition due to pore plugging, that is, internal limitation of diffusivity. The proposed model pre-
dicts the performance of residue HDT reactor with the same feedstock and catalyst at operation
temperatures between 385 C and 415 C. It describes the evolution of vanadium and sulfur con-
centration in products and deposits on catalyst surface in the three stages of catalyst deactivation.
For estimation of model parameters, it is highly recommended to have experimental data of the
three stages of catalyst deactivation. The predictions with the proposed model are in good agree-
ment with experimental data. By using more accurate methods and simplifying the calculation of
some parameters, the proposed model can describe the profiles of heteroatoms content in products
and on catalyst. Detailed sequence of parameter estimation and input data were described to help
researchers with the use of the proposed model.
2.6 Application of the Deactivation Model for Hydrotreating

of Heavy Crude Oil in Bench-Scale Reactor
Experimental data of a heavy crude oil HDT carried out in a bench-scale reactor at 380 C, 98 bar,
0.5 hours−1, and 890 m3/m3 H2/oil ratio for 500 hours were modeled by a quasi-dynamic
one-dimensional pseudohomogeneous TBR model.
2.6.1 Properties of Heavy Oil

The characterization of the heavy crude oil was performed according to ASTM standard methods
and is reported in Table 2.6.
2.6.2 Properties of the Catalyst

A tetra lobular catalyst was loaded to the TBR of the bench-scale pilot plant. A commercial HDT
catalyst was used for experiments, which is composed of NiMo supported on aluminum oxide
(NiMo/Al2O3). The main chemical and physical properties of the catalyst are presented in
Table 2.7. The catalyst pore distribution is concentrated in the ranges of 100–200 Å (49.44 vol.%)
2.6 Application of the Deactivation Model for Hydrotreating of Heavy Crude Oil in Bench-Scale Reactor 97
Table 2.6 Properties of heavy crude oil.
Property Value
Specific weight @20/4 C (ASTM D-1298) 0.9810

API gravity 12.74
Sulfur content (wt.%) (ASTM D-287) 4.997
Asphaltenes content (wt.%) (ASTM D-4294) 21.8
Viscosity at 37.8 C (cSt) (ASTM D-445) 11,549
Elemental analysis (ASTM D-7042)
C (wt.%) 83.28
H (wt.%) 10.91
Metals (by atomic absorption)
Ni (ppm) 85
V (ppm) 456
Table 2.7 Properties of catalyst and dimensions of reactor.
Property Value
Nominal size (in) 1/18

Average pore diameter (Å) 172.6
Particle diameter (mm) 9.95
2
Specific surface area (m /g) 197.2
Average pore volume (cm3/g) 0.85
Average pore radio (Å) 86.3
Bulk density (kg/m3) 533.9
Chemical composition (wt.%)
NiO/Ni 0.73/0.58
MoO3/Mo 3.27/2.18
Pore size distribution (vol.%)
<50 Å 2.32
50–100 Å 14.86
100–200 Å 49.44
200–500 Å 30.34
500–1000 Å 2.26
>1000 Å 0.78
Reactor bed length (cm) 44.5
Reactor diameter (cm) 2.54
and 200–500 Å (30.34 vol.%), which makes it suitable for processing heavy oils with high content of
metals and asphaltenes for long time of operation due to more available volume to accumulate
deposits without causing severe deactivation.
2.6.3 Bench-Scale Reactor

HDT experiments were carried out in an isothermal TBR. A general scheme of both the bench-scale
plant and the reactor are shown in Figures 2.19 and 2.20, respectively. The plant has four sections:
1) Feed section. A stainless-steel container within a heating system and temperature controller
over a mass quantifier scale, a controlled velocity gear pump, a feeding hydrogen module,
and a mass flow meter.
2) Reaction section. A stainless-steel tubular reactor heated by three electric resistances located
along the reactor length. The temperature profile is controlled through three axial thermocou-
ples installed at the bottom, middle, and top of the catalyst bed.
3) Separation section. Two high pressure separators with control temperature system and two low
pressure separators, which condensate liquefiable products, are further added to the hydro-
treated product stream. A venting line for safe draining of gases also added.
4) Measurement and analysis of gases section. A flow meter to register the flow of reaction gases at
the outlet of the plant and an online gas chromatograph.
2.6.4 Catalyst Activation

The catalyst was activated by the following three stages:
1) Drying. Water in catalyst was eliminated under nitrogen gas at 120 C. The common content of
water on alumina-supported catalysts is from 1 to 3 wt.% due to its high hygroscopy.
2) Soaking. Typically, naphtha, kerosene, or light straight-run gas oil (SRGO) is used to soak the
external surface area of the catalyst particles in the bed as much as possible. In this case, SRGO
was used. The presence of non-wetted areas on catalyst bed reduces the effectiveness of the
catalyst.
Chromatograph
Dry ice
Hydrogen
,
Hydrogen
Heavy crude storage
oil Sosa
Feed
container Light cuts trap
Acid gas
High pressure neutralizer
separator
Isothermal
reactor
Hydrotreated
products
Feed pump
Figure 2.19 Heavy crude oil HDT bench-scale reactor system.

Reactor center
thermocouple
13.5 cm
Input
Tubing 1/4’
51 cm
resistance 1
Electric
9 cm Inert particles
resistance 2
Electric
44.5 cm
167 cm
140 cm
resistance 3
Electric
3 cm Inert particles
Tubing 1/4’
32.5 cm
resistance 4
Output
Electric
13.5 cm
∅ = 2.54 cm
Figure 2.20 Trickle-bed reactor for heavy crude oil HDT.
3) Sulfiding. A sulfiding stream consisting of 98.7 vol.% SRGO and 1.3 vol.% dimethyl disulfide
(DMDS) was employed to convert metal oxides into metal sulfides.
The conditions for drying, soaking, and sulfiding steps are detailed in Table 2.8.
2.6.5 Operating Conditions

Experiments were carried out at 380 C, 98 bar, 0.5 hours−1 LHSV and 890 m3/m3 H2/oil ratio for
480 hours. The volume of catalyst bed was 200 mL. The diameter and length of the reactor are 2.54
and 44.5 cm, respectively. Product samples were recovered every 24 hours.
2.6.6 Characterization Methods

For sulfur content determination, a model SLFA 2100 Horiba X ray fluorescence equipment was
used according to ASTM D-4294. The quantity of asphaltenes was determined according to ASTM
D3279 method. The elemental analysis of feed and products streams was determined based on
Table 2.8 Conditions of catalyst activation.
Variables Drying Soaking Activation
Pressure (bar) 1.013 1.013 27.45

Temperature ( C) 120 150 260, 320
Hydrogen flow (m3/h) 0.380 0.070 0.1535
Diesel flow (mL/h) – 500 –
SRGO + DMDS (mL/h) – – 200
Time (h) 2 2 3 (@260 C)
10 (@320 C)
ASTM D5291 method with a 2400 Series II model Perkin Elmer elemental analyzer. The determi-
nation of textural properties through nitrogen physisorption was based on BET method from
adsorption isotherm in Nova 2000 model Quantachrome equipment. Obtained values of properties
were input data for modeling.
2.6.7 Parameter Estimation

The accuracy of the model has been demonstrated with literature data for HDT of residue.
The input data of the model are shown in Table 2.9, which consider the properties of the oil
and the catalyst. Parameter estimation was carried out in three steps. Table 2.10 shows the variation
of parameters values in each step. The model parameters were initialized with similar order of mag-
nitude of previous parameter estimation (Jurado and Ancheyta 2022). Initial values of reaction rate
coefficients for HDS and HDM were obtained through considering the experimental data in the first
Table 2.9 Values of input data for (application of the model).
Symbol Value Unit
ρcat 553 kg/m3

ρcoke 990 kg/m3
ρvanadium 2001.77 kg/m3
ρoil T avg 831.3 kg/m3
εs 0.484 m3cat m3R
CS0 0.0499 kgS/kgT
CV0 541 ppm
dp 9.95 m
PV 0.85 cm3pore gcat
QV 0 kgV/kgcat%
Qc 0 kgC/kgcat%
Ars 197 m2/gcat
ArC 197 m2/gcat
ArV 197 m2/gcat
Table 2.10 Parameter values (application of the model).
Values in each step of estimation
Symbol Initial First Second Third Unit
kS 1.512 0.3021 0.3021 0.3348 kg/m3h

kV 100.0 82.321 82.321 80.81 kg/m3h
−8 −8 −8 −8
kC 1 × 10 2.98 × 10 2.98 × 10 1.041 × 10 kg/m3h
kDC 1 × 10−8 9.23 × 10−7 9.23 × 1010−7 8.034 × 10−7 kg/m3h
αM 1.0 1.0 1.385 1.4766 Dimensionless
αc 1.0 1.0 1.256 1.2640 Dimensionless
βM 1.0 1.0 0.278 0.2556 Dimensionless
βc 1.0 1.0 1.463 1.4965 Dimensionless
δV 1.0 1.0 1.942 1.9641 Dimensionless
γV 1.0 1.0 0.502 0.5265 Dimensionless
ψ max 0.4 0.4 1.638 1.7999 m3V m3cat
−7 −7 −7 −7
DS0 1.0 × 10 1.0 × 10 1.0 × 10 3.51 × 10 m2/h
DV0 1.0 × 10−6 1.0 × 10−6 1.0 × 10−6 1.73 × 10−6 m2/h
100 hours of operation so that the first step of parameter estimation was to obtain the values of
reaction rate coefficients based on the set of experimental points by SQP method. The second step
of parameter estimation was to obtain the values of Greek letter parameters through SQP method.
In the last step, all the parameters were re-estimated through SQP, Powel, and Levenberg–
Marquardt methods.

2.6.8.1 Evolution of Sulfur and Metals Concentration in Products
Figure 2.21 clearly shows that the model predicted in an accurate manner the experimental data of
sulfur and metals concentrations in products during heavy crude oil HDT. The first shorter TOS
(<100 hours) and second longer TOS (>100 hours) stages of catalyst deactivation are also well
described with respect to experimental data. The rapid deposition of coke is present in the first
100 hours of TOS, which corresponds to the first stage of catalyst deactivation. Then, the second
stage can be observed, which is of more metal deposition dominant. The rate of coke deposition
is depleted at the end of the first stage of catalyst deactivation (first curvature). As the amount
of coke increases, the desorption of coke increases. Hence, there is a kinetic limited quantity of coke
in the catalyst. Once the coke content on catalyst reaches a maximum value, it starts to decrease,
and metals deposition governs the behavior of the profiles of sulfur and metals concentrations in
the product. The deposition of metals on catalyst is carried out in the remaining time of operation.
The horizontal tendency in both sulfur and metals concentration profiles is due to the balance
between coke formation and desorption. For the sulfur concentration in products, the trend is dif-
ferent than that for metals due to the different impact of metals deposits on HDS and HDM active
sites. It should be commented that the catalyst used is the first catalyst in a series of three HDT
catalysts for upgrading heavy oils, so that it is designed for metals removal. That is why its metal
retention capacity is uniform (constant HDM), while sulfur removal exhibits an increasing
3.5
Sulfur in products (wt.%)
3.0
2.5
2.0
222
Metals in products (ppm)
185
148
111
0 100 200 300 400 500

Time-on-stream (h)
Figure 2.21 Profiles of sulfur and metals content in products vs TOS. (lines) Calculated with the model,
(symbols) experimental data.
tendency. The third stage of catalyst deactivation is not observed because the maximum TOS is 480
hours. However, there would be the increasing tendency in both sulfur and metals concentration
profiles.
Although typical experimental data of heavy crude oil HDT unclearly indicates the change of
regime from coke formation to metals deposits, the understanding of the behavior may be improved
with the observation of calculated parameters, for example, impurities content on catalyst and the
sulfur and metals concentration in products along TOS.
2.6.8.2 Coke and Metals on Catalyst

The simulated profiles of COC and MOC on catalyst at different TOS along the reactor length can be
observed in Figures 2.22 and 2.23, respectively. Coke is formed uniformly along the catalyst bed
axial length in the first ten hours of the operation when metals deposition has no notorious effect.
The coke content profile along the reactor length increases rapidly during the first 100 hours of TOS
Top 0.0
0.2
0.4
0.6
0.8
Bottom 1.0
30 32 34 36 38
Coke-on-catalyst (kg/kgcat)
Figure 2.22 Evolution of calculated COC profiles along TOS. 50 hours (——), 100 hours (---), 150 hours ( ),
200 hours (—— ——), 300 hours (—— ——), and 500 hours (—— ——).
Top 0.0
0.2
0.4
0.6
0.8
Bottom 1.0
0 10 20 30 40 50 60 70 80 90 100
Metals-on-catalyst (kg/kgcat)
Figure 2.23 Evolution of calculated metals-on-catalyst profiles along TOS. 50 hours (——), 100 hours (---),
150 hours ( ), 200 hours (—— ——), 300 hours (—— ——), and 500 hours (—— ——).
until the coke desorption is favored to pursue the chemical equilibrium. The solid line profile indi-
cates the quantity of coke on catalyst bed at 50 hours of TOS. Later at 100 hours (dashed line), the
highest values of coke deposits on catalyst are reached. The values of coke deposits are depleted
after 100 hours of TOS and do not increase anymore with respect to TOS. A portion of the released
catalyst sites by coke desorption may be occupied by metals deposits. This can be observed at the
beginning of the second stage of catalyst deactivation in the concentration of sulfur and metals pro-
files since they coincide with the curve at 100 hours of TOS in Figure 2.21. An increasing tendency
in coke profile along the reactor length is because of the effect of metals deposits and the consump-
tion of hydrogen by HDS and HDM reactions. The coke on catalyst increases along the reactor
length since near the outlet hydrogen partial pressure is lower due to hydrogen consumption. It
is clearly observed that the COC is higher near the 100 hours of TOS than the end of operation.
As time passes, the increasing tendency of coke along reactor length is contrasted due to the accu-
mulation of metals deposits majorly near the inlet of the catalyst bed.
The high quantities of both COC and MOC are due to the high pore volume per catalyst particle.
Most of metals deposits are at the inlet of the catalyst bed since a high amount of metals in liquid
phase is removed in this section. The profile of MOC increases continuously during TOS.
Figure 2.24 shows that the profile of the conversion of HDS decreases continuously as TOS passes
because of the effect of COC and MOC. In the first 100 hours of operation, the major reduction of
HDS conversion is observed due to the rapid formation of coke. Afterward, the effect of metals dep-
osition can be observed. The decreasing behavior of calculated HDS conversion profile is due to the
poisoning of catalytic HDS sites by coke and metals during TOS. At 500 hours, HDS conversion of
lower than 5% is obtained in the first 20% of the reactor length due to the limitation by COC and
MOC near the inlet of the reactor. The experimental outlet values of HDS conversion are plotted
and compared with the outlet value of the calculated profiles at the corresponding TOS.
The highest part of the HDM conversion takes place in the first third part of the reactor length as
shown in Figure 2.25. The profile of HDM conversion along the reactor length decreases until 100
hours of TOS. Later, the HDM conversion profile maintains its values since the metals deposit accu-
mulation effect is not considerable on HDM reactions and the depletion of coke formation on cat-
alyst is slowed. This latter observation can also be proved with the low value of βM. The depletion of
Top 0.0
0.2
0.4
0.6
0.8
290 h 190 h 50 h
Bottom 1.0
150 h
0 10 20 30 40 50
HDS conversion (%)
Figure 2.24 Evolution of HDS conversion profiles along TOS. Calculated HDS conversion at 50 hours (——),
100 hours (---), 150 hours ( ), 200 hours (—— ——), 300 hours (—— ——), and 500 hours (—— ——) and outlet
experimental HDS conversion (○).
Nomenclature 105
Top 0.0
0.2
0.4
0.6
0.8
290 h
Bottom 1.0 150 h
0 10 20 30 40 50 60 70
HDM conversion (%)
Figure 2.25 Evolution of HDM conversion profiles along TOS. Calculated HDM conversion at 50 hours (——),
100 hours (---), 150 hours ( ), 200 hours (—— ——), 300 hours (—— ——), and 500 hours (—— ——) and outlet
experimental HDM conversion (○).
the calculated outlet HDM conversion along TOS agrees with the behavior of the experimental
values. In the first 20% of reactor length, the highest slope in HDM conversion agrees with the low-
est slope of HDS conversion at 500 hours of operation.
2.6.9 Conclusion
Experimental data of the HDT of a heavy crude oil in a bench-scale reactor were generated and
modeled at 380 C, 98 bar, 0.5 hours−1 and 890 m3/m3 H2/oil ratio during 500 hours of TOS. The
typical behavior of the first and second stages of catalyst deactivation in heavy oils was observed
for the profiles of sulfur and metals concentrations in products. The first stage of catalyst deacti-
vation due to coke deposition was accurately described with the deactivation model. The amounts
of COC and MOC and their effect on the concentration of sulfur and metals in products along TOS
were included in the model. The model shows good agreement with experimental data and offers
the advantage of a better understanding of the mechanism of the first and second stages of catalyst
deactivation.
Nomenclature
Symbols
A constant, depending on catalyst, feed stock, and temperature
A0 pre-exponential factor of reaction rate coefficient, kg/(m2 h)
Ao active sites fresh surface area, marea2/kgcat
A0C effective surface area of catalyst under use for coke deposition expressed, m2/kg
A0S effective surface area of catalyst under use for desulfurization deposition expressed,
m2/kg
Arc remanent surface area for coke deposition, marea2/kgcat

Ari remanent active superficial area for remotion of heteroatom i, marea2/kgcat
Ars fresh surface area of HDS active sites, marea2/kgcat
ArV hydrodevanadization active sites remanent surface area, marea2/kgcat
a constant
a1 deactivation rate for hydrocracking
a2 deactivation rate for desulfurization and demetallation
b ratio of desulfurization rate constant of the fresh core to that of the fouled shell
C local reactant concentration, kmol/m3
C0 carbon content of the catalyst, kg carbon/kg catalyst
CAS surface reactant concentration, kmol/m3
Cash,0 initial concentration of metal determined as ash, mass% (kg/kgoil) × 100
Ccoke equilibrium local coke concentration
Ccoke,0 initial concentration in used oil, mass%, (kg/kgoil) × 100
CF reactant concentration entering reactor, ppm
Ch concentration of hydrogen, kg/kg
Cm concentration of metals
Cmax
m maximum concentration of metals
Cm0 concentration of metals at particle surface
Cmp concentration of organometallic compounds in the liquid in the pores, mol/m3
Cp,i product concentration, wt.% or wt-ppt
Cs concentration of sulfur-containing molecules in the liquid in the pores, mol/m3
Cs0 concentration of sulfur at particle surface
CS sulfur concentration in products, kgS/kgl
Cs0 carbon content during first 50 hours of time-on-stream
Cv molar vanadium concentration, inside the pore, kmol/m3
CV vanadium concentration in products, kgV/kgl
Cvo molar vanadium concentration, outside the catalyst particle, kmol/m3
D0 fresh catalyst effective diffusivity
DA diffusivity in single capillary, m2/s
Dm0 effective diffusivity of the organometallic compounds at zero time, m2/s
DP particle effective diffusivity, m2/s
Dz diffusivity of metal-bearing species in the deposited layer, m2/s
d constant
dc distance from pellet center for coke, m
dv distance from pellet center for vanadium, m
dpe equivalent particle diameter, cm
dR inner diameter of reactor, cm
EA activation energy for reaction rate coefficient, cal/mol
Em activation energy for HDM, kcal/mol
Es activation energy for HDS, kcal/mol
Ev activation energy for HDV, kcal/mol
f(r) pore size distribution function
g constant
K adsorption rate constant, g of catalyst/g of feed
Nomenclature 107
k1 first-order observed reaction rate constant, hour−1

k0 demetallation rate constant for original active sites, second−l
kAi apparent rate constant in the initial rapid deactivation period
Kads adsorption constants of asphaltenes
kc1 intrinsic rate constant for coke formation, kg/m2 h
kc2 intrinsic rate constant for coke hydrogenation, kg/m2 h
kd deactivation constant, second−1
kd1 deactivation constants for hydrocracking
kd2 deactivation constants for desulfurization and demetallation
kfm rate constant of demetallation, kg/m2 h
kfs rate constant of desulfurization, kg/m2 h
ki reaction rate constant
km rate constant of fresh catalyst for demetallation, kg/m2 h
kms rate constant for demetallization in the shell, m3/kg s
ks rate constant of fresh catalyst for desulfurization, kg/m2 h
kSA reaction rate constant per unit surface area
kss rate constant for desulfurization in the shell, m3/kg s
kv first-order surface reaction rate constant for vanadium removal, m s−1
kz demetallation rate constant in the deposited layer, second−1
kn rate coefficient at η 0, h−1atm−1
L total bed length in plug-flow reactor, cm
l fractional bed length in plug-flow reactor
L
LHSV liquid-hourly space-velocity, hour−1
M molecular weight ratio, metal sulfides to metal
M0 maximum amount of metal deposits when the pore is completely filled, kg of metal/kg of
catalyst
Md present amount of metal deposits, kg of metal/kg of catalyst
Mm average molar mass of metals
MMS molecular weight of metal sulfide deposit
Mp molecular weight of metal deposit, kg/kmol
MO moles of reactant per unit volume of catalyst
MOC concentration of metals-on-catalyst, wt.%
Mw molecular weight of residue
m deactivation order
N Avogadro’s number, molecules/Kmol
n correlation coefficient or reaction order
np number of pores per unit surface area, m/s
P reactor pressure, kPa
Pv pore volume of fresh catalyst, m3/kg
qc mass quantity of coke by volumetric unity, kgcoke/mcat3
Qi dimensionless mass quantity of heteroatom i on catalyst, kgi/kgcat
Q quantity of contaminant on catalyst, wt.%
Q0 concentration of deposited metals corresponding to a monolayer, kmol/kg
Qc dimensionless quantity of coke on catalyst
r mean random pore radius, m
r0 equivalent radius of catalyst pellet, m

rA actual pore radius, m
ra average pore radius
Rc2 slow coke deposition rate, h−1
Rg ideal gases constant, cal/mol K
rHDM rate of hydrodemetallization reaction, mol cm−3 s−1
ri reaction rate of remotion of compound of heteroatom i from process stream, kgi/m3 h
rM radius of residue molecule, m
rm radius of molecule
rmin minimum pore radius
rMS metal sulfides deposition rate, h−1
rp radial position in catalyst pellet, m
rp0 pore radius of fresh catalyst, m
RPD remaining pore diameter
rpm catalyst pore radius which neglects coke thickness, m
rs rate of desulfurization, kg/m3 h
rv rate of vanadium removal, kg/m3 h
S specific surface area of catalyst
t any time-on-stream (>50 h)
tb time in which it is presented φ1 and φ2
td dimensionless time
tm maximal life of pore
t∞ catalyst life
T temperature, K
ul mass flux velocity, kgliq/m2 s
UMOC unit conversion factor of Eq. (2.61)
V a dimensionless time-measuring variable
Vdep deposition volume per mole vanadium, m3 kmol
W local mass of adsorbed poison per unit poison-free area, kg/m2
w cumulative feed-to-catalyst ratio, a pseudo-time coordinate, g of catalyst/g of feed
Wf fresh catalyst weight, g
WMS metal sulfides weight, g
Ws value of W corresponding to complete active site poisoning
X distance from particle center, m
X0 slab half-thickness, m
Xc mass fraction of carbon in the residue
XMOC fractional concentration of metals-on-catalyst
xs fraction of surface area not covered with coke
ya constant
yb constant
ym constant
yn constant
z axial reactor coordinate, cm
Nomenclature 109
Greek letters
α deactivation constant
α1 rate constant of diminution on sites concentration Type I
α2 rate constant of diminution on sites concentration Type II
αc proportional constant
aD dimensionless reactant concentration
αp relative pore plugging/active site poisoning tendency
αDK deactivation parameter for diesel to kerosene, hour−1
αFG deactivation parameter for diesel to kerosene, hour−1
αFN deactivation parameter for feed to naphtha, hour−1
αNG deactivation parameter for naphtha to gas, hour−1
αKN deactivation parameter for kerosene to naphtha, hour−1
αV,αc,βV,βc, and γ V effective deposition of vanadium and nickel parameters
β fitting parameter of Eq. (2.60)
βD dimensionless distance along the pore
γ fitting parameter of Eq. (2.60)
Γ relative rate constant of desulfurization
γD ratio of the instantaneous pore radius to the initial pore radius
δ fitting parameter of Eq. (2.60)
Δrpc thickness of monolayer deposited metal, m
ϵ number of metal sulfide molecules per molecule of reactant
0 initial catalyst bed porosity
B catalyst bed void fraction, cm3G + L cm3r
ϵp porosity of catalyst particle
t catalyst bed porosity at time t
εS solid fraction in reactor, mS3/mR3
Θ partial surface poisoning, dimensionless
ζ number of metal atoms per reactant molecule
η effectiveness factor
η0 initial effectiveness factor
ηm(t) relative activity of demetallation
ηs (t) relative activity of desulfurization
θ relative catalyst age
θC1 loss in catalyst porosity due to fast coke
θC2 loss in catalyst porosity due to slow coke
θM metal distribution factor
θMS loss in catalyst porosity due to metal sulfides
θT total loss in catalyst porosity due to pore-plugging
Λ relative rate constant of demetallation
λ0 ratio of molecular to pore radius defined
ξ fractional distance from the center of the pellet
ρ density oil, kg/m3
ρc density coke, kg/m3
ρcat density catalyst, kg/m3
ρf density of phase f at process conditions, gf cm3f
ρi density of deposit on catalyst, kgi/mi3

ρL density of deposited layer, kg/m3
ρm density of metal sulfides
ρmd density of metal deposited
ρoil feedstock density, kgoil/moil3
ρp adsorbed poison density, kg/m3
ρv density vanadium, kg/m3
τ space-time, hour
τD dimensionless time for the restricted-diffusivity
τf tortuosity factor
φ activity function
φ1 loss of activity due to site coverage
φ2 loss of activity due to sites coverage and pore blockage
φc dimensionless concentration of coke
φs steady-state catalyst activity
φsf activity function for sulfur removal
φm activity function for metal removal
φv dimensionless concentration of vanadium
Φ The Thiele modulus for system with restricted diffusivity
Øm The Thiele modulus for demetallation
ψ dimensionless volumetric quantity of deposit on catalyst, mi3/mcat3
ψ max dimensionless volumetric maximum quantity of deposit on catalyst,
mi3/mcat3
Ω fraction of initial porosity occupied by the solid deposited
Subscripts and Superscripts

H2 molecular hydrogen
i component designation number
j number of experimental data point
R reactor
S sulfur
V vanadium
Abbreviation
CSTR, continuous stirred tank reactor
CCR, Conradson carbon residue
COC, coke–on-catalyst
DMDS, dimethyl disulfide
EOR, end-of-run
HDM, hydrodemetallization
HDO, hydrodeoxygenation
HDS, hydrodesulfurization
HDT, hydrotreating
References 111
HDV, hydrodevanadization
LHSV, liquid-hourly space-velocity
MOC, metal-on-catalyst
MOR, middle-of-run
RDS, residue hydrodesulfurization
SOR, start-of-run
SQP, sequential quadratic programming
SRGO, straight-run gas oil
TBR, trickle-bed reactor
TOS, time-on-stream
USC, ultimate storage capacity
VGO, vacuum gas oil
References
Ahn, B.J. and Smith, J.M. (1984). Deactivation of hydrodesulfurization catalysts by metals deposition.
AIChE Journal 30: 739–746.
Alvarez, A. and Ancheyta, J. (2008). Modeling residue hydroprocessing in a multi-fixed-bed reactor
system. Applied Catalysis A: General 351: 148–158.
Alvarez, A., Ancheyta, J., Centeno, G., and Marroquín, G. (2011). A modeling study of the effect of reactor
configuration on the cycle length of heavy oil fixed-bed hydroprocessing. Fuel 90: 3551–3560.
Ancheyta, J. (2016). Deactivation of Heavy Oil Hydroprocessing Catalysts: Fundamentals and Modeling.
New Jersey: Wiley.
Ancheyta, J. and Lopez, F. (2000). Analysis of deactivation models based on time-on-stream (TOS). Theory
for fluid catalytic cracking process. Revista de La Sociedad Química de México 44: 183–187.
Ancheyta, J., Betancourt, G., Marroquin, G. et al. (2001). An exploratory study for obtaining synthetic
crudes from heavy crude oils via hydrotreating. Energy & Fuels 15: 120–127.
Ancheyta, J., Betancourt, G., Centeno, G. et al. (2002). Catalyst deactivation during hydro processing of
maya heavy crude oil: evaluation at constant operating conditions. Energy & Fuels 16: 1438–1443.
Ancheyta, J., Betancourt, G., Centeno, G., and Marroquín, G. (2003). Catalyst deactivation during
hydroprocessing of Maya heavy crude oil. (II) Effect of temperature during time-on-stream. Energy &
Fuels 17: 462–467.
Ancheyta, J., Rana, M., and Furimsky, E. (2005). Hydroprocessing of heavy petroleum feeds: tutorial.
Catalysis Today 109: 3–15.
Arbabi, S. and Sahimi, M. (1991a). Computer simulations of catalyst deactivation-I. Model formulation
and validation. Chemical Engineering Science 46: 1739–1747.
Arbabi, S. and Sahimi, M. (1991b). The effect of morphological, transport and kinetic parameters on the
performance. Chemical Engineering Science 46: 1749–1755.
Baghalha, M., Mohammadi, M., and Ghorbanpour, A. (2010). Coke deposition mechanism on the pores
of a commercial Pt–Re/γ-Al2O3 naphtha reforming catalyst. Fuel Processing Technology 91: 714–722.
Bartholomew, C.H. (1982). Catalysis reviews: science and reforming and methanation carbon deposition
in steam reforming and methanation. Catalysis Reviews: Science and Engineering 24: 67–112.
Bartholomew, C.H. (2001). Mechanisms of catalyst deactivation. Applied Catalysis A: General 212: 17–60.
Beaton, W.I. and Bertolacini, R.J. (1991). Resid hydroprocessing at Amoco. Catalysis Reviews 33: 281–317.
Beuther, H., Larson, O.A., and Perrotta, A.J. (1980a). The mechanism of coke formation on catalysts. In:
Studies in Surface Science and Catalysis, vol. 6. Issue C.
Bourseau, P., Muratet, G., Saillour, C., and Toulhoat, H. (1985). Evolution au cours du temps d’un grain
de catalyseur d’hydrodémétallisation: module et simulation numérique de résultats expérimentaux.
Revue de l’Institut Français Du Pétrole 40: 595–607.
Calderón, C.J. and Ancheyta, J. (2017). Dynamic modeling and simulation of a slurry-phase reactor for
hydrotreating of oil fractions. Energy & Fuels 31: 5691–5700.
Callejas, M.A., Martn ́ ez, M.T., Fierro, J.L.G. et al. (2001). Structural and morphological study of metal
deposition on an aged hydrotreating catalyst. Applied Catalysis A: General 220: 93–104.
Castañeda, L.C., Muñoz, J.A., and Ancheyta, J. (2014). Current situation of emerging technologies for
upgrading of heavy oils. Catalysis Today 220–222: 248–273.
Centeno, G. (2008). Catalyst deactivation during hydroprocessing of heavy oils. PhD thesis (In Spanish).
Instituto Tecnológico de Ciudad Madero, México
Chao, Y.C., Liaw, H.J., and Huang, H.P. (1991). A mathematical model for the catalyst deactivation in a
commercial residue hydrodesulfurization reactor system. Chemical Engineering Communications 104:
267–290.
Corella, J., Bilbao, R., Molina, J.A., and Artigas, A. (1985). Variation with time of the mechanism,
observable order, and activation energy of the catalyst deactivation by coke in the FCC process.
Industrial and Engineering Chemistry Process Design and Development 24: 625–636.
Dautzenberg, F., Klinken, J., Pronk, K. et al. (1978). Catalyst deactivation through pore mouth plugging
during residue desulfurization. In: Chemica Reaction Engineering, vol. 65, 254–267.
Elizalde, I. and Ancheyta, J. (2014a). Application of a three-stage approach for modeling the complete
period of catalyst deactivation during hydrotreating of heavy oil. Fuel 138: 45–51.
Elizalde, I. and Ancheyta, J. (2014b). Modeling catalyst deactivation during hydrocracking of
atmospheric residue by using the continuous kinetic lumping model. Fuel Processing Technology 123:
114–121.
Eser, S., Jenkins, R.G., Derbyshire, F.J., and Malladi, M. (1986). Carbonization of coker feedstocks and
their fractions. Carbon 24: 77–82.
Félix, G., Ancheyta, J., and Trejo, F. (2019). Sensitivity analysis of kinetic parameters for heavy oil
hydrocracking. Fuel 241: 836–844.
Ferreira, C., Marques, J., Tayakout, M. et al. (2010). Modeling residue hydrotreating. Chemical
Engineering Science 65: 322–329.
Flores, R.M. (2014). Chapter 1 – introduction and principles. In: Coal and Coalbed Gas (ed. R.M. Flores),
1–40. Boston.
Froment, G.F. and Bischoff, K.B. (1962). Kinetic data and product distributions from fixed bed catalytic
reactors subject to catalyst fouling. Chemical Engineering Science 17: 105–114.
Froment, G.F., Bischoff, K.B., and Wilde, J.D. (2010). Chemical Reactor Analysis and Design, 3e.
USA: Wiley.
Furimsky, E. and Massoth, F.E. (1999). Deactivation of hydroprocessing catalysts. Catalysis Today 52:
381–495.
Galiasso, R. (2007). Effect of recycling the unconverted residue on a hydrocracking catalyst operating in
an ebullated bed reactor. Fuel Processing Technology 88: 779–785.
Galiasso, R., Blanco, R., Gonzalez, C., and Quinteros, N. (1983). Deactivation of hydrodemetallization
catalyst by pore plugging. Fuel 62: 817–822.
Gary, J.H., Handwerk, G.E., and Kaiser, M.J. (2007). Petroleum Refining: Technology and Economics, 4e.
New York: Marcel Dekker, Inc.
Gianetto, A. and Specchia, V. (1992). Trickle-bed reactors: state of art and perspectives. Chemical
Gray, M., Zhao, Y., McKnight, C. et al. (1999). Coking of hydroprocessing catalyst by residue fractions of
bitumen. Energy & Fuels 13: 1037–1045.
References 113
Gualda, G.A. and Kasztelan, S.A. (1996). Initial deactivation of residue hydrodemetallization catalysts.
Journal of Catalysis 161: 319–337.
Hauser, A., Marafi, A., Stanislaus, A., and Al-Adwani, A. (2005). Relation between feed quality and coke
formation in a three-stage atmospheric residue desulfurization (ARDS) process. Energy & Fuels 19:
544–553.
Haynes, H. and Leung, K. (1983). Catalyst deactivation by pore plugging and active site poisoning
mechanisms. Chemical Engineering Communications 23: 161–179.
Hegedus, L. and McCabe, R. (1994). Catalyst poisoning. Catalysis Reviews 36: i–ii.
Idei, K., Yamamoto, Y., and Yamazaki, H. (1995). Kinetic model and analysis method of catalysts
deactivation in hydrodesulfization of light and heavy oils. Kagaku Kōgaku Ronbunshū 21: 972–983.
Idei, K., Yamamoto, Y., and Takehara, S. (1998). Generalized deactivation equation of catalyst in
hydrodesulfurization of light and heavy oils. Kagaku Kōgaku Ronbunshū 24: 653–659.
Idei, K., Takahashi, T., and Kai, T. (2003). Estimation of coke and metal deposition distribution within
hydrodesulfurization catalyst pore at the last stage of operation. Journal of the Japan Petroleum
Institute 46: 45–52.
Isaza, M., Pachon, Z., Kafarov, V., and Resasco, D. (2000). Deactivation of Ni–Mo/Al2O3 catalysts aged in
a commercial reactor during the hydrotreating of deasphalted vacuum residuum. Applied Catalysis A:
General 199: 263–273.
Jacob, S.M., Gross, B., Voltz, S.E., and Weekman, V.W. (1976). A lumping and reaction scheme for
catalytic cracking. AIChE Journal 22: 701–713.
Jurado, J. and Ancheyta, J. (2022). Reactor model for heavy oil hydrotreating with catalyst deactivation
based on vanadium and coke deposition. Energy & Fuels 36: 11132–11141.
Kam, E.K.T., Al-Shamali, M., Juraidan, M., and Qabazard, H. (2005). A hydroprocessing multicatalyst
deactivation and reactor performance model-pilot-plant life test applications. Energy & Fuels 19:
753–764.
Khang, S. and Mosby, J. (1986). Catalyst deactivation due to deposition of reaction products in
macropores during hydroprocessing of petroleum residuals. Industrial & Engineering Chemistry
Process Design and Development 25: 437–442.
Kobayashi, S., Kushiyama, S., Aizawa, R. et al. (1987). Kinetic study on the hydrotreating of heavy oil. 2.
Effect of catalyst pore size. Industrial & Engineering Chemistry Research 26: 2245–2250.
Kodama, S., Nitta, H., Takatsuka, T., and Yokoyama, T. (1980). Simulation of residue
hydrodesulfurization reaction based on catalyst deactivation model. Journal of The Japan Petroleum
Institute 23: 310–320.
Kohli, K., Prajapati, R., Maity, S.K. et al. (2016). Deactivation of hydrotreating catalyst by metals in resin
and asphaltene parts of heavy oil and residues. Fuel 175: 264–273.
Korsten, H. and Hoffmann, U. (1996). Three-phase reactor model for hydrotreating in pilot trickle-bed
reactors. AIChE Journal 42: 1350–1360.
Kressmann, S., Morel, F., Harl, V., and Kasztelan, S. (1998). Recent developments in fixed-bed catalytic
residue upgrading. Catalysis Today 43: 203–215.
Kumar, V.R., Balaraman, K.S., Rao, V.S.R., and Ananth, M.S. (1997). Modelling of hydrotreating process
in a trickle-bed reactor. Petroleum Science and Technology 15: 283–295.
Lee, L.-S., Chen, Y.-W., Huang, T.-N., and Pan, W.-Y. (1989). Four-lump kinetic model for fluid catalytic
cracking process. The Canadian Journal of Chemical Engineering 67: 615–619.
Marafi, A., Stanislaus, A., and Furimsky, E. (2010a). Kinetics and modeling of petroleum residues
hydroprocessing. Catalysis Reviews – Science and Engineering 52: 204–324.
Marafi, M., Stanislaus, A., and Furimsky, E. (2010b). Handbook of spent hydroprocessing catalysts, 1e.
short term catalyst deactivation. Fuel 100: 193–199.
Maxted, E.B. (1951). The poisoning of metallic catalysts. Advances in Catalysis 3: 129–178.
Mederos, F.S., Rodríguez, M.A., Ancheyta, J., and Arce, E. (2006). Dynamic modeling and simulation of
catalytic hydrotreating reactors. Energy & Fuels 20: 936–945.
Mederos, F.S., Elizalde, I., and Ancheyta, J. (2009). Steady-state and dynamic reactor models for
hydrotreatment of oil fractions: a review. Catalysis Reviews 51: 485–607.
Mederos, F.S., Ancheyta, J., and Elizalde, I. (2012). Dynamic modeling and simulation of hydrotreating of
gas oil obtained from heavy crude oil. Applied Catalysis A: General 425–426: 13–27.
Menon, P.G. (1990). Coke on catalysts-harmful, harmless, invisible and beneficial types. Journal of
Molecular Catalysis 59: 207–220.
Miki, Y., Yamadaya, S., Oba, M., and Sugimoto, Y. (1983). Role of catalyst in hydrocracking of heavy oil.
Mitchell, P.C.H. (1990). Hydrodemetallisation of crude petroleum: fundamental studies. Catalysis Today
7: 439–445.
Mochida, I., Zhao, X., Sakanishi, K. et al. (1989). Structure and properties of sludges produced in the
catalytic hydrocracking of vacuum residue. Industrial & Engineering Chemistry Research 28: 418–421.
Moghadassi, A., Amini, N., Fadavi, O., and Bahmani, M. (2011). Hydrocracking lumped kinetic model
with catalyst deactivation in arak refinery hydrocracker unit. Journal of Petroleum Science and
Technology Petroleum Science and Technology 1: 31–37.
Mohammed, J., Mahmoud, A., Hasan, Q., and Kam, K.T. (2006). A refined hydroprocessing multicatalyst
deactivation and reactor performance model pilot-plant accelerated test applications. Energy & Fuels
20: 1354–1364.
Monzón, A., Romeo, E., and Borgna, A. (2003). Relationship between the kinetic parameters of different
catalyst deactivation models. Chemical Engineering Journal 94: 19–28.
Nagaishi, H., Chan, E.W., Sanford, E.C., and Gray, M.R. (1997). Kinetics of high-conversion
hydrocracking of bitumen. Energy & Fuels 11: 402–410.
Newson, E. (1975). Catalyst deactivation due to pore-plugging by reaction products. Industrial and
Engineering Chemistry Process Design and Development 14: 27–33.
Nielsen, J.R. and Trimm, D.L. (1977). Mechanisms of carbon formation on nickel-containing catalysts.
Oliveira, L. and Biscaia, E. (1989). Catalytic cracking kinetic models. Parameter estimation and model
evaluation. Industrial & Engineering Chemistry Research 28: 264–271.
Oyekunle, L. and Ikpekri, O. (2004). Modeling of hydrodesulfurization catalysts. I. Influence of catalyst
pore structures on the rate of demetallization. Industrial & Engineering Chemistry Research 43:
6647–6653.
Oyekunle, L.O., Ikpekri, O.B., and Jaiyeola, A. (2005). Modelling of hydrodesulfurization catalysts: II.
Effects of catalyst pore structures on deactivation by metal deposits. Catalysis Today 109: 128–134.
Pereira, C.J. (1990). Metal deposition in hydrotreating catalysts. 1. A regular perturbation solution
approach. Industrial and Engineering Chemistry Research 29: 512–519.
Rajagopalan, K. (1979). Influence of catalyst pore size on demetallation rate. Industrial & Engineering
Chemistry Process Design and Development 18: 459–465.
Rana, M.S., Sámano, V., Ancheyta, J., and Diaz, J.A.I. (2007). A review of recent advances on process
technologies for upgrading of heavy oils and residua. Fuel 86: 1216–1231.
Rashidzadeh, M., Ahmad, A., and Sadighi, S. (2010). Studying of catalyst deactivation in a commercial
hydrocracking process (ISOMAX). Journal of Petroleum Science and Technology 1: 46–54.
Richardson, S.M., Nagaishi, H., and Gray, M.R. (1996). Initial coke deposition on a NiMo/γ-Al2O3
bitumen hydroprocessing catalyst. Industrial & Engineering Chemistry Research 35: 3940–3950.
References 115
Sau, M., Narasimhan, C.S.L., and Verma, R.P. (1997). A kinetic model for hydrodesulfurisation. In:
Studies in Surface Science and Catalysis, vol. 106 (ed. G.F. Froment, B. Delmon, and P. Grange),
421–435.
Shah, Y.T. and Mhaskar, R.D. (1976). Optimum quench location for a hydrodesulfurization reactor with
time-varying catalyst activity. Industrial & Engineering Chemistry Process Design and Development 15:
400–406.
Shah, Y.T., Mhaskar, R.D., and Paraskos, J.A. (1976). Optimum quench location for a
hydrodesulfurization reactor with time varying catalyst activity. Industrial & Engineering Chemistry
Process Design and Development 15: 400–406.
Shimura, M., Shiroto, Y., and Takeuchi, C. (1986). Effect of catalyst pore structure on hydrotreating of
heavy oil. Industrial & Engineering Chemistry Fundamentals 25: 330–337.
Skala, D., Saban, M., Orlovic, M. et al. (1991). Hydrotreating of used oil: prediction of industrial trickle-
bed operation from pilot-plant data. Industrial & Engineering Chemistry Research 30: 2059–2065.
Song, C., Nihonmatsu, T., and Nomura, M. (1991). Effect of pore structure of nickel-molybdenum/
alumina catalysts in hydrocracking of coal-derived and oil sand derived asphaltenes. Industrial &
Speight, J.G. (1999). The Desulfurization of Heavy Oils and Residua. New York: CRC Press.
Spry, J.C. and Sawyer, W.H. (1975). Configurational diffusion effects in catalytic demetallization of
petroleum fractions. Chemical Engineering Science 24: 16–20.
Tamm, P.W., Harnsberger, H.F., and Bridge, A.G. (1981). Effects of feed metals on catalyst aging in
hydroprocessing residuum. Industrial and Engineering Chemistry Process Design and Development 20:
262–273.
Thakur, D. and Thomas, M. (1985). Catalyst deactivation in heavy petroleum and synthetic crude
processing: a review. Applied Catalysis 15: 197–225.
Toulhoat, H., Szymanski, R., and Plumail, J. (1990). Interrelations between initial pore structure,
morphology and distribution of accumulated deposits, and lifetimes of hydrodemetallisation catalysts.
Toulhoat, H., Hudebine, D., Raybaud, P. et al. (2005). THERMIDOR: a new model for combined
simulation of operations and optimization of catalysts in residues hydroprocessing units. Catalysis
Today 109: 135–153.
Trimm, D.L. (1977). The formation and removal of coke from nickel catalyst. Catalysis Reviews 16:
155–189.
Trimm, D.L. (1983). Catalyst design for reduced coking (review). Applied Catalysis 5: 263–290.
Van, F., Nevicato, D., Pitault, I. et al. (1996). Fluid catalytic cracking: modelling of an industrial riser.
Applied Catalysis A: General 138: 381–405.
Verstraete, J.J., Le Lannic, K., and Guibard, I. (2007). Modeling fixed-bed residue hydrotreating
processes. Chemical Engineering Science 62: 5402–5408.
Voorhies, A. (1945). Carbon formation in catalytic cracking. Industrial & Engineering Chemistry Research
37: 318–322.
Ward, J.W. (1993). Hydrocracking processes and catalysts. Fuel Processing Technology 35: 55–85.
Weekman, V. (1968). A model of catalytic cracking conversion in fixed, moving, and fluid-bed reactors.
Industrial and Engineering Chemistry Process Design and Development 7: 90–95.
Wheeler, A. (1951). Reaction rates and selectivity in catalyst pores. Advances in Catalysis 3: 249–327.
116
Simulation of the Oxidative Regeneration of Coked Catalysts:

Kinetics, Catalyst Pellet, and Bed Levels
Sergey Zazhigalov 1,2, Osman Abdulla2, and Andrey Zagoruiko1,2
1
Boreskov Institute of Catalysis, Novosibirsk, Russia
2
Tyumen State University, Tyumen, Russia
3.1 Introduction
Catalysts are used in most petroleum refining and petrochemical processes and play a key role in
their execution. Currently, at least 90% of all chemical products in the world are produced by the
catalytic method (Meyers 2016; Speight 2016).
Most hydraulic oil refining processes use catalysts based on molybdenum and tungsten. This is
explained by their unique catalytic properties: The ability to accelerate target reactions of the proc-
ess and low sensitivity to sulfur- and nitrogen-containing compounds, which are strong catalytic
poisons for other types of hydrogenation catalysts.
Hydrotreating catalysts are a combination of oxides (or sulfides) of active components (Ni, Co,
Mo, W, etc.) supported on a carrier, usually the active aluminum oxide. The carrier in hydrotreating
catalysts not only plays the role of an inert diluent but also participates in the formation of active
phases and also serves as a structural promoter that creates a specific porous structure, optimal for
the processing of specific raw materials.
The main sign of a decrease in catalyst activity is an increase in the sulfur content in the hydro-
treating product (Marafi et al. 2017). At the same time, it loses its activity, and the degree of desul-
furization of products during its use decreases—normal aging of the catalyst occurs. The most
typical process leading to blocking of active centers is the deposition of carbon-containing com-
pounds (coke) on the catalyst. If coking occurs in a reaction medium containing sulfur and oxygen,
then these elements partially transform into coke.
An important part of modern oil refining and petrochemical processes in moving and stationary
catalyst beds are the processes of regeneration of spent catalysts (Kern and Jess 2005; Zhou et al.
2020; Ferella et al. 2016). Such processes are necessary to remove coke deposited on the catalyst
granules during its operation.
Among them, from an energy and environmental point of view, the most effective ones are oxi-
dative regeneration processes based on the oxidation of coke (Furimsky 1991). In such processes,
complete removal of coke and other undesirable components is possible; however, such processes
are highly exothermic. Considering that most oil refining and petrochemical catalysts have
moderate heat resistance (the maximum permissible temperatures during their processing should
not exceed 450–650 C, depending on the type of catalyst), very stringent requirements are imposed
on the thermal conditions of the oxidative regeneration process.
Experimental development of the technological foundations of the process, as well as
thermal management strategies, is extremely complex, resource-intensive, and to some extent,
risky. This is due to the fact that in laboratory conditions, it is almost impossible to create
adiabatic conditions corresponding to the conditions of an industrial process. Accordingly,
in laboratory-scale experiments, it is impossible to reproduce the real thermal regime of an
industrial process.
To solve this problem, a development strategy was proposed that combines experimental work
and mathematical modeling in the following sequence:
• experimental study of the kinetics of reactions occurring during the oxidative regeneration of a
spent catalyst;
•• experimental study of the oxidative regeneration process in laboratory and pilot-scale reactors;
preliminary mathematical modeling of the industrial oxidative regeneration process;
•• comparison of simulation results with the results of experiments at the pilot reactor;
verification of the model, making adjustments to the mathematical model and model parameters;
• final mathematical modeling of the industrial process.
The following is a description of experimental studies on the regeneration of a coked catalyst in a

bench-scale reactor and mathematical modeling of the process.
3.2 Process Chemistry and Laboratory Experiments
3.2.1 Catalyst and Proposed Reactions

To carry out the experiments, we used a coked cobalt–molybdenum catalyst for the hydrotreatment
of the diesel fraction, worked out under real industrial operation conditions at the Omsk Refinery.
The characteristics of the catalyst are given in Table 3.1 and its external appearance is demonstrated
in Figure 3.1.
The process involves reactions of oxidation of catalyst metal sulfides and oxidation of coke.
Table 3.1 Properties of spent hydrotreating catalyst.
(a) Appearance Black extrudate with trefoil cross section,

granule length ~ 4 mm, outer diameter ~ 1.3 mm
(b) Bulk density ~900 g/l
(c) Chemical composition: % wt.
molybdenum (IV) sulfide 16.9
cobalt sulfide 4.0
carrier (aluminum oxide) 68.2
coke, including: 10.9
carbon 82.2
sulfur 7.7
nitrogen 0.9
hydrogen 9.2
118 3 Simulation of the Oxidative Regeneration of Coked Catalysts: Kinetics, Catalyst Pellet, and Bed Levels
Figure 3.1 Appearance of the original spent catalyst.
Oxidation of sulfides of active metals in the presence of atmospheric oxygen begins already at
200 C. In this case, metal sulfides are converted into the corresponding oxides and sulfur dioxide
(SO2) is released into the gas phase:
2МоS2 + 7О2 2МоО3 + 4SО2 31
2NiS + 3О2 2NiO + 2SО2 32
2СоS + 3О2 2СоО + 2SO2 33
A certain amount of nickel and cobalt oxides in the presence of oxygen reacts with sulfur dioxide
to form the corresponding sulfates, which do not decompose at a regeneration temperature, usually
not exceeding 550 C. The interaction of nickel and cobalt oxides with oxygen and sulfur dioxide is
described by the following reactions:
2CoO + O2 + 2SO2 2CoSO4 34
2NiO + O2 + 2SO2 2NiSO4 35
Coke oxidation begins at temperatures around 300 C, with most of the coke burning out in the
range of 300–450 C.
In general, carbon combustion is characterized by reactions that are highly exothermic. The
chemical interaction of carbon with oxygen, with the formation of carbon oxides, proceeds accord-
ing to the following schemes:
С + О2 СО2 + 395 4 kJ mol 36
С + 1 2 О2 СО + 110 4 kJ mol 37
Further transformations of the resulting oxides proceed according to the following schemes:
СО + 1 2 О2 СО2 + 285 0 kJ mol 38

С + СО2 2СО − 172 2 kJ mol 39
It is also possible that coke carbon interacts with water vapor, since during coke oxidation, water
vapor is almost always present in the reaction zone:
С + Н2 О СО + Н2 + 41 kJ mol 3 10
The sulfur contained in coke is oxidized into sulfur dioxide,

S + О2 SO2 + 298 kJ mol 3 11
and further into sulfuric anhydride:
2SО2 + О2 2SO3 3 12
Coke nitrogen is oxidized to the corresponding oxides:
2N + O2 2 NO 3 13
N + O2 NO2 3 14
2NO + O2 2NO2 3 15
Hydrogen oxidizes to water,
2Н2 + О2 2Н2 O + 136 kJ mol 3 16
Catalysts used in hydrotreating are quite resistant to heat treatment and generally retain their
composition and structure. However, at temperatures above 550 C, sublimation of molybdenum
is observed, which forces the process to be carried out below the specified temperature. With a
sharp increase (decrease) in temperature, the strength of the catalyst deteriorates. To preserve it,
it is not recommended to change the temperature by more than 40 K per hour. A significant tem-
perature difference between the catalyst and the circulating gas can cause destruction to the catalyst
structure—the permissible temperature gradient between the gas and catalyst particles should not
exceed 150 C (Froment and Biscoff 1999).
The coke formed on catalysts is not homogeneous in composition. The heterogeneity of compo-
sition and structure determines its unequal reactivity to oxidation during the burning process. Thus,
when burning coke, two temperature maxima are clearly distinguished. The first corresponds to the
combustion of an amorphous coke, which is the most rich in hydrogen; and the second corresponds
to the combustion of a slow-burning component of coke, consisting of highly condensed pseudogra-
phite structures, the second maximum is observed at a higher temperature in the reaction zone. The
ratio of these coke components is determined by the severity of the catalyst operating mode.
3.2.2 Reaction Kinetics

Studies of the kinetics of the ongoing reactions were carried out using the TG-DTA-MS method
(thermogravimetry–differential thermal analysis–mass spectroscopy). They were performed on a
Netzsch STA 409 PC instrument coupled to an SRS UGA 200 mass spectrometer. The measure-
ments were carried out using corundum crucibles. The DTA sample holder together with the cru-
cibles was preheated to 1000 C in an air stream.
Before the experiment, the catalyst sample was crushed to a fraction of 0.25–0.5 mm to eliminate
the influence of intra-diffusion limitations in the catalyst granules during reactions. The sample
weight in each experiment was 76–78 mg.
After loading the sample and installing the crucible on the holder, the temperature was raised to
50 C and held for 1 hour to extinguish the inertia of the furnace and establish thermal equilibrium.
Thermal curves of the samples were recorded in the range from 50 to 450 C with a heating rate of
3 or 1 K/min in a mixture of argon and argon–oxygen mixture (10% oxygen, 90% argon). The argon–
oxygen mixture was supplied to the device furnace at a rate of 10 (or 20) ml/min, and at the same
time, a protective gas (argon) was supplied to the scales at a rate of 40 (or 20) ml/min. Correction
curves were recorded under similar conditions for each experiment. A typical type of experimental
results is shown in Figure 3.2.
(a)
(b)
Figure 3.2 Results of TG-DTA-MS analysis of spent catalyst samples. Volume ratio of argon–oxygen
mixture and argon flows is 1:4 (calculated oxygen concentration in the mixture is 4% vol.), temperature
rise rate is 3 K/min (a) and 1 K/min (b).
According to the experimental data, it is clear that during regeneration, the mass of the catalyst
decreases by ~19%, which approximately corresponds to the mass loss calculated on the basis of
information obtained from data on the composition of the spent catalyst. In this case, three tem-
perature zones can be distinguished:
a) Low-temperature region (up to 200–220 С), where the mass loss is ~4%. Judging by the mass
spectrometry data of the exhaust gases, it can be assumed that desorption of water, as well as
light hydrocarbon components of coke, occurs in this region.
b) Region from 220 to 300–330 С with mass loss of about 2–3%. In this area, judging by the com-
position of the exhaust gases, active oxidation of surface sulfides occurs.
c) From 300–330 to 500 С with a mass loss of about 12–13%. At this range, active oxidation of coke
occurs.
When constructing the kinetic model, the low-temperature region (a) was excluded from consid-
eration, since it is not associated with chemical transformations. To describe the rate of oxidation of
sulfides and coke, it was proposed to use kinetic equations of the general form,
dθ E
= −k0 exp − θn C m
O2 3 17
dt RT
where θ is the relative mass content of the surface component on the catalyst (kg per kg of catalyst),
k0 is the pre-exponent (kg per kg of catalyst per sec), E is the activation energy, T is temperature (K),
CO2 is the oxygen concentration (volume fraction), n and m are the orders of reaction rates for the
surface component and oxygen, respectively.
Since, on the basis of experimental information, it is impossible to separate the oxidation rates of
cobalt and molybdenum sulfides, and for both sulfides, the rate of their oxidation is described by
one summary equation. The initial content θ is taken based on the mass composition of the initial
catalyst (0.21 for sulfides and 0.11 for coke).
As calculations using proprietary software have shown, the best description is achieved with the
following set of model parameters:
Figure 3.3 shows a comparison of experimental results and data calculated using the proposed
model. It can be seen that overall good agreement is achieved.
On the contrary, in the future, the model can be significantly improved, in particular, by taking
into account the influence of oxygen concentration in more detail (which will require additional
experiments), as well as by taking into account the degree of coke compaction and its dynamic
change at different catalyst heating rates.
3.2.3 Experimental Setup

Experiments to study the oxidative regeneration of the catalyst were carried out in a catalytic reac-
tor with a fixed bed of granular catalyst. The regenerating flow is formed by mixing nitrogen taken
from the vessel and air passing through filters and a dryer. The feed rate is regulated by mass flow
controllers. The general setup diagram is shown in Figure 3.4.
The analysis of the composition of the inlet and outlet gas flows was carried out using an MRU
Vario Plus industrial gas analyzer. O2 and CO2 were recorded with an accuracy of 0.2%, whereas
CO, NO, SO2, and C3H8 with an accuracy of 5 ppm.
Regeneration was carried out until the detection of CO2, CO, NO, and SO2 in the outlet reaction
mixture ceased.
The experiments were carried out in a reactor equipped with an internal heating system for the
catalyst bed, which has good internal thermal insulation and is also designed for catalyst loadings
of up to 2.5 liters, which allows it to be operated in conditions close to adiabatic. However, as it
turned out later, even in a relatively large reactor, there are very noticeable temperature gradients
along the radius of the catalyst bed and the reactor as a whole, that is, its actual operating
mode cannot be correctly classified as adiabatic. In this regard, it is more correct to call it
quasi-adiabatic.
(a)
0
–0,001
Mass change rate, %/s
–0,002
· 4%O2 3K/min
–0,003 · 4%O2 1K/min

· 10%O2 1K/min
–0,004
–0,005
0 100 200 300 400 500 600
Temperature, °C
(b)
1
0,9
0,8
SO2 content,relative units
0,7 · 4%O2 3K/min

0,6 · 4%O2 1K/min
0,5
0,4
0,3
0,2
0,1
0
0 100 200 300 400 500
Temperature, °C
(c)
1
0,9
CO2 content,relative units
0,8
0,7
0,6 · 4%O2 3K/min
0,5 · 4%O2 1K/min

0,4
0,3
0,2
0,1
0
0 100 200 300 400 500
Temperature, °C
Figure 3.3 The rate of change in the mass of the catalyst (a), the relative content of SO2 (b), and CO2 (c) in the
outlet gases under various conditions of oxidative regeneration of the catalyst. Points – experiment,
lines – calculations.
Air Filter
Dryer
MFC
Valve
Reactor
Mixer Gas analyzer
Valve
MFC
N2
Figure 3.4 Scheme of an experimental setup for oxidative regeneration of catalysts.
Outlet
Sampling
Thermocouples out
880
360
200
Heater
Inlet
Figure 3.5 The reactor scheme and photo.
The general view of a quasi-adiabatic reactor is shown in Figure 3.5. Along the inner perimeter of
the reactor, there is a layer of thermal insulation of 5 cm thick, so the reaction zone does not have
thick steel walls, but is formed by a basket of thin steel mesh adjacent to the heat insulator (mineral
wool). The heater is made in the spiral form located in the lower part of the reactor, under the cat-
alyst bed, thus the supplied regenerating mixture is heated when passing through the area of the
heating spiral. Temperature control in the catalyst bed was carried out using eight thermocouples,
some of which were located along the axis of the reactor and some along its wall (Figure 3.6).
Apex 4 8
Upper part 3 7
Wall Center 2 6
1 5
Lower part
0
Inlet flow
Figure 3.6 Scheme of the catalyst bed and location of thermocouples.
3.2.4 Experiments
250 g of catalyst was loaded into the reactor and the reaction zone was a cylinder with radius and
height of 5 cm. The flow rate of the supplied mixture was 38 l/min. The mixture was supplied at
ambient temperature; the reaction was initiated by heating a fragment of the catalyst bed in its
inlet part (thermocouple 0). After reaching a temperature of more than 300 C, the heating was
turned off, and then the process was carried out in the mode of a moving thermal wave.
Regeneration in quasi-adiabatic reactor was carried out with a flow with variable oxygen content;
up to the 18th minute of the process, the oxygen content was 4%, then up to the 30th minute, 6%,
and after that 15%. The concentrations of the gas mixture components are shown in Figure 3.7.
Figure 3.8 shows the temperature dynamics at various points where thermocouples are located
along the bed. The maximum temperature in the bed was 775 C.
After regeneration, the catalyst was selected for analysis from different parts of the bed. The bed
was divided into lower part, middle, upper part, and apex. The middle of the bed was further divided
into central and wall regions.
The results of the chemical analysis are presented in Table 3.2, and the textural characteristics of
the selected samples are presented in Table 3.3.
It can be seen that, with the exception of the center of the bed middle part, where the highest
temperatures were observed, the catalyst was not completely regenerated. The apex corresponds
to the top 5 mm of the bed height and has the lowest temperature, so in this part of the bed only
about half of the coke contained in the catalyst is burned out.
Figure 3.9 shows an enlarged image of the transverse fracture of catalyst grains from the bottom
(inlet) and the top (outlet) of the bed. It is observed that in grains from the lower part of the bed,
the black color on the cut, corresponding to coke, is larger than in grains from the top. This cor-
responds to the carbon content analysis data in Table 3.2. It is also apparent that the transition
between the gray and black parts of the slice in Figure 3.9a is more pronounced, while in
Figure 3.9b it is more blurred. This suggests worse regeneration of the catalyst in the inlet part
of the bed and existence of a greater gradient in the distribution of coke along the grain in this bed
section. It may be caused by insufficient development of the heat wave of oxidative regeneration
in the very inlet part of the catalyst bed. Subsequent parts of the bed regenerate better due to the
increase in temperature at the regeneration front toward the exit of the reactor.
(a)
4% O2 6% O2 15% O2
14
12
Concentration, vol. %
10
O2
8 CO2
0
0 20 40 60 80 100 120 140
time, min
(b)
4% O2 6% O2 15% O2
3500
CO
3000
SO2
Concentration, ppmv
2500
2000
1500
1000
500
0
0 20 40 60 80 100 120 140
time, min
Figure 3.7 Experimentally measured evolution of O2 and CO2 (a), CO and SO2 (b) concentrations at the outlet
of the reactor in time.
4% O2 6% O2 15% O2
800
700
600
Temperature, °C
#0
500 #1
#2
400 #3
#4
300 #5
#6
200
#7
100 #8
0
0 20 40 60 80
time, min
Figure 3.8 The temperature dynamics in central (black) and wall (gray) bed parts.
Table 3.2 Catalyst composition and maximum temperature in different parts of the bed.
Bed part Mo, wt.% Co, wt.% C, wt.% H, wt.% N, wt.% S, wt.% Тmax, C
Lower (inlet) part 10.6 2.91 1.44 1.1 0.036 0.92 700
Middle part – center 11.4 2.89 — 0.98 — 0.16 775
Middle part – wall 11.0 2.74 1.5 1.14 — 0.47 650
Upper (outlet) part 11.4 2.90 1.0 1.17 0.03 0.57 550
Apex 9.75 2.54 5.6 1.3 0.13 1.54 300
Table 3.3 Textural characteristics of the catalyst in different parts of the bed.
Bed part Specific surface area, m2/g Pore volume, cm3/g Average pore size, A
Lower part 178 0.50 112

Middle part – center 140 0.56 160
Middle part – wall 189 0.52 110
Upper part 184 0.51 111
Apex 165 0.42 102
3.3 Mathematical Model
To simulate the process of oxidative regeneration in a fixed catalyst bed, a one-dimensional math-
ematical model was used, which takes into account the course of reaction processes, both heat and
mass transfers between the gas flow and the surface of the catalysts and inside the catalyst granules,
as well as the thermal conductivity of the bed (Zazhigalov et al. 2012; Reshetnikov et al. 2020).
The regeneration process is described by the oxidation of molybdenum and cobalt sulfides into
oxides (18, 19), as well as the oxidation of coke (20):
MoS2 + 3 5O2 MoO3 + 2SO2 3 18

CoS0 89 + 1 39O2 CoO + 0 89SO2 3 19
C + O2 CO2 3 20
The material balance of the concentrations of substances in the gas phase in the pores of catalyst
grains was described by parabolic equations that take into account intra-diffusion limitation:
∂CpSO2 ∂ 2 C pSO2 p
2 ∂C SO2
εC = DSO2 + + W form SO2 3 21
∂t ∂r 2 r ∂r
∂CpCO2 ∂ 2 C pCO2 p
2 ∂CCO2
εC = DCO2 + + W form CO2 3 22
∂t ∂r 2 r ∂r
∂CpO2 ∂ 2 CpO2 p
2 ∂CO2
εC = DO 2 + − W cons O2 3 23
∂t ∂r 2 r ∂r
(a)
(b)
Figure 3.9 Transverse catalyst fracture from the lower (a) and upper (b) parts of the catalyst beds.
C pN2 = 1 − CpSO2 − CpCO2 − C pO2 3 24
Equations for the rates of consumption and formation of gas components are as follows:
−E s 05
W form SO2 = 0 27 1 667 ρc k s0 exp θS C pO2 3 25
RT p
−Ec 05
W form CO2 = 1 87 ρc k C0 exp θC CpO2 3 26
RT p
W cons O2 = 1 72 W обр SO2 + W обр CO2 3 27
Material balance for sulfide and coke concentrations on the inner surface of grains:
∂θS −E s 05
= − k s0 exp θS C pO2 3 28
∂t RT p
∂θC −E c 05
= − k C0 exp θC C pO2 3 29
∂t RT p
The change in the concentrations of reagents in the gas mixture along the length of the bed, tak-
ing into account external mass transfer processes, was described by the following equations:
∂CSO2
u = βSO2 S C pSO2 r = R − C SO2 3 30
∂l
∂CCO2 p
u = βCO2 S C CO2 r = R − CCO2 3 31
∂l
∂CO2 p
u = βO2 S CO2 r = R − CO2 3 32
∂l
CN2 = 1 − CSO2 − CCO2 − C O2 3 33
The heat balance equations for the gas flow and grains took into account the heat exchange of the
gas phase with the catalyst, thermal conductivity along the bed, and heat release during chemical
reactions. It was assumed that the grain volume is isothermal.
∂T
ucp = αS T p − T 3 34
∂l
∂T p λ ∂2 T p
γp = − αS T p − T +
∂t ρc ∂l2
1 17 4
+ W sulf ox Q + Q + W coke ox Q3 dV 3 35
V V 21 1 21 2
For Eqs. (3.21)–(3.34), the corresponding initial and boundary conditions were set:
t = 0, 0 < r < R, 0 < l < L: (Initial conditions for pellet internals)
C pSO2 0, r, l = C pSO2 0 r, l , CpCO2 0, r, l = CpCO2 0 r, l , CpO2 0, r, l = CpO2 0 r, l , 3 36
θS 0, r, l = θS0 r, l , θC 0, r, l = θC0 r, l , 3 37
t = 0, r = R, 0 < l < L: (Initial conditions for pellet outer surface)

T p 0, R, l = T p0 l , 3 38
t > 0, r = 0, 0 < l < L: (Symmetry conditions in pellet center)
∂C pSO2 ∂C pCO2 ∂CpO2

= 0, = 0, = 0, 3 39
∂r ∂r ∂r
t > 0, r = R, 0 < l < L: (Mass exchange conditions on pellet external surface)
∂C pSO2 ∂Cp
DSO2 = βSO2 C SO2 − C pSO2 , DCO2 CO2 = βCO2 CCO2 − CpCO2 ,
∂r ∂r
∂C pO2
DO 2 = βO2 CO2 − CpO2 , 3 40
∂r
t > 0, r = R, l = 0: (Inlet conditions)
C SO2 t, R, 0 = C SO2 ,in t , CCO2 t, R, 0 = C CO2 ,in t ,
CO2 t, R, 0 = CO2 ,in t , T t, R, 0 = T in t , 3 41
∂T p
λ = 0, 3 42
∂l
t > 0, r = R, l = L: (Outlet conditions)
∂T p
λ =0 3 43
∂l
In the process of oxidative regeneration under consideration, the catalyst has a shape close to
cylindrical. The average grain size was 1.6 mm in diameter and 4 mm in height. To go to the spher-
ical pellet used in the mathematical model under consideration, it is necessary to go to the equiv-
alent pellet diameter.
The equivalent diameter of a cylindrical grain in the spherical approximation is calculated by the
following formula:
6V cyl 3dh
dsph = = 3 44
Scyl d + 2h
where d—diameter, h—cylinder height. For the size of the catalyst used, the equivalent diameter is
2 mm.
The model assumes that the oxidation of molybdenum and cobalt sulfides is described by one
summary Eq. (3.28), thus the concentration θS is the sum of the weight concentrations of molyb-
denum and cobalt sulfides.
At the initial time, the catalyst contains 17% wt. MoS2 and 4% wt. CoS0.89. This model assumes
that the weight ratio between sulfides will continue to be 17:4.
Equations (3.25)–(3.27) are obtained as follows. Due to the fact that the molar ratio of sulfides is
7:3, on multiplying Eqs. (3.18) and (3.19) by 0.7 and 0.3, we get the total as follows:
0 7MoS2 + 0 3CoS0 89 + 2 867O2 0 7MoO3 + 0 3CoO + 1 667SO2 3 45
It follows that the rate of formation of SO2 is equal to the following:
dθS
W form SO2 = − 1 667 3 46
dt
The rate of oxygen consumption can then be expressed as follows:
2 867
W cons O2 = W form SO2 + W form CO2 = 1 72 W form SO2 + W form CO2 3 47
1 667
However, for the correct transition from reaction rates in the solid phase to the rates of formation
and consumption of gas components, it is necessary to introduce the corresponding coefficient in
Eqs. (3.25) and (3.26). By definition of bulk density, 1 m3 of catalyst contains ρcθS kg of sulfides.
In this case, from 1 kg of molybdenum and cobalt sulfides (in a ratio of 17:4) in accordance with
Reactions (18) and (19), 0.27 m3 SO2 can be obtained. Consequently, from ρcθS kg of sulfides,
0.27ρcθS m3 SO2 is obtained. Thus, per 1 m3 of catalytic bed, there is 0.27ρcθS m3 SO2. This means
that the rate of SO2 formation can be written as follows:
dθS
W form SO2 = − 0 27 1 667 ρc , 3 48
dt
which corresponds to Eq. (3.25).
Carrying out similar reasoning with Eq. (3.20), we obtain Eq. (3.26).
The mathematical model under consideration takes into account changes in the bulk density of
the catalyst, heat capacity of the gas, and catalyst. Thus, the coefficients ρc, cp, and γ p are variables
during the regeneration process.
The change in catalyst bulk density ρc occurs due to a change in the mass of the catalyst due to
coke burnout and the transition of molybdenum and cobalt sulfides into oxides. In this case, the
mass of the carrier (Al2O3) remains unchanged in a given volume (9000.68 kg/m3); therefore,
the recalculation of bulk density obeys the following equation:
ρnew = 900 0 68 + ρold θC + ρold θS + ρold θO , 3 49
here θO is the relative mass content of molybdenum and cobalt oxides on the catalyst (kg/kg), which
is calculated based on the initial and current content of sulfides (θS).
The heat capacity of the gas mixture is calculated based on the current values of the concentra-
tions of the components:
cp = cSO2 cSO
p
2
+ cCO2 cCO
p
2
+ cO2 cO2 N2
p + cN2 cp , 3 50
here cSO CO2 O2 N2

p , cp , cp , and cp are heat capacities of the corresponding gas components (Table 3.4).
2
The heat capacities of the catalyst are also calculated based on the concentrations of the corre-
sponding components and their heat capacities:
17 4
γ p = θS γ + γ + θC γ C + θMoO γ MoO + θCoO γ CoO + θAl2 O3 γ Al2 O3 , 3 51
21 MoS 21 CoS
where the corresponding terms are determined in Table 3.5.
In the integro-differential Eq. (3.35), the thermal effects of Reactions (18) and (19), Q1 and Q2,
enter with coefficients 17/21 and 4/21, indicating the fraction occupied by molybdenum and
cobalt sulfides in θS. Woxid.sulf. and Woxid.coke. are the rates of oxidation of sulfides and coke and
correspond to the right-hand sides of Eqs. (3.28) and (3.29).
Heat effects of reactions are presented in Table 3.6.
Table 3.4 Heat capacities of gas components.
Component O2 N2 CO2 SO2
Heat capacity, kJ/st.m3/K 1.46 1.36 2.18 2.25
Table 3.5 The heat capacities of solid components.
Component MoS2 MoO3 CoS0.89 CoO Al2O3 C
Heat capacity, kJ/kg/K 0.467 0.661 0.579 0.727 1.41 1.73

Table 3.6 Heat effects of reactions.
Reaction kJ/kg
18 6644
19 4643
20 33,333
Table 3.7 Kinetic parameters of oxidative regeneration reactions.
Reactions ln k0 (kg per kg of catalyst per s) E/R (K) n m
18 + 19 12.41 10,108 ~1 ~0.5

20 10.55 11,279 ~1 ~0.5
The values of the kinetic parameters for Eqs. (3.28) and (3.29) are obtained from thermodynamic
calculations and presented in Table 3.7.
The coefficients of heat and mass transfer and the diffusion coefficient for the components of the
gas mixture were equated to the coefficients calculated for oxygen, which were calculated using the
following formulae (Aerov et al. 1979):
Nuλg ShO2 DO2
α= , β O2 = , 3 52
deq deq
where
1 μ
Nu = ARen Pr13 , ShO2 = ARen ScO32 , ScO2 = ,
ρg DO 2
udeq ρg 6 1−ε 4ε
Pr = 0 69, Re = ,S = , deq = dsph 3 53
με dsph 6 1−ε
The correspondence between the coefficients A and n and the Reynolds number, Re, is shown in
Table 3.8.
Viscosity, thermal conductivity, and gas flow density were calculated using approximations con-
structed on the basis of reference data in the form of functions of the reference temperature:
3
T ref2
μ = 1 551 10 − 6 N s m2,
T ref + 124
Table 3.8 The correspondence between the coefficients

A and n and the Reynolds number.
Re A n
Re < 2 0.515 0.85

2 ≤ Re ≤ 30 0.725 0.47
Re > 30 0.395 0.64
λg = 5 44 10 − 8 T ref + 12 23 10 − 6 kJ m s K,
353 6
ρg = kg m3
T ref
The reference temperature is 400 C.

The effective thermal conductivity of the bed is calculated using the following formula:
λ = 1 − ε λc + 10λg Re Pr, 3 54
where λc—thermal conductivity of catalyst supports (Al2O3): λc = 13 J/s/m/K.
3.4 Model Solution Method
In view of the complexity of the equations set and the various types of equations included in the
model, an analytical solution of the system is not possible. For the numerical solution of the model,
a mesh was introduced on the variables r, l, and t. The sequential construction of a model solution at
each time step made it possible to obtain a complete model solution at an arbitrary time interval.
To construct a discrete problem, the model equations were approximated by difference schemes.
Parabolic-type Eqs. (3.21)–(3.23) and (3.35) were approximated by an implicit difference scheme of
the first-order accuracy in the time variable t and the second in the coordinates r and l. As a result of
approximation, systems of algebraic equations with tridiagonal matrices were obtained, the solu-
tion of which was carried out by the tridiagonal sweep method. In this case, the integral that deter-
mines the heat release over the pellet volume, which is part of Eq. (3.35) and defined in its discrete
formulation at each time point and coordinate along the length of the bed, was calculated using the
trapezoidal rule:
R R
1 3 4 3
F r dV = 3πr 2 F r dr = 3 r 2 F r dr
V V 4πR3 0 3 R 0
3 nr 2
= r F r i + r 2i + 1 F r i + 1 Δr 3 55
2R3 i = 1 i
Equations (3.28)–(3.32) and (3.34) were approximated by an implicit difference scheme of

first-order accuracy.
The resulting discrete problem was solved by the following algorithm. The transition from time
layer t to t + τ was as follows. At the entrance of the bed (at l = 0), the concentrations of the com-
ponents and the temperature of the supplied gas mixture (C SO2 , C CO2 , CO2 , and T) are known. Based
on these values, solving the algebraic systems obtained from Eqs. (3.21)–(3.23) gives the concentra-
p p p
tions of gas components in the pores of the grain (C SO2 , CСO2 , and C O2 ) at each radial point on the
t + τ time layer. In this case, the values of surface concentrations of sulfides and coke (θS и θC) from
the previous time layer were used. Thus, having obtained the value of C pO2 at the t + τ layer, the
values of θS and θC are calculated. Consequently, the concentration values of all components
are determined at l = 0 on the t + τ layer. When moving to the next section along the length of
the layer (l = Δl), the values of the concentrations of the components and the temperature
p p p
of the gas flow (CSO2 , C CO2 , C O2 , T are calculated using the values C SO2 , CСO2 , CO2 , and T p from
the previous time layer. Thus, according to the algorithm described earlier, the values of
3.5 Modeling Results 133
p p p
C SO2 , CСO2 , CO2, θS, and θC are found for l = Δl. By performing this procedure, the required number
p p p
of times, the values of C SO2 , CСO2 , CO2 , θS , θC , CSO2 , C CO2 , C O2 , and T are found on the entire time
layer t + τ. After determining these values, solving the algebraic system obtained from Eq. (3.35)
gives the values of the catalyst temperature Tp over the entire time layer t + τ. Next, the transition
to the next time layer is carried out according to the algorithm outlined earlier.
When implementing the algorithm for numerical solution of the model, the step along the bed
length was taken equal to Δl = 10−3 m, and the time step τ was taken equal to 10−3 s.
The work verified the mathematical model based on an experiment in quasi-adiabatic reactor. In
the model, a mixture with oxygen content and temperature corresponding to the experiment was
supplied to a catalyst bed of 5 cm long.
3.5 Modeling Results
For a better description of the experimental data, an effective diffusion coefficient was chosen equal
to DO2 = 5 × 10−8 m2/s, which is obviously less than standard value calculated earlier (DO2 = 5 ×
10−7 m2/s) for a coked catalyst with pores of radius 12 nm (Reshetnikov et al. 2020). The change in
temperature profiles along the length of the bed over time during regeneration is depicted in
Figure 3.10 for DO2 = 5 × 10−8 m2/s. The experimental curves (black) are described quite well
by the model curves (gray), with the exception of the tail part of the bed, which is not sufficiently
heated during the experiments, which is also shown in Table 3.2. Figure 3.11 demonstrates the
dynamics of changes in SO2 and O2 concentrations at the outlet of the bed in the experiment
and model for both of the diffusivities.
It can be seen the case with DO2 = 5 × 10−8 m2/s describes the experiment with better precise than
with DO2 = 5 × 10−7 m2/s. The model SO2 concentration is sharper—the peak value is higher and
emission is shorter than the experimental one. This is consistent with the slightly lower oxygen
yield in the model. Figure 3.12 shows the coke content averaged by pellet volume in different
5 min
800 10 min
30 min
50 min
600
Temperature, °C
400
200
0
0.00 0.01 0.02 0.03 0.04 0.05
Bed length, m
Figure 3.10 Temperature along the catalyst bed in different time points in experiment (black) and model
(gray) during the oxidative regeneration.
(a)
15000
SO2 concentration, ppmv
Experiment
Model DO2 = 5.10–8 m2/s
10000
Model DO2 = 5.10–7 m2/s
5000
0
0 20 40 60 80
time, min
(b)
16
14
O2 concentration, vol.%
12
10
Experiment
8 Model DO2=5.10–8 m2/s
Model DO2=5.10–7 m2/s
6
0
0 20 40 60 80
time, min
Figure 3.11 Simulated and experimental changing of SO2 (a) and O2 (b) concentrations at the reactor outlet in
time during the oxidative regeneration.
bed points for DO2 = 5 × 10−8 m2/s. Here, one can observe a good correlation between the exper-
imental and model data, with the exception of the output bed region, which remained unregener-
ated during the experiment, as mentioned earlier.
3.6 Conclusion
At the moment, the proposed model provides a more or less accurate description of the experimen-
tal data and may be considered as essentially verified. At the same time, additional improvement of
the model is required to provide better description of experiments. The first improvement may
relate to the corrections of the kinetic model to better describe all observed regularities. Most likely,
3.6 Conclusion 135
6
Experiment
Coke concentration, wt.%
Model
5
0
0.00 0.01 0.02 0.03 0.04 0.05
Bed length, m
Figure 3.12 Experimentally measured and simulated average mass content of coke in catalyst pellets in the
end of regeneration procedure.
it will be necessary to account for at least two types of coke with significantly different oxidation
rates. Physically these differences may be related to the different chemical nature of the different
coke itself (“crumbly” and “dense” coke).
Another, more serious issue is a more accurate simulation of catalyst pellets under visible influ-
ence of mass transfer limitations. The modeling revealed that the value of the effective diffusion
coefficient, calculated within the conventional approach, resulted in significant disagreement
between calculated and experimental data. A much better description is provided under the
decrease of diffusion coefficient by an order of magnitude.
The first possible explanation is based on the hypothesis that the worse diffusion is caused by
coke deposition in all pores; in this case, the diffusivity will increase with the decrease of coke con-
tent in the catalyst. It was demonstrated that this approach does not provide the appropriate
improvement of the description as well.
Another explanation includes the assumption on significant deposition of coke not only in rel-
atively large transport pores but also in micropores, which may become almost or completely
blocked by coke. Obviously, the apparent kinetics of coke and sulfide oxidation in these micropores
may be significantly different from the bulk one, and the effective diffusion coefficient may be sig-
nificantly lower than that calculated for large pores. The decrease of diffusivity may be attributed to
the decrease in Knudsen diffusion due to the decrease in effective size of pores partially filled with
coke as well as to complete blocking of pore mouths, causing the “bottleneck” complication of dif-
fusion at the micropore entrance.
The impact of nonuniform coke distribution among pores of different sizes on the regeneration
process was recently theoretically predicted and analyzed (Zhdanov 2022). This assumption is also
in good correspondence with the fact that such unusual decelerating of oxidation is not observed
at all in kinetic experiments at fine catalyst powder, where the complication of diffusion in micro-
pores does not influence the reaction parameters. In our opinion, this approach deserves further
development.
3.7 Notation
Ci, gas flow component concentration, mole fraction;

Cpi , gas component concentration in the pellet pores, mole fraction;
cip , cp, gas component and overall heat capacity, J/kg/K;
Di, effective diffusion coefficients in pellet, m2/s;
dsph, equivalent diameter of a cylindrical grain in the spherical approximation, m;
Es, activation energy, J/mol;
k0, pre-exponential factor, 1/s;
l, bed length coordinate, m;
nr, number of radial points;
Nu, Nusselt number;
Pr, Prandtl number;
Re, Reynolds number;
Sc, Schmidt number;
Sh, Sherwood number;
Qi, thermal effects, kJ/kg;
r, pellet radial coordinate, m;
R, pellet radius, m;
S, outer specific surface area, m2/m3;
T, Tp, gas and bed temperature, K;
t, time coordinate, s;
u, linear gas velocity, m/s;
V, pellet volume, m3;
Wi, rate of formation or consumption, 1/s;
α, heat transfer coefficient, J/m2/K/s;
βi, mass transfer coefficients, m/s;
γ i, γ p, surface compounds and catalyst heat capacity, J/kg/K;
Δl, bed length discrete step, m;
εc, catalyst porosity;
θi, surface concentration, mass fraction;
λ, λc, λg, effective bed, catalyst, and gas thermal conductivities, W/m/K;
μ, gas dynamic viscosity, kg/m/s;
ρc, ρg, catalyst bed and gas density, kg/m3;
τ, time step, s.
Abbreviations
TG, thermal gravimetry

DTA, differential thermal analysis
MS, mass spectrometry
References 137
Acknowledgment
This work was partially supported by the Ministry of Science and Higher Education of the Russian
Federation within the state assignment for the Boreskov Institute of Catalysis (project FWUR-2024-
0037) and by the Tyumen Oblast Government, as part of the West-Siberian Interregional Science
and Education Center’s project No. 89-DON (3).
References
Aerov, M.E., Todes, O.M., and Narinskii, D.A. (1979). Vessels with fixed granular bed. Hydraulic and heat
operation fundamentals. Leningrad, Chemistry.
Ferella, F., Innocenzi, V., and Maggiore, F. (2016). Oil refining spent catalysts: a review of possible
recycling technologies. Resources, Conservation and Recycling 108: 10–20.
Froment, G.I. and Biscoff, K.B. (1999). Chemical Reactor Analysis Design, Seconde. New York: Wiley.
Furimsky, E. (1991). Effect of coke and catalyst structure on oxidative regeneration of hydroprocessing
catalysts. Fuel Processing Technology 27: 131–147.
Kern, C. and Jess, A. (2005). Regeneration of coked catalysts—modelling and verification of coke burn-off
in single particles and fixed bed reactors. Chemical Engineering Science 60 (15): 4249–4264.
Marafi, M., Stanislaus, A., and Furimsky, E. (2017). Handbook of Spent Hydroprocessing Catalysts,
Seconde. Elsevier.
Meyers, R. (2016). Handbook of Petroleum Refining Processes, Fourthe. New York: McGraw Hill.
Reshetnikov, S.I., Petrov, R.V., Zazhigalov, S.V., and Zagoruiko, A.N. (2020). Mathematical modeling of
regeneration of coked Cr-Mg catalyst in fixed bed reactors. Chemical Engineering Journal 380: 122374.
Speight, J.G. (2016). Handbook of Petroleum Refining. Boca Raton: Taylor&Francis.
Zazhigalov, S.V., Chumakova, N.A., and Zagoruiko, A.N. (2012). Modeling of the multidispersed
adsorption-catalytic system for removing organic impurities from waste gases. Chemical Engineering
Science 76: 81–89.
Zhdanov, V. (2022). Kinetics and percolation: coke in heterogeneous catalysts. Journal of Physics A:
Mathematical and Theoretical 55 (17): 1–16.
Zhou, J., Zhao, J., Zhang, J. et al. (2020). Regeneration of catalysts deactivated by coke deposition:
A review. Chinese Journal of Catalysis 41 (7): 1048–1061.
138
Modeling of Unsteady-State Catalytic and Adsorption–Catalytic

Processes: Novel Reactor Designs
Sergey Zazhigalov 1,2, Andrey Elyshev 2, and Andrey Zagoruiko1,2
1
2
4.1 Introduction
The protection of atmospheric air from the harmful effects of anthropogenic factors is subject to
state supervision worldwide. The existing system of such supervision provides, among other mea-
sures, the organization of state, industrial and public control over emissions from industrial enter-
prises. Development of new scientific and technical approaches to improve the efficiency of
purification of exhaust gases from volatile organic compounds (VOCs) (hydrocarbons, alcohols,
acids, ethers, aldehydes, ketones, etc.) in industrial enterprises is part of such a system.
The problem of reducing VOC emissions from anthropogenic sources (Garcia et al. 2014;
Gonzalez-Velasco et al. 2014; Ojala et al. 2011; Spivey 1987), raised in the 20th century, remains
relevant even at present times. The environmental consequences of air pollution with VOC
emissions are associated with a destructive effect on the ozone layer and increase in the amount
of photochemical smog (Gonzalez-Velasco et al. 2014), as well as with a high level of toxicity
of few VOCs. In connection with the increasing production capacities in areas such as energy,
transport, chemistry and petrochemistry, mechanical engineering, wood processing, printing,
production of building and finishing materials, and so on, research aimed at solving this problem
is becoming part of a strategy to ensure the safety of life on earth.
Despite the existence of a large number of developed technologies for the neutralization of VOCs,
they cannot be applied always and everywhere. The efficiency and applicability of these technol-
ogies should be evaluated considering their VOC abatement efficiency, capital and operation costs,
and especially their energy consumption. The latter factor makes not only economic but also envi-
ronmental sense, as it usually involves the use of additional fuels, which leads to an increase in the
carbon footprint of technologies, and sometimes leads to the direct formation of toxic secondary
atmospheric air pollutants.
For emissions with high concentrations of VOCs, it is cost-effective to use technologies based on
recovery of organic substances from waste gases with their recycling back to the technological proc-
ess (e.g., absorption, adsorption, condensation, and membrane methods). Despite technological
advantages of these methods in application to gaseous wastes with a relatively high VOC content
(usually above 1% mass), they become unfeasible in case of lean gases processing, when the cost of
the recycled VOCs is significantly lower than the operating expenses for such recovery.
For diluted gases, it is expedient to use thermal and, especially, catalytic oxidation processes of
VOCs by atmospheric oxygen to nontoxic products: carbon dioxide and water vapor (Matros et al.
1991). Although this method produces additional emissions of the resulting CO2 to the atmosphere,
the overall environmental effect is strongly positive.
Thermal methods, in absence of catalysts, require extremely high temperatures (often around
1000 C), leading to limited combustion efficiency, and potentially resulting in the formation of
secondary pollutants, such as CO, NOx, soot, and so on. The concentration of VOC in waste gases
is often insufficient for autothermal combustion; so such approaches usually require the consump-
tion of additional fuel, worsening the economic and environmental efficiency.
Catalytic oxidation of VOCs has several advantages over thermal methods: a lower temperature
for the oxidation reaction, higher efficiency of oxidation of VOCs, and most importantly, significant
decrease of energy consumption in comparison with noncatalytic combustion. Reviews of the
results of the neutralization of VOCs on catalysts of various types, such as noble metals or transition
metal oxides on various supports, metal-containing zeolites, and so on, are quite widely presented
in the scientific literature (Evaraert and Baeyens 2004; Li et al. 2009; Spivey 1987). To date, catalytic
methods are considered as the most promising for the neutralization of VOCs in the waste gases
from various industrial facilities.
Catalytic oxidation of VOCs occurs at elevated temperatures, typically from 100–150 up to
500–600 C. However, waste gases very often have ambient or slightly elevated initial temperature,
insufficient for complete VOC oxidation. The flow rate of these gases may be significant, and more
energy may be required for their preheating. Therefore, the overall energy efficiency of catalytic
processes is crucial for the efficiency of heat recuperation through the heat transfer from outlet
purified gases to the inlet gaseous stream.
Modern improvement of catalytic processes for the neutralization of VOCs goes both in the direc-
tion of creating new effective catalysts for the complete oxidation of organic substances and in the
line of creating new options for organizing the oxidation process. Currently, catalytic reverse-flow
processes (Matros and Bunimovich 1996), which use the very efficient regenerative heat exchange,
are capable of autothermally (i.e. without the supply of additional energy) processing gases with a
VOC content of 0.6–0.8 g/m3 and higher and have the best economic and environmental indicators
for processing low-concentration emissions. For gases with lower
VOC content, additional energy and capital costs are required.
The reverse-flow process implies a periodic change in the direc-
tion of supply of the mixture being cleaned in two apparatuses.
This tactic allows to save most of the heat released during the
oxidation of impurities inside the catalytic bed (Figure 4.1).
Purification of emissions with extremely low impurity con-
centrations, that is, below 0.1 g/m3, remains a serious and
widespread practical problem (Garcia et al. 2014). From the view-
point of compliance with the criteria of energy efficiency, ease,
and safety in use and maintenance, adsorption–catalytic pro-
cesses (ACPs) based on the combination of the principles of
adsorption and catalytic technologies are of particular interest
(Zagoruiko et al. 1996). Despite wide variety of variants of such
processes proposed in the literature, their common essence lies
in the fact that during the periodic process, adsorption and regen-
eration stages alternate (Robinson 1971). There are a significant Figure 4.1 Reverse-flow process
number of such processes, which are a simple combination of scheme.
140 4 Modeling of Unsteady-State Catalytic and Adsorption–Catalytic Processes: Novel Reactor Designs
Cleaned gas adsorbers with catalytic reactors. In this case, VOCs are initially
adsorbed in the adsorbent bed, and then the sorbent is regenerated
(usually by steam or hot air) and the desorbed impurities are catalyti-
cally oxidized in a separate apparatus.
Such a multistage scheme with the separation of the adsorbent and
catalyst requires relatively large energy and capital costs, especially
for the neutralization of large volumes of low-concentration industrial
emissions. An interesting alternative to the schemes described above is
ACP based on the adsorption of VOC directly on the surface of a deep
oxidation catalyst (which also acts as an adsorbent) at ambient temper-
ature. When VOC impurities appear at the exit from the bed, the
adsorption stage is completed, and the bed is regenerated. In this case,
the gas supplied to the bed is heated using a heater located in front of the
Outside reactor inlet (Figure 4.2). Upon completion of regeneration, the adsorb-
Heater ent–catalyst bed is again ready for the first stage, which is adsorption.
ACPs at different times and in different versions of their technological
Feed gas organizations were considered and developed by UOP (Stanley and
Ritscher 1979) and Haldor Topsoe (Schoubye 1979), as well as by several
Figure 4.2 Adsorption–
catalytic process with scientific institutes in the USSR.
external heater scheme. Adsorption–catalytic and catalytic reverse-flow processes, with any
method of their organization, have significant advantages over tradi-
tional catalytic technologies for the deep oxidation of VOCs, especially
in cases of processing low-concentration gases. The use of ACP makes it possible to eliminate
energy costs for the constant heating of gases to be purified, supplied to the apparatus with a cat-
alyst. Their important advantage is high technological flexibility associated with the possibility of
processing gases with varying initial concentrations and types of VOCs (up to the processing of a
mixture of various VOCs in exhaust gases).
However, as shown by practical experience on the example of pilot plants and the study of ACPs
by means of mathematical modeling (Zagoruiko et al. 1996; Vernikovskaya et al. 1999; Salden and
Eigenberger 2001), they also have certain disadvantages:
• Insufficiently low power consumption when heating the reactor from an external heater, asso-
ciated with inefficient energy consumption for heating the inlet gas duct and the inlet part of the
reactor vessel;
• The possibility of partial desorption of nonoxidized VOCs or products of their incomplete oxida-
tion during catalyst heating, leading to a decrease in the degree of gas purification. The release of
VOCs during regeneration can also lead to bursts, when for a short period of time the concen-
tration of VOCs in the exhaust gases may even exceed their initial concentration in the trea-
ted gases;
• The possibility of ejection of a gas stream with a high temperature (up to 300–500 C) into the
exhaust gas duct during regeneration. Such a hot gas discharge may slat for a relatively short
time, but it can cause significant damage to gas communications, especially in ventilation sys-
tems where gas ducts are not designed to operate at high temperatures.
At different times, various ways to improve the ACP were proposed; for example, gradual “step-
wise” heating of the catalyst during regeneration, regeneration in a filtration combustion wave,
optimization of the size of catalyst–adsorbent pellets, application of technological schemes using
a postreactor and spiral reactor with a specific system of internal heat exchange (Retallick
4.2 Novel Reactor Designs for Catalytic Reverse-Flow and Adsorption–Catalytic Processes 141
1996), the implementation of catalyst regeneration under conditions of a heat wave moving of com-
bustion of adsorbed VOCs, propagating toward the motion of the gas flow due to the thermal con-
ductivity of the catalyst bed.
Next, a proposed way to overcome disadvantages of the ACP is by changing the designs of the
reactors used (Zazhigalov et al. 2016, 2017, 2018, 2023b); and the effect of asymmetric reactor
designs on the efficiency and stability of the catalytic reverse-flow process (Zazhigalov et al.
2023a) is also estimated. Evaluation of the effectiveness of the proposed modifications of the appa-
ratuses is based on mathematical modeling of processes in the commercial software, COMSOL
Multiphysics.
4.2 Novel Reactor Designs for Catalytic Reverse-Flow

and Adsorption–Catalytic Processes
4.2.1 Unsteady-State Catalytic Reverse-Flow Process

The stability and efficiency of the reverse-flow process are the two most important indicators that
determine it and can vary significantly depending on the conditions of the gas mixture supply.
Thus, an asymmetric mode of operation can lead to a significant decrease in stability and arise,
for example, as a result of different intervals between switching the direction of the mixture supply
or asymmetry of conditions, both at the inlet to the apparatus and at the initial time.
However, one of the most important factors affecting the asymmetry of the operating mode of the
process is the asymmetric arrangement of the apparatuses themselves. As a rule, such asymmetry is
determined by the nonstandard, nonaxial, location of the inlet or outlet ducts. Such designs can
noticeably change the velocity field in the system, the distribution of reagent concentrations,
and bed temperature. In this case, some areas of the reactors may remain unused in the process
and reduce its overall efficiency.
The following reactor designs (Figure 4.3) will show the different behavior of process parameters
in such systems as examples. Thus, reactors with axial, lateral, tangential, tangential–lateral, lat-
eral–axial, lateral–orthogonal gas mixture supplies will be considered. This figure shows one of
(a) (b) (c) (d) (e) (f)
Figure 4.3 Computational geometry (one chamber) for process schemes with axial (a), lateral (b),
tangential (c), tangential–lateral (d), lateral–axial (e), and lateral–orthogonal (f ) gas inlet/outlet ducts.
the two devices, that is, half of the system being modeled. The catalyst bed has cylindrical shape
with 0.5 m in diameter and 1 m in height in all cases.
4.2.2 Adsorption–Catalytic Process

Two important circumstances determine the energy consumption for the ACP—the pumping of the
mixture to be purified through the reactor and the periodic regeneration of the bed, during which it
is required to heat a certain amount of catalyst to the oxidation reaction temperature ( 300 C). It is
shown that heating only the input region of the bed allows for efficient regeneration, as the thermal
front moving in the direction of the flow allows the oxidation of impurities in the entire subsequent
bed part in an autothermal mode. Due to the fact that the heating element operates for a short time,
relative to the entire adsorption cycle, the energy consumption in the process remains quite low. If
we talk about systems with large volumes of gases to be purified, then heating such volumes to the
required temperature, even for a short period, can lead to quite large energy costs. It should also be
noted that in this case both the inlet gas duct and the
Cleaned gas
inlet of the reactor vessel are heated. Of course, such
systems require a lot of heater power and must be
adapted to peak high power demand.
The problems described above can be solved by
few modifications in the system design. The location
of the heating system directly in the inlet part of the
adsorbent–catalyst bed makes it possible to heat not
the supplied gas, but a small volume of the catalyst
bed (Figure 4.4).
An evaluation comparison of various purification
Outside Inside technologies using toluene as an example is pre-
Heater Heater sented in Figure 4.5. Specific energy consumption
Feed gas
includes both costs of pumping the mixture and
Figure 4.4 Adsorption–catalytic process with required heating. It can be seen that the catalytic
external (left) and internal (right) heater reverse-flow process and ACP have their best perfor-
schemes. mances when their inlet toluene concentrations
Specific energy consumption Figure 4.5 Specific energy

consumption for different VOC
Thermal oxidation
106 abatement technologies on the toluene
example (Adapted from Zagoruiko
et al. 2021).
Catalytic oxidation
105
with
ACP
J/m3
ter
hea
Catalytic r nal
reverse-flow exte
104
process
ACP with
103 internal heater
10 100 1000 10,000

Inlet toluene concentration, ppmv
4.2 Novel Reactor Designs for Catalytic Reverse-Flow and Adsorption–Catalytic Processes 143
range below 1000 ppmv. Meanwhile, the location of the heating element inside the bed ensures
significantly lower energy consumption (up to 1 order in the range of concentrations less than
500 ppmv that are relevant for ACP). Undoubtedly, properties of a particular adsorbent–catalyst,
the stream to be purified, the type of impurity, and the organization of the process will affect
the quantitative arrangement of the curves on this graph; however, one can qualitatively speak
of the undeniable advantage of ACP in the low concentration region, especially, when the heating
initiator is located directly in the bed.
Nevertheless, in the case of a large inlet section of the apparatus, it may also be necessary to use a
sufficiently powerful internal heater. This problem can be solved using the inlet part of the appa-
ratus, in the form of a truncated cone (Figure 4.6), then the thermal front from a small area at the
inlet propagates to the full cross section of the reactor, and it moves not only in axial, but also in
radial directions. In such a system, the dependence of energy consumption for regeneration on size
of the reactor and volumetric flow rate of the feed mixture is minimized. An analogy of such a proc-
ess can be drawn with bringing a match to a combustible substance, where the size of the object
does not determine the required power of the match, and the process after ignition proceeds in an
autothermal mode.
With the described organization of the process, after the passage of a thermal wave, the bed again
acquires the temperature of gases to be purified, at which the adsorption of impurities occurs; thus,
the process is continuous, that is, there is no need to stop the supply of the mixture during regen-
eration. In the case of a system with a truncated cone inlet, the question arises which cone angle
must be used in the design to ensure stable operation of the system. In addition to the cylindrical
reactor, apparatuses with conical angles from 20 to 70 of the same volume (Figure 4.6) are also
considered by the method of mathematical modeling.
As described above, one of the problems limiting the application of the ACP is the burst emissions
of VOCs and high-temperature stream during the regeneration process. The internal location of the
heater, obviously, has almost no effect on these indicators. Meanwhile, the average degree of puri-
fication in the process can remain high and the duration of burst emissions can be a low fraction of
the process cycle time, but the peak concentration of VOCs can exceed both the concentration in
the treated stream and the limit values for a specific substance.
Changing the system design by dividing it into several separated independent adsorption–
catalytic systems allows solving the described problem. In this case, it is necessary to provide con-
ditions under which all these systems will have a common inlet and outlet of the gas flow, but sep-
arate heating elements for regeneration. Such a system can be implemented not only in separate
reactor vessels, but also in one apparatus (Figure 4.7). In this implementation, the gas is supplied to
all sections and distributed between them in a natural way. However, as each of the sections has its
20° 30° 40° 50° 60° 70°
α Cone angle (α)
Figure 4.6 Reactors with truncated cones entrance part schemes.

Cleaned gas
Cleaned gas
Heater
Heater Feed gas
Feed gas
Figure 4.7 Multisectional adsorption–catalytic process schemes with sections in one reactor (left) and in
distinct parallel apparatuses (right).
own heating element, regeneration will be carried out in them in turns. The time interval between
regenerations is determined individually based on the properties of a specific system.
It is obvious that the concentration of VOCs leaving the regenerated section does not differ much
from the concentration in a single-section system. But the concentration of VOCs leaving the
remaining sections is close to zero, as these sections operate in adsorption mode. In the total outlet
stream, the concentration of VOCs is reduced by mixing these streams, and a decrease in the peak
temperature in the effluent gases is also achieved. Such a scheme makes it possible to reduce the
peak values of VOC concentration in the outlet stream, without changing the total amount of non-
oxidized impurities and, accordingly, the average degree of purification. However, it must be taken
into account that the section operating in the regeneration mode has a higher temperature than the
others, and therefore, has a higher hydraulic resistance, which limits the flow through it. Such a
redistribution of flows probably makes it possible to reduce the total amount of impurities that have
not been neutralized. In the case when sections are located in the same apparatus, in addition to the
redistribution of flows, each of them has thermal contact with neighboring sections; therefore, the
thermophysical properties of the dividing partition will be important here and will also determine
the course of the process.
Here, cylindrical apparatuses with one, two, and five sections in one body will be considered and
the influence of the number of sections on the main process indicators will be evaluated. Figure 4.8
shows the bed cross sections for these reactors.
Figure 4.8 Cross section of the bed with one, two, and five sections.
4.3 Mathematical Models of the Processes
Obviously, the considered designs of reactors require the setting of three-dimensional mathemat-
ical models, because the ongoing processes have an inhomogeneous radial distribution in the reac-
tor. However, some modifications can also be considered in a two-dimensional axisymmetric
formulation, as the geometry and conditions of the problem are axially symmetrical. Such geometry
in the model can significantly save computing resources. Basically, in the literature, models of both
the reverse-flow process and ACP are found in a simplified one-dimensional formulation. When the
radial distribution of process parameters is taken into account, it is also highly desirable for the
distribution of the velocity field and pressure in the system to be taken into account. For example,
the time-dependent Navier–Stokes Eqs. (4.1) and (4.2) for a free medium and the Brinkman
Eqs. (4.3) and (4.4), which are a modification of Darcy’s law taking viscous friction into account
for a porous medium that presented the catalyst bed.
∂u T 2μ
ρ +ρ u ∇ u=∇ −pI + μ ∇u + ∇u − ∇ uI , 41
∂t 3
∂ρ
+ ∇ ρu = 0 42
∂t
1 ∂u 1 1 2μ
ρ + 2ρ u ∇ u=∇ −pI + μ ∇u + ∇u T
− ∇ u I − uμK − 1 , 43
εp ∂t εp εp 3εp
∂ρ
εp +∇ ρu = 0, 44
∂t
The heat balance can be represented by two equations for the temperature of the gas and catalyst;
however, it is quite a reasonable simplification to use a single temperature model. In practice, the
phase temperatures are quite close, especially for small grains, and the model includes the following
Eq. (4.5), which takes into account the convective heat transfer, effective thermal conductivity of
the bed, and heat released during chemical reactions.
∂T
1 − εp ρc C c + εp ρC P + ρCP u ∇T − ∇ 1 − εp λc + εp λ ∇T = Q 45
∂t
The presented system of equations, of course, should be supplemented by the mass balance,
which is represented by slightly different equations for the systems under consideration. In both
cases, toluene was used as a model VOC.

The mass balance for the nonstationary catalytic reverse-flow process took into account diffusion
and convective mass transfer in the flow, as well as changes in the concentrations of the supplied
VOC and oxygen during the oxidation reaction. At low concentrations of reacting compounds, the
model of diluted species can be applied, which does not take into account the change in the con-
centration of solvent compounds (4.6).
∂ci
εp −∇ Di ∇ci + uci = Ri , 46
∂t
It was assumed that only the toluene oxidation reaction occurs in the bed
C7 H8 + 9O2 7CO2 + 4H2 O 47
In each reactor, the catalyst was loaded into a cylindrical mold with a volume of 0.195 m3, and the
residence time was 3.1 seconds. During the entire process, the toluene concentration and the inlet
gas mixture temperature were 1000 ppmv and 20 C, respectively. The process began with the fact
that the gas was supplied to a heated reactor, which had a temperature of 400 C. At a certain point,
when the heat wave approached the exit from the second reactor, the direction of the supply chan-
ged to the opposite. As a result of such alternating cycles of supply, the system either decays, losing
energy, or comes to a stable operating cyclic mode, where the temperature and concentration
dynamics in each cycle repeats that of the previous one. The maximum possible cycle times can
be chosen as a criterion for assessing the stability and energy efficiency of the process.
Reaction (4.7) was assumed to have first orders in the participating components and the following
kinetic parameters.
R = kctol co2 , 48
where k = k0 exp(−E/(Rg T)) and k0 = 3 107 mol/(m3 s), E = 62,900 J/mol.

The void of the bed was taken equal to 0.4, which corresponds to an average spherical filling, and
the radius of the catalyst pellets for calculating the hydraulic permeability according to the Kozeny–
Carman formula (4.9) was 2 mm. It was supposed to use the commercial copper–chromia–
alumina deep oxidation catalyst ICT-12-8 (manufactured by Katalizator Co., Novosibirsk, Russia)
in the model. The pellet porosity coefficient was 0.5, and the values of density (3900 kg/m3),
heat capacity (900 J/kg K), and thermal conductivity (27 W/m K) of the catalyst material were
taken according to the aluminum oxide data.
R2 ε3p
K= 2 m2 49
45 1 − εp

In ACP, it was assumed that there were no reactions in the flow, and the conversion occurred in the
pores of the pellet, on inner surface of the adsorbent–catalyst. The mass balance (4.10) took into
account the mass transfer with the surface of the pellets, where the exchange condition (4.11)
was set, as well as the diffusion transfer in the pores (4.12) with the corresponding reaction rates
on the inner surface (4.13).
∂ci
εp − ∇ Di ∇ci + uci = hD,i ci − cpe,i Sb , 4 10
∂t
∂cpe,i
Dpe,i = hD,i ci − cpe,i at r = R, 4 11
∂r
∂cpe,i 1 ∂ ∂cpe,i
εpe − 2 − r 2 Dpe,i = Rpe,i , 4 12
∂t r ∂r ∂r
∂csurf ,i
a = Rsurf ,i 4 13
∂t
The process of VOC adsorption and subsequent surface regeneration are quite complex and con-
sist of various stages, but it can also be described quite accurately by the following simplified
scheme.
A+ O A 4 14
A P 4 15
P + ν O2 CO2 + H2 O + O 4 16
It is assumed that, initially, the gas phase compound A is physically adsorbed on the oxidized sites
of the inner surface of the catalyst [O] with the formation of surface compound [A]. Then, the
reversibly adsorbed compound [A] passes into the strongly bound irreversible chemisorbed form
[P]. And substance [P] can already react with oxygen from the gas phase with the formation of car-
bon dioxide, water, and surface sites available for the subsequent adsorption of VOCs.
At this reaction rates (4.14)–(4.16) were presented as follows:
W 1 = k1 cpe, A csurf , O − k − 1 csurf , A , 4 17
W 2 = k2 csurf , A , 4 18
W 3 = k3 cpe, O csurf , P 4 19
And the reaction rates for the bulk and surface species are as follows:
Rpe, A = −W 1 , 4 20
Rpe, O = −ν W 3 , 4 21
Rsurf , A = W 1 − W 2 , 4 22
Rsurf , P = W 2 − W 3 , 4 23
Rsurf , O = −W 1 + W 3 4 24
Initially, the adsorbent–catalyst bed does not contain VOC impurities and the entire inner surface
is filled with compound [O]. Next, a gas stream of 100 ppmv enters the bed containing compound A,
which is gradually adsorbed in the bed, and when 5–10 ppmv VOC appears at the outlet, regener-
ation is carried out. Regeneration is initiated by turning on the internal heating element in the inlet
part of the bed. In this case, for a system with several sections, the heater is turned on in turn at time
intervals inversely proportional to the number of sections in the reactor. Thus, the time interval
between switching on the heating element in each of the sections is equal to the period between
regenerations in a single-section system.
The truncated cone inlet system was also compared with a cylindrical system having a diameter
and a height of 50 cm. The conical inlet reactors had the same volume as the cylindrical reactor, so
the height of their cylindrical part was changed in such a way as to maintain its diameter. The parti-
tioned system was a cylinder with a diameter of 25 cm, and a height of 1 m, and the thickness of the
separating partition was 1 cm.
The heated area was in the bed’s entrance and cylindrical in shape with a diameter of 12 cm and
height of 10 cm, which corresponded to about 1 kg of catalyst. In a system with several sections, the
heating area occupied the entire cross section and had a height of 5 cm. The specific power of the
heating element in all systems was 4 kW/l, and the time taken was 300 seconds.
For the cone entrance and multisectional (further in brackets) systems, the pellet radius, weight
adsorption capacity, gas flow rate, and residence time were set as 4 (3) mm, 1.5 (2) wt.%, 0.03
(0.005) m3/s, and 3.3 (10) s, correspondingly.
The unit volume catalyst adsorption capacity was calculated according to its awt wt.% meaning as
follows:
awt ρc
a= mol m3 4 25
M VOC
Table 4.1 The heat effects, activation energies, and preexponential factors of the reactions (Zazhigalov
et al. 2012).
i 1 −1 2 3
Qi, kJ/mol 49.3 × 103 — 1.4 × 106 2.86 × 106

Ei, kJ/mol 0 37.3 37.3 75.8
k 0i , 3
mol/m /s 106 1/s 1.47 × 109 4680 1.53 × 106 1/s
Table 4.2 Thermal properties of wall materials (Pfundstein et al. 2008/Walter de Gruyter GmbH).
Mineral wool Expanded clay Steel
Density, kg/m3 100 500 7850

Thermal conductivity, W/(m K) 0.05 0.2 44.5
Heat capacity, J/(kg K) 1000 840 475
And the kinetic parameters and heat effects for Reactions (4.14)–(4.16) are in Table 4.1.
The mass transfer coefficients were calculated in COMSOL using the equation:
Shi Dpe,i 1 3
hD,i = , Shi = 0 94 Re 1 2 Sci 4 26
R
The molecular diffusion coefficients for toluene and oxygen were calculated from the following
equations:
3 2 3 2
T T
DA = 7 7 10 −6 m2 s, DO = 1 78 10 −5 m2 s 4 27
273 273
The effective diffusion coefficients in pellet pores that are used in Equations (4.11) and (4.12)
depend more on pores’ properties and may vary in wide range. Based on various estimates, here
they are taken to be 50 times smaller than the molecular ones.
As mentioned above, of course, the partition material between the sections in a multisectional
system has a direct impact on the process, and so, three different materials are considered here,
mineral wool (by default), expanded clay, and steel with thermal properties indicated in
Table 4.2. In all corresponding figures, the gas moves from bottom to top.
4.4 Results

As stated earlier, depending on the duration of the cycle, the reverse-flow process can either sta-
bilize or die out due to large loss of heat. For each of the considered reactor geometries, the max-
imum cycle time for stable implementation of the process was calculated. The distribution of the
temperature field (top) and toluene concentration (bottom) is presented in Figures 4.9 and 4.10 for
4.4 Results 149
600
500
400
300
200
100
°C
1000
900
800
700
600
500
400
300
200
100
0
0 min 19 min 38 min 57 min 76 min ppmv
Figure 4.9 The temperature field (top) and toluene concentration (bottom) distribution during the cycle
for the system with axial feed.
600
500
400
300
200
100
°C
1000
900
800
700
600
500
400
300
200
100
0
Figure 4.10 The temperature field (top) and toluene concentration (bottom) distribution during the cycle
for the system with lateral feed.
axial and lateral feeds, respectively. The presented distributions correspond to the cycle durations of
76 and 60 minutes, and the time point 0 minute is the beginning of the cycle. It can be seen that in
the middle of the cycle, the distribution is naturally symmetrical to its beginning and end.
In case of axial feed, small radial gradients are observed, but they are almost symmetrical, as here
the reactor is axially symmetrical, and only the connecting tube introduces a slight asymmetry. As
for the lateral feed, significant radial gradients are observed. So at the beginning of the cycle, the
60-minutes cycles 65-minutes cycles

800 800
100 100 750
750
Maximal temperature,°C
Maximal temperature,°C
Toluene conversion, %
Toluene conversion, %
700 700
80 80
650 650
60 600 60 600
550 550
40 40
500 500
20 450 20 450
400 400
0 0
350 350
0 100 200 300 400 0 100 200 300 400 500
Time, min Time, min
Figure 4.11 Toluene conversion and maximal temperature dynamics for the 60- and 65-minutes cycles in
systems with lateral feed.
unevenness is observed mainly in the upper part of the reactors, and as the gas moves along the bed,
the unevenness becomes more noticeable in that part of the system into which the gas is supplied.
If you look at the graphs (Figure 4.11) of toluene conversion and maximum bed temperature for
cycles of 60 and 65 minutes, you can see that the former ensures stable operation of the system,
because successive cycles repeat the previous one, and the second leads to attenuation after 8 cycles.
The average conversion of toluene per cycle in the first case is 99.2%.
With regard to the tangential arrangement of feed ducts, the following distribution of tempera-
ture and concentration in the central plane (Figure 4.12) can be observed for a cycle of 88 minutes.
However, the most interesting is the temperature distribution in the axial planes, which is shown in
Figure 4.13 for a half-cycle. Here, one can observe the movement of the heat wave not only along
the bed, but also a slight twist around the central axis, which theoretically can have a favorable
effect on the efficiency of the process.
600
500
400
300
200
100
°C
1000
900
800
700
600
500
400
300
200
100
0
Figure 4.12 The temperature field (top) and toluene concentration (bottom) distribution during the cycle for
the system with tangential feed.
4.4 Results 151
600
500
400
300
200
100
°C
0 min 11 min 22 min 33 min 44 min
Figure 4.13 The temperature field distribution in axial planes during the half of 88-minutes cycle for the
system with tangential feed.
The behavior of the system with a tangential–lateral arrangement of ducts is close enough to the
behavior with a tangential arrangement. Similarly, a reactor with lateral inlets and an axial con-
necting duct gives temperature and concentration distributions close to that of a reactor with lateral
ducts. A system with a mutually perpendicular arrangement of lateral ducts forces the gas to move
in a not-quite straight direction, thus, showing greater stability than a system with a standard lateral
arrangement. This fact is reflected in Table 4.3, which shows the cycle duration at toluene conver-
sion as higher, 99%, including for a one-dimensional model often used to represent the reverse-flow
process.
The 1D model does not take radial gradients into account, so it shows longer stable cycles
with toluene conversion higher than 99% (97 minutes). The reactor with lateral ducts shows
the worst stability among the modifications presented (61 minutes), the axial location of the
ducts allows for longer cycles (80 minutes). One of the reasons is more uniform radial distribu-
tion of temperatures in the second case. Furthermore, the wider spread of residence time,
coupled with uneven temperature distribution in the lateral feed reactor, also results in uneven
distribution of toluene oxidation rates. This causes a correspondingly uneven heat release in the
bed, and the interaction of these facts leads to a decrease in process efficiency, toluene conver-
sion, and thermal stability.
Table 4.3 The cycle duration at toluene conversion >99% with different
connecting ducts.
Type of inlet/outlet ducts Cycle duration at toluene conversion

connection >99%, min
1D 97
Lateral 61
Axial 80
Tangential 80
Tangential–lateral 77
Lateral–axial 63
Lateral–orthogonal 68
The lateral–axial (63 minutes) and lateral–orthogonal (68 minutes) systems are ranked between
those described above, and the main inhomogeneities in them arise in the part of the reactor where
the gas is supplied. It can be concluded that the design of the region connecting the reactor parts has
little effect on the process. Therefore, the behavior of these variants is rather close to the behavior of
a system with a lateral arrangement of branch ducts. The lateral–orthogonal system has few advan-
tages, as it causes the gas flow to slightly change its trajectory, twisting slightly.
The systems with tangential feed turned out to be the most stable (77 and 80 minutes). Tangential
systems provide stable cycles of up to 88 minutes with a toluene conversion of about 94%, which the
axial type of supply cannot provide. The paths of the supplied mixture and the heat wave form a
kind of spiral, which also allow us to speak of a more uniform radial distribution in the bed.
Undoubtedly, many factors influence the behavior of the process in all systems considered,
including the size of the catalyst granules used. Therefore, for systems with lateral and axial gas
supplies, in addition to grains with a radius of 2 mm, variants with radii of 1 and 4 mm were also
considered. The size of the particles used affects the hydraulic resistance in the bed and the spatial
distribution of process parameters. Figure 4.14 shows the temperature distribution for different par-
ticle radii. Large pellets lead to a lower hydraulic resistance, and due to this, the temperature front
in the bed is somewhat washed out. The lower permeability at fine particles indicates a more uni-
form temperature distribution in the bed, with the maximum cycle time being almost independent
of the system design (Figure 4.15). However, as particle radius increases, the axial feed system first
shows better behavior, and then the lateral feed system shows better stability. This is due to the fact
that at low pressure drop, the gas with axial flow moves predominantly along the central axis.
600
500
400
300
200
100
°C
1 mm 2 mm 4 mm
Figure 4.14 Temperature distribution at the end of the stabilized cycle for 1-, 2-, and 4-mm pellet radius in
axial and lateral systems.
4.4 Results 153
Maximal cycle duration, min Maximal temperature,°C Average pressure drop in bed, kPa
100 750
14
90 700 12
80 10 Axial
650
70 8 Lateral
600
60 6
550
50 4
40 500 2
30 450 0
1 2 4 1 2 4 1 2 4
Pellet radius, mm Pellet radius, mm Pellet radius, mm
Figure 4.15 The maximal stable cycle duration, maximal temperature, and average pressure drop depending
on pellet radius in axial and lateral reactor geometries.

4.4.2.1 Reactor with Truncated Cone Entrance
In ACP, the duration of the adsorption stage was determined by the moment of appearance of tol-
uene at the exit from the bed in an amount of 5–10 ppmv. First of all, the simulation of the process in
a cylindrical bed was carried out, where gas is supplied to the entire section, and regeneration is
carried out by a heating element located in the center of the bed entrance.
The Figure 4.16a shows how toluene moves through the bed, penetrating into the pores of the
pellets, adsorbs, and forms a reversibly sorbed compound [A] (Figure 4.16b). Over time, [A] trans-
forms into the chemisorbed form [P] (Figure 4.16c). The temperature at the adsorption stage
changes insignificantly and is in the range of 20–23 C.
After the adsorption cycle is completed, the bed is regenerated by turning on the heating element.
When a thermal wave moves along the bed (Figure 4.17a), part of the reversibly adsorbed impurities
can be desorbed into the gas phase (in Figure 4.17b, the concentration peak reaches 5000 ppmv) and
re-adsorbed further along the bed (Figure 4.17c), and part of it can pass into the compound [P]
(Figure 4.17d). Oxidation of adsorbed VOCs leads to a significant release of heat during the reaction,
which forms a temperature front that ensures further oxidation of VOCs over the bed and can exist
due to the released heat. It can be seen that the thermal wave moves mainly in the axial direction—
the direction of the gas flow—and almost does not propagate in the radial direction. After the pas-
sage of the heat wave, approximately 85% of the volume of the entire bed remained unregenerated.
The oxygen is not consumed completely for oxidation, that is, its concentration in the mixture does
not fall below 17 vol.%; this is important to avoid the formation of products of partial oxidation of
toluene, possible under oxygen-lean conditions.
It can be assumed that with such an organization, the process cannot be sufficiently efficient and
provide a high degree of purification. Because after the first regeneration most of the bed remained
unregenerated, the duration of the second adsorption stage was only 10,000 seconds, as toluene
appeared quite quickly at the outlet. Such a duration of the adsorption stage cannot provide a suf-
ficient amount of accumulated impurities for carrying out regeneration in an autothermal mode.
The duration of the third adsorption stage is 5000 seconds, and in the fourth cycle, the concentra-
tion of toluene at the outlet immediately exceeds to 10 ppmv, so the adsorption is inefficient due to
the large part of the unregenerated bed. Figure 4.18 shows the distribution of adsorbed toluene in
the bed after each of the three cycles.
(a)
100
90
80
70
60
50
40
30
20
10
0
A, ppmv
(b)
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
[A]
(c)
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
[P]
25,000 s 50,000 s 75,000 s 91,000 s
Figure 4.16 Toluene concentration (a), reversibly [A] (b), and irreversibly [P] (c) adsorbed toluene surface
fraction averaged on pellet volume in cylindrical bed at the adsorption stage.
The process proceeds in a completely different way in a reactor with an inlet in the form of a
truncated cone. The adsorption and regeneration steps are shown in Figures 4.19 and 4.20 using
the example of a 70 cone angle reactor. At the adsorption stage (Figure 4.19), the movement of
toluene over the bed and the formation of compounds on the inner surface of the adsorbent is fun-
damentally the same as for the cylindrical bed.
4.4 Results 155
Figure 4.17 The temperature (a)

distribution (a), toluene concentration 600
(b), reversibly [A] (c), and irreversibly [P]
500
(d) adsorbed toluene surface fraction
averaged on pellet volume in cylindrical 400
bed at the regeneration stage of the
300
first cycle.
200
100
T, °C
(b)
5000
4500
4000
3500
3000
2500
2000
1500
1000
500
0
A, ppmv
(c)
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
[A]
(d)
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
[P]
300 s 1000 s 1700 s 2500 s
Figure 4.18 Adsorbed toluene distribution 0.7

[A] + [P] in the end of cycles.
0.6
0.5
0.4
0.3
0.2
0.1
0
[A]+[P]
After 1st cycle 2nd 3rd
100 Figure 4.19 Toluene concentration in the

90 bed with truncated cone entrance with 70
angle at the adsorption stage of the first cycle.
80
70
60
50
40
30
20
10
0
A,
10,000 s 30,000 s 60,000 s 90,000 s ppmv
However, during the regeneration stage, the heat wave propagates from a narrow inlet, covering
most of the bed (Figure 4.20a). It can be seen from Figure 4.20b–d that after the passage of the heat
wave, only parts of the bed at the inlet remained unregenerated (because the region is cooled by the
supplied cold air) and along the walls of the conical part.
With repeated cycles of adsorption and regeneration, the area of the inactive zone does not
expand, it occupies 13% of the bed volume, and the process stabilizes (Figure 4.21).
Undoubtedly, the value of the cone angle required for a cyclically repeating ACP depends on the
parameters of the bed and the supplied mixture. For example, with the parameters under consid-
eration, an angle of 60 cannot provide stable repeating cycles (Figure 4.22). The nonregenerated
area expands and after the eighth cycle, adsorption cannot be carried out, because the concentra-
tion of toluene at the outlet of the bed exceeds 10 ppmv. Increasing the angle to 80 makes it pos-
sible to reduce the portion of the unregenerated area to 5%. Changing the process parameters affects
the optimal value of the cone angle; however, it is qualitatively shown that the presented geometric
modification of the process makes it possible to efficiently conduct ACP when heating at the stage of
regeneration of a small area at the entrance to the bed.
4.4.2.2 Multisectional Reactor

Modeling of a multisectional system was carried out on the example of reactors with one, two, and
five sections. For a single-section system, the duration of the adsorption stage (until the appearance of
toluene at the outlet in an amount of 10 ppmv) was 200,000 seconds. Accordingly, for a system with
several sections, the interval between regenerations was 2 × 105/N, where N is the number of sections.
The Figure 4.23a shows the concentration of toluene in the gas phase at the adsorption stage
for a one-section system. Similar to the previously considered process, toluene moves through
the bed and adsorbed on the inner surface of the pellet (Figure 4.23b and c). Concentrations of
surface substances averaged on pellet volume give a general idea of their content in the bed; as
for the distribution in the pellet, Figure 4.24 gives a more detailed picture. Here, using the
example of a particle lying at the bed entrance, it is shown how initially a region with the
reversibly adsorbed compound [A] is formed near the outer surface of the pellet, which trans-
forms into the form [P].
After completion of the adsorption stage, as in cases described earlier, the bed is regenerated. As a
result, a heat wave (Figure 4.25) is formed in the bed, which ensures the oxidation of adsorbed
4.4 Results 157
Figure 4.20 The temperature distribution (a), (a)

toluene concentration (b), reversibly [A] (c), and 600
irreversibly [P] (d) adsorbed toluene surface
500
fraction averaged on pellet volume in the bed
with truncated cone entrance with 70 angle at 400
the regeneration stage of the first cycle.
300
200
100
T, °C
(b)
4000
3500
3000
2500
2000
1500
1000
500
0
A, ppmv
(c)
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
[A]
(d)
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
[P]
300 s 800 s 1300 s 2000 s
impurities. As before, part of the reversibly adsorbed impurities can be desorbed into the gas flow
and reduce the overall degree of purification, which is calculated as the ratio of the amount of neu-
tralized impurities to the total amount that entered the bed. Qualitatively, the distribution and
dynamics of the concentrations of substances and temperature repeat those described for the sys-
tem in Section 4.4.2.1.
1 Figure 4.21 Adsorbed toluene distribution

[A] + [P] in the end of cycles for the bed with
0.9
truncated cone entrance with 70 angle.
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
[A]+[P]
After 1st 4th 7th cycle
1 Figure 4.22 Adsorbed toluene distribution

[A] + [P] in the end of cycles for the bed with
0.9 truncated cone entrance with 60 angle.
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
[A]+[P]
After 1st 5th 8th cycle
Surface concentrations [A] and [P] in the pellet during heating change according to the
Figure 4.26. After few minutes, the pellet surface becomes regenerated, and adsorption can proceed
once again as before.
In a multisection system, when performing regenerations in turn, there is obviously some start-
up mode of operation, because a short adsorption stage may not allow autothermal regeneration
after the heating element is turned off. But after two complete cycles of regeneration of all sections,
the system comes to a stable mode and each cycle repeats the previous one. An example of the dis-
tribution of concentrations of surface compounds for stabilized operation of a two-section system is
shown in Figure 4.27a and b. Here, regeneration is carried out in the right section, and at the end of
the half-cycle, the distribution of concentrations is symmetrical with respect to its beginning.
In a multisection system, the sections have thermal contact through a separating wall, and
Figure 4.27c shows the temperature distribution in such a system. The maximum temperature
in the nonregenerated section is relatively low and about 80 C, because the wall material, mineral
wool, has a low thermal conductivity. However, such an increase in temperature accelerates the
4.4 Results 159
Figure 4.23 Toluene concentration distribution in (a)

gas phase (a), reversibly [A] (b), and irreversibly [P] (c) 100
adsorbed toluene surface fraction averaged on pellet 90
volume along the bed at the adsorption stage for
one-sectional system. 80
70
60
50
40
30
20
10
A, ppmv
(b)
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
(c) [A]
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0 0.7 × 105 1.4 × 104 2 × 105 s [P]
desorption of toluene, which affects its concentration in the gas phase—it can slightly exceed the
supplied 100 ppmv, but the toluene almost does not appear at the outlet, as it is successfully
adsorbed further along the bed. At the same time, a rise in temperature also activates the reaction
of the transition of toluene to the chemisorbed form, which can increase the final degree of
purification.
Adsorption stage Figure 4.24 [A] and [P] concentrations in

0.8 pellet at the reactor inlet at the adsorption
[A] stage for one-sectional system.
Time:
0.6
Surface fraction
70,000 s
140,000 s
0.4 200,000 s [P]
0.2
0.0
0.0 0.6 1.2 1.8 2.4 3.0
Pellet radius, mm
500 Figure 4.25 Temperature distribution along the

bed volume during the regeneration for one-
450 sectional system.
400
350
300
250
200
150
100
50
300 1500 2500 4000 s T, °C
In a system with a larger number of sections, the process proceeds principally in the same way;
Figure 4.28 shows the distribution of average concentrations [A] and [P] for a 5-section system.
Here, regeneration will be carried out in the area with the highest concentration of adsorbed
toluene and then in the remaining sections according to the arrow direction with a period of
2 × 105/5 = 4 × 104 seconds.
The purpose of dividing the reactor into several sections was to reduce the peak concentrations
of toluene and peak outlet gas temperature. The Figure 4.29a shows the concentration of toluene
at the outlet of the reactor for a different number of sections. As mentioned earlier, for
4.4 Results 161
Figure 4.26 [A] and [P] concentrations in Regeneration stage

pellet at the reactor inlet at the
0.8
regeneration stage for one-sectional [A]
system. Time:
0s
0.6
Surface fraction
150 s
300 s
0.4 400 s
[P]
0.2
0.0
0.0 0.6 1.2 1.8 2.4 3.0
Pellet radius, mm
multisection systems, the first two complete cycles are the starting point for reaching stable values.
During the first cycle, regeneration may not always take place in an autothermal mode, and in the
second cycle, the concentration in the outgoing emissions may be slightly higher than expected in
stable mode, as the bed has a larger amount of adsorbed toluene (part of it remained from the first
cycle). Similarly, a reduction in peak temperatures in the outlet stream in multisectional systems is
achieved (Figure 4.29b).
With the natural separation of the gas flow between sections, its rate in the regenerated
section should be lower than in the others, as it has a higher temperature and pressure drop. This
reduces the amount of emissions from the regenerated section, and one can observe a decrease in
peak concentrations more than proportional to the number of sections. This also confirms the
dependence of the degree of purification on the number of sections, presented in Table 4.4. A larger
number of sections lead not only to an improvement in the above important indicators but also to a
decrease in the peak pressure drop during the regeneration of one of the sections. However, given
that too many sections can lead to excessive complexity of the design, we can talk about their opti-
mal number—a compromise between process performance and the cost of manufacturing and
operating the reactor.
Another factor limiting the number of sections is the increase in the specific heat transfer area
with an increase in the number of sections. Thus, the heat from the regenerated section more
actively flows into the adjacent sections and causes redistribution of the concentrations of surface
compounds, which can have an ambiguous effect on the average degree of purification. Of course,
the material of the partition between the sections, namely, its thermal conductivity, also has an
important influence here. The Table 4.5 shows data for three different materials—mineral wool,
expanded clay, and steel for a two-section system.
If we compare expanded clay and mineral wool, then the first one has five times higher thermal
conductivity, which also affects the performance. Thus, for expanded clay, the maximum temper-
ature in the regenerated section decreased from 440 to 390 C, and in the neighboring section it
increased from 80 to 170 C. At the same time, the peak concentration of toluene at the outlet
and the average degree of purification are almost equal in both cases. Steel baffles have a very high
(a) Figure 4.27 The reversibly [A] (a),

0.5 irreversibly [P] (b) adsorbed toluene surface
0.45 fraction averaged on pellet volume and
temperature (c) during the established half-
0.4
cycle in two-sectional bed.
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
(b) [A]
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
[P]
(c)
450
400
350
300
250
200
150
100
50
T, °C
0 1500 3000 4500 100,000 s
thermal conductivity and the standard turn-on time of the heating element of 300 seconds does not
allow reaching the ignition temperature due to high heat exchange with the adjacent section.
A significantly longer operation of the heater leads to the simultaneous ignition of both sections
and the behavior of the process becomes similar to the behavior of a single-section system, which
makes the idea of sectioning meaningless.
4.4 Results 163
0.5
0.4
0.3
0.2
0.1
0
[A] [P]
Figure 4.28 Reversibly [A] and irreversibly [P] adsorbed toluene averaged on pellet volume distribution in
entrance section. Circular arrow—direction of the regeneration sequence.
1 section
Outlet toluene concentration, ppmv
2500 2 sections
5 sections
2000
1500
1000
500
0
0 2 × 105 4 × 105 6 × 105
Time, s
500
400
Outlet temperature, °C
300
200
100
0 2 × 105 4 × 105 6 × 105

Time, s
Figure 4.29 Outlet toluene concentration and gas temperature time dependence for different number of
sections in the reactor (Zazhigalov et al. 2023b / American Chemical Society).
Table 4.4 The process parameters depending on the number of sections (Zazhigalov et al. 2023b/American
Chemical Society).
Number of sections, N 1 2 5
Average VOC abatement degree, % 70.9 78.3 82.1

Peak outlet toluene concentration, ppmv 2870 990 290
Peak outlet gas temperature, C 482 204 103
Average pressure drop, Pa 19.71 19.67 19.68
Maximum pressure drop, Pa 54.15 31.60 23.43
Table 4.5 The process indicators depending on the wall material in a two-sectional reactor (Zazhigalov et al.
2023b/American Chemical Society).
Mineral wool Expanded clay Steel
Average VOC abatement degree, % 78.3 78.1 74.5

Peak outlet toluene concentration, ppmv 990 980 2200
Peak outlet gas temperature, C 204 269 411
4.5 Conclusion
One of the directions for increasing the efficiency of processes for neutralizing waste gases from
impurities of VOCs is, among other things, improving the design of reactors and the organization
of the process. The study of the proposed modifications using the method of mathematical mod-
eling allowed us to draw the following conclusions.
In the ACP, placing the heating system directly in the inlet part of the adsorbent–catalyst bed
makes it possible to reduce specific energy consumptions by 1–2 orders of magnitude, while orga-
nizing the inlet part of the reactor in the form of a truncated cone increases the overall efficiency of
the process in this case. Dividing the system into separate sections with an independent heating
system allows not only to reduce the peak concentrations of unconverted impurities in the exhaust
gases and outflow temperature, but also to increase the process average degree of purification from
impurities.
It was shown for the reverse-flow process that both the concentration and temperature radial
gradients present in the bed do not permit the use of a classical one-dimensional model for a
detailed description of the process. Among the various arrangements of the inlet/outlet ducts of
the reactors, the best performance was obtained for their tangential arrangement, where the ther-
mal wave generated in the bed has a slight spiral twist, which ensures more complete bed
utilization.
In general, it can be stated that the use of the CFD method for modeling nonstationary catalytic
process and ACP opens up new opportunities for optimizing the design of reactors for their imple-
mentation with account of complicated dynamic phenomena, especially significant and specific in
3D geometrical representation.
Acknowledgments 165
4.6 Notation
[A] – product of reversible adsorption;

[O] – oxidized site at the catalyst surface;
[P] – product of VOC irreversible chemisorption on internal pellet surface;
A – VOC molecule;
a – catalyst adsorption capacity, mol/m3;
awt – catalyst weight adsorption capacity, wt.%;
ci – gas flow component (VOC or oxygen) concentration, mol/m3;
cpe,i – gas component (VOC or oxygen) concentration in the pellet pores, mol/m3;
csurf,i – surface compound, surface fraction;
Cc, CP – catalyst and gas heat capacity, J/kg/K;
Di, Dpe,i – molecular and effective in pellet diffusion coefficients, m2/s;
Ei – activation energy, kJ/mol;
hD,i – mass transfer coefficients, m/s;
K – bed permeability, m2;
ki, k 0i – reaction rate constant and pre-exponential factor, mol/m3/s or 1/s;
MVOC – VOC molar mass, kg/mol;
N – number of sections;
Q – heat source, W/m3;
Qi – thermal effects, kJ/mol;
r – pellet radial coordinate, m;
R – pellet radius, m;
Rpe,i – reaction source term for gas components, mol/m3/s;
Rsurf,i – reaction source term for surface compounds, mol/m3/s;
Sb – active specific surface area, m2/m3;
Shi – Sherwood number;
T – bed temperature, K;
u – gas velocity vector, m/s;
εp, εpe – bed and catalyst porosity;
λc, λ – thermal conductivity of the catalyst and gas, W/m/K;
μ – gas dynamic viscosity, kg/m/s;
ρ – gas density, kg/m3
ρc – catalyst bed density, kg/m3
Abbreviations
ACP—adsorption–catalytic process
VOC—volatile organic compound
Acknowledgments
This work was partially supported by the Ministry of Science and Higher Education of the
Russian Federation within the state assignment for the Boreskov Institute of Catalysis (project
FWUR-2024-0037) and by the Tyumen Oblast Government, as part of the West-Siberian
Interregional Science and Education Center’s project No. 89-DON (3).
References
Evaraert, K. and Baeyens, J. (2004). Catalytic combustion of volatile organic compounds. Journal of
Hazardous Materials 109: 113–139.
Garcia, T., Solsona, B., and Taylor, S.H. (2014). The catalytic oxidation of hydrocarbon volatile organic
compounds. In: Handbook of Advanced Methods and Processes in Oxidation Catalysis (ed. D. Duprez
and F. Cavani). Imperial College Press.
Gonzalez-Velasco, J.R., Aranzabal, A., Pereda-Ayo, B. et al. (2014). Catalytic oxidation of volatile organic
compounds: chlorinated hydrocarbopns. In: Handbook of Advanced Methods and Processes in
Oxidation Catalysis (ed. D. Duprez and F. Cavani). Imperial College Press.
Li, W., Wang, J., and Gong, H. (2009). Catalytic combustion of VOCs on non-noble metal catalysts.
Matros, Y.S. and Bunimovich, G.A. (1996). Reverse-flow operation in fixed bed catalytic reactors.
Catalysis Reviews – Science and Engineering 38 (1): 1–68.
Matros, Y.S., Noskov, A.S., and Chumachenko, V.A. (1991). Catalytic Purification of Industrial Waste
Gases. Novisibirsk: Nauka.
Ojala, S., Pitkaaho, S., Laitinen, T. et al. (2011). Catalysis in VOC abatement. Topics in Catalysis 54:
1224–1256.
Pfundstein, M., Gellert, R., Spitzner, M., and Rudolphi, A. (2008). Insulating Materials: Principles,
Materials, Applications. München: Birkhäuser.
Retallick, W.B., 1996. Air cleaner capable of catalytic regeneration. US Patent No.5487869.
Robinson, E., 1971. Treatment of gaseous effluent. UK Patent No. 1582441.
Salden, A. and Eigenberger, G. (2001). Multifunctional adsorber/reactor concept for waste air
purification. Chemical Engineering Science 56: 1605–1611.
Schoubye, P., 1979. Method and apparatus for the removal of oxidizable pollutants from gases. UK Patent
application 2051761.
Spivey, J. (1987). Complete catalytic oxidation of volatile organics. Industrial & Engineering Chemistry
Research 26 (11): 2165–2180.
Stanley, P.B., Ritscher, J.S., 1979. Novel combustion process for an organic substrate. USA Patent No.
4234549.
Vernikovskaya, N.V., Zagoruiko, A.N., Chumakova, N.A., and Noskov, A.S. (1999). Mathematical
modeling of unsteady-state operation taking into account adsorption and chemisorption processes on
the catalyst pellet. Chemical Engineering Science 54: 4639.
Zagoruiko, A.N., Kostenko, O.V., and Noskov, A.S. (1996). Development of the adsorption-catalytic
reverse-process for incineration of volatile organic compounds in diluted waste gases. Chemical
Engineering Science 51 (11): 2989.
Zagoruiko, A.N., Bobrova, L., Vernikovskaya, N., and Zazhigalov, S. (2021). Unsteady-state operation of
reactors with fixed catalyst beds. Reviews in Chemical Engineering 37 (1): 193–225.
Zazhigalov, S.V., Chumakova, N.A., and Zagoruiko, A.N. (2012). Modeling of the multidispersed
adsorption-catalytic system for removing organic impurities from waste gases. Chemical Engineering
Science 76: 81–89.
Zazhigalov, S.V., Mikenin, P.E., Lopatin, S.A. et al. (2016). An improved adsorption–catalytic process for
removing volatile organic compounds from exhaust gases. Catalysis in Industry 8 (3): 231–241.
Zazhigalov, S., Mikenin, P., Pisarev, D. et al. (2017). Modifications of the adsorption-catalytic system for
organic impurities removal. Chemical Engineering and Processing: Process Intensification 122: 538–549.
References 167
Zazhigalov, S., Chumakova, N., and Zagoruiko, A. (2018). Adsorption-catalytic process for removal of
volatile organic compounds from lean waste gases: optimization of the adsorbent-catalyst bed
geometry. Chemical Engineering and Processing: Process Intensification 132: 1–10.
Zazhigalov, S., Elyshev, A., and Zagoruiko, A. (2023a). Catalytic reverse-flow oxidation process in
reactors of various designs: axial, side and tangential gas inlet. Chemical Engineering Research and
Design 191: 364–374.
Zazhigalov, S., Elyshev, A., and Zagoruiko, A. (2023b). Mathematical modeling of adsorption-catalytic
process with internal heater in multi-sectional arrangement. Industrial & Engineering Chemistry
Research.
168
Molecular Reconstruction of Complex Hydrocarbon Mixtures

for Modeling of Heavy Oil Processing
Nikita Glazov and Andrey Zagoruiko
Boreskov Institute of Catalysis SB RAS, Prospekt Akademika Lavrent’yeva, 5, Novosibirsk, Novosibirsk Oblast, Russian Federation
5.1 Introduction
Humanity needs fuel. Energy sources are crucial for today’s human society, and the oil refinery is
still an essential part of the modern economy. One of the challenging tasks it is facing right now is
the efficient processing of heavy petroleum cuts, which becomes even more severe with each pass-
ing year. Many technological advancements, both small and big, have been made to solve this
problem.
Of course, mathematical modeling is also widely employed in tackling this task on various levels,
from the atomic level where the quantum chemistry methods unveil the details of catalytic pro-
cesses to the level of the whole plant with different planning procedures.
This particular chapter is focused on one narrow application of mathematical modeling for a very
common problem that emerges during the building of kinetic models for oil processing: the com-
position representation problem and a particular set of methods for solving it. These methods are
often grouped together under the umbrella term “molecular reconstruction” (Ren et al. 2019).
Unlike the lumping strategy, molecular reconstruction often represents mixtures in terms of
molecules/well-studied mixtures, rather than abstract compounds, which has a certain advantage.
For example, models that use lumping approach to represent mixture can violate conservation of
elements, which is not the case with reconstructed composition. Although, molecular reconstruc-
tion methods are often distinguished from lumping strategies, some of them can work with pseudo-
components as well.
5.2 The Problem
The number of chemical species in the heavy petroleum cuts is usually far beyond the capabilities of
the current analytical methods. It means that whenever we want to build a kinetic model for an oil
refinery process, we cannot simply rely on the composition of the products and feedstock (because it
is usually unavailable). There is always a gap between the real composition and the information we
have about it. To bridge this gap, we can use math. Notice that in case of the heavy petroleum cuts
not only abundance of different molecules is unavailable but also the list of molecules in the
mixture. Even the exact number of molecules is unknown.
This task might seem insurmountable or even impossible, however, this challenge could be over-
come with the understanding that the composition we need might be just “close” to the real one
instead of being the actual composition. The word “close” here is used in the vaguest way possible.
We will see that the exact meaning in which the composition is close to the real one varies from
method to method, but in general, reconstructed composition is better be seen as a very good model
mixture that imitates everything we know about the sample rather than actual composition.
5.3 Illustration
To better understand what different molecular reconstruction methods imply, we will be consid-
ering the following toy example: a mixture of normal paraffines from C10 to C20, and the only piece
of information available would be the average carbon content in the mixture. This example is rather
arbitrary and does not have (to the authors’ knowledge) any meaningful application. However, it is
sufficiently complicated to demonstrate some meaningful aspects of molecular reconstruction
methods. Moreover, there is nothing particularly interesting in carbon content, and it all can be
easily generalized.
Let α be the average carbon content in the mixture, αi, Mi, and xi be the carbon content, molar
mass, and the molar fraction of the ith molecule, respectively. It is easy to see that any set of positive
molar fractions that satisfies the following two equations represent a mixture with the average car-
bon content, exactly α.
x i M i αi − α = 0 51
xi = 1 52
This system has 2 equations and 11 unknowns, which is an extremely ill-posed problem. Pro-
blems in real application are even less well-posed.
5.4 Reconstruction by Entropy Maximization (REM)
From the earlier small analysis, it is clear that the system has numerous degrees of freedom. One
way of approaching this problem is to notice that if we consider all possible solutions, there are
many solutions with concentrations close to uniform values, than solutions with extreme values.
The way to formalize it mathematically is Shannon entropy with Lagrange multiplier.
ε x, μ, λ = − x i ln x i + μ 1 − xi + λ 0 − x i M i αi − α 53
Notice that the framework of entropy maximization is very useful and general for distribution
finding. Exponential, normal, beta, gamma distributions, and many more, including even Cauchy
distribution, can be obtained through entropy maximization under certain constraints (Lisman and
Zuylen 1972).
Application of this approach for molecular reconstruction was performed by Hudebine and
Verstraete (2011). The reader is encouraged to read this work for extra details that were left out
of this section.
The extremum of (3) will satisfy Equations (5.1) and (5.2) by construction. Moreover, it has a
semi-analytical solution.
170 5 Molecular Reconstruction of Complex Hydrocarbon Mixtures for Modeling of Heavy Oil Processing
∂ε
= − ln x i − 1 − μ − λM i αi − α = 0 54
∂x i
ln x i = − 1 − μ − λM i αi − α 55
xi = e − 1+μ
e − λM i αi − α
56
e1 + μ xi = e − λM i αi − α
57
Z = e1 + μ = e − λM i αi − α
58
1 − λM i αi − α
xi = e 59
Z
ε λ = ln Z 5 10
Equation (5.4) is the partial derivative of Equation (5.3) with respect to xi. This expression is set to
be 0 to find the extremum of the original function. Equations (5.5) and (5.6) are results of algebraic
transformation to isolate xi. Equation (5.7) results from the summation of all concentration and
factoring out exp(1 + μ) as a common factor. The sum of all components is 1 due to material bal-
ance. With the sum of exponents renamed (Equation 5.8), Equation (5.6) can be written as Equa-
tion (5.9). The Equation (5.10) is the result of substituting Equation (5.9) into the original entropy
Equation (5.3).
For practical application, the extremum for the natural logarithm of Z occurs at the same point as
for Z itself. Notice that the problem went from 13 unknowns (11 molar fractions + 2 Lagrange mul-
tipliers) to 1 unknown (1 Lagrange multiplier). For this toy example, it might not look that impres-
sive, but it is a huge time-saver when the number of components goes beyond thousands and tens of
thousands.
In a more general case, there might be more constraints, but as long as they are linear, the semi-
analytical solution exists and reduces the number of unknowns to the number of linear constrains.
N N J N
ε x, μ, λ = − x i ln x i + μ 1 − xi + λj f j − x i f i,j 5 11
i=1 i=1 j=1 i=1
N J
Z= exp − λj f i,j 5 12
i=1 j=1
J
ε λ = ln Z + λj f j 5 13
j=1
Equation (5.11) is the generalization of Equation (5.3); where fi,j is the weight with which xi con-
tributes to the property fj, just like Mi (αi − α) in (1) is the weight with which xi should sum to 0.
Notice that when at least one of the fj values is nonzero, the final expression for entropy (13) has
an additional sum compared to (10). However, it is always possible to reformulate fi,j in such a way
to make all values of fj zero as shown (14).
N N N N N N N
fj − x i f i, j = f j xi − x i f i, j = xi f j − x i f i, j = x i f j − f i, j = x i f ∗i, j
i=1 i=1 i=1 i=1 i=1 i=1 i=1
5 14
Entropy maximization procedure can be carried out not only for the exact value but also for the
variance, thus making the final distribution closer to normal. The mathematical machinery for
deriving the semi-analytical solution is very similar to what was shown before.
N
2
σ2 = xi f − f i 5 15
i=1
The simple semi-analytical solution also exists for “uncertain/relax” constraints. The details
behind the reason why the suggested method works are beyond the scope of this chapter and curi-
ous readers are advised to look for information about a Gaussian prior. Turns out, the way to
include “uncertain” constraints for entropy models is by adding logarithm of the Gaussian function,
with fk and σ k being the mean and standard deviation, respectively.
2 2
1 f − x i f i,k 1 f k − x i f i,k
ln exp − k = − + const 5 16
2πσ 2k 2σ 2k 2 σ 2k
The normalization factor does not depend on xi and for any given value, σ k is just a constant that
does not change the extremum for xi, so it can be dropped. With this modification, the entropy func-
tion has the following form (Equation 5.17):
N N K N 2
1
ε x, μ = − x i ln x i + μ 1 − xi + − fk − x i f i,k 5 17
i=1 i=1 k=1
2σ 2k i=1
The similar procedure to Equations (5.4)–(5.10) yields,

K
1 2 fi
ε ϵ = ln Z + ϵ − ϵk 5 18
k=1
2 k σk
with
N K f i,j
Z= exp ϵk 5 19
i=1 k=1
σk
N
fk − x i f i,k
i=1
ϵk = 5 20
σk
All mentioned ways of putting constraints can work together, and the most general form of
entropy maximization problem (that has a simple semi-analytical solution) is as follows:
N N
ε x, μ, λ, υ = − x i ln x i + μ 1 − xi
i=1 i=1
J N
+ λi f j − x i f i,j 5 21
j=1 i=1
N 2
K fk − x i f i,k
1 i=1
+ −
k=1
2 σ 2k
M N
2 2
νm σm − xi f − f i
m=1 i=1
which can be simplified to the following:
J K M
1 2 fi
ε λ, ϵ, υ = ln Z + λj f j + ϵ − ϵk + νm σ 2m 5 22
j=1 k=1
2 k σk m=1
With
N J K M
f i,j 2
Z= exp − λj f i,j + ϵk − νm f i,m − f m 5 23
i=1 j=1 k=1
σk m=1
After finding the maximum of (22), the concentration could be found as follows:
J K M
1 f i,j 2
xi = exp − λj f i,j + ϵk − νm f i,m − f m 5 24
Z j=1 k=1
σk m=1
Notice an important distinction between exact constraints with standard deviation and uncertain
constraints. The first one guarantees that concentration averages to a specific value, and distribu-
tion is close to a normal one. The second does not restrict either average or standard deviation;
instead, it ensures that the calculated average value is not too far from the desired value. The gen-
eral formula might be confusing because it uses σ m as exact constraints on the standard deviation
and σ k as a way to quantify the penalty for discrepancy for inexact constraints. It is important to
understand the difference between them.
The derivation of this equation and much more can be found in the works of Hudebine and Ver-
straete’s (2011). This paper also has expressions for fi,j for different analytical methods.
Let us consider our toy example, and see how this method deals with it. We can find the optimal
value of the Lagrange multiplier for different values of average carbon content (Figure 5.1).
10
5
λ
–5
0.845 0.846 0.847 0.848 0.849 0.850

α
Figure 5.1 Optimal values of Lagrange multiplier for different values of average carbon content.
For comparison, Figure 5.2 shows the distribution for different values of the Lagrange multiplier.
It is clear that for λ > 0; the distribution is skewed toward lighter paraffins; and the bigger the value,
the stronger the effect. For λ = 0; the distribution become uniform and for negative values, it skewed
in the other direction.
The quotient of optimal normalized residue and standard deviation for uncertain constraints works
very similar to Lagrange multipliers for exact constraints. Figure 5.3 shows it for different values of
carbon content and standard deviation. It is easy to see the vertical line that corresponds to the fixed
point (uniform distribution), meaning that increasing standard deviation does not affect it. For values
0,6
0,5
0,4
λ=5
C
0,3
λ = 2.5
λ=0
0,2 λ = –4
0,1
0
10 11 12 13 14 15 16 17 18 19 20
Length
Figure 5.2 Concentration distributions under exact constraints corresponding to λ = 5, 2.5, 0, and −4.
2
σ
0.845 0.846 0.847 0.848 0.849 0.850

α
Figure 5.3 Contour of optimal value of quotient of normalized residue for different carbon contents and
standard deviations.
0,8
0,7
0,6
0,5
σ = 0.01
C
0,4
σ = 0.34
0,3 σ = 1.8
0,2
0,1
0
10 11 12 13 14 15 16 17 18 19 20
Length
Figure 5.4 Distributions for the same value of carbon content (0.8447) and different standard deviations.
closer to edges, it is clear that high standard deviation cases correspond to low standard deviation
cases near the uniform distribution. It just confirms the expected behavior of the system: increasing
standard deviation results in a more uniform distribution (Figure 5.4).
Entropy maximization is a powerful technique, but it should be noted that the simplified semi-
analytical solution does not exist in case of nonlinear constraints, that is, the same procedure will
not exclude molar fraction dependences.
Fortunately, various properties can be encoded exactly or approximately (e.g. density/specific
gravity is known for being nonlinear) in linear constraints framework. Although some important
properties cannot be effectively approximated with linear constraints, Octan number is a notable
example.
Entropy maximization method allows one to find the concentration of a predetermined set of
molecules. Each molecule, in this case, can be considered an explicit assumption of the mixture
composition. Because even qualitative composition is often unavailable, it is hard to apply this
method for heavy petroleum cuts.
That is why the method is often paired with other methods that can produce a detailed
composition.
The other issue with this method is that it usually produces exponential distribution (unless var-
iance is also constrained), which can result in the prevalence of few compounds, making the por-
tion of other compounds negligible, which is usually not the case. To ameliorate this problem, it is
usually expected that molecules used to reconstruct the mixture are generally very similar to the
mixture (Hudebine and Verstraete 2011).
5.5 Stochastic Reconstruction (SR)
One method that can provide us with molecules is the stochastic reconstruction method. It was first
introduced by Neurock et al. (1994).
The first insight into the stochastic reconstruction is the realization that we do know something
about the sample. It might not seem much, but we know it is a hydrocarbon mixture. We generally
have an idea where this mixture came from, what we can expect to see, and what is unlikely to
occupy a sizable portion of components. It is possible to greatly reduce complexity by choosing few
molecular families that (to our mind) represent the bulk of the mixture/very important part of the
mixture for the model we build.
For example, we usually can expect paraffins, iso-paraffins, naphthenes, aromatics, and some
heteroatomic compounds in heavy petroleum cuts. If we know that mixture came from hydrotreat-
ment, it might be reasonable to neglect concentrations of heteroatomic compounds. Again, think-
ing of the reconstructed composition as a very good model mixture helps to understand why other
parts are ignored.
Each molecule family can be characterized with several numerical values. They are often
referred to as “structure attributes.” In a way, structure attributes are very similar to basis vectors
that span vector space (chemical compounds). For example, all normal paraffins can be uniquely
described by the length of their carbon chain. To represent monosubstituted iso-paraffins, one
can use the length of the main chain, length of the branch, and the position of the main chain
to which the branch is attached to. It is also possible to use, for example, the total number of car-
bon atoms in the molecule instead of the length of the main carbon chain/length of the branch.
That is to say that the representation is not unique. It is worth mentioning that, at this stage, it is
possible to “lump” together different molecules by ignoring certain structure attributes. For
example, the location of the branch might not be very important for the model we build and,
in this case, we can ignore it, thus making characterization of iso-paraffin much easier. This part
of the method is very similar to the structure-oriented lumping (SOL) approach that we are to
discuss later.
It is possible to narrow down the possible molecules in the mixture even further by restricting the
possible numerical values of the chosen structure attributes. For example, it would be rather unrea-
sonable to assume that methane can occupy a sizable portion of a vacuum gas oil sample.
The second idea on the way to discover the stochastic reconstruction method is the futility of any
attempts to describe the composition in an actually detailed way due to the sheer number of mole-
cules that can be found in the mixture with a reasonable number of parameters. For the model to be
usable, it must be relatively simple.
Even after greatly reducing the number of possible structures, it is hard to describe their abun-
dance in the mixture accurately without turning their concentration into variables. It is the same
problem we faced before: The number of unknowns is far greater than the number of equations. In
this approach, however, instead of maximizing the entropy to find the appropriate distribution, a
different strategy is used: The structure attributes are assumed to follow a certain type of statistical
distribution. Moreover, they are assumed to be independent.
In reality, of course, the distributions of structural attributes in different families are not inde-
pendent. There is no reason to expect that, for example, the length of an aliphatic chain attached
to an aromatic core in general does not depend on the number of aromatic rings in it. However, with
this assumption, the problem becomes much more approachable.
The same goes for using specific types of statistical distributions. The distributions of different
structural attributes are often “smooth” in petroleum cuts (Wu and Zhang 2010), but there is no
rule to guarantee it in general. That is the reason why the distribution for this task is often chosen
to be gamma or discrete. Gamma distribution is known for being very flexible (exponential and χ 2
distributions are particular cases of it!) with a relatively small number of parameters (only 2).
The final touch to reach the classical stochastic reconstruction is to generate molecules stochas-
tically with the Monte Carlo approach. The number of molecules generated is often chosen to be
5000–50,000 to balance computational complexity/accuracy (Lopez Abelairas et al. 2016) (the com-
putational complexity increases linearly with the number of molecules, but the standard deviation
of the error deceases as a square root). Concentration in this case is just assumed proportional to the
number of times the molecule appeared in the mixture.
After all these simplifications, the unknowns in the model are no longer concentrations, but the
distribution parameters. Finding them is hindered by the stochastic nature of the problem. Genetic
algorithms and simulated annealing are the most common optimization techniques for this task, as
they do not rely on derivatives. Lopez Abelairas et al.’s study (2016) cites other source (which we
could not find) that claims that genetic algorithms are better suited to this task because they are less
dependent on the initial guess compared with simulated annealing. Some works use particle swarm
optimization approaches for this problem.
To wrap it up, to reconstruct a mixture with SR, the first thing is to choose structure attributes and
corresponding statistical distributions. Generate a mixture with initial (usually random) para-
meters and evaluate mixture properties. It allows one to use an optimization method to minimize
the discrepancy between properties of the real and reconstructed mixtures. After optimization, the
parameters can be found to generate a mixture that resembles the real one.
One important question was overlooked in the previous discussion: properties estimation and
mixing rules. This approach works well with group contribution methods (e.g. Constantinou
and Gani 1994; Skander and Chitour 2006; Ghasemitabar and Movagharnejad 2016) and a great
review on this topic in general (Su et al. 2017) for properties estimation (mainly boiling point
and density). Unlike the entropy maximization approach, mixing rules can be nonlinear in this
method (i.e. the procedure does not spike in difficulty when nonlinear constraints are added). Dif-
ferent mixing rules and group contribution methods can greatly affect the result of reconstruction
(Deniz et al. 2018).
For demonstration purposes, we will consider our toy example. We will assume gamma distri-
bution for the length of normal paraffins as it often happens in other works (Lopez Abelairas
et al. 2016; De Oliveira et al. 2013; Alvarez-Majmutov and Chen 2017). As we have just two para-
meters, it is possible to represent the average carbon content visually as a contour for different
parameters.
Notice that there are different ways to parametrize the gamma distribution. To make it easier to
understand, we used expected value and variance. As shown in the following, the conversion from
them to shape and scale/rate parameter is easy.
α α
E X = kθ = Var X = k θ2 = 2 5 25
β β
E X 2 1 Var X
α=k= θ= = 5 26
Var X β E X
The gamma distribution is a continuous distribution, so naturally we need to round off the results
to the closest integer. Under normal circumstances, we expect that molecules are in predetermined
range and follow their distributions, so the tails outside of the range are normally just rounded to
the closest inbound value. For example, in the following plot, the probabilities for values above
20 are added to the probability of exactly 20. It will create some artifacts near the boundaries.
For the first plot (Figure 5.5), we did not use the stochastic approach to generate molecules, and
instead used all possible molecules. Their abundance was set to be the probability calculated with
cumulative function distribution to show the “ideal” case without distortion caused by the stochas-
ticity of the method. This approach might be harder to perform for a bigger system.
10
0.847034
0.847918
0.848801
0.849243
8
Variance
4
0.846150
0.845708
0.848359
0.847476
0.845266
0.846592
12 14 16 18
Expected value
Figure 5.5 The average carbon content in a mixture of normal paraffins from C10 to C20 distributed with
gamma distributions of different variance and expected values.
It is quite evident that the expected value affects the average carbon content much stronger than
the variance. We can obtain a similar graph using the Monte Carlo approach with different values of
N (number of molecules/mixture). The lines of a constant level in the contours in that case
(Figure 5.6) look fuzzy due to the Monte Carlo sampling.
If we were reconstructing a mixture of normal paraffins with a specific value of carbon content,
any point on the constant level line would fit. We can see that if the optimization is carried out with
low N, the resulting parameters could be much further from the “ideal” case.
This method allows one to represent mixtures in terms of molecules, no matter how complicated
they are. On the contrary, constraining structural attribute distributions to a certain type of statis-
tical distributions makes the system much more rigid. The other significant drawback of this
method is optimization procedure that usually takes a lot of time. Different approaches, including
usage of artificial neural networks, have been employed to solve this problem (Deniz et al. 2017).
Another important concern is connected to the trustworthiness of such compositions. It is impor-
tant to understand, that some parts of the output might be completely unconstrained by the avail-
able analytical data. As a result, some structure attribute values are just remnants of the initial guess
for distributions (Glazov et al. 2021), which can potentially affect kinetic model step.
The procedure described here is the basis for other stochastic methods. They can vary in objective
functions, optimization procedures (e.g. Dantas et al. (2023) uses particle swarm optimization strat-
egy), and molecule representation methods (e.g. Guan et al. (2022) combine stochastic method with
SOL-inspired framework or Zhao et al. (2022) use certain structure descriptors). Some works even
attempt to represent mixture with very small ( 10 to 100) number of molecules (Campbell and
Klein 1997).
The core concept, however, (viewing molecules as a combination of structure attributes which
can be described with probability density functions) is mostly unchanged in those works.
10 10
0.845162
8 8
0.845613
318
Variance
6 Variance 6
0.848
9
921
0.84
0.846064
4 4
0.848769
0.846515
0.847416
0.846966
0.847867
2 2
12 14 16 18 12 14 16 18
Expected value Expected value
10 10
8 8
Variance
Variance
6 6
4 4
2 2
12 14 16 18 12 14 16 18
Expected value Expected value
Figure 5.6 The average carbon content in a mixture of normal paraffins from C10 to C20 distributed with gamma distributions of different variance and expected
values obtained by the Monte Carlo method with N = 50,000, 5000, 500, and 50 molecules/mixture (from left to right, from top to bottom).
5.6 SR-EM 179
5.6 SR-EM
One way to solve problems of two previous methods is by combining them together. Stochastic
reconstruction method can be used as a basis for molecular library and entropy maximization
for fine-tuning the component concentrations. This idea was purposed in (Hudebine and Verstraete
2004).
This combination overcomes problems of individual methods. The entropy maximization
approach requires an appropriate initial guess for composition, which can be provided by stochastic
reconstruction. On the contrary, stochastic reconstruction is computationally demanding, unlike
entropy maximization. The entropy maximization can be used to reconstruct similar samples with
relative ease. The downside of this combination is that entropy maximization operates only with
linear constraints (or, to be precise, the task became much more complicated when nonlinear con-
straints are imposed).
Let us test this method with our toy example. It starts the exact same way as stochastic recon-
struction. Let us say, at some point of optimization, the following distribution was obtained
(Figure 5.7).
This distribution corresponds to α = 0.8470. Now, imagine that the actual value we need is
0.8480. We can use REM to alter the distribution (Figure 5.8). The resulting distribution does
not follow gamma distribution exactly, but it is still recognizable. For target values even further,
the initial distribution might be completely overridden (Figure 5.9). It is the result of the men-
tioned phenomenon when exponential nature of the exact constraint entropy maximization
can turn a mixture into few components. To this end, it might be best to restrict standard devi-
ation (Figure 5.10).
Notice that in this case, entropy maximization process can be simplified even further due to rep-
etition in the molecules. The exponents for all identical molecules are exactly the same, so obvi-
ously, the summation of exponents for all components is equivalent to the summation of
exponents for unique components multiplied by the number of time that molecule appeared in
the mixture. Moreover, dividing by the same factor, all exponents do not affect the position of
0.15
0.10
C
0.05
0.00
10 11 12 13 14 15 16 17 18 19 20
Length
Figure 5.7 A concentration distribution after SR (N = 50,000), α = 0.8470.

0.100
0.075
C
0.050
0.025
0.000
10 11 12 13 14 15 16 17 18 19 20
Length
Figure 5.8 The concentration profile after entropy maximization to make α = 0.8480.
0.15
0.10
C
0.05
0.00
10 11 12 13 14 15 16 17 18 19 20
Length
Figure 5.9 The concentration profile after entropy maximization to make α = 0.8490.
the extremum, so the number of times a certain molecule appeared in the mixture (ni) during SR is
just N times bigger than the concentration (ci) assigned by SR.
N N∗ N∗
exp −λf i = ni exp −λf i ci exp −λf i 5 27
i=1 i=1 i=1
ci exp −λf i
ci = 5 28
cj exp −λf j
5.7 Structure-Oriented Lumping (SOL) Method 181
0.15
0.10
C
0.05
0.00
10 11 12 13 14 15 16 17 18 19 20
Length
Figure 5.10 The concentration profile after entropy maximization to make α = 0.8490 with σ 2 = 0.15.
where c is the concentration of a molecule after entropy maximization, and c is the concentration
before entropy maximization.
5.7 Structure-Oriented Lumping (SOL) Method
A different approach to molecular reconstruction of hydrocarbon mixtures was developed by

Quann et al. (Quann 1998; Quann and Jaffe 1992). The key idea is to lump molecules together
on the molecular level. It is very similar to the way molecules in SR could be lumped together.
In this framework, molecules are represented with vectors of structural increments. In the original
paper, 22 structural increments were proposed (Table 5.1). Later (Jaffe 2005) an extended list of
increments was created to describe metal-organic compounds in vacuum residua. This framework
can be tailored to other systems too. Tian et al.’s study (2010) used only 7 increments to describe the
feedstock, whereas Hou et al.’ (2023) used 38 increments.
The basic structure increments are A6, A4, A2 – aromatic rings; N6, N5, N4, N3, N2, and
N1 – naphthenic rings; R – total alkyl-group length; IH – incremental hydrogen to specify the
Table 5.1 Structural increments and increment stoichiometry.
A6 A4 A2 N6 N5 N4 N3 N2 N1 R br me H A_A S RS AN NN RN O RO O=
C 6 4 2 6 5 4 3 2 1 1 0 0 0 0 −1 0 −1 −1 0 −1 0 0
H 6 2 0 12 10 6 4 2 0 2 0 0 2 −2 −2 0 −1 −1 1 −2 0 −2
S 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0
N 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0
O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1
degree of unsaturation; br – number of branches; me – number of methyl groups in R; AA – biphenyl

bridge; NS, NN, and NO – sulfur, nitrogen, and oxygen in naphthenic rings; RS, RN, and RO – mer-
captan, amin, and alcohol group, respectively; AN – nitrogen in aromatic ring; and KO – ketone or
aldehyde.
The way of determining abundance of different structures in the mixture is not clearly described
in the original papers, except “non-negative Lagrange multiplier method, starting with an equimo-
lar composition distribution” (Jaffe 2005). Ren et al. (2019) speculated that the objective function in
their work might be negative entropy, sum of squares, or other criteria.
Where SOL method shines the most is the computing of reaction networks. It is possible to
describe many important reactions of oil processing (cracking, hydration/dehydration, etc.) in this
framework with a set of rules, which is particularly suitable for computers. This approach allows
generating huge reaction networks for complex systems. It was shown in the original paper.
5.8 State Space Representation Method
Another approach to this problem, which takes a bit different route to represent mixtures, is the
state space representation method (Mei et al. 2017). The basic idea is to represent feedstock as a
linear combination of well-known mixtures. The output for kinetic models for the sample is
assumed to be the linear combination of known mixtures. This method is not very useful for heavy
petroleum cuts as for now, but it might find its use in the future.
Let us start from an n-dimensional vector space with pure components being its orthonormal
basis. In that case, they can be represented as follows:
T
Component 1 e1 = 1 0 … 0
T
Component 2 e2 = 0 1 … 0 5 29
…
T
Component n en = 0 0 … 1
Given these vectors, any mixture composed of these pure components can be represented as a
column-vector as well.
T
ν= w1 w2 … wn 5 30
with wi being the concentration of ith component. Another (more general) way of looking is to
think of wi as being the projection of the mixture vector on the ith basis vector.
Using L1 norm, any mixture can be represented as a normalized vector:
n
ν = wi 5 31
i=1
ν
ν= 5 32
ν
Pure component basis is not the only viable option. It is also possible to represent mixtures in the
basis of other (independent) mixtures ν1, ν2, …, νn.
5.9 Molecular Type-Homologous Series Matrix 183
n n
νj νj νj n
ν j=1 j=1
ν= = = = hj ν j 5 33
ν n n
νj νj j=1
j=1 j=1
where hj is the blending ratio of the jth mixture νj to the final mixture.
Now, if we consider N well-defined mixtures (=vectors), we can rewrite Equation (5.33) in a
matrix form as follows:
V =W H 5 34
where V (n × N) is the sample matrix, W (n × n) is the basis fraction matrix, and H (n × N) is the
blending matrix.
The simplification comes with the realization that we can approximate V with m (less than n)
basis mixtures in a subspace without losing too much information.
The task of finding mixtures representation turns into a matrix decomposition. However, the
values of blending ratio and concentration must be positive and therefore “standard” decomposi-
tion strategies from linear algebra, such as singular value decomposition, do not work. Fortunately,
nonnegative matrix factorization can solve this problem.
This approach was used for the naphtha reconstruction. 59 samples from a plant were collected
and represented in terms of 35 lumps. Authors obtained 21 basis fractions through nonnegative
factorization. That way, they managed to predict the product yield of pyrolysis of said fractions.
Generally, the feedstock that plants use may vary, but it is not entirely random. This method
might be useful to find a (relatively small) set of basic fractions that can span the feedstock subspace
with certain accuracy.
5.9 Molecular Type-Homologous Series Matrix
Another framework to model the composition is known as molecular type-homologous series

matrix method (MTHS). In this case, the mixture is represented with a matrix. Columns of the
matrix correspond to different homologous series, and rows are used for different carbon numbers.
The entry of the matrix denotes the abundance of a lump from the given homologous series and
carbon number. A modified approach that works with boiling range instead of carbon number
(Gomez-Prado et al. 2008) also exists and arguably better suited for heavy petroleum cuts.
From the earlier construction, it is clear that each entry in the matrix could be pure components
and pseudo-components alike. Choosing the properties for lumps in this case might be especially
challenging. Wu and Zhang (2010) used an interesting approach to this problem: First, they enu-
merate all isomers which fall into the same entry, and then they estimated their properties with
group contribution methods. With this information, they found isomer distribution based on ther-
modynamic equilibrium assumption, and this equilibrium blend is considered the lump.
As for the methods of determining the abundance of different molecules, there are three main
approaches (to the best of our knowledge).
One method assumes that lump distribution within each homologous series can be approximated
with a statistical distribution. The distribution of choice is usually gamma distribution. This method
is very similar to what we did in stochastic reconstruction to demonstrate an “ideal” case. Except,
there are more than one homologous series. Gamma distribution parameters are found by
minimizing deviation experimental data and calculated. In this work, the squared relative error was
used (Liu et al. 2019).
Second method treats the sample as a blend of well-characterized mixtures. In Aye and Zhang’s
(2005), they represent mixtures with at least eight other mixtures. The blending proportions are
found by minimizing the sum of square of relative errors.
The third method relies on empirical correlations among carbon number and molecular weight,
boiling point, and density. This method is hard to use for heavy petroleum cuts.
5.10 Conclusion
Molecular reconstruction is an umbrella term used for different methods, which allows one to rep-
resent hydrocarbon mixtures based on limited information. These methods do not normally find
the real composition, but they can be useful for kinetic models. In that regard, molecular recon-
struction builds a very good model mixture.
There are two main strategies for choosing components that constitute the mixture: deterministic
and stochastic. Generally speaking, stochastic strategies require more computations, but the diffi-
culty of using them grows slower than for deterministic one. That is part of the reason why the
stochastic approach (at least to the best of authors’ knowledge) is used more often for heavy
fractions.
As for determining abundance of component in the mixture, it is usually an optimization prob-
lem. The properties of individual constituents are found first (though experiments or group contri-
bution methods). The objective function controls how close the blend of component is to the
sample. The concentrations are rarely used as the variables in the objective function; instead, an
assumption is often made about the connection between them (e.g. concentrations follow a certain
distribution, the whole mixture is a linear combination of other mixtures, the distribution maxi-
mize entropy, etc.).
Acknowledgment
This work was supported by the Ministry of Science and Higher Education of the Russian Feder-
ation within the state assignment for the Boreskov Institute of Catalysis (project FWUR-2024-0037).
References
Alvarez-Majmutov, A. and Chen, J. (2017). Stochastic modeling and simulation approach for industrial
fixed-bed hydrocrackers. Industrial and Engineering Chemistry Research 56: 6926–6938.
Aye, M.M.S. and Zhang, N. (2005). A novel methodology in transforming bulk properties of refining
streams into molecular information. Chemical Engineering Science 60: 6702–6717.
Campbell, D.M. and Klein, M.T. (1997). Construction of a molecular representation of a complex
feedstock by Monte Carlo and quadrature methods. Applied Catalysis A: General 160: 41–54.
Constantinou, L. and Gani, R. (1994). New group contribution method for estimating properties of pure
compounds. AIChE Journal 40: 1697–1710.
References 185
Dantas, T.S.S., Noriler, D., and Huziwara, K.W. (2023). A multi-population particle swarm optimization
algorithm with adaptive patterns of movement for the stochastic reconstruction of petroleum fractions.
Computers & Chemical Engineering 174: 108221.
De Oliveira, L.P., Trujillo Vazquez, A., Verstraete, J.J., and Kolb, M. (2013). Molecular reconstruction of
petroleum fractions: application to various vacuum residues from different origins. Energy & Fuels 27:
3622–3641.
Deniz, C.U., Yasar, M., and Klein, M.T. (2017). Stochastic reconstruction of complex heavy oil molecules
using an artificial neural network. Energy and Fuels 31 (11): 11932–11938.
Deniz, C.U., Yasar, S.H.O., Yasar, M., and Klein, M.T. (2018). Effect of boiling point and density
prediction methods on stochastic reconstruction. Energy and Fuels 32 (3): 3344–3355.
Ghasemitabar, H. and Movagharnejad, K. (2016). Estimation of the normal boiling point of organic
compounds via a new group contribution method. Fluid Phase Equilibria 411: 13–23.
Glazov, N., Dik, P., and Zagoruiko, A. (2021). Effect of experimental data accuracy on stochastic
reconstruction of complex hydrocarbon mixture. Catalysis Today 378: 202.
Gomez-Prado, J., Zhang, N., and Theodoropoulos, C. (2008). Characterisation of heavy petroleum
fractions using modified molecular-type homologous series (MTHS) representation. Energy 33:
974–987.
Guan, Y.M., Guan, D., Zhang, C. et al. (2022). Diesel molecular composition and blending modeling
based on SU-BEM framework. Petroleum Science 19: 839–847.
Hou, L., Ye, L., Qin, X. et al. (2023). Predicting the physicochemical properties of molecules in petroleum
based on structural increments. Industrial & Engineering Chemistry Research 62: 7744–7756.
Hudebine, D. and Verstraete, J.J. (2004). Molecular reconstruction of LCO gasoils from overall petroleum
analyses. Chemical Engineering Science 59: 22–23.
Hudebine, D. and Verstraete, J.J. (2011). Reconstruction of petroleum feedstocks by entropy
maximization. application to FCC gasolines. Oil and Gas Science and Technology 66: 437–460.
Jaffe, S.B. (2005). Extension of structure oriented lumping to vacuum residual. AIChE Annual Meeting,
Conference Proceedings 08066: 9578.
Lisman, J.H.C. and Zuylen, M.C.A. (1972). Note on the generation of most probable frequency
distributions. Statistica Neerlandica 26: 19–23.
Liu, L., Hou, S., and Zhang, N. (2019). Incorporating numerical molecular characterization into pseudo-
component representation of light to middle petroleum distillates. Chemical Engineering Science: X 3:
100029.
Lopez Abelairas, M., De Oliveira, L.P., and Verstraete, J.J. (2016). Application of Monte Carlo techniques
to LCO gas oil hydrotreating: molecular reconstruction and kinetic modelling. Catalysis Today 271:
188–198.
Mei, H., Cheng, H., Wang, Z., and Li, J. (2017). Molecular characterization of petroleum fractions using
state space representation and its application for predicting naphtha pyrolysis product distributions.
Chemical Engineering Science 164: 81–89.
Neurock, M., Nigam, A., Trauth, D., and Klein, M.T. (1994). Molecular representation of complex
hydrocarbon feedstocks through efficient characterization and stochastic algorithms. Chemical
Quann, R.J. (1998). Modeling the chemistry of complex petroleum mixtures. Environmental Health
Perspectives 106: 1441–1448.
Quann, R.J. and Jaffe, S.B. (1992). Structure-oriented lumping: describing the chemistry of complex
hydrocarbon mixtures. Industrial and Engineering Chemistry Research 31: 2483–2497.
Ren, Y., Liao, Z., Sun, J. et al. (2019). Molecular reconstruction: recent progress toward composition
modeling of petroleum fractions. Chemical Engineering Journal 357: 761–775.
Skander, N. and Chitour, C.E. (2006). A new group-contribution method for the estimation of physical
properties of hydrocarbons. Oil & Gas Science and Technology 57: 369–376.
Su, W., Zhao, L., and Deng, S. (2017). Group contribution methods in thermodynamic cycles: Physical
properties estimation of pure working fluids. Renewable and Sustainable Energy Reviews 79: 984–1001.
Tian, L., Wang, J., Shen, B., and Liu, J. (2010). Building a kinetic model for steam cracking by the method
of structure-oriented lumping. Energy and Fuels 24: 4380–4386.
Wu, Y. and Zhang, N. (2010). Molecular characterization of gasoline and diesel streams. Industrial and
Zhao, G., Yang, M., Du, W. et al. (2022). A stochastic reconstruction strategy based on a stratified library
of structural descriptors and its application in the molecular reconstruction of naphtha. Chinese
Journal of Chemical Engineering 51: 153–167.
187
Modeling of Catalytic Hydrotreating Reactor for Production

of Green Diesel
Alexis Tirado1, Fernando Trejo2, and Jorge Ancheyta1,3
1
Department of Petroleum Engineering, Kazan Federal University, Kazan, Russia
2
Instituto Politécnico Nacional, CICATA–Legaria, Ciudad de México, Mexico
3
Instituto Mexicano del Petróleo, Ciudad de México, Mexico
6.1 Introduction
One of the most crucial worldwide research projects in the energy field is developing alternative
fuels to reduce the use of fossil fuels because of the high levels of harmful gases released into
the atmosphere. Biofuels are considered a promising alternative because they can be produced
from a variety of biomass sources (vegetable oils, animal fats, and even agricultural wastes) and
are not harmful to the environment (Kubičková and Kubička 2010; Sotelo-Boyas et al. 2012). These
feedstocks are subjected to hydrotreating processes in order to produce high-quality green fuels.
Nevertheless, this technology type still presents some challenges, such as high exothermicity,
hydrogen consumption, and yields of undesirable by-products (Furimsky 2013; Sivasamy et al.
2009). The use of mathematical models is essential for the design, scaling, and optimization of these
processes focused on the conversion of vegetable oils to market-driven transportation renewa-
ble fuels.
This chapter presents a study of the kinetic and reactor modeling of hydrotreating vegetable oils
for the production of renewable fuels. The different approaches considered in the literature and
their results are discussed. Finally, a case study of dynamic reactor modeling is presented in order
to provide some insights and considerations on its development, solution, and scaling-up, as well as
the interpretation of the results obtained.
6.2 Conversion of Vegetable Oils into Renewable Fuels
Vegetable oils or lignin-derived bio-oils contain compounds such as fatty acid (FA) esters, trigly-
cerides (Tg), phenols, carboxylic acids, and ketones, which are refined to obtain oils rich in Tg capa-
ble of being used in the energy industry. Tg present in refined vegetable oils consist of saturated and
unsaturated FAs with carbon atom numbers from C4 to C24 and more commonly between C16 and
C18 (Snåre et al. 2006, 2007), which turn them into adequate substitutes for diesel-type fuels. Nev-
ertheless, the high oxygen content in refined vegetable oils (10–12 wt.%) increases thermal and
188 6 Modeling of Catalytic Hydrotreating Reactor for Production of Green Diesel
chemical instability, high viscosity, corrosiveness, immiscibility with fossil fuels, low heating value,
and tendency to polymerize. Therefore, removal of oxygen atoms is necessary to obtain high-quality
fuels (Noriega et al. 2020).
Catalytic hydrotreating (HDT) has been used for years in the oil industry to increase the hydro-
gen-to-carbon (H/C) ratio as well as remove sulfur, nitrogen, and metals, among other heteroatoms
in diverse petroleum refining streams (atmospheric and vacuum residues, gas oils, and naphtha)
through a series of chemical reactions carried out on catalysts under a hydrogen-rich atmosphere
(Ancheyta et al. 2010). This process may involve multiple catalytic reaction systems, including
fixed-bed, ebullated-bed, and slurry-bed options, and it is crucial due to the environmental
restrictions that apply during the combustion of fossil fuels. Such regulations include the emission
of sulfur oxides (SOx), nitrogen oxides (NOx), particulate matter (PM), and volatile organic
compounds (VOCs) like aromatics and olefins. Therefore, this process is appropriate to remove
the oxygen atoms in vegetable oils through hydrodeoxygenation (HDO) and decarboxylation/
decarbonylation (deCO2/deCO) reactions, obtaining high-quality fuels rich in aliphatic chains
called “renewable fuels” or “green fuels.” The performance of the hydrotreating process is mainly
influenced by the temperature, hydrogen/oil ratio, pressure, catalyst, and reaction time. However,
the elimination of oxygen leads to the production of water during HDO reactions, which may
impact the catalyst activity, whereby bio-oil conversion into fuels is a processing challenge
(Furimsky 2013; Saidi et al. 2014).
Kalnes et al. (2008) summarized some advantages and disadvantages of ultra-low sulfur diesel
(ULSD) from petroleum, biodiesel, renewable diesel, and synthetic diesel produced by the
Fischer–Tropsch synthesis. Table 6.1 shows that ULSD from petroleum has a considerable number
of heteroatoms and aromatic compounds enhancing the formation of greenhouse gases. On the
other hand, synthetic diesel and biodiesel are abundant in oxygenated compounds, displaying
the undesirable properties previously mentioned. In contrast, renewable diesel derived from
vegetable oils is rich in iso- and n-alkanes and free of aromatic and oxygenated compounds. It is
notable that renewable diesel, synthetic diesel, and biodiesel have zero sulfur content and emit
fewer greenhouse gases than petroleum diesel. Although the boiling point properties are
comparable, diesel has a higher cetane number. It is concluded from these data that the properties
of renewable diesel are better than those obtained by other processes. In addition, the investment
cost of the process to produce renewable diesel can be reduced by using infrastructure utilized in oil
refineries, such as the hydrotreating units (Bezergianni and Dimitriadis 2013).
Table 6.1 Comparison of diesel properties obtained from different processes.
Property Diesel (ULSD) Biodiesel Renewable diesel Synthetic diesel (Fischer–Tropsch)
Oxygen (wt.%) 0 11 0 0
Specific gravity 0.84 0.88 0.78 0.77
Sulfur content (ppm) <10 <1 <1 <1
Heating value (MJ/kg) 43 38 44 44
Cloud point ( C) –5 –5 to 15 –20 to 20 –
Cetane index 51–55 50–55 70–90 >75
H/C 0.162 0.176 0.176 —
Polyaromatics (wt.%) <8 0 0 —
Stability Standard Good Excellent Good
6.2 Conversion of Vegetable Oils into Renewable Fuels 189
6.2.1 Commercial Production of Renewable Diesel

In recent decades, there has been a rapid increase in global energy demand, prompting researchers
to address this challenge and decrease greenhouse gas emissions. Currently, catalytic hydrotreating
technologies are utilized by industrial refineries to produce renewable fuels, including diesel,
kerosene, naphtha, and liquefied petroleum gas (LPG), using various raw materials. Neste
Corporation developed the NExBTL technology, through which it has reported biofuel production
volumes of 2 million tons per year (Douvartzides et al. 2019; Kubička and Tukač 2013).
Nonetheless, the production of renewable fuels is still limited. NExBTL process is a technology
compatible with current refinery equipment for producing renewable diesel and jet fuel from
feedstocks such as used cooking oil, palm oil, and waste animal fat, resulting in a 40–80% lifecycle
reduction in CO2, depending on the feedstock type (Baldiraghi et al. 2009; Pelemo et al. 2020). On
the other hand, the Ecofining process is a technology developed by Honeywell UOP and Eni SpA
companies. The design of the process usually varies in the type of catalyst, feedstocks, and operating
conditions, although there are similarities in the process configuration, as shown in Figure 6.1a.
Through these, high-quality fuels suitable for regular engines are obtained, meeting environmental
specifications either directly or mixed with petroleum distillates (Baldiraghi et al. 2009; Kalnes et al.
2008). In addition, companies such as Haldor Topsøe, ConocoPhillips, and Petrobras have
developed hydrotreating technologies by co-processing renewable feedstocks and petroleum
distillates, as shown in Figure 6.1b (Zhang et al. 2018). Therefore, the process is focused on reducing
the heteroatom content while dropping the carbon footprint. However, hydroprocessing of this
feedstock type is more exothermic than petroleum-derived streams, causing a high heat release
and hydrogen consumption. In industrial practice, the heat loss in these units is negligible
compared with the heat released by exothermic reactions. As a result, the behavior of these units
is remarkably close to an adiabatic operation. The problem for industrial units is not achieving
(a)
Cleanup Deoxygenation Hydroisomerization Separation
Pretreatment HDO/deCOx
Light Gases
CO2
Naptha - Green Gasoline
Feedstocks Jet Fuel
Vegetable Oils Green Diesel
Animal Fats
Hydrogen Water
(b)
Gas products
normal saturated hydrocarbons C1-C4
Green Propane
Vacuum Gas Oil 80-95% Remaining Hydrogen
Hydrogen
Green Diesel (CN 60-90)

normal and isomer saturated hydrocarbons C5-C30
Cycloparaffins, aromatics
Vegetable Oils 5-20% Water
Figure 6.1 Schematic representation of (a) standalone renewable feedstock hydroprocessing plant and
(b) co-processing hydroprocessing plant for renewable feedstock and petroleum distillates (Douvartzides et al.
2019/ MDPI /CC BY). ∗Other hydrocarbons include products from hydrocracking, polymerization, and
cyclization.
complete feedstock conversion but rather controlling the high levels of reaction heat released. An
increase in reactor temperature beyond permissible limits enhances the generation of hot spots
that affect the catalyst activity, change the reaction product yields, and damage the reactor
infrastructure. Additionally, keeping the temperature constant in a large-scale trickle-bed reactor
is complicated since the removal or addition of heat to the reactor walls is considered economically
unfeasible.
Haldor Topsøe revamped a mild hydrocracking unit to carry out the co-processing of light gas oil
with 30 vol.% of raw tall diesel (RTD). This unit has a capacity of approximately 10,000 b/d and
reports that the reaction heat increased with the increasing concentration of RTD.
In hydrotreating units for petroleum-derived streams, a proper quench system effectively
manages the high heat released. The catalytic bed volume is divided into small beds separated
by quench zones to mix the hot feedstock streams from the previous catalytic bed with cold streams.
This ensures a controlled mixture before it passes to the next bed (Egeberg et al. 2009, 2011). Based
on the rate and amount of reaction heat released, the catalytic bed must be configured with multiple
catalytic beds. The necessary number of quench zones is adjusted not to exceed a permissible limit
temperature because of mechanical restrictions and excessive feedstock hydrocracking.
The design and operation of these processes present notable challenges as they rely on vast
experimental data and prior comprehensive knowledge of the design variables acquired in labora-
tory-scale hydrotreating reactors. In addition, modeling an industrial reactor is a practical tool for
ensuring proper start-up and stable operation throughout the process. However, these types of
reactor models are scarce because they require the incorporation of aspects such as deactivation
of catalysts, hydrogen consumption, pressure drop, and reaction heat released to describe the
commercial unit performance (Donnis et al. 2009; Scharff et al. 2013).
6.3 Hydrotreating Kinetic Models and Reaction Pathways
Elucidation of reaction schemes and determination of kinetic parameters are important steps for
catalyst screening, as well as designing and scaling-up of processes. Due to the complexity of the
hydrotreating process of vegetable oils, which involves a large number of chemical compounds, and
chemical and physical phenomena occurring within the reactor, mathematical models are widely
utilized in predicting the reactor performance, determining optimal processing conditions, and
carrying out the design and scale-up of the process for industrial application. Nevertheless, kinetic
and reactor models related to vegetable oils hydrotreating are still scarce in the literature, where
most of the reports are focused on experiments involving model compounds (Hickman et al.
2004; Tirado et al. 2018). Table 6.2 summarizes the operating conditions, feedstock, reactor, and
catalyst types reported in kinetic studies focused on producing green fuels by the hydrotreating
process.
6.3.1 Model Compounds

The development of kinetic models focused on the hydrotreating of vegetable oil is based on accurate
and reliable experimental data obtained during studies of reaction mechanisms and catalyst
performance. Model compounds such as FAs and Tg serve as representative samples of complex
systems such as vegetable oils (Boda et al. 2010; Chen et al. 2013; Kubička et al. 2009; Kubička
and Kaluža 2010; Snåre et al. 2006). Thus, kinetic models consider possible reaction pathways
and their rates by determination of kinetic parameters under certain operating conditions
(Furimsky 2013; Hermida et al. 2015; Mortensen et al. 2011). Figure 6.2 depicts diverse reaction
Table 6.2 Kinetic studies performed for the production of renewable fuels through the hydrotreating process.
Reactor Temperature Stirring Conversion

Feedstock type ( C) Pressure (bar) Time (h) rate (rpm) Catalyst/solvent (%) Reference
(R)
Ethyl stearate Semi-batch 270–360 17–40 5Pd/C /n-dodecane 41–100 Snåre et al. (2007)
25 ml Ar/min
(5% H2)
Stearic acid Semi-batch 260–290 13–14.5 6 Ni/γ–Al2O3 (O)/n-dodecane 50–95 Kumar et al. (2014)
25 ml Ar/min
(5% H2)
Oleic acid Semi-batch 320–380 20 (10% H2) 1 2000 FMoOx/Zeol(R) 40–100 Ayodele et al. (2015)
Waste cooking oil Batch 300–375 90 8 1000 Unsupported CoMoS(S) 91–99 Zhang et al. (2014)
Methyl heptanoate Batch 250–330 60–100 6 700 Rh/ZrO2(R)/n-hexadecane 15–90 Bie et al. (2015)
Methyl palmitic Batch 250–330 60–100 1 700 Rh/ZrO2(R)/n-hexadecane Bie et al. (2016)
Tripalmitin– Batch 280–360 30 6 1200 15NiAl(R)/n-dodecane 40–100 Yenumala et al. (2017)
tristearin
(R) (R)
Stearic acid TOFA Semi-batch 300 7–30 6 1200 5Ni/γ–Al2O3 , 5Ni/Y–H–80 , 99 Jeništová et al. (2017)
5Ni/SiO2(R), Pd/C(R)
Methyl palmitate Trickle bed 290 30 0.083–3.75 min — Ni2P/SiO2 34–82 Shamanaev et al. (2019)
Stearic acid Batch 275–325 40–70 10–180 min 500–1000 NiMo/Al2O3 Arora et al. (2019)
Rapeseed oil Fixed bed 300–380 10 2.7–9.8 h–1 (LHSV) — NiCu/CeO2–ZrO2(R) 10–100 Selishcheva et al. (2014)
Solvolytic oil Batch 300 80 1 1000 NiMo/γ–Al2O3(O–R–S), Pd/C (R), MoS2(S), 43–94 Grilc et al. (2014b)
NiMo/Al2O3–SiO2(R), Pd/γ–Al2O3(R)/Tetralin
Solvolytic oil Batch 200–350 20–80 1 250–1000 NiMo/γ–Al2O3(S)/n-dodecane 83–85 Grilc et al. (2014a)
Jatropha oil Fixed bed 300–320 80 2–12 h–1 (LHSV) — CoMo/MTS(S), H2/HC =1200–1500 l/l 65–70 Sharma et al. (2012)
Jatropha oil Fixed bed 320–360 80 1–8 (LHSV) — CoMo/Al2O3(S), H2/HC = 1500 Nl/l 89–98 Anand and Sinha (2012)
Jatropha oil Fixed bed 340–420 80 0.5–12 (WHSV) — NiW/SiO2–Al2O3(O), H2/HC = 500–2500 Nl/l 29–100 Anand et al. (2016a)
Palm oil Fixed bed 335–365 30–60 15–45 min — NiMo/γ– Al2O3(S) 73–100 Vélez Manco (2014)
Castor oil methyl Batch 395–425 50 2–4 750 — 44–86 Mederos et al. (2021)
esters
State of catalyst: O, oxide form; R, reduce form; S, sulfide form.

(a) k7 Olefinic C17 Aromatic C17 (b) k2 Octadecanal Heptadecane (c) k1
k6 Stearic acid
k5 k3
k1 k1 Octadecene Octadecane
Ethyl stereate Stearic acid Oleic acid k3
k4 Stearic acid Octadecanol
k2 k4 Pentadecane
k2 Bio fuels
k3 n-heptadecane k5 Hexadecane
k1
k2 (e) k2 (f) k6
(d) HDO
Methyl heptanoate Heptanol
k1 Stearic acid Octadecanal Heptadecene Heptadecane
Glycerides Intermediates k3 k6 Other k1 k2 k4
k7 hydrocarbons*
k4 k8 k3
k3 HDCx Octadecanol Octadecene Octadecane Heptanoic acid Hexane
k5 k5
Pentadecane Heptyl heptanoate
(g) (h) ODOL-HDOL (i)
k3 ODE-HXDE
k2 k4 k1 k3
Stearic
Methyl k3 Palmitic k1 Stearyl alcohol Heptadecane
acid
palmitate Hexadecanal SA-PA ODAL-HDAL
acid k2
OD-HXD
k3
k4 k2 Octadecane
HPD-PD
Hexadecanol
(j) (l) k3 Octadecanol Octadecane
k5
k1 k2 k3 k4 Stearic k6 k8 k9 k1
Methyl Methyl Methyl Methyl Stearyl Stearyl
Octadecane Stearic Octadecanal
linolenate linoleate oleate stearate acid aldehyde alcohol acid
k5 k7 k4
k2
Heptadecane Heptadecane
(k) (m)
k3 Methyl k4 k6 k8 k9 Stearic k2 k3
Methyl Palmitic Octadecanal Octadecanol
Palmitoleate palmitate acid Hexadecanal Hexadecanol Hexadecane acid
k1 k4
k5 k7
Heptadecane
Pentadecane Octadecane
Figure 6.2 Reaction schemes for the hydrotreating of model compounds.

schemes proposed for studying reaction mechanisms as well as the prediction of the conversion and
production of compounds.
Snåre et al. (2007) investigated the kinetics of ethyl stearate over a commercial Pd/C
catalyst. The reaction scheme, presented in Figure 6.2a, indicates the ethyl stearate (A)
conversion to form its corresponding FA (B) and paraffins (C). However, dehydrogenation
reactions can form olefins (D) and eventually aromatics by cyclization (E). The Langmuir–
Hinshelwood mechanism accurately predicted the experimental concentration profiles for both
reactants and products. The adsorption coefficients of some product gases (i.e. CO2, CO, eth-
ylene, etc.) were neglected. Additionally, the surface reactions are rate-limiting instead the
adsorption steps:
kj Ci
rj = for j = 1 − 7 61
1 + K A CA + K B CB + K C CC + K D CD + K E CE
where kj is an apparent reaction coefficient involving the reaction rate and equilibrium adsorption
coefficients (k j = k j × K i of the diverse reaction pathways j, and Ci is the concentration of reactant
species i. Based on the kinetic rate expressions, the mass balances of components are expressed by a
set of ordinary differential equations (ODEs) in the function of the catalyst bulk density (ρB) and
reaction time (t):
1 dC i
= rj 62
ρB dt
The result obtained with the model confirmed that the adsorption coefficients can be disregarded.
Therefore, Eq. (6.1) follows a first-order power-law model.
Kumar et al. (2014) performed a kinetic study for the HDO of stearic acid with a Ni/γ–Al2O3
catalyst proposing a reaction pathway (Figure 6.2b) that considers octadecanol as an intermediate
product, which undergoes dehydrogenation and decarbonylation reactions to finally produce
heptane and one mole of CO. Alternatively, octadecanol is dehydrated and form octadecene, which
is hydrogenated to octadecane. The formation of hexadecane and pentadecane is not entirely clear,
and it has been speculated that they are the end products of octadecanol. As octadecanal was not
detected in the HDO of stearic acid, it was assumed to be rapidly converted into heptadecane.
Experiments under different stirring rates (1000–2000 rpm) showed negligible changes in the
stearic acid conversion, indicating the reaction follows the intrinsic kinetics (i.e. the transfer at
the gas–solid and liquid–solid interfaces is neglected). Nevertheless, resistance due to molecule
diffusion inside the pores could not be evaluated. Therefore, kinetic parameters include the internal
diffusional resistance effect. The reaction rates were expressed by the power-law model a first
reaction order concerning the concentration of reactant species and a hydrogen excess in the
reaction system.
dC i
= kj Ci 63
dt
The kinetic results indicated a large conversion of alcohols to aldehydes. The highest activation
energy was observed for k4 and k5, indicating that the formation of pentadecane and hexadecane
predominated at higher temperatures. A similar model was used to determine the activity and
product yields using a zeolite-based fluoride-ion functionalized molybdenum-oxalate catalyst
(FMoOx/zeol) in the oleic acid (OA) HDO (Ayodele et al. 2015). The model considers the
hydrogenation of the double bond in the OA to form stearic acid, followed by the subsequent
elimination of an oxygen atom, as shown in Figure 6.2c. The production of the C18 biofuel is insig-
nificant as its formation occurs with difficulty. The reaction rate and mass balance equations were
developed with similar assumptions to determine the kinetic coefficients using the linear Arrhenius
equation to calculate the activation energies (EA, energy quantity required to perform a chemical
reaction) and pre-exponential factors (A0, total number of collisions per time unit leading to
reaction).
EA
kj = A0 e − RT 64
EA
ln k j = ln A0 − 65
RT
where R is the universal ideal gas constant. The results indicated that the hydrogenation rate of OA
is much higher than the rate of removal of oxygen atoms in stearic acid.
Zhang et al. (2014) proposed a kinetic model for hydrotreating waste cooking oil glycerides over
an unsupported CoMoS catalyst. The kinetic parameters calculated using a power-law model
showed that glycerides are rapidly transformed into FAs through desterification reactions. FAs
are continuously converted into alkanes and other products to form alcohols/aldehydes as
intermediates, which are converted via HDO or hydrodecarboxylation/hydrodecarbonylation
(HDCx) reactions (Figure 6.2d). According to the kinetic parameters obtained, the removal of
oxygen atoms took place through HDCx reactions. In addition, the kinetic model includes a
catalytic deactivation term, which is discussed in a later section.
Alcohol predominated as the main intermediate, as observed by its large reaction rate coefficient
compared with the aldehyde formation in the HDO of FAs (Figure 6.2e). Nevertheless,
polymerization and cyclization reactions occur in parallel.
Most of the kinetic models assume an excess of hydrogen relative to the stoichiometric
requirement in the reaction system. Thus, the hydrogen concentration does not change
significantly and it is held as constant in the reaction rate equations. Nonetheless, some kinetic
studies consider the hydrogen pressure effect in order to analyze its effect on the reaction system.
Bie et al. (2015) developed a reaction network during HDO of methyl heptanoate on Rh/ZrO2
catalyst. The simplified route shown in Figure 6.2f was modeled by both power-law and
mechanistic models. In the first model, six-reaction steps for reaction j, the concentration of species
i, and the hydrogen pressure at its corresponding reaction order were modeled:
m
r j = kj Ci PH2j 66
Since the reaction pathway 5 is considered as reversible, its mathematical expression is
represented by the van’t Hoff equation, considering the temperature-dependent equilibrium
constant keq:
C D C H2 O
r5 = k5 CB CC − 67
k eq
k eq = 0 338e1136 RT
68
The molar balance of each species ni in both gas and liquid phases is expressed considering the
mass transfer effects in the gas–liquid interphase (aGL) by the following batch reactor model:
dnGi
= − V R N GL,i aGL 69
dt
dnLi
= W Ri V L + V R N GL,i aGL 6 10
dt
where VR and VL represent the effective reactor and the liquid phase volumes, W indicates the mass
of the catalyst, and Ri is the reaction rate of each species based on the stoichiometry.
In another study, Bie et al. (2016) used a Rh/ZrO2 catalyst during HDO of methyl palmitate. The
oxygen vacancy sites on ZrO2 are active for oxygenated species, and the dihydrogen activation
through the dissociative adsorption carried out by the active sites on reduced state explains the
bifunctionality of noble/transition metal-based catalysts. During the reaction, palmitic acid and
methane are produced, as shown in Figure 6.2g, following a similar mechanistic model as
previously mentioned. For the reaction rate r1, hydrogen concentration was considered to better
describe the effect of the pressure on the reaction kinetics.
k1 CA K H2 CH2
r1 = 6 11
K B CB K C CC
1+ + 1+ K H2 C H2
K H2 C H2 K H2 CH2
Reaction rates r2–r3 follow the same mechanism hypothesis reported (Bie et al. 2015), which
share a common rate of determination step (the final addition of a hydrogen atom in the reactant
species):
k j K B K H2 C B C H2
rj = 6 12
K B CB K C CC
1+ + 1+ K H2 C H2
K H2 CH2 K H2 C H2
Finally, the reaction rate r4 considers the reversible dehydrogenation–decarbonation of
hexadecanol in pentadecane:
k4 K C CC
r4 = 6 13
K B CB K C CC
1+ + K H2 C H2
K H2 C H2 K H2 CH2
It was stated that the kinetic model can be adequately fitted to experimental data on a wide range
of reaction conditions and conversions, unlike other models based on power-law kinetics.
Recently, the HDO of Tg was investigated using a mixture of tripalmitin (TP) and tristearin (TS)
over Ni/Al2O3 (Yenumala et al. 2017). HDO experiments showed that TP and TS were converted
instantaneously to FAs and propane, as also observed by Zhang et al. (2014) who reported that
Tg were rapidly converted. The scheme presented in Figure 6.2h shows the kinetic model based
on the experimental results, where the deoxygenation pathway is mainly observed. Tg formed
intermediate aldehydes and alcohols, which were subsequently converted into alkanes and water
through reduction/dehydration reactions carried out in unsaturated sites of sulfhydryl groups. In
addition, the alkane production is enhanced by decarbonylation reactions because the Ni/Al2O3
catalyst is weakly acidic. The reaction rate is modeled using a first-order power-law with respect
to compounds in the liquid phase and considering the hydrogen partial pressure (PH2 ) during
hydrogenation reactions. For stearic or palmitic acid (A) as reactant, the reaction rate equation is:
dCA
− = k 1 CA PH2 6 14
dt
The conversion of aldehydes (B) to alcohols (D) is a reversible reaction that utilizes the
equilibrium constant (keq) as follows:
dC B
= k1 C A PH2 − k 2 C B − k 3 CB PH2 + k 3 k eq C B 6 15
dt
dCD
= k 3 CB PH2 − k3 k eq C D − k 4 C D 6 16
dt
To precisely determine the hydrogen partial pressure at any given reaction time, it is necessary to
subtract the amount of stoichiometric hydrogen consumed by Tg, C, D, and E according to:
V L 3CTG,0 + CC + 2CD + 3CE RT

PH2 = P0H2 − 6 17
VG
where P0i and VG are the initial pressure of component i and the volume of the gas phase. The
decarbonylation reaction (k2) predominated over the HDO path (k3), as shown by the kinetic
coefficient values. However, similarities in the rate coefficients k1, k2, and k3 were observed for both
types of Tg, with only k4 showing slight variation.
Jeništová et al. (2017) investigated the influence of hydrogen pressure on the HDO of stearic acid
and tall oil fatty acids (TOFAs). It was shown that the higher the hydrogen pressure, the higher the
FA conversion. The reaction mechanism is similar to previous studies reported elsewhere (Zhang
et al. 2014), with some differences in the pathways in which oxygen atoms are subtracted, as
observed in Figure 6.2i. The Langmuir–Hinshelwood model was utilized to analyze the effect of
hydrogen pressure on the hydrogenolysis reactions (reactions 1 and 3). It is assumed that each reac-
tion rate involves non-competitive adsorption of hydrogen and organic species. For instance, the
rate for reaction 1 (r1) is:
k 1 C A P H2
− r1 = 6 18
1 + K A CA 1 + K H PH2
Different catalysts (Ni–H–Y–80, Ni/SiO2, Pd/C) were used on the HDO of stearic acid, where the
kinetic parameters show that decarboxylation/decarbonylation (DCx) reactions predominated.
Hachemi and Murzin (2018) developed kinetic studies to analyze the FAs methyl esters and Tg
HDO using Ni/H–Y–80 and Pd/C catalysts. Reaction schemes for C18 and C16 are depicted in
Figure 6.2j and Figure 6.2k, respectively. The results based on the Langmuir–Hinshelwood
mechanism indicated that molecules having different numbers of carbon atoms underwent similar
pathways. Reaction rate equations were developed considering that the absorption of acid species is
significantly higher than those of the aldehydes, alcohols, alkanes, etc. The Ni-based catalyst
promoted the HDO reaction, as indicated by the kinetic parameters (k6 > k5), while the Pd catalyst
enhanced the DCx reactions (k5 > k6). It was also found that for both catalysts, the carbon chain
length of the feedstock influences the kinetic parameters.
Arora et al. (2019) investigated the kinetics of stearic acid HDO to explore the fundamental
chemistry and mechanisms involved in the reactions with a NiMo/Al2O3 catalyst (Figure 6.2m).
A model based on the Langmuir–Hinshelwood type kinetics was developed, which predicted well
trends in variation of selectivities with the reaction conditions, including intermediate compounds
such as octadecanol and octadecanal as well as predicting phenomena like inhibiting effects of the
FA. The evaluation of the external mass transfer limitations was performed by experimental results
varying stirring speeds (500–1000) and the Carberry number. Additionally, the mass transfer
resistance in catalyst pores was evaluated by analytical correlations, which presented a moderate
influence on the observed rate of reactions for experiments with stearic acid feedstock. However,
the results indicated a strong effect of film and pore mass transfer resistances concerning
experiments with C18O feedstock.
Recently, Pedroza et al. (2022) performed some research on the kinetics of the HDO of stearic acid
over a NiMo/Al2O3 catalyst to obtain hydrocarbons in the diesel range. The reaction scheme
presented in Figure 6.2n was proposed, and diverse kinetic models were evaluated based on
assumptions to the adsorption step, using Eley–Rideal and Langmuir–Hinshelwood mechanisms.
The surface reaction-limited model based on competitive adsorption with non-dissociate H2
adsorbed presented the best results and was used to simulate the process.
6.3.2 Vegetable Oils

Due to the wide variety of chemical species present in real feedstock for the hydrotreating process
(such as vegetable oils and petroleum distillates), some assumptions are considered to simplify the
calculation of kinetic parameters, i.e. chemical compounds with similar properties are grouped in
the so-called lumps or pseudocomponents in order to reduce the number of kinetic parameters to be
determined. Kinetic studies of vegetable oil hydrotreating consider various reaction schemes due to
differences in feedstocks and methods for grouping their chemical compounds (Zhang et al. 2009).
These models have been analyzed and discussed in the literature (Tirado et al. 2018) and mentioned
in this section, where the most relevant kinetic models for the simulation of vegetable oil
hydrotreating reactors are studied. Figure 6.3 depicts diverse schemes used to analyze the reaction
pathways that take place under diverse operating conditions.
(a) Light (C5–C8) (d) Light (C5–C8)

k4 k4 k6
k3 Middle (C9–C14) k3 Middle (C9–C14)
Triglycerides Triglycerides k5
k2 Heavy (C15–C18) k2 Heavy (C15–C18)
k1 k1
Oligomerized (>C18) Oligomerized (>C18)
(b) (e)
Light (C5–C8) Light (C5–C8)
k4 k7 k4 k6
k3 Middle (C9–C14) k3 Middle (C9–C14) k7
Triglycerides k6 Triglycerides k5
k2 Heavy (C15–C18) k2 Heavy (C15–C18)
k1 k5 k1
Oligomerized (>C18) Oligomerized (>C18)
(c)
k4 Light (C5– C8)
k2
Triglycerides Heavy (C15– C18)
k1 k3 Middle (C9– C14)
Oligomerized (>C18)
(f)
Triglycerides
k1 Oligomerized (> C18)

k5 k6
k7
Heavy (C15–C18)
k2 k9
k8
Middle (C9–C14)
k3
k10
Light (C5–C8)
k4
Figure 6.3 Reported kinetic model for hydrotreating of vegetable oils (Anand et al. 2016a; Anand
and Sinha 2012; Sharma et al. 2012; Tirado and Ancheyta 2019).
Sharma et al. (2012) developed a kinetic model for Jatropha oil hydrotreating using a CoMo/MTS
catalyst. A single lump included the Tg, free fatty acids (FFAs), and some traces of diglycerides and
monoglycerides, where different reaction pathways of conversion toward oligomerized compounds
(>C18), heavy (C15−C18), middle (C9−C14), and light hydrocarbons (C5−C8) were evaluated in the
temperature range of 300−320 C. The model assumed a first-order reaction for the different lumps.
The reaction scheme is depicted in Figure 6.3a, and the values of reaction rate coefficients are listed
in Table 6.3. However, the model did not account for gas production pathways. The reaction rate
coefficients show that the reaction rate of Tg to oligomerized products is much higher than other
reactions at 300 C. The reaction rate coefficients for producing heavy compounds increase by
20 times as the temperature rises from 300 to 320 C. On the other hand, the high acidity of the
CoMo/MTS catalyst increased the formation of light and middle hydrocarbons through cracking
reactions. Experimental data indicated only primary reactions, and the rate of reaction of oligomer-
ization is much higher compared with other reaction pathways.
Anand and Sinha (2012) reported a set of kinetic models for the hydrotreating of Jatropha oil
using a CoMo/γ–Al2O3 catalyst. Diverse reaction pathways were studied in a temperature range
of 320−360 C and pressure of 80 bar. The kinetic parameters obtained for each sequence of reac-
tions are presented in Table 6.3. The results showed that at 320 C (low temperature), Tg underwent
consecutive side reactions with naphtha and kerosene compounds as by-products (C5−C8 and
C9−C14, respectively) in small amounts because of lower acidity and metal support interaction com-
pared with the CoMo/MTS catalyst reported by Sharma et al. (2012). At 340 C (mild temperature),
lighter compounds were produced as thermal cracking impacts all pseudocomponents, and even
diesel (C15−C18), kerosene (C9−C14), and non-stable oligomerized compounds underwent cracking
reactions. At 360 C, comparable pathways are pursued, resulting in stable oligomerized products.
Most of the kinetic models reported for vegetable oil hydrotreating assume mass transfer effects to
be negligible. Nevertheless, experimental studies are not reported to support this assumption.
Anand et al. (2016a) conducted a kinetic and thermodynamic study on the hydrotreating of Jatropha
oil to determine the diffusion limitations, kinetic reaction mechanisms, and thermodynamics using
a NiW/SiO2−Al2O3 catalyst in a fixed-bed reactor. The results indicated that intraparticle diffusion
was negligible since the effectiveness factor was 1. Additionally, the model results at 340 C indi-
cated that Tg follow similar kinetic mechanisms, and the reaction rate coefficients have some tenden-
cies as those reported by Anand and Sinha (2012) at 320 C with a CoMo/Al2O3 catalyst. On the other
hand, the reaction rate changes slightly at a high temperature (420 C) since diesel is also converted to
naphtha fraction due to the presence of acidic sites on the catalyst and thermal cracking.
Despite the sophistication of these kinetic models, the calculated kinetic parameters obtained do
not guarantee to be the optimal ones, i.e. the values that minimize the difference between the exper-
imental and calculated data. When kinetic parameters are inaccurately calculated and not opti-
mized, a model that demonstrates good fit may either disregard reaction pathways that are
being carried out or incorporate those that should not be considered. This inappropriate definition
of possible reaction schemes has a strong effect on predicting the yields and other reaction para-
meters that cause poor design of commercial-scale reactors (Alcázar and Ancheyta 2007). Tirado
and Ancheyta (2019) recalculated the kinetic parameters using experimental data reported by
Anand et al. (2016a) and Anand and Sinha (2012) under different operating conditions and catalyst
types. The estimation of the kinetic parameter was performed using a proposed reaction mechan-
ism shown in Figure 6.3f, which involves all reaction pathways reported in the literature for the
hydrotreating of vegetable oils as well as a series of statistical analyses based on the algorithm pres-
ent in Figure 6.4 and ensuring that the kinetic parameters converge to the global minimum of the
objective function for each reaction scheme by taking into account the sum of the squares of
the errors (SSE) as an objective function to optimize the values. Experimental data obtained during
Table 6.3 Kinetic parameters determined during the HDO of vegetable oils using reaction schemes presented in Figure 6.3.
Reaction pathway
Condition Tg–Prod. Tg–Ol Tg–Hv Tg–Md Tg–Lh Ol–Hv Ol–Md Hv–Ol Hv–Md Hv–Lh Md–Lh
CoMo/MTS Model of Sharma et al. (2012)

300 (a) 7.39 7.27 0.07 0.02 0.03 — — — — — —
320 (a) 14.35 13.25 1.24 0.11 0.04 — — — — — —
CoMo/Al2O3 Model of Anand and Sinha (2012)
320 (c) 17 7 10 0.3 0.3 — — — — — —
340 (b) 17 13 3 0.2 1.1 5 — — 0.4 — 3
360 (d) 24 2 19 2 1 — — — 0.1 0.1
Ea (kJ/mol) 26 87 43 127 47 — — — — — —
A0 31.9 × 1002 14.5 × 1004 62.6 × 1005 4.2 × 1010 13.7 × 1004
NiW/SiO2Al2O3 Model of Anand et al. (2016a)
340 (c) 1.06 0.2 0.85 0.04 0.04 — — — — — —
420 (e) 23.4 2.7 18.8 1.2 0.6 — — — 0.2 0.1 0.1
CoMo/Al2O3 Model of Tirado and Ancheyta (2019)
360 (f ) 23.3 0.317 24.85 1.95 0.82 14.37 0.03 2.09 0.1 0 0.268
NiW/SiO2Al2O3 Model of Tirado and Ancheyta (2019)
420 (f ) 22.62 2.054 18.494 1.257 0.815 2.662 0 0.553 0.172 0.06 0.094
Initialization of parameters
(Monte Carlo method)
Reaction rate
Nonlinear optimization Estimated kinetic
process parameters
Experimental data
Residual analysis Sensitivity analysis
No Is global minimum of
New initial guess
objective function same
based on sensitivity
for all parameters?
Yes
Successful nonlinear
optimization
Figure 6.4 General approach for optimization of parameter values of kinetic models (Alcázar and Ancheyta
2007 / with permission of Elsevier).
the hydrotreating of Jatropha oil at 360 C with a CoMo/Al2O3 catalyst and 420 C with a
NiW/SiO2–Al2O3 catalyst were used to perform the calculation of kinetic parameters based on reac-
tion schemes reported by Anand et al. (2016a) and Anand and Sinha (2012) as well as that proposed
(Figure 6.3f ). It was reported that at 360 C with a CoMo/Al2O3 catalyst, the reactor scheme in
Figure 6.3d with six kinetic parameters showed the best fit with experimental data, while at
420 C with a NiW/SiO2–Al2O3 catalyst, the best reaction scheme was the Figure 6.3e with seven
parameters (Anand et al. 2016a; Anand and Sinha 2012). The comparison between the obtained
kinetic values of the models shows that some reaction pathways are similar. However, certain reac-
tion pathways that were not considered exhibited noteworthy reaction rates. For example, the
transformation of oligomerized compounds into large hydrocarbons (k5) occurs at a rapid reaction
rate. This behavior is expected because the oligomerized compounds fraction groups together with
oxygenated compounds, such as FAs, aldehydes, and alcohols, undergo deoxygenation reactions
and eventually form heavy hydrocarbons (Huber et al. 2007; Zhang et al. 2014). On the other hand,
hydrocarbons under severe temperatures can undergo polymerization reactions, resulting in the
formation of compounds with high molecular weight (k7). However, the values of k6 for the
conversion of oligomerized compounds to middle hydrocarbons were negligible, indicating low
selectivity toward this reaction pathway. It is corroborated that the yield of light hydrocarbons from
heavy hydrocarbons (k9) is zero at low temperatures with CoMo/Al2O3 catalyst, but it starts at 420
C with NiW/SiO2–Al2O3 catalysts. The proposed model gave better fits of the objective function
values (SSE) and the determination coefficient (R2) than those obtained with literature models.
Both models predicted higher decomposition of Tg k Tg0 for CoMo/Al2O3 catalyst compared with
the NiW/SiO2–Al2O3 catalyst.
(a)
20 80
15
Yield fraction, %
70
10
60
5
0 50
0 2 4 6 8
(b) LHSV, h–1
30 80
60
Yield fraction, %
20
40
10
20
0 0
0 3 6 9 12
WHSV, h–1
Figure 6.5 Comparison of experimental and calculated data with kinetic parameters of Anand et al. (––)
and with those obtained by Tirado and Ancheyta (2019) / with permission of Elsevier ( ) at (a) 360 C and
(b) 420 C: Tg (○), >C18 (□), C5–C8 (◊), C9–C14 (∗), C15–C18 (Δ).
Figure 6.5 shows the comparison of the experimental data and the calculated profiles of
each model with their corresponding kinetic parameters. It is noteworthy that the heavy
hydrocarbon yield rises, which is consistent with the LHSV increase as reported by Anand
et al. (2016a) and Anand and Sinha (2012). This trend is well-predicted by the proposed model
by Tirado and Ancheyta (2019), while the previous model predicts a maximum value around
LHSV of 4, and then the yield decreases. This behavior demonstrates that a more detailed
kinetic model allows for better prediction of the experimental yields and their trends
concerning the reaction time.
In order to ensure that kinetic parameters provide the best fit for their corresponding reaction
scheme, sensitivity analysis for the different models is depicted in Figure 6.6 and Figure 6.7. It
is observed in Figure 6.6 that the kinetic coefficients reported by Anand et al. (2016a) and Anand
and Sinha (2012) do not reach the global minimum, particularly k1 and k5 of scheme A7 and k6 and
k7 of scheme A8, which present the minimum value of the objective function further to 0% distur-
bance. In contrast, Figure 6.7 demonstrates that the sensitivity analysis of kinetic parameters fits
perfectly at 0% disturbance of the objective function, reaching the global minimum.
(a)
0.0100 0.014
0.013
0.0096
0.012
0.011
0.0092
SSE
0.010
0.0088 0.009
0.008
0.0084
0.007
0.0080 0.006
–20 –10 0 10 20 –20 –10 0 10 20
(b) 0.028
0.034
0.028
0.032
0.030 0.027
0.028
SSE
0.027
0.026
0.026
0.024
0.026
0.022
0.020 0.025
–20 –10 0 10 20 –20 –10 0 10 20
Perturbation, % Perturbation, %
Figure 6.6 Sensitivity analysis for the reaction rate coefficients reported by Anand et al. (Anand et al. 2016a;
Anand and Sinha 2012) at (a) 360 C and (b) 420 C: k1 (●), k2 (○), k3 (Δ), k4 (▲), k5 (■), k6 (♦), k7 ( ).
(a)
0.0041
0.0035
0.0039
0.0034
0.0037
0.0033
SSE
0.0035
0.0032
0.0033
0.0031
0.0030 0.0031
0.0029 0.0029
–20 –10 0 10 20 –20 –10 0 10 20
(b) 0.027
0.0189
0.0187 0.025
0.0185
0.023
0.0183
SSE
0.0181 0.021
0.0179
0.019
0.0177
0.0175 0.017
–20 –10 0 10 20 –20 –10 0 10 20
Perturbation, % Perturbation, %
Figure 6.7 Sensitivity analysis for the reaction rate coefficients calculated by Tirado and Ancheyta (2019) /
with permission of Elsevier at (a) 360 C and (b) 420 C: k1 (●), k2 (○), k3 (Δ), k4 (▲), k5 (+), k6 (∗), k7 (◊), k8 (□), k9
( ), k10 (♦).
Table 6.4 Statistical parameters from regression analysis and degree of confidence.
CoMo/Al2O3 NiW/SiO2Al2O3
Catalyst 360 C 420 C
Anand and Tirado and Anand Tirado and

Sinha (2012) Ancheyta (2019) et al. (2016a) Ancheyta (2019)
Statistical
parameters Figure 6.3d Figure 6.3f Figure 6.3e Figure 6.3f
R2 0.9971 0.9991 0.9901 0.9932

Slope 0.9916 1.0034 0.9861 0.9906
Intercept 0.586 0.3545 0.329 0.2387
Residual (+) 12 8 15 14
Residual (–) 13 17 15 –16
Range 0.0707 0.0337 0.1505 0.0902
Parity plots of the calculated and experimental yield values for both operating conditions
indicated that the values obtained by Tirado and Ancheyta (2019) presented a better fit concerning
the diagonal line. Regression analysis was used to verify this point, and the results are shown in
Table 6.4. The slope value and determination coefficient calculated for the results of the proposed
model are closer to 1, while the intercept value is closer to 0 compared with those obtained for the
data in the literature.
Residual plots for both operating conditions show no discernible trend among residual
values, indicating that the models are not overestimating or underestimating experimental
data. The balance of positive and negative residual values further supports this conclusion,
as reported in Table 6.4. On the other hand, the range between the lowest negative residual
value and the highest positive residual value shows a much lower value for the proposed
model at both operating conditions compared with the original models, indicating a greater
difference between the calculated and experimental values. Based on the results obtained, it is
demonstrated that the proposed kinetic model presents an appropriate estimation of reaction
rate coefficients, which show a better fit with experimental data compared with kinetic
schemes d and e shown in Figure 6.3. Additionally, the achievement of the global minimum
confirmed by the sensitivity analysis makes the proposed model more accurate (Sámano et al.
2020; Tirado and Ancheyta 2019).
Kinetic studies of hydrotreating conducted on model compounds and vegetable oils have resulted
in a plethora of reaction schemes owing to the diverse reaction mechanisms that arise from the
complex feedstocks, catalysts, and operating conditions employed. Most of them achieve similar
conclusions, indicating that Tg and ester molecules of FAs undergo instantaneous hydrogenolysis
reactions to produce intermediate acid compounds (Hachemi and Murzin 2018). Depending on
these parameters, it is possible to remove oxygen atoms via HDO, decarboxylation, and decarbo-
nylation. The development of kinetic studies requires vital experimental considerations to obtain rep-
resentative information about the reactive system and its subsequent application in reactor models and
simulations. Assessment of external mass transfer and internal diffusional resistance is determined to
estimate the intrinsic reaction kinetics. In addition, it has been observed that the mathematical
algorithm of methods used to estimate the kinetic parameters significantly influences the prediction
of the behavior of the diverse components considered during the kinetic modeling. Table 6.5
Table 6.5 Summary of methodologies used for estimation of kinetic parameters.
Solution
Parameter method/
Kinetic model number Objective function software Optimization algorithm
Snåre et al. (2007) 7 RSS =

exp
C i,t − C cal
2 Odessa ODE —
i,t
i t solver
exp
Kumar et al. 5 RSS = Ci − Ccal
i
4th Order Modified Levenberg−
(2014) i Runge–Kutta Marquardt
Zhang et al. (2014) 8 Least squares Runge–Kutta Trust-region-reflective
2
Bie et al. (2015) 6 RSS = wexp
i − wcal
i
Backward Levenberg−Marquardt
i j difference
Yenumala et al. 5 exp 2 4th Order Nelder–Mead
ER = wi − wcal
i
(2017) i Runge–Kutta
Jeništová et al. 5 — Backward Simplex and
(2017) difference Levenberg–Marquardt
Hachemi and 10 — Backward —
Murzin (2018) difference
Shamanaev et al. — OF = wexp − wcal
2 Gear method Basin-hopping
i i
(2019) i j algorithm
Tirado and 10 2 4th Order Levenberg−Marquardt
exp
SSE = yi,j − ycal
i,j
Ancheyta (2019) i j Runge–Kutta
Vélez Manco 6 exp
Ci − Ccal
2 MatLab —
i
(2014) MSE =
i 2
Mederos-Nieto 4 exp
2 — Levenberg–Marquardt
SSE = yi,j − ycal
i,j
et al. (2021) i j
summarizes the characteristics of the algorithms used to estimate the kinetic parameters for
renewable fuel production through the hydrotreating process. The diversity in the number of
kinetic parameters depends mainly on the required complexity of the model and the capability
to analyze and quantify different compounds. Among the different numerical methods for the
solution of ODEs and parameter optimization, the Runge–Kutta and the Levenberg−Marquardt
are the most used since they have proven accuracy for highly complex systems. In addition, the
selection of the objective function plays a key role during nonlinear optimization. The SSE is the
most used objective function since it has demonstrated an adequate minimization of the exper-
imental and calculated data. Different objective functions can be used considering weighting fac-
tors, absolute differences between calculated and experimental information, and correlation
coefficients. Recently, the average absolute error (AAE) has demonstrated a satisfactory balanced
weighting to predict components with different magnitude orders in nonlinear optimization algo-
rithms (Félix et al. 2022, 2023).
6.4 Models for Catalytic Deactivation
One of the main challenges to overcome during the hydrotreating of vegetable oils is catalyst deac-
tivation, as it is argued that coke deposition is the main cause of deactivation. Compounds contain-
ing oxygen atoms exhibit high reactivity in oligomerization reactions, resulting in coke formation.
Table 6.6 Deactivation model of the hydrotreating process for renewable fuels.
Mathematical
Model expression Observations
Sebos et al. k HDOt Catalytic activity (α) is defined as the ratio of reaction rate coefficient at
α=
(2009) k HDOref any time t concerning a reference time.
Activity coefficient is set as reference (α = 1) at first 100 h of operation.
While catalyst activity declines slowly up to 300 h with a loss of
about 13%.
Bykova et al. r i = α k C i Pm
H2 Catalytic activity decreases linearly with concentration.
(2013) α = 1 − γCB γ is a temperature-dependent parameter, which is a linear function of
γ = c1T − c2 temperature, where c1 and c2 are constants.
The coefficient γ determines the concentration at which the reaction
slows down sharply.
Boldrini et al. φ1 = α1 Deactivation coefficients are associated with the number of catalyst
(2014) β = β0(1 − ct) reuses, with β = 1 for the first use of the catalyst.
φ2 = α2 Two independent deactivation phenomena: one associated with the
β = β0 exp(−ct) reaction time, and the other one with the batch number.
φ3 = α3 For the model, αj presented a value of 1.93 × 1026. For the rest of the
β0 models, parameter a showed similar values. The value of parameter b for
β=
1 + ct the different batches (values between 0.83 and 0.13 for Batch 2 and 10.
For the low Pd loading, β presented lower values, which could be
attributed to the extremely low Pd/oil ratio.
Zhang et al. dαj The decay rate was assumed to follow a second-order law according to
= k dj α2j
(2014) dt the activity parameter. The sum of both decay coefficients (kd1 and kd2)
provides the total deactivation rate.
A catalytic activity function was applied to calculate the deactivation
rate of both HDO and HDCx routes using specific decay constants kd1
and kd2, respectively. The decay coefficients indicated a loss of activity of
up to 80%. HDCx routes contributed mainly on deactivation (kd2 =
0.4475), while the HDO route had no significant effect (kd1 ≈ 0).
Hasanudin rt Catalytic activity is defined as ratio of reaction rate at particular time (t)
α=
et al. (2019) r ref to reaction rate of fresh catalyst as expressed.
A second-order for deactivation rate was obtained. Deactivation rate
constant is highly affected by hydrocracking temperature. Catalyst tends
to deactivate rapidly, which shortens its lifetime at high temperatures.
Frequency factor (A) and deactivation energy (Ed) are 2.75 × 1012 and
54.27 kJ K−1 mol−1, respectively.
The investigation into the decarboxylation of FA in a tubular reactor has shown that high concen-
trations of stearic acid contribute to catalyst deactivation, which is reflected in a lower value of the
corresponding rate constant (Snåre et al. 2007).
Table 6.6 presents the characteristics of the deactivation models reported to produce renewable
fuels by the hydrotreating process. These models have presented different approaches for the anal-
ysis of the catalytic activity. However, more experimental investigations are required to obtain a
better understanding of the phenomena that affect the catalytic activity.
6.5 Reactor Modeling for Vegetable Oil Hydrotreating
Reactor mathematical models are useful and indispensable tools to carry out the scaling-up design
and control of reaction units at laboratory and industrial scales. These models consist of mass balance
equations and energy balance equations (when the system operates under a non-isothermal state)
among different phases in the reacting system (gas, liquid, and solid) to provide a clear idea of the
phenomena that take place in the reactor and predict the yield of reactants and reaction products.
Diverse reactor models have been developed to simulate the performance of the hydrotreating
process for petroleum distillate streams. Mederos et al. (2009b) reviewed the phenomena observed
during the hydrotreating of petroleum distillates in fixed-bed reactors. The authors studied the
phenomena of mass and energy transfer between gas−liquid and liquid−solid interfaces. Addition-
ally, they examined the intraparticle diffusion, axial and radial diffusion, accumulation, convection,
and generation phenomena in this system type. The consideration of all these phenomena requires
some assumptions to simplify the solution of the mathematical models based on experimental data
or analytical criteria to discard the influence of these phenomena on the reaction system. Therefore,
the complexity of a mathematical equation system of a reactor depends on the prediction
requirements and characteristics of the reactive system. Using simple models, convergence with
experimental data is not ensured, while overly complex mathematical models may require
complicated mathematical solutions and long time-consuming for simulation. For this reason,
there are limited models available to illustrate the reactor performance during the hydrotreating
of vegetable oils. Furthermore, some of them do not consider the energy balance to predict the
thermal profile of the reactor (Forghani and Lewis 2016; Vélez Manco 2014). The first models were
focused on the hydrogenation of unsaturated compounds present in bio-oils (Cabrera and Grau
2008; Chen et al. 1981; Fillion et al. 2002; Holser et al. 2002; Takeya et al. 1997). Vélez Manco
(2014) established a conceptual industrial reactor design for vegetable oil hydrotreating based
on experiments with refined, bleached, and deodorized palm oil (RBDPO) and kinetic studies
performed. The reactor model aimed to analyze two cases to avoid hot spots higher than 370 C,
by considering ideal flow and avoiding excessive product cracking (i.e. reactive system is not
affected by gradients of concentration or temperature). The molar balance equation for each
compound i considers the temperature and liquid residence time based on the catalytic bed length
(z) and reactor cross-sectional area (A):
dCi Aρb
= ri 6 19
dz V0
Energy balance was expressed to consider the exothermal behavior of HDO reaction based on the
overall heat-transfer coefficient (Ua) and differential of heat of reaction (ΔHij) by the equation:
q
Ua Ta − T + r ij ΔH ij
dT i=1
= m 6 20
dW
F i C pi
i=1
Additionally, the Ergun equation was used to calculate the pressure drop in packed beds:
dP G 1−ε 150 1 − ε μ
= − + 1 75G 6 21
dz ρb dpe ε3 dpe
Two situations were analyzed, and in case (a), the feedstock and hydrogen are injected
concurrently at the top of the reactor, whereas in case (b), a hydrogen-quenching stream near
the reactor inlet is considered. The energy balance depicted a temperature peak placed at 0.55 m
due to the energy released by exothermic reactions. The pressure drop, though negligible in both
cases, was higher in case (b). However, case (b) was deemed to be the optimal choice as the liquid
temperature increased by only 9 C, resulting in a reduction in the catalyst weight of 800 kg to
achieve an alkane/RBDPO ratio comparable to that of case 1. Nevertheless, the reactor model
was considered to be free from mass transfer effects.
Forghani et al. (2014) reported a two-dimensional model of a trickle-bed reactor based on the
two-film theory to determine the mass and energy transfer coefficients. The reactor model uses
the kinetic model reported by Anand and Sinha (2012). For the representation of the reactor math-
ematical equations, the following assumptions have been considered for the reactor: (a) a plug-flow
behavior is assumed; (b) radial concentration gradients are negligible; (c) axial heat dispersion is
not present; (d) isothermal behavior is achieved. With these assumptions, the mass balance equa-
tions consider evaporation for ideal mixing for each lump (Tg, LC, HC, MC, and OC) as:
∂CLi k L as A
uL A = vi r i + i x i P∗i − PG 6 22
∂z Ru T G i
∂PG k L as A
uG A i
= − i x i P∗i − PG 6 23
∂z Ru T G i
∂PG
H2 k L as ARu T G PG
= − H2 H2
− C LH2 6 24
∂z UG H H2
The heat balance was taken into account to evaluate temperature variations along the catalytic
bed for both liquid and gas phases as well as for the reactor wall:
Liquid phase:
∂T L
F G CLp = r i ΔH i A − hG as A T G − T L − 2πr in εhL T W − T L 6 25
∂z
Gas phase:
∂T G
F G CG = hG as A T G − T L − 2πr in 1 − ε hL T W − T G 6 26
p
∂z
Reactor wall:
∂2 T W 1 ∂ ∂T W
λax + λra =0 6 27
∂z2 r ∂r ∂r
The system of differential equations was solved with a set of boundary conditions, obtaining the
concentration and temperature profiles shown in Figure 6.8. The temperature profile of the liquid
phase increases at the reactor inlet, while the temperature of the gas phase, which is fed in coun-
tercurrent, increases linearly through the bed due to the constant physical properties of the gas at
this temperature range. The main product was diesel (HC), and the evaporation of the liquid pro-
ducts was negligible. In addition, the model fitted well with the experimental data.
Mederos et al. (2020) developed dynamic, plug-flow, one-dimensional, and heterogeneous math-
ematical models for the catalytic hydrocracking of Jatropha oil using a commercial CoMo catalyst.
Multiphase trickle-bed reactor models were used to predict the dynamic behavior of micro-scale
and industrial-scale reactors in order to understand the influence on the liquid molar concentra-
tion, temperature profiles, and partial pressure. The system of equations is described in detail in
Section 6.7. Additionally, the pressure drop in countercurrent flow in a packed bed was modeled
by an equation of the Carman–Kozeny type. The simulations fit well with the experimental data
displaying significant findings at the industrial scale. The industrial-scale predictions revealed that
the most substantial gradients in temperature occur in the first half of the catalytic bed, resulting in
the highest Tg conversion to green diesel because of the availability of reactants in this zone. Fur-
thermore, there is an increase in the effectiveness factor beyond the mid part of the catalytic bed
caused by the cooling effect of the ascending gas phase through the reactor, which avoids reaching
high temperatures.
(a)
480
460
440
T (°C)
420
400
380
0 0.2 0.4 0.6 0.8 1
X/L0
(b)
2.5
2
Ci (mol/m3)
1.5
0.5
0
0 0.2 0.4 0.6 0.8 1
X/L0
Figure 6.8 (a) Predicted temperature profiles: ( ) wall, (––––) liquid, ( ) gas. (b) Predicted
concentration profiles of triglycerides in a hydrocracking reactor: (●) Tg, (◊) HC, (□) OC, (▲) MC, (○) LC
(Adapted from Forghani et al. 2014).
Computational fluid dynamics (CFD) has been used to model different reactors to produce
renewable fuel by hydrotreating vegetable oils using cooling fluids and considering thermal
equilibrium between phases to give a single reactor temperature profile. Muharam et al. (2017)
modeled a trickle-bed reactor for the hydrotreating of 5 wt.% triolein in dodecane. The results
showed a small increase in the temperature along the catalytic bed (almost 10 C). Other simula-
tions conducted on slurry bubble column reactors reported minimal issues with heat removal,
requiring only a small cooling area (Muharam and Adevia 2018; Muharam and Putri 2018). In
the same way, other studies were developed to analyze the effects of operating conditions and heat
transfer on the hydrodynamics of the process (Kiran 2015; Mendoza Sépulveda 2013).
6.5.1 Deviation from Ideal Flow Pattern

Adequate knowledge of phenomena occurring along with chemical reactions must be accurate
to carry out the design and scaling-up of reactors. During laboratory tests, the collected data
can be affected by mass and temperature gradients (deviation of the ideal pattern flow, effi-
ciency of catalytic particle wetting, and wall effects), resistance to mass and heat transfer at
gas–liquid and liquid–solid interfaces, as well as in the interior of the catalytic particle. The
extensive knowledge of these phenomena on the reactive system allows an appropriate inter-
pretation of experimental data and results from a mathematical model. If these phenomena
are unavoidable, their quantification is necessary to obtain reliable and replicable data
(Mederos et al. 2009a).
• Ideal plug-flow
One of the main mistakes made during the modeling and scaling-up of trickle-bed reac-
tors is to assume in advance that feedstocks and products present a plug-flow behavior.
A deviation from this pattern would deviate the measurement results, causing a misunder-
standing of experimental data. Experimental studies have shown that the reactor-to-particle
size ratio (LB/dpe) strongly influences axial and radial dispersion phenomena (Gierman
1988). Therefore, by controlling this ratio, the effect of these phenomena in reactors can
be minimized. Mederos et al. (2009a) reviewed a series of analytical criteria to evaluate
the effects of deviation from ideal flow patterns in packed beds. A criterion reported by
Bischoff and Levenspiel (1962) has been widely used to evaluate the effect of radial
dispersion in packed beds as follows:
LB u L dR
> 0 04 6 28
dpe εL DLr
On the other hand, axial dispersion effects are present in packed beds. An appropriate
selection of dimensions for the catalytic bed can greatly mitigate the impact of axial dispersion.
It has been reported that to avoid back-mixing effects in the trickle-bed reactor, the
catalyst bed length (LB) should be at least 100 times the catalyst particle size (dpe) (Perego
and Peratello 1999).
• Catalyst wetting efficiency

The distribution of the liquid phase through the catalytic bed influences the performance of
reactors since an incomplete wetting of catalytic particles would limit the conversion of the
feedstock. In commercial units, the surface velocity of the liquid phase is high enough to com-
pletely wet the catalyst particles. However, in laboratory-scale reactors, the superficial velocity is
much smaller, causing incomplete wetness (Carruthers and DiCamillo 1988). An even distribu-
tion of the liquid phase in trickle flow is observed when the liquid flow is mostly affected by the
frictional force rather than the gravity force, causing an even flow through the cross-section of the
catalytic bed wetting the interstitial channels available and improving contact with the catalytic
surface. Otherwise, a partial wetting of the catalyst is attained and affects the performance of the
system as follows:
dP dz flow 180μL 1 − εL 2 uL
= >1 6 29
dP dz gravity ρL d2pe gεL 4
• Wall effects
Wall effects are occasioned by the inadequate distribution of the liquid phase through the
cross-section of the catalytic bed. The liquid phase displaces toward reactor walls, where the fluid
velocities are higher and the conversions are lower due to the smaller amount of catalyst particles
available in this part of the reactor. Different researchers have related these phenomena to the
reactor-to-particle diameter ratio (dR/dpe); however, it is not an easy task to find the optimum
minimum value to avoid the wall effects because the method of catalyst loading affects the liquid
distribution. Therefore, different criteria have been developed based on experimental data
reported (Froment et al. 1990; Mederos et al. 2009). To consider that a TBR presents negligible
wall effects, the following criterion can be considered since the minimum value of (dR/dpe) ratio is
one of the most rigorous reported in the literature:
dR
> 20 6 30
dpe
6.6 The Importance of Modelling Reactors for Vegetable

Oil Hydrotreating
Mathematical models focused on catalytic hydrotreating of vegetable oil are scarce (Tirado et al.
2018). The development of these model types is fundamental for process design and optimization
due to:
•• It is necessary for scaling up, controlling, and automation of process of catalytic reactors
Process optimization is only possible if the reactor model adequately predicts the variation of the
concentration over time, as in a dynamic model
• Make recommendations for industrial plants and improve the process efficiency based on
accurate simulations
• Make modifications in operating conditions by making changes in the feed properties and/or to
achieve the product specifications
• Predict the dynamic behavior during the start-up, shutdown, and disturbances in the process
and/or establish adequate control strategies
• Describe the influence of operating variables on the catalytic bed length and time-on-stream
6.7 Study Case for the Development of Dynamic Reactor Model
This section describes in detail a methodology for the development of dynamic reactor mathemat-
ical models for the hydrotreating of vegetable oil in order to provide some insights and considera-
tions on its development, solution, scaling-up, and interpretation of the results obtained. The
reactor models were configured to simulate the process of Jatropha oil hydrotreating with a
NiW/SiO2–Al2O3 catalyst. Table 6.7 summarizes the physical and chemical properties of the
catalyst and vegetable oil.
6.7.1 Equations and Assumptions for Hydrotreating Reactor Modeling

A series of dynamic mathematical models was developed to simulate the performance of a reactor
for vegetable oil hydrotreating. The reactor models consist of a set of partial differential equations
(PDEs) describing the mass and heat balances in the multiphase system (gas, liquid, and solid) over
time-on-stream and spatial variable “z” since the reactants and products follow a concurrent
downflow. The mass and heat balance equations are based on a reactor model for vacuum gas
oil hydrotreating reported by Korsten and Hoffmann (1996). Mederos and Ancheyta modified this
reactor model to analyze the behavior of the reactor in a dynamic regime (Mederos et al. 2006).
Firstly, a reactor model was adjusted with characteristics of an isothermal bench-scale reactor
using experimental data obtained during Jatropha oil hydrotreating. Then, the trickle-bed reactor
model was modified according to the dimensions of a pilot-scale unit to evaluate the hydrodynamic
Table 6.7 Properties of the feedstock and catalyst for hydrotreating

process.
Property Value
Vegetable oil Jatropha oil

3
Density (g/cm ) 0.9163
Molecular weight (g/mol) 867
Sulfur content (ppm) 4
Composition (wt.%)
Free fatty acids 1.7
C16:0 (Palmitic) 19.5
C18:0 (Stearic) 7.9
C18:1 (Oleic) 45.4
C18:2 (Linoleic) 27.2
Elemental composition (wt.%)
C 77.0
H 13.6
O 9.4
Distillation curve (vol. %)
IBP–250 C —
250–380 C 0.20
380 C–FBP 99.8
Catalyst 4% NiO, 24% WO3
Support SiO2–Al2O3
Surface area (m2/g) 233
Pore volume (ml/g) 0.32
Mean pore diameter (nm) 10.2
Surface acidity (mmol/g) 0.77
IBP, initial boiling point; FBP, final boiling point.
effect by carrying out the scaling-up process. Additionally, the design and simulation of a commer-
cial vegetable oil hydrotreating unit were carried out to examine the dynamic behavior of the proc-
ess and to establish operational strategies.
The following considerations were made during the development of the reactor model:
1) Reactors operate in a dynamic regime

2) The properties of gas and liquid, such as density, viscosity, superficial velocities, and holdups,
are kept constant along the catalytic bed. The same consideration is made for the properties of
the catalytic particles and bed (size, porosity, etc.), wetting efficiency, and bed void fraction
3) Organic compounds do not undergo vaporization or condensation
4) Chemical reactions occur when the liquid phase diffuses into the catalyst particles
5) Catalyst activity is constant due to short time-on-stream (<100 h)
6) Liquid phase fills the catalyst pores with capillary forces
7) Analytical criteria are used to quantify the axial and radial dispersion, wetting efficiency, and
diffusion limitations within the catalytic particle
8) The heats of reaction are invariant to the reactor temperature
The unsteady-state mass balance equations were formulated for each phase. The flow rate for
compound i in the gas phase is influenced by convection and mass transfer phenomena, while
the gas–liquid equilibrium is calculated by Henry’s law. Therefore, the continuity equation for
compounds in the gas phase is defined by:
εG ∂pG uG ∂pG pG
i
= − i
− K Li aL i − C Li
RT G ∂t RT G ∂z Hi
where i = H2 , C3 H8 , C4 H10 , CO, CO2 , H2 O 6 31
Consequently, the solubility of gaseous compounds in the liquid phase is expressed as:
∂CLi ∂C L pG
εL = − uL i + K Li aL i − C Li − f w K Si as C Li − CSi
∂t ∂z Hi
i = H2 , C3 H8 , C4 H10 , CO, CO2 , H2 O 6 32
Organic compounds are assumed not to be subject to evaporation/condensation effects based on
simulations performed under severe operating conditions (420 C) (Forghani et al. 2014), by which
the mass balance equation for the compounds in the liquid phase is:
∂CLi ∂C L
εL = − uL i − f w K Si as CLi − C Si
∂t ∂z
i = Tg, Ol, Hv, Md, Lh 6 33
In studies carried out during the hydrotreating of vegetable oil, it was observed that
deoxygenation and cracking reaction are negligible in the absence of a catalyst (Anand et al.
2016b). Therefore, it is feasible to assume that the chemical reaction takes place on the wet catalyst
surface. In addition, the consumption/generation term is influenced by the effectiveness factor, and
the mass and energy transfer terms are affected by hydrodynamic effects (Lange et al. 2004).
N RL
∂CSi
S 1− B = − f w K Si as C Li − C Si + φ υi,j r i,j
∂t j=1
i = Tg, Ol, Hv, Md, Lh 6 34

In addition, the catalyst deactivation is considered negligible as the yields of hydrotreating
products were constant up to 100 h of continuous operation.
Since most of the chemical reactions occurring during vegetable oil hydrotreating are highly
exothermic, the reactor model includes energy balance equations for each phase to analyze
the changes in temperature when operating in non-isothermal operating conditions. The
unsteady-state heat balance equations for gas, liquid, and solid phases are termed as follows:
∂T G ∂T G
εG ρG CpG = − uG ρG CpG − hGL aL T G − T L 6 35
∂t ∂z
∂T L ∂T L
εL ρL CpL = − uL ρL CpL − hGL aL T G − T L − f w hLS as T L − T S 6 36
∂t ∂z
N RL
∂T S
εS ρS CpS = f w hLS as T L − T S + ηj vi,j r i,j − ΔH R 6 37
∂t j=1
6.7.2 Kinetic Model for Hydrotreating of Vegetable Oil

Based on the approach reported by Tirado and Ancheyta (2019) for the estimation of optimal kinetic
parameters, where the huge number of chemical components inside the HDT system was classified
using a single 5-lump reaction scheme (Figure 6.3f ), the kinetic model was used under different
experimental conditions for the hydrotreating of Jatropha oil with a NiW/SiO2–Al2O3 catalyst:
340–420 C of reaction temperature, a hydrogen pressure of 80 bar, and a WHSV of 0.5–12 h−1.
The calculation of the kinetic parameters was performed assuming first-order rate equations repre-
sented by the scheme shown in Figure 6.3f:
dyTg
= − k 1 + k2 + k3 + k 4 yTg 6 38
dτ
dyOl
= k 1 yTg + k7 yHv − k 5 + k6 yOl 6 39
dτ
dyHv
= k2 yTg + k5 yOl − k 7 + k8 + k9 yHv 6 40
dτ
dyMd
= k 3 yTg + k 6 yOl + k 8 yHv − k 10 yMd 6 41
dτ
dyLh
= k4 yTg + k9 yHv + k 10 yMd 6 42
dτ
The system of ODEs (6.38)–(6.42) was solved by a Runge–Kutta method subjected to some con-
straints based on the global conversion of compounds. Additionally, each kinetic coefficient must
be a positive term (ki ≥ 0) and follow the Arrhenius equation. The Levenberg–Marquardt algorithm
was used for optimizing the reaction rate coefficients (k1 to k10). The initial guess values required for
the nonlinear optimization were determined using the Monte Carlo method, which evaluates ran-
dom values for each kinetic parameter at a time and selects those that provide the least value of the
sum of squared errors (SSE):
N data 2
SSE = yexp − ycalc 6 43
i=1
Values of pre-exponential factors (A0) and activation energies (EA) were determined by the Arrhe-
nius equation and reported in Table 6.8. In addition, the heats of reaction for each representative
reaction pathway were calculated from the heats of formation of the species involved, as shown in
Table 6.8. These values are similar to those reported during the hydrotreating of renewable feed-
stocks (Anand et al. 2016a; Mederos et al. 2019; Vélez Manco 2014).
6.7.3 Hydrogen Consumption and Gas Generation

Hydrogen consumption is a critical parameter to determine because of its large impact on the
design and operation of industrial reactors. In addition, it is well known that renewable feedstock
often has high hydrogen consumption during the HDT process, so the hydrogen cost is relevant to
the feasibility of this technology. The prediction of hydrogen consumption, gases (C3H8, C4H10, CO,
and CO2), and H2O formation is performed by a method based on reaction pathways reported for
the compounds from each lump (Kubička et al. 2009; Mohammad et al. 2013; Sinha et al. 2014).
Considering the set of representative reaction pathways shown in Table 6.9 and the reaction rate
law, the global reaction rate of hydrogen and gas products is obtained.
Table 6.8 Kinetic parameters and heat of reactions for vegetable oil hydrotreating.
MJ
Reaction j A0 (h−1) EA (kJ/mol) ΔH LR
mol
Tg–Ol (1) 9.8723 × 109 128.4918 −1.0235
Tg–Hv (2) 157.5590 12.7951 −1.2620a
−0.8039b
−0.9284c
Tg–Md (3) 1.1004 × 1021 277.9029 −1.0574
Tg–Lh (4) 3.4066 × 10 9
127.7768 −1.0574
Ol–Hv (5) 127.1923 20.7014 −0.0795a
−0.0732b
−0.0317c
Ol–Md (6) — — —
12
Hv–Ol (7) 6.1409 × 10 173.1821 0.0299
Hv–Md (8) 1.3894 × 1019 260.8500 −0.0424
Hv–Lh (9) 0.0424 2.1775 −0.0430
Md–Lh (10) 0.2817 6.5759 −0.0202
a
HDO, hydrodeoxygenation.
b
HCO, hydrodecarbonylation.
c
HCO2, hydrodecarboxylation.
Table 6.9 Representative chemical reactions during the hydrotreating of vegetable oil.
Chemical reaction Kinetic expression
C57H110O6 + 3H2 3C17H35COOH + C3H8 r1 = k1CTg

C57H110O6 + 12H2 3C18H38 + C3H8 + 6H2O
C57H110O6 + 6H2 3C17H36 + C3H8 + 3H2O + 3CO r2 = k2CTg
C57H110O6 + 3H2 3C17H36 + C3H8 + 3CO2

C57H110O6 + 6H2 3C8H18 + 3C9H20 + C3H8 + 3CO2 r3 = k3CTg
C57H110O6 + 6H2 3C8H18 + 3C9H20 + C3H8 + 3CO2 r4 = k4CTg
C17H35COOH + 3H2 C18H38 + 2H2O
C17H35COOH + H2 C17H36 + H2O + CO r5 = k5COl
C17H35COOH C17H36 + CO2
C18H38 + C8H18 C26H54 + H2 r7 = k7CHv
C18H38 + H2 C12H26 + C6H14 r8 = k8CHv
C17H36 + H2 C8H18 + C9H20 r9 = k9CHv
C12H26 + H2 C8H18 + C4H10 r10 = k10CMd
ri = υij r j 6 44
where υ is the stoichiometric coefficient for i = H2, C3H8, C4H10, CO, CO2 and H2O. Since the
amount of H2 is in excess during the hydrotreating process, each reaction pathway summarized
in Table 6.9 does not take it into account, rather hydrogen partial pressure is grouped in the reaction
rate coefficient.
6.7.4 Solution of Reactor Models

The method of lines was used to solve each reactor model (Schiesser 1991), where the PDEs were
transformed into ODEs by discretization in the axial directions using the backward finite differ-
ence method. Thus, the catalyst bed length was divided into N equidistant segments since the
gradient in z is equal to the catalyst bed length divided by the number of lines (N). The
fourth-order Runge–Kutta method is used to solve the set of ODEs. The initial conditions and
boundary constraints to predict the profiles of partial pressure, concentration of the pseudocom-
ponents, and temperature profiles of the phases under non-isothermal conditions are given as
follows:
• Initial conditions:
For t = 0, at z = 0
pG G
i = pi 0
i = H2
pG
i =0 i = C3H8, C4H10, CO, CO2, H2O
C Li = CLi 0
i = Tg, H2
C Li = 0 i = C3H8, C4H10, CO, CO2, H2O, Ol, Hv, Md, Lh
C Si = 0 i = H2, C3H8, C4H10, CO, CO2, H2O, Tg, Ol, Hv, Md, Lh
at z > 0
pi,G = (pi,G)0 i = H2, C3H8, C4H10, CO, CO2, H2O
C Li = 0 i = H2, C3H8, C4H10, CO, CO2, H2O, Tg, Ol, Hv, Md, Lh
Boundary constraints:
For t > 0, at z = 0,
pG G
i = pi 0
i = H2, C3H8, C4H10, CO, CO2, H2O
C Li = CLi 0 i = Tg, H2
The initial Tg concentration in the liquid phase was calculated according to the weight fraction of
Tg (wTg) in the vegetable oil:
ρL 0
CLTg = wTg 0
6 45
0 MWL
The feedstock gas was considered to be pure hydrogen, and the initial hydrogen concentration in
the liquid phase is calculated by Henry’s law (Tirado et al. 2018).
pG
H2
CLH2 = 0
6 46
0 H H2
where the Henry’s law coefficient (H H2) determines the H2 solubility in vegetable oil and is based on
an Arrhenius-type equation, using a pre-exponential constant (H0) equal to 8.84 MPa m3/kmol and
an apparent activation energy of absorption (ΔEA) of –5000 kJ/kmol reported for the solubility of
hydrogen in soybean oil (Fillion and Morsi 2000):
− ΔEA
H H2 = H 0 exp 6 47
RT
On the other hand, the solubility of H2O and reaction product gases was determined by the Peng–
Robinson equation of state (Tirado et al. 2018).
Mass- and heat-transfer parameters in interfaces (gas–liquid and liquid–solid), the properties of
vegetable oils as well as gases, were calculated through correlations reported in the literature.
Table 6.10 presents a summary of these correlations.
Table 6.10 Correlations to estimate model parameters.
Property Correlation References
Liquid–solid specific aS = (6/dpe)(1 − ϵB) Korsten and

surface area Hoffmann (1996)
Solid density ρB Tarhan (1983)
ρS =
1 − ϵB
Solid holdup εS = 1 − ϵB Tarhan (1983)
Liquid holdup εL = 0.185 ϵB aS χ
0.333 0.22
Tarhan (1983)
for bench-unit εL = 9.9Re0.333Ga−0.333 Tarhan (1983)
for commercial-unit
Gas holdup εG = ϵB − εL Tarhan (1983)
Catalyst porosity S = ρSVg Tarhan (1983)
Bed void fraction 2 Froment et al.
dR dpe − 2
B = 0 38 + 0 073 1 + 2 (1990)
dR dpe
Molecular weight MW = 3 xiMWi + 38.0488 Anand et al.
(2010)
Molar diffusivity vL 0 267 TL Tyn and Calus
DMi = 8 93x10 − 8 (1975a)
vL 0 433 μL
Liquid viscosity log(μL)T = − 0.6298 + [273.15/(T + 88.81)] Rodenbush et al.
P
0 181
(1999)
− 0 0102 + 0 04042 μL
μL = μL T 10 1000 T
Liquid molar volume vL = 0.285vc1.048 Tyn and Calus

(1975b)
1
Liquid–solid mass k i aS GmL nS
μL 3
α = − 577 8dpe + 184 3
Goto and Smith
transfer coefficient = αS ; nSS = − 0 5886dpe + 0 7018 (1975)
DMi μL ρL DMi
1 1
for bench-unit ki S GmL 2 μL 3
=18
for commercial-unit DMi aS a S μL ρL DMi
Liquid side mass-transfer α2 1 α = − 26 23dpe + 13 63 Goto and Smith

k i aL GmL μL 2 for dpe > 2 91cm α1 = − 0 08197dpe + 0 4339
coefficient at gas–liquid = α1 ; 2
α1 = − 7 598dpe + 8 211 (1975)
DMi μL ρL DMi for dpe ≤ 2 91cm α2 = − 0 08442dpe + 03854
interface
Thiele modulus for n−1 Froment et al.
1 Vp n + 1 ρS k in CSi
irreversible reactions Φj = (1990)
ϕS Sp 2 DLei
Effective diffusion ϵS 1 Froment et al.
DLei = (1990)
τ 1 DLMi + 1 DLKi
Effectiveness factor tanh Φj Froment et al.
for Φj < 3 ηLj = (1990), Mederos
Φj
1 et al. (2012)
for Φj ≥ 3 ηLj =
Φj
6.8 Analysis and Discussion of Results
6.8.1 Criteria to Ensure Ideal Behaviors in Trickle-Bed Reactor

The influence of hydrodynamic phenomena on the three-phase system was analyzed and
quantified by analytical criteria and correlations developed using experimental information.
In this way, the changes in partial pressure, temperature, and concentration of the compounds
involved in the process can be accurately predicted and simulated as a function of the oper-
ating time and length of the catalytic bed. Operating conditions and dimensions of different
reactor scales used to evaluate and analyze the reactor mathematical model are summarized in
Table 6.11.
• Ideal plug-flow
Analytical criteria were used to evaluate the effect of axial and radial dispersion on the reaction
system at operating conditions. The results of these phenomena for bench-scale and pilot-scale
reactors are listed in Table 6.12. The results of the commercial-scale reactor are not disclosed
because the effect of hydrodynamic phenomena is negligible due to the high liquid velocities
in these units. It is understood that both bench-scale and pilot-scale reactors operate under ideal
flow conditions because axial and radial dispersions are neglected. Although the particle size
used in the pilot-scale reactor is 10 times larger than the bench-scale, the LB/dpe ratio for both
scales resulted to be of similar magnitude due to the greater length of the catalytic bed in the
pilot-scale reactor (12.8 times higher than that in the bench-scale reactor).
Table 6.11 Properties of catalyst and operating conditions for different reactor scales.
Scale Bench Pilot Commercial
Mode Isothermal Isothermal Adiabatic

Catalyst mass (g) 2 100 41,514,783.2
Catalyst shape Crushed Extrudates Extrudates
Equivalent particle diameter, dpe (cm) 0.023 0.254 0.254
Bed length, LB (cm) 3.5 45 853.44
Reactor diameter, dR (cm) 1.3 3 304.8
Bed volume (cm3) 4.64 318 62,272,143.7
Inlet temperature ( C) 420 420 340
Pressure (MPa) 8 8 8
WHSV (h–1) 1 1 3
H2/feed ratio (Nl/l) 1500 2500 1500
–04 –02
Liquid mass flow (g/s) 5.55 × 10 2.77 × 10 34,595.6
uL cm3L cm2R s 1.01 × 10–03 7.27 × 10–03 7.71 × 10–01
K STg s − 1 8.24 × 10–05 1.51 × 10–03 1.00 × 10–02
aS cm2S cm3R 126.20 11.86 11.20

εL cm3L cm3R 0.430 0.288 0.390
Table 6.12 Analytical criteria to evaluate hydrodynamic effects in packed-bed reactor.
Criterion Bench-scale Pilot-scale
Radial dispersion 152.17 > 4.74 × 10–05 177.16 > 3.26 × 10–02
LB uL dR Fulfilled Fulfilled
> 0 04
dpe εL DLr
Axial dispersion 152.17 > 100 177.16 > 100
LB Fulfilled Fulfilled
> 100
dpe
Catalyst wetting 0.0298 <1 0.0933 <1
dP dz flow 180μL 1 − εL 2 uL Not fulfilled Not fulfilled
= >1 0.435 0.539
dP dz gravity ρL d2pe gεL 4
fw
Wall effects 56.52 > 20 11.81 > 20
dR Fulfilled Not fulfilled
> 20
dpe
• Catalyst wetting efficiency

Evaluation of catalyst wetting efficiency for the different reactor scales was carried out through
the friction and gravity forces ratio. Results in Table 6.8 indicate that at the operating conditions,
bench-scale and pilot-scale reactors are more influenced by the force of gravity, which indicates
an incomplete wetting of the catalytic bed (Mederos et al. 2009a). Therefore, as wetting efficiency
is an essential scale-up parameter, it was estimated using the correlation reported by El-Hisnawi
et al. (1982).
f w = 1 617 Re L0 146 GaL − 0 071 6 48
This correlation was obtained with experimental data reported for superficial velocities of liq-
uid phase in the range of 0.55 × 10−3 ≤ uL ≤ 7.5 × 10−3. For bench-scale and pilot-scale reactors,
values of 1.01 × 10−3 and 7.27 × 10−3 are obtained, respectively, which are adjusted to the estab-
lished range. ReL is the Reynolds number and GaL is the Galileo number. Both of them are cal-
culated as follows:
dpe uL ρL
Re L = 6 49
μL
3
dpe gρ2L
GaL = 6 50
μ2L
The result of the calculation of wetting efficiency for the bench-scale reactor was 0.435, while
for the pilot-scale reactor was 0.539. The wetness of the catalytic bed in the pilot-scale reactor is
10% higher than at the bench-scale. These results are consistent with the experimental data
reported in the literature, where laboratory-scale reactors give wetting efficiencies in the range
of 0.12–0.6 (Bhaskar et al. 2004; Satterfield 1975). This efficiency increases as the surface velocity
of the liquid phase does until it reaches values similar to those used in commercial-scale reactors,
which are close to 1. Wetting efficiency results were plotted in Figure 6.9 along with data exper-
imentally obtained during hydrotreating of heavy crude oil carried out in a pilot-scale reactor
with LB = 147.3 and dR = 2.94 cm ratio and hydrotreating of gas oil in a reactor with LB = 66.5
cm and dR = 3 cm (Alvarez and Ancheyta 2012; Korsten and Hoffmann 1996). It is observed that
1
0.9
0.8
0.7
0.6 Pilot-scale
fw
0.5 Bench-scale
0.4
0.3
0.2
0.1
0
0.001 0.01 0.1 1 10 100
GmL, kg/(m2 s)
Figure 6.9 Catalyst wetting efficiency as a function of GmL: (Δ) Alvarez and Ancheyta (2012), (●) Korsten
and Hoffman (1996), and (◊) dynamic HDT simulation (Tirado and Ancheyta 2020b).
the results of catalyst wetting efficiency obtained in this work are within the reported literature
and follow the expected tendency for superficial mass flow velocity of the liquid phase.
• Wall effects
The wall effects were evaluated for both bench-scale and pilot-scale reactors. It was found that
the bench-scale reactor fulfilled the criterion (dR/dpe = 56.52 > 20), while the pilot-scale reactor
did not meet the criterion (dR/dpe = 11.81 < 20).
The model considers the contact surface between liquid and solid phases through catalyst wet-
ting efficiency. Therefore, wetting efficiency and wall effects are complementary parameters that
help to understand the phenomena occurring in the reactor.
• Catalyst effectiveness factor

The effectiveness factor calculations for the crushed catalyst showed negligible internal diffu-
sion limitations (>0.96) for each reaction pathway at a reaction temperature of 420 C, hydrogen
pressure of 8 MPa, and WHSV of 1 h–1. On the other hand, commercial-size catalysts presented
lower values of effectiveness factor. Reaction pathways 2 and 5 have the greater diffusion limita-
tions in catalytic particles, as they showed the lowest values of effectiveness factor (0.814 and
0.953, respectively) because these pathways present the highest reaction coefficients
(k2 = 18.49 and k5 = 2.66), which could indicate a large formation of products, preventing the
diffusion of reactive species into the catalytic pore. The effectiveness factor of the other reaction
pathways shows negligible internal diffusion limitations (>0.96).
6.8.2 Dynamic Profiles of Feedstock and Products of a Bench-Scale Reactor

for Catalytic Hydrotreating of Vegetable Oil
The reactor mathematical model considered the characteristics of a bench-scale unit to simulate the
hydrotreating process of Jatropha oil and analyze the behavior of each lump involved in the reac-
tion scheme. The heat balance equations were omitted from the model as the reactor operates at
isothermal conditions. The yield profiles of each lump along catalyst bed length under steady-state
(a)
100
80
60
Yield, %
40
20
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Catalytic bed length, cm
(b)
90
80
70
60
Yield, %
50
40
30
20
10
0
0 2 4 6 8 10 12
Space-velocity, h–1
Figure 6.10 Comparison of predicted (lines) and experimental (symbols) yields at (a) WHSV of 1 h−1
along the catalytic bed, and (b) varying space-velocity: Tg (—, ○), Ol (— —, ♦), Hv (– –, Δ), Md (–––, ∗)
and Lh (– –, ◊) (Tirado and Ancheyta 2020a / with permission of Elsevier).
conditions are shown in Figure 6.10a. It is observed that in the reactor top, the unconverted Tg
form mainly heavy hydrocarbons due to the high Tg reaction rate (k1 = 2.05, k2 = 18.49) at the
studied operating conditions. The formation of heavy hydrocarbons is enhanced and reaches a
maximum yield at one-third of the catalytic bed length until the concentration of Tg is insufficient
to continue the reaction. Therefore, heavy hydrocarbon undergoes mostly conversion reactions
(k8 and k9). The yield profiles of each lump fit well with the experimental data at the end of the
catalytic bed. In the same way, Figure 6.10b shows that when varying space–velocity, the pre-
dicted values for each lump at the reactor bottom agree well with experimental data obtained
at the operating conditions.
Figure 6.11 shows the dynamic profiles of each lump along the catalytic bed length to analyze the
behavior of feedstock and products according to chemical reactions advance. Simulations show that
the Tg reached half of the catalytic bed in a short time (100 s), whereas they are almost completely
converted in 600 s at steady-state conditions. On the other hand, product concentrations increase
according to time-on-stream advances. Even after Tg are fully converted, chemical reactions con-
tinue to take place mainly due to the conversion of heavy hydrocarbons. This behavior agrees with
information obtained during the vegetable oil hydrotreating, which is attributed to the high reac-
tion temperature and acidic nature of the catalyst (Anand and Sinha 2012; Zhang et al. 2014).
7.0E–04
6.0E–04
Triglycerides, mol/cm3
5.0E–04
4.0E–04
3.0E–04
2.0E–04
1.0E–04
0.0E+00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
5.0E–04
4.5E–04
4.0E–04
Heavy, mol/cm3
3.5E–04
3.0E–04
2.5E–04
2.0E–04
1.5E–04
1.0E–04
5.0E–05
0.0E+00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
6.0E–05
5.0E–05
Middle, mol/cm3
4.0E–05
3.0E–05
2.0E–05
1.0E–05
0.0E+00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Bed length, cm
Figure 6.11 Concentration profiles of product pseudocomponents along the catalytic bed at WHSV of 3 h–1:
100 s (— —), 300 s (– –), 600 s (——) and 1200 s (–♦–) (Tirado and Ancheyta 2020a / with permission of
Elsevier).
3.5E–05
3.0E–05
2.5E–05
Light, mol/cm3
2.0E–05
1.5E–05
1.0E–05
5.0E–06
0.0E+00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
9.0E–05
8.0E–05
Oligomerized, mol/cm3
7.0E–05
6.0E–05
5.0E–05
4.0E–05
3.0E–05
2.0E–05
1.0E–05
0.0E+00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Bed length, cm
Figure 6.11 (Continued)
The predicted hydrogen pressure and concentration profiles along the catalytic bed are depicted
in Figure 6.12, whereas the hydrogen concentration decreases rapidly at the reactor top. When the
hydrogen reaction rate decreases, mass transfer at the gas–liquid interface predominates, increas-
ing hydrogen concentration in the liquid phase. This behavior is similar to that reported during
catalytic hydrotreating of gas oil due to the balance between mass transfer and reaction rate
(Mederos et al. 2012). However, the molecular weight and higher complexity of Tg hinder the
hydrogen solubility in vegetable oil. The hydrotreating of vegetable oil consumes large amounts
of hydrogen (1640 scf/bbl) compared with that reported during gas oil hydroprocessing (800 scf/
bbl) (Ancheyta 2011) because of the high demand for hydrogen in the cracking of Tg molecules
(Guzman et al. 2010).
6.8.3 Validation of Hydrotreating Reactor Model with Pilot Plant Data

Once the accuracy of the dynamic model had been verified, the reactor model was scaled up to a
pilot-scale unit size to validate the ability of the model by varying the dimensions of the reactor and
catalytic particle size. The simulation of the pilot-scale reactor models was conducted at isothermal
8.12
Hydrogen pressure, MPa
8.1
8.08
8.06
8.04
8.02
0 0.5 1 1.5 2 2.5 3 3.5
1.0E–04
Hydrogen, mol/cm3
8.0E–05
6.0E–05
4.0E–05
2.0E–05
0.0E+00
0 0.5 1 1.5 2 2.5 3 3.5
Figure 6.12 Hydrogen pressure and concentration in liquid phase along the catalytic bed at WHSV of 12 h–1:
50 s (–––), 100 s (— —), 200 s (– –), 300 s (—).
conditions to compare the predicted yields of each lump. The reactor specifications are summarized
in Table 6.11, whereas the predicted yield profiles along the catalytic bed length at steady-state con-
ditions are shown in Figure 6.13. The products obtained in experimental studies were lumped into
naphtha (<C9 hydrocarbons), kerosene (C9–C14 hydrocarbons), and diesel (C15–C18 + >C18). The
results show that the simulations are in good agreement with the experimental values. Although
the yield profiles seem quite comparable to the results obtained in the bench-scale reactor, there are
indeed some important differences:
• Triglyceride conversion is slightly faster in the pilot-scale reactor than in the bench-scale reactor.
For instance, for a bench-scale reactor, the unconverted Tg yield is 31.82% at 10% of catalytic bed
length, while for a pilot-scale reactor, it is 22.68%.
• The heavy hydrocarbon yield is lower for pilot-scale reactors along the catalytic bed length. This
behavior is attributed to the diffusion limitations suffered by the catalytic particles, especially
reaction pathways 2 and 5, which showed the lowest effectiveness factor values under the studied
operating conditions (0.814 and 0.953, respectively).
• The formation of middle hydrocarbons is preferred in the pilot-scale reactor since the reaction
pathways involved during the production of these pseudocomponents are not influenced by
internal diffusion limitations (effectiveness factor greater than 0.98 in both scales).
100
90
80
70
60 Hv+Ol
Yield, %
50 Hv
40
30 Md
20
Lh
10
Ol
0 Tg
0 10 20 30 40 50 60 70 80 90 100
Catalytic bed length, %
Figure 6.13 Comparison of predicted (lines) and experimental (symbols) yields in isothermal operation
of pilot-scale reactor: Tg (—, ○), Ol (––, ♦), Hv (– –, Δ), Md (– –, ◊), Lh (–––,∗) and Hv+Ol ( , □)
(Tirado and Ancheyta 2020b/American Chemical Society).
To measure the deviation among the predicted yields against experimental data, the mean abso-
lute error (MAE) was calculated. The results indicated that for bench-scale and pilot-scale reactors,
the MAE values were 2.75 and 1.06, respectively. It validates the ability of the model to correctly
predict the behavior of each lump during vegetable oil hydrotreating.
Triglyceride conversion profiles were compared between both reactor scales to compare the
differences between them, and the predicted profiles of the bench-scale reactor and pilot-scale
reactor are illustrated in the figure at the top and bottom, respectively. In addition, catalytic
bed length was normalized for an appropriate comparison of both reactor scales. Simulation
of the pilot-scale reactor was developed at higher WHSV to magnify predicted profiles since,
at WHSV of 1, the triglyceride conversion profiles practically overlap. Dynamic triglyceride con-
centration profiles at 420 C, 8 MPa, and WHSV of 3 h–1 are shown in Figure 6.14. The triglyceride
concentration increases as the feedstock flows through the catalyst particles until the catalytic
bed becomes as wet as possible. Triglyceride concentration profiles shown in the pilot-scale reac-
tor indicate lower values along the catalyst bed than those of the bench-scale reactor, which is
caused by a smaller fraction of liquid present in the pilot-scale reactor (εL = 0.288) compared with
that in bench-scale (εL = 0.450). This parameter directly influences the dynamic term in the
model, and it is affected by the liquid–solid interfacial area, which has a lower value in the
pilot-scale reactor (aS = 11.86) compared with the bench-scale reactor (aS = 126.20), since liquid
holdup is inversely proportional to the equivalent particle diameter as follows (Froment et al.
1990; Shah 1979):
εL = 0 185 ϵB aS 0 333 χ 0 22 6 51
6 1 − ϵB
aS = 6 52
dpe
On the other hand, the smaller the particle size, the higher the liquid holdup by capillary pressure
(Carruthers and DiCamillo 1988; Macías and Ancheyta 2004).
(a)
7.00E–04
6.00E–04
5.00E–04
4.00E–04
3.00E–04
2.00E–04
1.00E–04
0.00E+00
0 20 40 60 80 100
(b)
7.00E–04
6.00E–04
5.00E–04
4.00E–04
3.00E–04
2.00E–04
1.00E–04
0.00E+00
0 20 40 60 80 100
Figure 6.14 Dynamic concentration profiles of triglycerides along the catalytic bed at 420 C, 8 MPa and
WHSV of 3 h–1 (a) bench-scale and (b) pilot-scale: 60 s (— —), 180 s (—— ——), 300 s (– –), 480 s (———)
(Tirado and Ancheyta 2020b/American Chemical Society).
6.8.4 Dynamic Simulation of a Non-isothermal Reactor

To study the variations of the temperature profiles at non-isothermal conditions, the mass and heat
balance equations for each phase are solved simultaneously.
6.8.4.1 Comparison of Non-isothermal Model with Experimental Results in Isothermal Reactor

An appropriate comparison of reactor performance at isothermal and non-isothermal conditions
was performed by calculating the average temperature (Tav) of the liquid phase profile. This param-
eter is determined by integrating the predicted temperature profile in each simulation step as
follows:
N
TN + TN + 1
2
T av = i=0 6 53
N
where TN is the liquid temperature at point N of the catalytic bed length. The results showed that an
inlet temperature of 400 C was required for WHSV between 0.5 and 3 to achieve an average
460 100
LIquid temperature, °C
Triglycerides yield, %
450 80
440
60
430
40
420
20
410
400 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
Catalytic bed length, cm Catalytic bed length, cm
100 30
80
Middle yield, %
Heavy yield, %
20
60
40
10
20
0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
25 20
Oligomerized yield, %
20
15
Light yield, %
15
10
10
5
5
0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
Figure 6.15 Liquid temperature profile and product yields in non-isothermal mode at inlet temperature
of 400 C (solid lines) and isothermal mode at 420 C (dotted lines). Symbols indicate experimental data
at 420 C, 80 bar, WHSV of 3 h–1, and 1500 Nl/l H2/feed ratio. (○) Tg, (♦) Ol, (Δ) Hv, (∗) Md, (◊) Lh (Tirado
et al. 2020 / with permission of Elsevier).
temperature similar to isothermal conditions (420 C). Figure 6.15 shows the yield profiles for each
lump under both operation modes when reaching steady-state conditions. Based on these beha-
viors, the following observations arise:
• The temperature profile of the liquid phase increased rapidly at the reactor top due to the fast
conversion of triglyceride, which is the most exothermic chemical reaction occurring in the proc-
ess. This behavior agrees with information reported by Forghani et al. (2014) during the modeling
of a hydrocracking reactor for triglyceride conversion into green fuel.
1.000 1.000
0.995
0.998 0.990
Effectiveness factor
0.985
0.996 0.980
0.975
0.994 0.970
0.965
0.992 0.960
0 20 40 60 80 100
Figure 6.16 Simulated effectiveness factor profiles for each reaction pathway along the catalytic bed
at inlet temperature of 400 C, 80 bar, and WHSV of 3 h–1. (Δ) η1, (–––) η2, (– –) η3, (○) η4, (□) η5, (◊) η7,
(—) η8, (+) η9, (∗) η10 (Tirado et al. 2020 / with permission of Elsevier).
• Triglyceride presents a lower concentration along the catalytic bed for non-isothermal mode
compared with isothermal operation since the increase in temperature in the reactor top
enhances the conversion rate of the feedstock.
• Heavy hydrocarbon yield under non-isothermal operating conditions is affected by the slight
increase in temperature in the upper part of the reactor, where the hydrocracking reactions
mainly occur.
• The formation of middle and light hydrocarbons from Tg (k3 and k4) has high activation energy
(EA3 = 277.90 kJ/mol and EA3 = 127.77 kJ/mol), which combined with the higher temperature at
the reactor inlet for non-isothermal operation improves the middle hydrocarbon yield
• Production of oligomerized compounds gives a maximum yield of around 31% (1.15 cm) of the
catalytic bed length for non-isothermal mode, while increasing linearly during the isothermal
behavior. The formation of oligomerized products is enhanced due to the high activation energy
of Tg conversion (EA1 = 128.49 kJ/mol).
Figure 6.16 depicts the effectiveness factor profiles for each reaction pathway obtained in the
bench-scale reactor under non-isothermal operating conditions. It is noted that within the catalytic
bed, the impact of internal diffusion limitations is negligible (η > 0.95) because of using powdered
catalysts (Perego and Peratello 1999). The results show that even under a sharp increase in temper-
ature, the internal diffusion limitations are not significant in sprayed catalysts. The lowest value of
the effectiveness factor was indicated for the reaction pathway (η3 = 0.963) in the upper part of the
reactor (7% of the reactor length), while the other reaction pathways present values higher than
0.99. Diminution of effectiveness factors is due to the increase in products that hinder the diffusion
of species by the increase in temperature in the reactor top.
6.8.4.2 Comparison of Bench-Scale and Pilot-Scale Reactor Under Non-isothermal

Operating Condition
Figures 6.17 and 6.18 show the predicted temperature profiles along the catalytic bed for gas and li-
quid phases, respectively. The simulations were developed at similar operation conditions (temper-
ature of 420 C, 8 MPa, and WHSV of 1 h–1) for both reactor sizes. It is observed that the pilot-scale
(a)
440
Gas temperature, °C
435
430
425
420
0 10 20 30 40 50 60 70 80 90 100
(b)
620
580
540
500
460
420
0 10 20 30 40 50 60 70 80 90 100
Figure 6.17 Dynamic gas temperature profiles at non-isothermal conditions for (a) bench-scale and
(b) pilot-reactor scale: ( ) 3 min, (– – –) 10 min, (— —) 30 min, (– –) 60 min, (– –) 120 min, (—) 240 min.
(a)
520
Liquid temperature, °C
500
480
460
440
420
0 10 20 30 40 50 60 70 80 90 100
(b)
700
660
620
580
540
500
460
420
0 10 20 30 40 50 60 70 80 90 100
Figure 6.18 Dynamic liquid temperature profiles at non-isothermal conditions for (a) bench-scale
and (b) pilot-reactor scale: ( ) 3 min, (– – –) 10 min, (— —) 30 min, (– –) 60 min, (– –) 120 min, (—) 240 min.
reactor exhibits greater temperature increment in the liquid phases at the upper zone of the cata-
lytic bed compared with the bench-scale reactor (85 C and 260 C, respectively). This behavior in
the temperature profile of the liquid phase corresponds to the profile reported by Forghani et al.
(2014), and it is associated with the exothermal decomposition of Tg. This statement agrees with
the temperature profile behavior as zero levels of triglyceride have been detected in the latter half
of the catalytic bed, resulting in a decrease in temperature and subsequent stabilization.
However, the results are different from those reported for the hydrotreating of straight-run petro-
leum distillates (Guo et al. 2008; Mederos et al. 2012) and diluted triglyceride (Muharam et al. 2017;
Muharam and Putri 2018) feedstocks, obtaining smaller increments in the temperature reactor
(15–30 C). On the other hand, the temperature profile of the gas increases along the catalytic
bed by transfer of heat between the gas–liquid interfaces. Solid-phase temperature profiles are
not shown for bench-scale reactors; however, they behave similarly to liquid-phase profiles show-
ing negligible resistance at the liquid–solid interface, while for the pilot-scale reactor, a small dif-
ference was obtained (2 C). This resistance to energy transfer is caused by the dimensions of the
commercial catalyst used. The temperature profiles of the pilot-scale reactor exhibited noticeable
increments in each stage, with the potential for harmful reactions, reactor and particle damages,
and difficulties to reach stable temperatures in the latter half of the catalytic bed. Moreover, the
abrupt rises in temperature pose challenges in achieving a steady state due to increased feedstock
flows and the catalyst amount employed.
The results of the simulations of vegetable oil hydrotreating under non-isothermal conditions
provide a clear idea of the temperature behavior of each phase along the catalytic bed. To carry
out the process on a commercial scale, it is crucial to establish adequate temperature control. This
will help prevent hot spots in the reactor (Alvarez and Ancheyta 2012; Tirado and Ancheyta 2020b).
6.8.5 Dynamic Simulation of an Adiabatic Commercial Reactor

The reactor model for hydrotreating was focused on the design of a commercial-scale unit to exam-
ine the effect of the characteristics of the reactor as well as control of temperature and yield of the
desired reaction products. The characteristics of the commercial-scale reactor were taken from an
existing hydrotreating unit of petroleum-derived streams (Mederos et al. 2012; Rodriguez and
Ancheyta 2004). The set of PDEs that constitute the reactor model was modified to provide a real-
istic industrial environment. Figure 6.19 presents the mass balance equations of the three-phase
system, including the thermal effect, hydrogen consumption, product gas formation, and superfi-
cial velocity of the gas phase (uG).
Similar to mass transfer phenomena, the heat balance equations to simulate the adiabatic oper-
ation in commercial units and quench stream equations for temperature control are presented in
Figure 6.20. The necessary mass flow rate of cold hydrogen quench gas (q) to decrease the temper-
ature of the hot flow of gas (g) and liquid (l) from the previous bed is calculated through an energy
balance in the quench zone, where Tq, Tin, and Tout are the quench, the inlet, and the outlet tem-
peratures of the quench zone, respectively (Alvarez et al. 2007; Alvarez and Ancheyta 2008).
The dynamic reactor model was solved with the kinetic parameters obtained from the bench-
scale unit. The high liquid velocities encountered in commercial units provide a complete wetting
efficiency and negligible hydrodynamic effects.
Reference data were obtained through a simulation of the commercial hydroprocessing unit
without temperature control. The high heat release that was presented in laboratory-scale reactors
indicates that the operating conditions must be carefully selected to have an operating range with-
out exceeding a permissible temperature limit. For this reason, mild operating conditions have been
Compounds in gas phase

εG 𝝏pG
i 1 𝝏(uG pG
i
) pG
i
= KLi aL CLi
RTG 𝝏t RTG 𝝏z Hi
i = H2, H2O,CO2,CO
𝝏uG RTG εGP 𝝏TG pG uG 𝝏TG

𝝏z
=
P RT 2G 𝝏t
∑ KLi aL i
Hi
CLi +
TG 𝝏Z
i
Gas Compounds in solid phase

NRL
𝝏CS
∈S (1–∈B)
𝝏t
i L
= fwKiS as (Ci – Ci ) +
S
∑ L
ηj υi,j ri,j
Catalyst j=1
i = H2, H2O, CO2, CO, Tg, Ol, Hv, Md, Lh
Gaseous compounds in liquid phase Liquid

𝝏CLi 𝝏CLi pG L S
Solid
i
εL = –uL + KLi aL CLi fwKiS as (Ci – Ci )
𝝏t 𝝏z Hi
i = H2, H2O,CO2,CO
Compounds in liquid phase

𝝏CLi 𝝏CLi L S
εL = –uL KiS as (Ci – Ci )
𝝏t 𝝏z
i = Tg, Ol, Hv, Md, Lh
Figure 6.19 Mass balance equations of the commercial-scale fixed-bed reactor model (Tirado et al. 2022 /
with permission of Elsevier).
Liquid phase
𝜕TL 𝜕TL
εLρLCpL = –uLρLCpL – hGLaL(TG – TL)
𝜕t 𝜕z
Gas phase – fwhLSaS(TL – TS)
𝜕TG 𝜕TG
εGρGCpG = –mGCpG – hGLaL(TG – TL)
𝜕t 𝜕z gout lout
Solid phase
NRL
𝜕TS
= fwhLSaS(TL – TS) +∑ ηj υi,j ri,j(– ΔHR)
L
εSρSCpS
𝜕t j=1
Quench stream, q
T T
g ʃT in CpG dT + l ʃT in CpL dT
out out
q= –
ʃ Tin Cp dT
Tq q
gin lin
q + lout + gout = lin + gin
Figure 6.20 Heat balance equations of the commercial-scale fixed-bed reactor model (Tirado et al. 2022 /
with permission of Elsevier).
(a)
520
720 s
490 600 s
480 s
460
430 300 s
400
180 s
370
60 s
340
(b)
640
590
540
1200 s
720 s
490 600 s
480 s
300 s
440 180 s
390
60 s
340
0 10 20 30 40 50 60 70 80 90 100
Figure 6.21 Dynamic temperature profiles along the catalytic bed length for (a) gas phase and (b) liquid
phase at inlet temperature of 340 C, 8 MPa, WHSV of 3 h−1, and 1500 Nl/l oil H2/oil ratio (Adapted from Tirado
et al. 2021).
chosen for the commercial unit, as an increase in the feedstock inlet temperature will also increase
the reaction temperature and cause the system to overheat (Alvarez and Ancheyta 2012; Fan et al.
2020; Tirado et al. 2020). Based on this information, the simulation was carried out under the fol-
lowing operating conditions: inlet temperature of the feed of 340 C, WHSV of 3 h–1, 8 MPa, and
1500 Nl/l oil H2/oil ratio. Figure 6.21 depicts the temperature profiles of the gas and liquid phases
along the catalytic bed. A sharp rise in temperature is observed from the start-of-run, especially in
the liquid phase, considering that at 180 s a temperature of 470 C was reached in the upper part
of the reactor, while at steady-state conditions (1200 s), the maximum temperature reached was
590 C. The variation in gas velocity influences the predicted temperature profiles since results
obtained when simulating a pilot-scale reactor show higher temperature profiles and longer times
to reach a steady state. On the other hand, the temperature difference calculated is higher than
those reported during the hydrotreating of oil-derived streams (Alvarez-Majmutov and Chen
2014; Fan et al. 2020; Mederos et al. 2012).
The Preem Gothenburg refinery reported that during the co-processing of straight-run gas oils
(SRGO) with 30 vol.% of RTD, the heat released increased; while in the absence of RTD, the
Delta-T was about 30 C, it was around 60 C when the feedstock contained 30 vol.% of RTD
(Egeberg et al. 2011; Egeberg et al. 2009). Therefore, for 100% renewable feedstock, higher
temperature differences can be expected as those obtained in this study. Even under non-severe
conditions, triglyceride conversion causes a high heat release that raises the temperature of the liq-
uid phase. It increases the feedstock conversion rate even to exceed the temperature under steady-
state conditions as observed at 300 s. In addition, the reduction of the triglyceride concentration and
the energy transfer at the gas–liquid phase reduces the heat released in the second half of the reac-
tor. Temperature profiles show that the system begins to overheat before the steady-state conditions
are reached. It highlights the importance of a dynamic model capable of establishing the appropri-
ate configuration of catalytic beds and the start-up time of quench injections. For industrial prac-
tice, a vegetable oil hydrotreating unit should be designed and operated in such a way that the
temperature rise is controlled from the start-of-run.
6.8.5.1 Configuration of Hydrogen Quenching

In order to establish strategies for temperature control in a hydrotreating reactor of residual oil,
Alvarez and Ancheyta carried out a series of simulations varying the operating conditions and set-
tings for temperature control in the reactor. The simulations indicated that quenching must be
started before or as soon as the hydrocarbon front reaches the quench zones to avoid overheating
and to keep the temperature under control. The results showed the temperature control was ade-
quate, as well as the hydrogen consumption, and sour gas emission. Thus, control strategies are
developed for different hydroprocessing technologies (Alvarez and Ancheyta 2012). Based on infor-
mation reported by Alvarez and Ancheyta for the temperature control of a residue oil hydrodeme-
tallization unit, a series of strategies are considered for the configuration of hydrogen quench
injections along the axial direction of the reactor. The sharp temperature rise and the high velocity
of the liquid flow through the catalytic bed were the main issues found in the development of
effective configurations, with a mean residence time of 456 s given by the interstitial velocity
uL = uL εL . A series of simulations (case 1) where the quenching injections start as soon as
the liquid temperature reaches the maximum permissible temperature were developed.
Figure 6.22 shows the liquid-phase profiles with quenching injections upon reaching the set limit
temperature of 400 C. The catalyst bed was divided into five catalytic beds with four quench zones
to distribute the total heat released below the permissible limit. The location along the catalytic bed
length and start time for each quench injection are shown in Table 6.13. It is observed that the liq-
uid-phase temperature gradually increases until cooling occurs, and the steady state is reached.
Even after rapid cooling, some profiles show transient temperature variations that do not exceed
the allowable limit. However, the heat released at the top of the reactor causes the length of the
catalytic beds to become shorter as it approaches the inlet of the reactor.
In a second case study, the cooling injections started simultaneously when the liquid-phase tem-
perature reached the allowable limit (400 C) at the first quenching zone, to examine the change in
the behavior of the temperature of the liquid phase. Figure 6.22 shows that the temperature profiles
of the liquid phase are kept under control. This configuration allows reducing the Delta-T in the
third and fourth catalytic beds (from 60 to 40 C) and changes the location of the fourth quench
zone, as shown in Table 6.13. However, starting quenching injections when the temperature has
not yet risen causes a sharp cooling of the catalytic bed areas and longer times to reach steady state
(1200 s).
6.8.5.2 Liquid-Phase Yields and Gas Composition

The yields of products in the liquid phase and gas compositions were analyzed based on the con-
figuration of case 1, since it seems to be the most suitable strategy for temperature control in this
type of process. Figure 6.23a depicts the yield profiles of the liquid pseudocomponents along the
(a)
420
400
300 s 480 s 600 s

380
180 s 720 s
360
60 s
340
(b) 420
400 1200 s
380 720 s
600 s
360 480 s
180 s 300 s
340
320
300
280
0 10 20 30 40 50 60 70 80 90 100
Figure 6.22 Dynamic temperature profiles of liquid phase with quench injection (a) starting when the liquid
temperature reaches the set limit temperature (400 C) and (b) starting simultaneously when the feedstock
is fed to the reactor (Tirado et al. 2022 / with permission of Elsevier).
Table 6.13 Hydrogen quenching configurations for each quenching zone.
Configuration Injection Localization (cm) Injection time (s)
Case 1 1 73.15 (8.57%) 95

2 158.49 (18.57%) 150
3 292.6 (34.28%) 325
4 634 (74.28%) 620
Case 2 1 73.15 (8.57%) 95
2 158.49 (18.57%) 95
3 292.6 (34.28%) 95
4 463.3 (54.28%) 95
Numbers in parenthesis indicate the percentage of the catalytic bed from top to bottom of the reactor.
(a)
10 100
Tg
8 80
Hv
6 60
Yield, %
Ol
4 40
Lh
2 20
Md
0 0
(b) 100 2
98 H2 1.8
Molar composition, %
96 1.6
94 1.4
92 1.2
90 H2O 1
88 0.8
86 C3H8 0.6
CO2
84 0.4
82 CO 0.2
C4H10
80 0
0 10 20 30 40 50 60 70 80 90 100
Figure 6.23 Profiles of (a) pseudocomponents yields in the liquid phase and (b) molar composition
of each component in the gas phase along the catalytic bed under steady state at inlet temperature of
340 C, 8 MPa, WHSV of 3 h−1, and 1500 Nl/l oil H2/oil ratio (Tirado et al. 2022 / with permission of Elsevier).
catalytic bed. The yield of unconverted Tg decreases until a conversion of 98% is reached at the
outlet of the reactor, while the yield of heavy hydrocarbons makes it possible to obtain a 90% yield
of diesel-type fuel (C15–C18). On the other hand, low-boiling-point fuels (middle and light hydro-
carbon) gave lower yields because of the mild severity conditions. The oligomerized compounds are
the most temperature-sensitive lumps due to the stepwise increase throughout the reactor.
Table 6.14 shows that the yields at the outlet of the reactor follow a trend with the data measured
with a catalyst powdered in an isothermal bench-scale reactor, despite the variations in tempera-
ture and intraparticle effects at the industrial scale.
Figure 6.23b shows the molar composition profiles of each gaseous compound in a steady state.
The gas composition at the inlet of the reactor is pure hydrogen, because the evaporation effects of
the vegetable oil are negligible, and the feed gas does not contain lighter hydrocarbons. It tends to
decrease along the reactor by the hydrogen consumption and product gas formation. However,
staggered fluctuations occur at points where the quench zones are located due to an increase in
hydrogen composition and a decrease in liquid phase temperature that diminishes the hydrogen
6.9 Conclusions 235
Table 6.14 Comparison of experimental data with calculated yields.
Experimental
Pseudocomponent 340 C 420 C Predicted
Tg unconverted 72.75 0.05 2.85

Ol (>C18) 7.65 9.49 3.45
Hv (C15–C18) 19.58 70.19 90.32
Md (C9–C14) 0.01 11.96 1.16
Lh (<C9) 0.01 8.32 2.22
solubility in the vegetable oil. Therefore, the dissolved hydrogen in the liquid phase is transferred to
the gas phase. Under the studied configuration, the hydrogen consumption calculation was 353.15
Nm3/m3, which agrees with the information reported for the hydroprocessing of renewable feed-
stock in industrial units with values of 300–400 Nm3/m3 (Egeberg et al. 2009; Guzman et al. 2010;
Verdier et al. 2019), while the amount of hydrogen for hydroprocessing of petroleum-derived
streams is 50–150 Nm3/m3 (Ancheyta 2011). Therefore, higher make-up hydrogen and quench
gas streams are needed even when co-processing quite small amounts of vegetable oil.
The results show that CO2 and H2O are the main product gases formed because HDO and decar-
boxylation are the predominant reactions, while CO has a lower composition. C3H8 continuously
increases throughout the catalytic bed due to the triglyceride conversion. On the other hand, C4H10
formation is insignificant because the hydrocracking reaction does not take place for a long extent.
6.9 Conclusions
During the development of kinetic models for the generation of renewable fuels, it is necessary
to consider diverse experimental and mathematical assumptions to obtain intrinsic kinetics.
Otherwise, it is appropriate to consider that the kinetic parameters are affected by transfer lim-
itations at the interfaces of the reactant system. Mass and heat transfer effects at the gas–liquid
and solid–liquid interfaces, as well as inside the catalyst pores, affect the experimental data.
Therefore, the influence of these effects on kinetic parameters and the development of reactor
models are evaluated. The selection of a reaction scheme is crucial to obtain relevant informa-
tion from the system. In addition, the optimization of kinetic parameters through complex non-
linear algorithms, mathematical constraints, and verification of the results through statistical
analysis is necessary.
Different approaches have been developed to analyze and predict catalytic decay during hydro-
treating feedstock capable of producing renewable fuels. However, the analyses of these approaches
with feedstocks are still scarce and require further investigation.
The development and scaling of reactor models involve hydrodynamic phenomena related to
plug-flow pattern, catalyst wetting efficiency, wall effects, and internal catalyst diffusion. These
effects can be determined by analytical criteria in order to achieve adequate prediction of the
diverse components in the reaction system during the scale-up of reactor models and their valida-
tion with experimental data.
The predicted temperature profiles depict an increase in the upper part of the catalytic bed attrib-
uted to triglyceride conversion, while the second half of the catalytic bed undergoes thermal sta-
bilization. This increase in temperature enhances the decomposition of Tg and promotes the
hydrocracking reactions. Simulations performed for industrial-scale reactors show high tempera-
ture rises along the catalytic bed are unacceptable during the operation of a commercial hydropro-
cessing unit. Therefore, the operating conditions of a renewable feedstock at a commercial-scale
hydroprocessing unit must be carefully selected to control the temperature rise throughout the
reactor. These results demonstrate the necessity of using quenching streams in the reactor to con-
trol the exothermicity of reactions. Different configurations were simulated to control the reactor
temperature and the reaction product yields. From a practical point of view, the results show that
during the hydroprocessing of renewable feedstock, the consumption of hydrogen and heat
released are much higher than those reported during petroleum distillates HDT. A greater focus
on developing mathematical models is required to optimize some process aspects. Hydrogen con-
sumption and quenching configurations are crucial for the choice of high-quality renewable fuels.
In addition, it is possible to configure the location and the start-up time of the quenching zones to
distribute the high heat of reaction throughout the reactor without exceeding a predetermined tem-
perature control.
References
Alcázar, L.A. and Ancheyta, J. (2007). Sensitivity analysis based methodology to estimate the best set of
parameters for heterogeneous kinetic models. Chemical Engineering Journal 128: 85–93.
Alvarez, A. and Ancheyta, J. (2008). Simulation and analysis of different quenching alternatives for an
industrial vacuum gasoil hydrotreater. Chemical Engineering Science 63: 662–673.
Alvarez, A. and Ancheyta, J. (2012). Transient behavior of residual oil front-end hydrodemetallization in
a trickle-bed reactor. Chemical Engineering Journal 197: 204–214.
Alvarez, A., Ancheyta, J., and Muñoz, J.A.D. (2007). Comparison of quench systems in commercial fixed-
bed hydroprocessing reactors. Energy and Fuels 21: 1133–1144.
Alvarez-Majmutov, A. and Chen, J. (2014). Modeling and simulation of a multibed industrial
hydrotreater with vapor-liquid equilibrium. Industrial and Engineering Chemistry Research 53:
10566–10575.
Anand, M. and Sinha, A.K. (2012). Temperature-dependent reaction pathways for the anomalous
hydrocracking of triglycerides in the presence of sulfided Co-Mo-catalyst. Bioresource Technology 126:
148–155.
Anand, K., Ranjan, A., and Mehta, P.S. (2010). Predicting the density of straight and processed vegetable
oils from fatty acid composition. Energy and Fuels 24: 3262–3266.
Anand, M., Farooqui, S.A., Kumar, R. et al. (2016a). Kinetics, thermodynamics and mechanisms for
hydroprocessing of renewable oils. Applied Catalysis A: General 516: 144–152.
Anand, M., Farooqui, S.A., Kumar, R. et al. (2016b). Optimizing renewable oil hydrocracking conditions
for aviation bio-kerosene production. Fuel Processing Technology 151: 50–58.
Ancheyta, J. (2011). Modeling and Simulation of Catalytic Reactors for Petroleum Refining. New
Jersey: Wiley.
Ancheyta, J., Trejo, F., and Rana, M.S. (2010). Chemical Transformation During Hydroprocessing of Heavy
Oils. Boca Raton: CRC Press.
Arora, P., Grennfelt, E.L., Olsson, L., and Creaser, D. (2019). Kinetic study of hydrodeoxygenation of
stearic acid as model compound for renewable oils. Chemical Engineering Journal 364: 376–389.
References 237
Ayodele, O.B., Farouk, H.U., Mohammed, J. et al. (2015). Hydrodeoxygenation of oleic acid into n- and
iso-paraffin biofuel using zeolite supported fluoro-oxalate modified molybdenum catalyst: kinetics
study. Journal of the Taiwan Institute of Chemical Engineers 50: 142–152.
Baldiraghi, F., Di Stanislao, M., Faraci, G. et al. (2009). Ecofining: new process for green diesel production
from vegetable oil. Sustainable Industrial Chemistry 427–438.
Bezergianni, S. and Dimitriadis, A. (2013). Comparison between different types of renewable diesel.
Renewable and Sustainable Energy Reviews 21: 110–116.
Bhaskar, M., Valavarasu, G., Sairam, B. et al. (2004). Three-phase reactor model to simulate the
performance of pilot-plant and industrial trickle-bed reactors sustaining hydrotreating reactions.
Bie, Y., Kanervo, J.M., and Lehtonen, J. (2015). Hydrodeoxygenation of methyl heptanoate over Rh/ZrO2
catalyst as a model reaction for biofuel production: kinetic modeling based on reaction mechanism.
Industrial and Engineering Chemistry Research 54: 11986–11996.
Bie, Y., Lehtonen, J., and Kanervo, J. (2016). Hydrodeoxygenation (HDO) of methyl palmitate over
bifunctional Rh/ZrO2 catalyst: Insights into reaction mechanism via kinetic modeling. Applied
Catalysis A: General 526: 183–190.
Bischoff, K.B. and Levenspiel, O. (1962). Fluid dispersion-generalization and comparison of
mathematical models-II comparison of models. Chemical Engineering Science 17: 257–264.
Boda, L., Onyestyák, G., Solt, H. et al. (2010). Catalytic hydroconversion of tricaprylin and caprylic acid as
model reaction for biofuel production from triglycerides. Applied Catalysis A: General 374: 158–169.
Boldrini, D.E., Damiani, D.E., and Tonetto, G.M. (2014). Mathematical modeling of the partial
hydrogenation of vegetable oil in a monolithic stirrer reactor. AlChe 60: 3524–3533.
Bykova, M.V., Zavarukhin, S.G., Trusov, L.I., and Yakovlev, V.A. (2013). Guaiacol hydrodeoxygenation
kinetics with catalyst deactivation taken into consideration. Kinetics and Catalysis 54: 40–48.
Cabrera, M.I. and Grau, R.J. (2008). Advanced concepts for the kinetic modeling of fatty acid methyl
esters hydrogenation. International Journal of Chemical Reactor Engineering 6: 1–37.
Carruthers, J.D. and DiCamillo, D.J. (1988). Pilot plant testing of hydrotreating catalysts. Influence of
catalyst condition, bed loading and dilution. Applied Catalysis 43: 253–276.
Chen, A.H., Mclntire, D.D., and Allen, R.R. (1981). Modeling of reaction rate constants and selectivities in
soybean oil hydrogenation. Journal of the American Oil Chemists’ Society 58: 816–818.
Chen, N., Gong, S., Shirai, H. et al. (2013). Effects of Si/Al ratio and Pt loading on Pt/SAPO-11 catalysts in
hydroconversion of Jatropha oil. Applied Catalysis A: General 466: 105–115.
Donnis, B., Egeberg, R.G., Blom, P., and Knudsen, K.G. (2009). Hydroprocessing of bio-oils and
oxygenates to hydrocarbons. Understanding the reaction routes. Topics in Catalysis 52: 229–240.
Douvartzides, S.L., Charisiou, N.D., Papageridis, K.N., and Goula, M.A. (2019). Green diesel: biomass
feedstocks, production technologies, catalytic research, fuel properties and performance in
compression ignition internal combustion engines. Energies 12: 809.
Egeberg, R.G., Michaelsen, N.H., and Skyum, L. (2009). Novel hydrotreating technology for production of
green diesel. Technical Info Haldor Topsøe 1–20.
Egeberg, R., Knudsen, K., Nyström, S. et al. (2011). Industrial-scale production of renewable diesel.
Petroleum Technology Quarterly 16: 1–7.
El-Hisnawi, A.A., Dudukovic, M.P., and Mills, P.L. (1982). Trickle-bed reactors: dynamic tracer tests,
reaction studies, and modeling of reactor performance. ACS Symposium Series 421–440.
Fan, X., Li, D., Feng, X. et al. (2020). Modelling and simulation of industrial trickle bed reactor
hydrotreating for whole fraction low-temperature coal tar simultaneous hydrodesulfurisation and
hydrodenitrification. Fuel 269: 117362.
Félix, G., Ríos, J.J., Tirado, A. et al. (2022). Monte carlo and sensitivity analysis methods for kinetic parameters
optimization: application to heavy oil slurry-phase hydrocracking. Energy and Fuels 36: 9251–9260.
Félix, G., Tirado, A., Al-Muntaser, A. et al. (2023). Catalytic mechanism and kinetics. In: Catalytic In-situ
Upgrading of Heavy and Extra-Heavy Crude Oils (ed. M.A. Varfolomeev, C. Yuan, and J. Ancheyta),
309–381. New Jersey: Wiley.
Fillion, B. and Morsi, B.I. (2000). Gas-liquid mass-transfer and hydrodynamic parameters in a soybean oil
hydrogenation process under industrial conditions. Industrial and Engineering Chemistry Research 39:
2157–2168.
Fillion, B., Morsi, B.I., Heier, K.R., and Machado, R.M. (2002). Kinetics, gas−liquid mass transfer, and
modeling of the soybean oil hydrogenation process. Industrial Engineering Chemistry Research 41:
697–709.
Forghani, A.A. and Lewis, D.M. (2016). Hydro-conversion of oleic acid in bio-oil to liquid hydrocarbons:
an experimental and modeling investigation. Journal of Chemical Technology and Biotechnology 91:
655–663.
Forghani, A.A., Jafarian, M., Pendleton, P., and Lewis, D.M. (2014). Mathematical modelling of a
hydrocracking reactor for triglyceride conversion to biofuel: model establishment and validation.
International Journal of Energy Research 38: 1624–1634.
Froment, G.F., Bischoff, K.B., and De Wilde, J. (1990). Chemical Reactor—Analysis and Design, 3rde.
Belgium: Wiley.
Furimsky, E. (2013). Hydroprocessing challenges in biofuels production. Catalysis Today 217: 13–56.
Gierman, H. (1988). Design of laboratory hydrotreating reactors scaling down of trickle-flow reactors.
Applied Catalysis 43: 277–286.
Goto, S. and Smith, J.M. (1975). Trickle-bed reactor performance. AIChE Journal 21: 706–713.
Grilc, M., Likozar, B., and Levec, J. (2014a). Hydrodeoxygenation and hydrocracking of solvolysed
lignocellulosic biomass by oxide, reduced and sulphide form of NiMo, Ni, Mo and Pd catalysts. Applied
Catalysis B: Environmental 150–151: 275–287.
Grilc, M., Likozar, B., and Levec, J. (2014b). Hydrotreatment of solvolytically liquefied lignocellulosic
biomass over NiMo/Al2O3 catalyst: reaction mechanism, hydrodeoxygenation kinetics and mass
transfer model based on FTIR. Biomass and Bioenergy 63: 300–312.
Guo, J., Jiang, Y., and Al-Dahhan, M.H. (2008). Modeling of trickle-bed reactors with exothermic
reactions using cell network approach. Chemical Engineering Science 63: 751–764.
Guzman, A., Torres, J.E., Prada, L.P., and Nuñez, M.L. (2010). Hydroprocessing of crude palm oil at pilot
plant scale. Catalysis Today 156: 38–43.
Hachemi, I. and Murzin, D.Y. (2018). Kinetic modeling of fatty acid methyl esters and triglycerides
hydrodeoxygenation over nickel and palladium catalysts. Chemical Engineering Journal 334:
2201–2207.
Hasanudin, H., Rachmat, A., Said, M., and Wijaya, K. (2019). Kinetic model of crude palm oil
hydrocracking over Ni/Mo ZrO2–pillared bentonite Catalyst. Periodica Polytechnica Chemical
Engineering 64: 238–247.
Hermida, L., Abdullah, A.Z., and Mohamed, A.R. (2015). Deoxygenation of fatty acid to produce diesel-
like hydrocarbons: a review of process conditions, reaction kinetics and mechanism. Renewable and
Sustainable Energy Reviews 42: 1223–1233.
Hickman, D.A., Weidenbach, M., and Friedhoff, D.P. (2004). A comparison of a batch recycle reactor and
an integral reactor with fines for scale-up of an industrial trickle bed reactor from laboratory data.
Holser, R.A., List, G.R., King, J.W. et al. (2002). Modeling of hydrogenation kinetics from triglyceride
compositional data. Journal of Agricultural and Food Chemistry 50: 7111–7113.
Huber, G.W., O’Connor, P., and Corma, A. (2007). Processing biomass in conventional oil refineries:
production of high quality diesel by hydrotreating vegetable oils in heavy vacuum oil mixtures. Applied
References 239
Jeništová, K., Hachemi, I., Mäki-Arvela, P. et al. (2017). Hydrodeoxygenation of stearic acid and tall oil
fatty acids over Ni-alumina catalysts: Influence of reaction parameters and kinetic modelling.
Chemical Engineering Journal 316: 401–409.
Kalnes, T.N., Marker, T., Shonnard, D.R., and Koers, K.P. (2008). Green diesel production by
hydrorefining renewable feedstocks. Biofuels Technology 2: 7–11.
Kiran, G.A.R. (2015). CFD Modeling and Simulations of Catalytic Hydrotreatment of Bio-Oil. Guwahati.
Korsten, H. and Hoffmann, U. (1996). Three-phase reactor model for hydrotreating in pilot trickle-bed
reactors. AIChE Journal 42: 1350–1360.
Kubička, D. and Kaluža, L. (2010). Deoxygenation of vegetable oils over sulfided Ni, Mo and NiMo
catalysts. Applied Catalysis A: General 372: 199–208.
Kubička, D. and Tukač, V. (2013). Hydrotreating of triglyceride-based feedstocks in refineries. Advances
in Chemical Engineering 42: 141–194.
Kubička, D., Šimáček, P., and Žilkova, N. (2009). Transformation of vegetable oils into hydrocarbons over
mesoporous-alumina-supported CoMo catalysts. Topics in Catalysis 52: 161–168.
Kubičková, I. and Kubička, D. (2010). Utilization of triglycerides and related feedstocks for production of
clean hydrocarbon fuels and petrochemicals: a review. Waste and Biomass Valorization 1:
293–308.
Kumar, P., Yenumala, S.R., Maity, S.K., and Shee, D. (2014). Kinetics of hydrodeoxygenation
of stearic acid using supported nickel catalysts: effects of supports. Applied Catalysis A: General 471:
28–38.
Lange, R., Schubert, M., Dietrich, W., and Grünewald, M. (2004). Unsteady-state operation of trickle-bed
reactors. Chemical Engineering Science 59: 5355–5361.
Macías, M.J. and Ancheyta, J. (2004). Simulation of an isothermal hydrodesulfurization small reactor
with different catalyst particle shapes. Catalysis Today 98: 243–252.
Mederos, F.S., Rodriguez, M.A., Ancheyta, J., and Arce, E. (2006). Dynamic modeling and simulation of
catalytic hydrotreating reactors. Energy and Fuels 20: 936–945.
Mederos, F.S., Ancheyta, J., and Chen, J. (2009a). Review on criteria to ensure ideal behaviors in trickle-
bed reactors. Applied Catalysis A: General 355: 1–19.
Mederos, F.S., Elizalde, I., and Ancheyta, J. (2009b). Steady-state and dynamic reactor models for
hydrotreatment of oil fractions: a review. Catalysis Reviews 51: 485–607.
Mederos, F.S., Ancheyta, J., and Elizalde, I. (2012). Dynamic modeling and simulation of hydrotreating of
gas oil obtained from heavy crude oil. Applied Catalysis A: General 425: 13–27.
Mederos, F.S., Elizalde-Martínez, I., Hernández-Altamirano, R. et al. (2019). Hydrotreating model
comparison of raw castor oil and its methyl esters for biofuel production. Chemical Engineering and
Mederos, F.S., Elizalde-Martínez, I., Trejo-Zárraga, F. et al. (2020). Dynamic modeling and simulation of
three-phase reactors for hydrocracking of vegetable oils. Reaction Kinetics, Mechanisms and Catalysis
131: 613–644.
Mederos, F.S., Santoyo-López, A.O., Hernández-Altamirano, R. et al. (2021). Renewable fuels production
from the hydrotreating over NiMo/γ-Al2O3 catalyst of castor oil methyl esters obtained by reactive
extraction. Fuel 285: 119168.
Mendoza Sépulveda, C.C. (2013). Simulación CFD de la transferencia de calor en un reactor de
hidrotratamiento de aceites vegetales de segunda generación. Medellín.
Mohammad, M., Kandaramath Hari, T., Yaakob, Z. et al. (2013). Overview on the production of paraffin
based-biofuels via catalytic hydrodeoxygenation. Renewable and Sustainable Energy Reviews 22:
121–132.
Mortensen, P.M., Grunwaldt, J., Jensen, P.A. et al. (2011). Applied catalysis A : general A review of
catalytic upgrading of bio-oil to engine fuels. Applied Catalysis A, General 407: 1–19.
Muharam, Y. and Adevia, R.T. (2018). Modelling and simulation of a slurry bubble column reactor for
green fuel production via hydrocracking of vegetable oil. Web of Conferences 67: 1–6.
Muharam, Y. and Putri, A.D. (2018). Simulation of hydrotreating of vegetable Oil in a slurry bubble
column reactor for green diesel production. International Journal of Technology 9: 1168–1177.
Muharam, Y., Nugraha, O.A., and Leonardi, D. (2017). Modelling of a hydrotreating reactor to produce
renewable diesel from non-edible vegetable oils. Chemical Engineering Transactions 56: 1561–1566.
Noriega, A.K., Tirado, A., Méndez, C. et al. (2020). Hydrodeoxygenation of vegetable oil in batch reactor:
experimental considerations. Chinese Journal of Chemical Engineering 28: 1670–1683.
Pedroza, M.A.P., Segtovich, I.S.V., Sermoud, V.d.M., and da Silva, M.A.P. (2022). Hydrodeoxygenation of
stearic acid to produce diesel–like hydrocarbons: kinetic modeling, parameter estimation and
simulation. Chemical Engineering Science 254: 117576.
Pelemo, J., Inambao, F.L., and Onuh, E.I. (2020). Potential of used cooking oil as feedstock for
hydroprocessing into hydrogenation derived renewable diesel: A review. International Journal of
Engineering Research and Technology 13: 500–519.
Perego, C. and Peratello, S. (1999). Experimental methods in catalytic kinetics. Catalysis Today 52:
133–145.
Rodenbush, C.M., Hsieh, F.H., and Viswanath, D.S. (1999). Density and viscosity of vegetable oils.
Journal of the American Oil Chemists’ Society 76: 1415–1419.
Rodriguez, M.A. and Ancheyta, J. (2004). Modeling of hydrodesulfurization (HDS),
hydrodenitrogenation (HDN), and the hydrogenation of aromatics (HDA) in a vacuum gas oil
hydrotreater. Energy and Fuels 18: 789–794.
Saidi, M., Samimi, F., Karimipourfard, D. et al. (2014). Upgrading of lignin-derived bio-oils by catalytic
hydrodeoxygenation. Energy and Environmental Science 7: 103–129.
Sámano, V., Tirado, A., Félix, G., and Ancheyta, J. (2020). Revisiting the importance of appropriate
parameter estimation based on sensitivity analysis for developing kinetic models. Fuel 267: 117113.
Satterfield, C.N. (1975). Trickle-bed reactors. AIChE Journal 21: 209–228.
Scharff, Y., Asteris, D., and Fédou, S. (2013). Catalyst technology for biofuel production: Conversion of
renewable lipids into biojet and biodiesel. OCL—Oilseeds and Fats, Crops and Lipids 20: 2–5.
Schiesser, W.E. (1991). The numerical method of lines: integration of partial differential equations. Texas:
Elsevier.
Sebos, I., Matsoukas, A., Apostolopoulos, V., and Papayannakos, N. (2009). Catalytic hydroprocessing of
cottonseed oil in petroleum diesel mixtures for production of renewable diesel. Fuel 88: 145–149.
Selishcheva, S.A., Lebedev, M.Y., Reshetnikov, S.I. et al. (2014). Kinetics of the hydrotreatment of
rapeseed oil fatty acid triglycerides under mild conditions. Catalysis in Industry 6: 60–66.
Shah, Y. (1979). Gas-Fluid-Solid Reactor Design. New York: McGraw-Hill International Book Co.
Shamanaev, I.V., Deliy, I.V., Aleksandrov, P.V. et al. (2019). Methyl palmitate hydrodeoxygenation over
silica-supported nickel phosphide catalysts in flow reactor: experimental and kinetic study. Journal of
Chemical Technology Biotechnology 94: 3007–3019.
Sharma, R.K., Anand, M., Rana, B.S. et al. (2012). Jatropha-oil conversion to liquid hydrocarbon fuels
using mesoporous titanosilicate supported sulfide catalysts. Catalysis Today 198: 314–320.
Sinha, A.K., Sibi, M.G., Naidu, N. et al. (2014). Process intensification for hydroprocessing of vegetable
oils: experimental study. Industrial Engineering Chemistry Research 53: 19062–19070.
Sivasamy, A., Cheah, K.Y., Fornasiero, P. et al. (2009). Catalytic applications in the production of
biodiesel from vegetable oils. ChemSusChem 2: 278–300.
Snåre, M., Kubičková, I., Mäki-Arvela, P. et al. (2006). Heterogeneous catalytic deoxygenation of stearic
acid for production of biodiesel. Industrial Engineering Chemistry Research 45: 5708–5715.
Snåre, M., Kubičková, I., Mäki-Arvela, P. et al. (2007). Production of diesel fuel from renewable feeds:
kinetics of ethyl stearate decarboxylation. Chemical Engineering Journal 134: 29–34.
References 241
Sotelo-Boyas, R., Trejo-Zarraga, F., and Hernandez-Loyo, F.d.J. (2012). Hydroconversion of triglycerides
into green liquid fuels. Hydrogenation 8: 135–152.
Takeya, K., Konishi, H., Kawanarl, M. et al. (1997). Kinetic model for soybean oil hydrogenation using
loop reactor. Food Science and Technology International 3: 10–16.
Tarhan, M.O. (1983). Catalytic Reactor Design. New York: McGraw-Hill.
Tirado, A. and Ancheyta, J. (2019). Defining appropriate reaction scheme for hydrotreating of vegetable
oil through proper calculation of kinetic parameters. Fuel 242: 167–173.
Tirado, A. and Ancheyta, J. (2020a). Modeling of a bench-scale fixed-bed reactor for catalytic
hydrotreating of vegetable oil. Renewable Energy 148: 790–797.
Tirado, A. and Ancheyta, J. (2020b). Scaling up the performance of a reactor model for hydrotreating
vegetable oil from bench-scale to pilot-scale reactors. Industrial and Engineering Chemistry Research
59: 21712–21719.
Tirado, A., Ancheyta, J., and Trejo, F. (2018). Kinetic and reactor modeling of catalytic hydrotreatment of
vegetable oils. Energy Fuels 32: 7245–7261.
Tirado, A., Trejo, F., and Ancheyta, J. (2020). Simulation of bench-scale hydrotreating of vegetable oil
reactor under non-isothermal conditions. Fuel 275: 117960.
Tirado, A., Trejo, F., and Ancheyta, J. (2021). Prediction of temperature profiles for catalytic
hydrotreating of vegetable oil with a robust dynamic reactor model. Industrial Engineering Chemistry
Research 60: 13812–13821.
Tirado, A., Alvarez-Majmutov, A., and Ancheyta, J. (2022). Modeling and simulation of a multi-bed
industrial reactor for renewable diesel hydroprocessing. Renewable Energy 186: 173–182.
Tyn, M.T. and Calus, W.F. (1975a). Diffusion coefficients in dilute binary liquid mixtures. Journal of
Chemical and Engineering Data 20: 106–109.
Tyn, M.T. and Calus, W.F. (1975b). Estimating liquid molal volume. Processing 21: 16–17.
Vélez Manco, J.F. (2014). Conceptual Design of a Palm Oil Hydrotreatment Reactor for Commercial Diesel
Production [Escuela de Química y Petróleos]. Medellin.
Verdier, S., Alkilde, O.F., Chopra, R. et al. (2019). Hydroprocessing of Renewable Feedstocks-Challenges
and Solutions. Denmark: Haldor Topsoe A/S.
Yenumala, S.R., Maity, S.K., and Shee, D. (2017). Reaction mechanism and kinetic modeling for the
hydrodeoxygenation of triglycerides over alumina supported nickel catalyst. Reaction Kinetics,
Mechanisms and Catalysis 120: 109–128.
Zhang, S., Yan, Y., Li, T., and Ren, Z. (2009). Lumping kinetic model for hydrotreating of bio-oil from the
fast pyrolysis of biomass. Energy Sources, Part A: Recovery, Utilization, and Environmental Effects 31:
639–645.
Zhang, H., Lin, H., Wang, W. et al. (2014). Hydroprocessing of waste cooking oil over a dispersed nano
catalyst: kinetics study and temperature effect. Applied Catalysis B: Environmental 150–151: 238–348.
Zhang, B., Wu, J., Yang, C. et al. (2018). Recent developments in commercial processes for refining bio-
feedstocks to renewable diesel. Bioenergy Research 11: 689–702.
242
Modeling of Slurry-Phase Hydrocracking Reactor

Cristian Calderón1 and Jorge Ancheyta2
1
Tecnologico de Monterrey, School of Engineering and Sciences, Mexico State, Mexico
2
Instituto Mexicano del Petróleo, Mexico City, Mexico
7.1 Introduction
There are different types of reactors commonly used in the petroleum industry where the hydro-
cracking (HCR) of heavy oils is one of the main processes for converting a heavy carbonaceous feed-
stock to lower-boiling point products. Those reactors are fixed-bed reactors (FBR), ebullated-bed
reactors (EBR), and slurry-phase reactors (SPRs). SPRs are more reliable to achieve high conver-
sions and have shown superiority, especially in the treatment of hydrocarbons containing sulfurous
compounds in exceedingly large quantities as well as large amounts of metals, carbon, and asphal-
tenes (Scheffer et al. 1998; Speight 2004; Bellussi et al. 2013).
SPR involves mixing of the feed oil with dispersed catalysts and hydrogen, whose purpose is the
inhibition of coke formation by hydrogenating the coke precursor and removing heteroatoms
(Zhang et al. 2007). Also, the catalyst acts as a supporter of coke, which reduces the coking of
the reactor wall. At present there are several technologies for slurry-phase HCR processes on a pilot
scale and even in industrial applications, which were reviewed by several authors (Bellussi et al.
2013; Zhang et al. 2007; Sahu et al. 2015) and are summarized in Table 7.1. Although SPRs have
been used in different applications, in the particular case of HCR, detailed modeling and other
aspects of the reactors are scarce because they are owned by manufacturers (Ancheyta 2013).
7.1.1 Characteristics of Slurry-Phase Reactors for Hydrocracking

7.1.1.1 Type of Reactors
SPRs are three-phase reactors that consist of a solid-phase catalyst (known as additive) suspended
in a liquid in batch mode or it may move co-currently or counter-currently to the gas flow
(Figure 7.1a). Generally, heterogeneous catalysts, typically transition metals (such as Mo, W, Fe,
or other elements), are used in the process. The purpose of catalyst and hydrogen is the inhibition
of coke formation by hydrogenating the coke precursor and removing heteroatoms. A catalyst with
high activity will result in a high yield of light fuel oil and a low yield of coke. In SPR, the catalyst is
added to the heavy oil and then the slurry is mixed with hydrogen in the reactor, typically operating
at high temperature and pressure. Finally, the products leaving the reactor are separated before
they are fractionated (Kriz et al. 2009).
Table 7.1 Main technologies for slurry-phase hydrocracking.
Pressure Temperature
Technology (MPa) ( C) Catalyst
VEBA-combi-cracking 15–27 440–485 Red mud (iron-containing material) and a fine

(VCC) process coke powder of Bovey coals.
M-coke technology by 17 440 Phosphormolybdic acid and molybdenum
Exxon Mobil naphthenate.
HDH technology by 7–14 420–480 An inexpensive natural ore located in Venezuela
INTEVEP based on iron or molybdenum.
CANMET process 10–15 440–460 Iron sulfate.
SOC technology by Ashi 20–22 420–460 A highly dispersed superfine powder and a
Industrial highly active transition-metal catalyst. With two
components: Mo and an ultra-fine particle (black
carbon).
Aurabon technology by 20–22 450 Vanadium oxide and Iron sulfate.
UOP
EST technology by ENI 13–16 430 Highly dispersed oil-soluble molybdenum
catalysts.
HCAT by HTIG 13–15 380–525 Synthesized molecular catalysts from metal–
organic precursors such as iron pentacarbonyl or
molybdenum 2-ethylhexanoate.
Fushun Research Institute 10–15 380–460 Synthesized oil-soluble catalyst.
of Petroleum and
Petrochemicals
Another configuration of SPR very common in industry is the slurry bubble column. Slurry bub-
ble column reactors (SBCR) are multiphase systems in which the gas feed stream is continuously
bubbled into the slurry phase (Figure 7.1b). In the simplest mode of operation, the liquid phase is
stationary while gas is sparged through the vessel. The reactor is an empty vessel placed vertically,
with a relationship between length-to-diameter ratio of at least 5 due to large ratios promoting
higher conversions but also high pressure drop and low ratios favor higher gas throughputs. There
are different internal configurations with devices to promote mass transfer (Zehner and Kraume
2005). SPR operating as a stirred-tank reactor (STR) is shown in Figure 7.1c. This vessel in batch
or semi-batch configurations is typically used to perform experiments at a laboratory scale due to
the low amount of reactants, small equipment, and easy operation characteristics. Most of these
reactors are used to explore the properties of catalysts and to obtain kinetic parameters for a given
kinetic model; however, mass transfer limitations, internal configuration of the reactor, or details of
operating characteristics are not evaluated (Angeles et al. 2014).
Due to the particular operating conditions of heavy oil HCR, the typical configuration is in the
SPR form where the feed oil is mixed with the gas before entering the reactor. Some of the general
advantages and disadvantages of SPR for HCR of heavy oil are given below.
The advantages are as follows:
•• Nearly isothermal operation.

Easy temperature control.
• Good interphase contacting.

244 7 Modeling of Slurry-Phase Hydrocracking Reactor
(a) Gas (b) Gas (c)
Slurry Slurry
Gas
Hydrogen
Slurry
Slurry
Slurry
Slurry
Hydrogen
Hydrogen
Figure 7.1 Configurations of SPR: (a) Slurry-phase reactor, (b) bubble column reactor, and (c) stirred-tank
reactor.
•• Large catalyst area.

Large liquid holdup.
•• High conversion rates.

Operational flexibility.
•• Low-pressure drop.
Low construction and operational costs.
•• Easy addition of catalyst.

Uniform catalyst distribution inside the reactor.
• The catalyst can promote cracking and restrain deposition during the catalytic reactions.
The disadvantages are as follows:
•• Difficult separation of solids and liquids.

Uncertain scale-up.
•• Plugging.
Catalyst sedimentation and agglomeration.
• Formation of coke during reactions.
While this type of reactor offers numerous advantages, particular attention must be paid to poten-
tial issues such as plugging, which can arise from catalyst agglomeration or sedimentation at the
reactor’s bottom. This problem could be avoided if the flow conditions as well as process lines and
the equipment assure that the catalyst is completely suspended. Also, using a catalyst of small par-
ticle size with similar density to the liquid phase would cause the particles to follow the motion of
the liquid avoiding catalyst sedimentation. Another problem that causes plugging in HCR systems
is the coke formation during the reactions. While it is indubitable that hydrogen plays an important
role during the reaction inhibiting coke formation during the thermal cracking process, the catalyst
is also responsible for avoiding the coke formation and prolonging the coke induction period.
Therefore, in order to avoid this problem, improving catalyst characteristics is of significant
importance.
7.1.1.2 Catalyst Properties

Highly dispersed catalyst particles have different advantages over the supported catalysts com-
monly used in hydroconversion and hydrotreating of crude oils; for example, they are less suscep-
tible to deactivation during heavy oil upgrading allowing for the elimination of catalyst pore
plugging issues, increase the accessibility of highly dispersed active sites by large size reactant mole-
cules, and minimize diffusion control process during the reaction (Angeles et al. 2014; Kennepohl
and Sanford 1996). Also, it was recently found that the catalyst deactivation in slurry-phase hydro-
conversion, in which unsupported catalysts are used, is likely different from the deactivation
observed on supported catalysts (Rezair and Smith 2013).
There are two types of catalysts for slurry-phase HCR, heterogeneous solid powder catalysts and
homogeneously dispersed catalysts classified as water-soluble catalysts and oil-soluble catalysts
(Angeles et al.; Khulbe et al. 1989). In heterogeneous solid powder catalysts, the active catalytic
phase is a solid mixed with the feed at the beginning of hydroprocessing. The homogeneously dis-
persed precursors consist of a catalyst added to the feed in the form of a precursor (water-soluble or
oil-soluble non-catalytic compound) that transforms into the catalytic active phase after an inter-
mediate step of activation in situ or during reaction conditions. Catalytic precursors have the advan-
tage that prior being added to the reaction mixture, a dissolution can be made (in water or in oil
according to the water-soluble or oil-soluble nature of the precursor). Then, the dissolution is mixed
with the feed to form a dispersion where the precursor gets well distributed all over the reaction
mixture. The former is no longer used because of the difficulty of separation and equipment wear
caused by the high dosage (Noguera et al. 2012). On the other hand, water-soluble catalysts are
preferred over oil-soluble catalysts due to their lower cost (Li et al. 1999). Homogeneous dispersed
catalysts use transition-metal compounds typically molybdenum, cobalt, iron, and nickel as
naphthenates or multi-carbonyl compounds. Molybdenum compounds are preferred to be used
as homogeneously dispersed catalysts due to their high hydrogenation activity (Ancheyta et al.
2005a,b; Domokos et al. 2006; Liu et al. 2009; Zhu and Zhou 2011; Todorova et al. 2012).
Panariti et al. (2000) studied the effect of operating conditions over a wide range of catalyst loads,
0–5000 ppm. It was observed that at any level of reaction severity, the coke formation increased at
high catalyst concentrations. To avoid this situation, a low catalyst load of around 50–250 ppm is
recommended. Ortiz-Moreno et al. (2012) studied the effect of loads from 330 to 1000 ppm in HCR
of Maya crude oil at mild conditions in a slurry batch reactor. It was first observed that catalyst load
makes no difference in the thermal process; however, the product distribution during the HCR of
heavy crude can be directed by changing the amount of catalyst and the operating temperature. On
the other hand, with variations in catalyst loads, it was also possible to analyze the thermal and
catalytic HCR on the liquid fractions and the different products obtained as shown by Ortiz-Moreno
et al. (2012). It was concluded that the catalytic HCR could be divided into two general stages
according to the conversion of vacuum residue (VR): below 50%, VR conversion is dominated
by the catalytic reactions and the catalyst is able to inhibit the formation of both coke and asphal-
tenes; and above 50%, VR conversion is dominated by thermal reactions. In this stage the oil phase
becomes incompatible to asphaltenes due to the decrease of resins and the increase of light ends,
promoting the aggregation of the most dealkylated asphaltenes and its subsequent transformation
to coke. The duration of the first stage is expected to be determined by the catalyst load, temperature
of operation, and the initial ratio of light/heavy fractions in the feed. These results show that the
optimal catalyst concentration of catalyst would depend on different variables such as operating
conditions, as well as external factors; use of recycle, catalyst separation, economic balance, pitch
specification, and even environmental issues.
7.1.2 SPR Modeling

In the modeling of SPRs due to the small diameter of the catalyst particles (dp ≤ 150 μm), it is
assumed that there is a very good homogeneity between the solid and liquid phases. As a result,
it is considered that the catalyst is part of the liquid forming a pseudo-homogeneous phase (slurry
phase), and it is necessary to consider the profile of solid concentration as well as the settling veloc-
ity of the particles (Ancheyta 2013; Martínez et al. 2010). The slurry column hydrodynamics is very
complex to model; however, if experimental or phenomenological correlations for the hydrody-
namic variables are available, the analysis could be simpler. The dispersion and interfacial heat
and mass transfer fluxes, which often limit the overall chemical reaction rates, are closely related
to the fluid dynamic of the system through the liquid–gas contact area and the turbulence proper-
ties of the flow.
The development of mathematical models describing the behavior of catalytic processes invol-
ving three phases is complicated because it is necessary to consider various aspects, from the dis-
persion and interfacial gas–liquid and liquid–solid heat and mass transfer fluxes, which are closely
related to the fluid dynamic and turbulence properties of the flow, to pore diffusion and reaction
kinetics (Martínez et al. 2010). However, the knowledge of the system behavior in steady-state or
dynamic operation is essential in determining proper reactor design and scale-up, and in correctly
interpreting data in research and pilot-plant work as well as for the trial and start-up period and
optimizing the operating conditions in manufacturing.
7.1.2.1 Classification
Froment et al. (2011) proposed a general classification for adiabatic and non-adiabatic FBR. This
classification goes from the simplest model, which considers plug flow, to the most complicated,
which involves axial and radial dispersion as well as interfacial and intrafacial gradients. In this
classification, the catalyst is big enough to be considered as one phase. It also takes into account
the bulk gas-phase temperature and concentration. These variables could be the same at the solid
surface (pseudo-homogeneous models) or different (heterogeneous models) as shown in Figure 7.2.
For two-phase bubble columns and SBCR, the most popular classification is based on the N + 1
model proposed by Tomiyama (1998) and Tomiyama and Shimada (2001). This model considers
N + 1 phases, where one phase corresponds to the slurry phase and the N phases correspond to
gas bubbles of different sizes. While this classification is similar to the one proposed by Froment
et al. (2011), it is based on the system hydrodynamics and not on the state variables T and C. There-
fore, the classification presented in Figure 7.3 involves both classifications, in which it is first nec-
essary to define the flow rate of the gas phase, either a simple model with one gas phase or a N gas
phases. Once the flow of the gas phase is established, the next step is to determine if the reaction is
carried out in the liquid phase (pseudo-homogeneous) or on the surface and inside of the catalyst
(heterogeneous). This would depend on the structure and size of the catalyst; for example, with
non-porous catalysts with very small particle diameters, the reaction takes place practically in
the liquid phase. However, for large particle sizes and porous catalysts, the reaction is carried
out in the solid.
Continuous Model
Pseudo-Homogeneous Heterogeneous
(C=Cs, T=Ts) (C≠Cs, T≠Ts)
One Dimensional Two Dimensional One Dimensional Two Dimensional
Axial Radial Interfacial Intrafacial Radial

Plug Flow
Mixing Mixing Gradients Gradients Mixing
Figure 7.2 Classification of reactor models.
Continuous Slurry-Phase Models
Single gas phase model N gas phases model

(G+SL) (G1+G2+...+GN+SL)
Pseudo-Homogeneous Heterogeneous
(C=Cs, T=Ts) (C≠Cs, T≠Ts)
One Dimensional Two Dimensional One Dimensional Two Dimensional
Axial Radial Interfacial Intrafacial Radial

Plug Flow
Mixing Mixing Gradients Gradients Mixing
Figure 7.3 Classification of slurry-phase models based on hydrodynamics.
The flow regimes in bubble and slurry bubble columns are classified according to the superficial
gas velocity. Two types of flow regimes are commonly observed in slurry and bubble reactors:
homogeneous (bubbly flow) regime and heterogeneous (churn turbulent flow) regime. Another
regime is called the slug flow regime, which has only been observed in small-diameter laboratory
columns at high gas flow rates as can be observed in Figure 7.4.
CHURN-TURBULENT REGIME
SLUG FLOW REGIME
Gas Velocity (m/s)
TRANSITION RANGE
HOMOGENEOUS BUBBLE FLOW REGIME
COLUMN DIAMETER (m)
Figure 7.4 Regime flow for bubble and slurry-phase reactors (Adapted from Deckwer, et. al. (1980)).
The homogeneous flow regime is assumed when the high pressure of the system and low gas
velocities prevent bubble coalescence. Then the single-phase model can be used. The homogeneous
flow is established when the difference between the density of solids and liquid is small or if the
liquid viscosity is high. Also operating at low gas flow rates where small bubbles (SB) of gas
(1–10 mm) are uniformly distributed into the slurry phase (L + S). The bubble size and uniformity
depend on the properties of the liquid, the design of the gas distributor, and the column diameter,
which are defined for a uniform bubble size distribution in the axial and radial direction of the
vessel.
N gas-phase models are used when the churn turbulent flow regime is found inside the reactor.
This flow is presented when there is a large gas fraction in a system with a high gas and low liquid
velocity and is frequently observed in large-diameter columns at an industrial scale. Typically, at
the beginning of this regime, the number of bubbles is low, which is called as ideally separated
bubble flow. In this type of flow, the bubbles do not interact with each other directly or indirectly,
but as the number of bubbles increases, they start colliding with each other and their size gets
reduced. Suddenly a situation comes when they tend to coalesce to form cap bubbles, and the
new flow pattern formed is called churn turbulent flow, which has a broad size distribution. This
model considers N bubble types plus the slurry phase; therefore, each bubble type would have their
respective mass, heat, and momentum equations in order to describe the system dynamics. Never-
theless, previous studies based on the extend of the two-phase (“dilute” and “dense” phases) model
proposed by van Deemter (1961) have shown that in a slurry bubble column operating in the churn
turbulent flow regime the gas phase can be split up in a “large” bubble population and a “small”
bubble population (27–29). In this heterogeneous system, SB combine in clusters to form large bub-
bles (LB) (20–70 mm). These LB travel up through the column at high velocities (1–2 m/s), in a
more or less plug-flow manner. Though this is still a very complicated model, it would lead to a
considerable simplification to the fluid dynamic equations of the system (Kantarci et al. 2005;
Deckwer et al. 1980).
Due to the high temperature and pressure conditions in heavy oil HCR, large amount of hydro-
gen, as well as oil and catalyst properties, the typical operating regime is the homogeneous flow.
Therefore, when modeling these systems, one of the main considerations is to establish a homoge-
neous gas flow inside the reactor. Hence in this review, only models concerning homogeneous gas
flow will be discussed.
7.1.2.2 Model Complexity

There is no general rule to select the level of sophistication of a reactor model. The complexity of a
model is closely related to the purpose of the investigation. In most cases, a typical strategy is select-
ing the simplest model and adding complexity as the error between experimental and calculated
data is reduced. The simplest models are built under ideal considerations neglecting gradients
in concentration in all spatial directions (perfect mixing) or recognizing them only in the principal
flow direction (plug flow), also a common assumption is a uniform concentration of the solids
throughout the reactor. These models can be used in SPR operated as STR. On the other hand, while
slurry and bubble column performance often can be fitted with axial dispersion models (ADMs),
decades of research have failed to produce a predictive correlation for the axial dispersion coeffi-
cient. The so-called computational fluid dynamics (CFD) models solve the fluid dynamics with a
numerical solution of the Navier–Stokes equations and a set of partial differential equations
(PDEs). These are the most sophisticated models but also the most difficult to solve due to their
high nonlinearity, even two-phase transient simulations require high computational cost
(Pfleger and Becker 2001). Most CFD models applied to slurry and bubble columns are focused
on the hydrodynamics of the system, neglecting mass and heat transfer effects in some cases, which
clearly represents a limitation to reactive systems. From the point of view of reaction engineering, a
complex model for the field of velocity inside the reactor would not offer advantages if the profile of
concentrations of substances involved is described poorly.
The reliability of a model, either ideal, ADM, or CFD, is a function of the validation method such
as testing it with independent data and thermodynamic properties at the reaction conditions.
7.1.2.3 Models for Slurry Reactors

Homogeneous Bubble Flow Regime Models Reactor models that feature a practical way to design two/
three-phase reactors and that provide a basic understanding of the chemical process on the semi-
industrial or even industrial scale have been published only rarely in the usual scientific literature
(Jung et al. 2010), and SPRs are not the exception. Unfortunately, only CFD models are often imple-
mented in as much detail as possible for the simulation of slurry HCR reactors, no matter that the
complex modeling and computational effort required are extremely time-consuming and costly.
However, it has been found that for simple reaction systems such as hydrogenation, three-phase
slurry reactor dynamics can be well represented by ADMs where deviation from plug flow is
described using an axial dispersion coefficient. For example, de Toledo et al. (2001) proposed
two dynamic non-isothermal ADMs applied to describe the dynamic behavior of the reactor during
the hydrogenation of O-cresol on Ni/SiO2 catalyst. The authors conclude that the most complete
model that includes internal mass transfer reproduces the dynamic behavior of the reactor in a bet-
ter way; however, it is possible to use the simplified model if there is not enough experimental infor-
mation about catalyst properties, in fact this model is the most frequently used in different reactions
involving different types of catalyst particles. Both models have been used in different applications
of control and optimization (Bahroun et al. 2010; Melo et al. 2005; Costa et al. 2007; Li et al. 2010;
Bahroun et al. 2010; Mariano et al. 2011). Shahrzad et al. (2009) presented a dynamical ADM that
depicts CO2 hydrate formation in a slurry bioreactor. The model considers mass transfer phenom-
ena between gas and liquid as well as liquid to solid phases. Also, the particle velocity is considered
along with the term of reaction. Chen et al. (2010) proposed a steady state ADM for the direct
dimethyl ether (DME) synthesis process from syngas in which the dispersion and velocity of solid
particles were introduced and mass transfer resistance was ignored in liquid–solid phase. The
authors concluded that the particle and reactor diameters are the two main factors influencing con-
centration distribution uniformity. On the other hand, for HCR reactions, it has been tested that the
field of velocity does not influence on a great way the conversion along the reactor (Matos and
Nunhez 2006). Therefore, it is expected that an ADM with adequate kinetic and hydrodynamic
parameters would be enough to describe slurry-phase HCR reactors.
The work by Parulekar and Shah (Parulekar and Shah 1980) attends in a practical way the HCR
in SPRs. This model consists of gas, liquid, and solid mass balance equations formulated under the
following considerations:
•• Isothermal operation.
Gas phase is assumed to behave according to the ideal gas law.
•• All reactions are purely catalytic under the conditions of operation considered.
The ADM is assumed to be applicable in the case of non-volatile liquid components.
•• The gas phase is assumed to move as plug flow.

The reactor is of concurrent up-flow type.
• All the global reaction rates are expressed in terms of the gas-phase concentration of hydrogen
because it is difficult to measure the liquid-phase (oil in the case of HCR) concentration of
hydrogen.
They proposed a kinetic model based on three components (AL, heavy oil; BL, light oil; and CG,
volatile product) and an HCR reaction, all represented by a simple power-law approach Eqs. (7.1–
7.4). Unfortunately, the authors do not give information about catalyst properties.
AL BL r 1 = k 1 C A C S 71
AL C G r 2 = k 2 C A C H2 C S 72
BL C G r 3 = k 3 C B C H2 C S 73
H2 + L HL r 4 = k4 C H2 CS 74
The model equations are as follows:
d U G C H2
Gas phase H2 = − k4 CH2 C S εL 75
dz
d dC A d U L CA
Liquid phase A DLa − − k 1 C A CS εL − k 2 CA CH 2 CS εL MM C = 0 76
dz dz dz
d dC B d U L CB
Liquid phase B DLa − + k1 CA C S εL − k 3 C B CH 2 C S εL MM C = 0 77
dz dz dz
d dCS d
Solid distribution CS DSa − U SL − 1 − εG U tS C S = 0 78
dz dz dz
The major complication in solving these equations is that all hydrodynamic parameters (εG, εL,
εS, DSa, and DLa) and superficial velocities (UG, UL, and USL) vary axially. Therefore, dynamic equa-
tions for axial velocities are needed. Thus, it was assumed that superficial velocities vary as a func-
tion of the reactive absorption rate into the catalyst and product desorption rate (Eqs. 9 and 10).
dU G
CGT = C H2 CS εL k 2 CA + k 3 CB − k 4 79
dz
dU L
ρL = C H2 CS εL k 4 MM H2 − k 1 CA + k2 CB MM C 7 10
dz
Though this model is limited by particle sizes between 75 and 250 μm with low solid loading
(εS ≤ 0.15) and a simple kinetic model, it is possible to observe the behavior in steady state for
the different variables like pressure, temperature, and dimensions of the reactor. How they give
an increase in product yield and a maximum in product yield can be found by increasing the rest of
the analyzed variables. Thus, the optimum operating conditions in the slurry-phase HCR reactor can
be established. Other quantitative results show the behavior of solid concentration along the reactor.
In this manner, it is observed that for low particle diameters, the solids distribution along the reactor
remains practically constant for a given liquid velocity. On the contrary, when particle diameter is
increased, an accumulation of particles is noted at the bottom of the reactor. Although the profile
of solids concentration could vary as a function of liquid velocity and column diameter, these vari-
ables have no great influence on the particle distribution. Certainly, this is a simple model; however, it
could be used as a reference for future works in slurry-phase HCR considering more detail kinetic
models as well as transitory state, gas–liquid mass transfer, or even more complete ADMs.
Carbonell and Guirardello (1997) studied an isothermal reactor based on the continuity and
momentum balance equations using an Eulerian approach applied in the HCR of heavy oils in
a slurry-phase reactor operating at severe conditions of temperature (350–500 C) and pressure
(7–25 MPa). They considered that the thermal cracking reactions do not interfere with the hydro-
dynamics of the system due to the high mixing of the liquid phase and the low gas absorption rate.
Under these conditions, a single-bubble class model could be assumed. Their simulations were per-
formed in two steps: first, fluid dynamic variables were determined as a function of radial position,
and then oil cracking conversion was obtained. The power-law kinetics was based on experimental
evidence. Feed and products are divided into several boiling range lumps, where each lump is con-
sidered a single chemical species with a single cracking rate constant. The model consists of the
following assumptions:
•• Steady-state operation.
A symmetric turbulent axial flow.
•• Single-bubble model.
A zero model turbulence in the mixing bed.
•• A parabolic radial distribution for gas holdup.

Dominant convection in the axial dispersion.
•• Dominant dispersion in the radial dispersion.

First-order irreversible reactions.
The hydrodynamic equations for the slurry and the gas phases are:
1 d duSL dP
Slurry phase εSL rμeff
SL − εSL − gρSL + F = 0 7 11
r dr dr dz
1 d duG dP
Gas phase εG rμeff
G − εG − gεG ρG − ρSL − F = 0 7 12
r dr dr dz
where F is the interfacial drag force (N/m3 reactor) proposed by Torvik and Svendsen (1990), v axial
velocity (m/s), r radial position, ε holdup (dimensionless), μeff effective viscosity (Pa s), and RC
radius of the reactor. The mass balance is:
dCi
ε L ρL u L = αi kρL C i + βm,i kρL Ci 7 13
dz
Despite isolating the hydrodynamic behavior of the system from the chemical reactions, concord-
ance with experimental data was obtained, although a more detailed model, with a more complex
study of the kinetics, is required for scaling up and improvements in the operation.
Matos and Guirardello (2002) presented an isothermal CFD model that includes momentum
equations, based on the kinetic network of six lumps presented by Krishna and Saxena (1989)
for thermal HCR reactions. The lumps are heavy and light aromatics, naphthenes, and paraffins.
The HCR is assumed to be thermal, and the catalyst is only used to avoid coke formation and for the
removal of heteroatoms. The main assumptions are a pseudo-homogeneous system (SL + G) where
the solid distribution is not considered due to the very small diameter and the small terminal veloc-
ity of the particles, the gas phase is composed mainly of hydrogen, and oil is completely saturated
with dissolved hydrogen. Due to hydrogen being found in excess inside the reactor, the reaction is
independent of the hydrogen consumption. Because of the lack of experimental information on
HCR systems, this model was validated by comparing different experimental data available in
the literature for the air–water system. Though this model can describe the behavior of an SPR
in a good manner, it is quite complicated to solve because it includes the velocity distribution
for gas and slurry phases along the reactor. However, a simple approximation can be obtained
by an ADM considering average velocities and the proper parameter values as shown in Figure 7.5.
Matos and Nunhez (2006) presented a fluid dynamic model based on the same considerations as
in Matos and Guirardello (2002) using the same kinetic model of Krishna and Saxena (1989), but in
this case, the work is focused on how the feed input affects the flow fields and the reactor conver-
sion. Although unsupported by experimental data, their theoretical results show that the fluid
dynamic fields inside the bubble column reactor do not affect considerably the reactor conversion.
Recently, Matos et al. (2009) presented a CFD model to estimate operating conditions that min-
imize sulfur and organometallics concentrations before the petroleum is processed. In the modeling
process, it is assumed that the difference between the catalyst and the slurry phase is negligible,
preventing catalyst settling. Therefore, a fluid dynamic model capable of suspending the catalyst
was deemed sufficient. The model considers the petroleum to be composed of pseudo-components,
and HDS and hydrodemetallization (HDM) reactions have a first-order kinetics decomposition con-
trolled by diffusion and occurring on the catalyst pores. As chemical components in the crude mix-
ture present very low reaction rates, there is low gas consumption and the mass transfer between
the phases is negligible. However, the main contribution in this work is that the model includes a
60
AH
AL
NH
50 NL
PH
PL
40
Percentage %
30
20
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
z
Figure 7.5 Axial dispersion model kinetics reported Adapted from Krishna and Saxena (1989). AH, Heavy
aromatics; AL, light aromatics; NH, heavy naphthenes; NL, light naphthenes; PH, heavy paraffins; PL, light
paraffins.
kinetic expression for the activity reduction of the catalyst as a function of the catalyst pore
characteristics.
Heterogeneous Bubble Flow Regime Models Operation at low gas velocities in a homogeneous
bubbly flow regime was considered to be preferred in industrial SPRs. Furthermore, most of
the literature models have been focused on operation at relatively low superficial gas velocities
(below 0.10 m/s). However, the growth of industrial demand caused the plants to operate at
higher production rates, which caused the processes to involve increasing operating conditions
such as gas and liquid flows. As a consequence, this generates heterogeneous flow conditions
inside the reactor.
So far, the latest study for SBCR modeling is presented by Khadem-Hamedani et al. (2015). They
showed a comparison between models in homogeneous and heterogeneous flow regimes for the
hydrodesulfurization (HDS) of diesel fuel. The mass and energy balances, as well as the model para-
meters, are based on the considerations proposed by Mills et al. (1996) and de Swart and Krishna
(2002). Both models reproduce the reactor behavior in a good way; however, as it was expected, the
more detailed model (heterogeneous model) shows better results in the validation with experimen-
tal data. These results are of great importance because they demonstrate that the parameters pro-
posed by Mills et al. (1996) and de Swart and Krishna (1997) generated for F–T synthesis for
heterogeneous regime can be used in HDS of diesel, which have similar conditions of pressure
and temperature to HCR.
7.2 Proposed Generalized Model
7.2.1 Equations for the Generalized Model

Most mass balance equations as well as motion equations are derived from the general equation of
continuity and motion, and slurry bubble columns are not the exception. While there are different
ways of representing these equations, good approximations of the behavior of this type of reactor
have been obtained using an Eulerian approach. The fundamental form of the multi-fluid continu-
ity equation for phase f has been presented by several authors (Tomiyama and Shimada 2001; Pfle-
ger and Becker 2001; Jakobsen et al. 2005; Pfleger et al. 1999). In this review, the equations are
presented for each component in the f phase, thus the reaction rate and convective mass transfer
are also considered. The equations are:
∂ εf C fi F −1
f 1 ,f f
f f
+∇ εf C i uf =∇ εf Df ∇Ci + Ni +1
+ ri 7 14
∂t f =1
∂ εf ρf Cpf T f
+∇ εf ρf Cpf T f uf =∇ εf λf ∇T f + μΦ + ΔH r + q T f , T W 7 15
∂t
+ q T f , T I + q T f , T Ssur
This governing set of PDEs consists of the continuity equations for the N + 1 phases for i com-
ponents and the energy and momentum equations for N + 1 phases. It holds that:
∂ εf ρf
+∇ εf ρf uf =0 7 16
∂t
F
εf = 1 7 17
f
εf uf = U f 7 18
Eq. (7.14) shows the classical terms of transitory state, convective and molecular fluxes, as well as
F −1 f 1 ,f f +1
the terms f = 1N i which is related to the convective fluxes between phase resistances for
each component, and the reaction rate r i. In energy Eq. (7.15), appear the terms related to the tran-
sient state temperature, heat exchange caused by convective flow, conduction, viscous dissipation
energy (usually negligible, except in systems with large velocity gradients) (Bird et al. 2002), the
generation term due to the chemical reaction, the heat fluxes, function of the temperature of each
phase, the temperature of fluid exchange, interfaces, and solid temperatures. Being strict, this set of
equations consists of three general equations, for the gas, liquid, and solid phases, but for SPRs, it is
commonly considered to simplify the gas and liquid phases as a continuous phase in which the
temperature is the same due to the large area of contact and favorable transfer coefficients (de
Toledo et al. 2001).
As mentioned before, this is a very complex PDE system, particularly for HCR of heavy petroleum
where various reactions occur simultaneously, for example, HDS, HDM, hydrodenitrogenation
(HDN), and hydrodeasphaltenization (HDA). However, some considerations can be made in order
to simplify this model. Taking into account that the model of SPRs is similar to EBR (Martínez et al.
2010) and the reacting system is the same as that presented by Mederos et al. (2009), some of their
general criteria can be adapted. For example, though the values of mass and heat transfer coeffi-
cients are distinct, the expressions for mass and heat transfer resistances between phases are
expected to be the same. Also due to the chemical reactions being merely catalytic, the conversion
will be directly associated with the concentration of catalyst particles inside the reactor. However,
unlike EBR in SPR, the solid catalyst is found in the wake of the liquid forming an L–S suspension.
This unique characteristic generates a greater degree of difficulty in the approach of the mathemat-
ical model, since system hydrodynamics becomes more complicated. For the bubble column slurry
operation, two suspension states may exist: namely, complete suspension in which all particles are
in suspension, commonly described by the axial sedimentation-dispersion model (SDM), and
homogeneous suspension, in which the particle concentration is uniform throughout the reactor
(Chaudhari and Ramachandran 1980). In the proposed generalized model, the SDM is taken into
account. Other considerations are as follows:
•• Heterogeneous bubble flow, composed of SB and LB.

The liquid–solid (L + S) suspension behaves as a homogeneous phase due to the small catalyst
particle size, whereas the gas phase bubbles up through the suspension.
• Liquid and gas properties (mass and heat dispersion coefficients, specific heats, densities, and
viscosities) are constant along the whole catalytic bed.
• Catalyst properties (porosity, size, activity, effectiveness, etc.) are constant along the whole cat-
alytic bed.
•• Due to the small particle size, wetting efficiency is considered to be the unity.
Inside the catalyst particle, mass and heat-effective diffusivity coefficients may also be assumed
constant.
• The liquid phase is assumed to be in thermal equilibrium with the gas phases. Then the energy
equation can be formulated for the continuous “f” phase defined as (SB + LB + L).
A detailed description of each term given in Tables 7.2 and 7.3 is shown by Mederos et al. (2009),
except for the solid dispersion terms that are presented below.
Table 7.2 Generalized mass balance equations.
Accumulation Convective Axial dispersion Radial dispersion G–L G–S SB-LB
(A) Gas ∂εSB C SB ∂εSB uSB CSB ∂ ∂CSB ∂ 2

εSB C SB 1∂ εSB CSB C SB C SB − U LB − U SB
K
CSB − CLB
phase SB
i
= ± i
+ εSB DSB i
+ DSB i
+ i − KLa i
− CLi − kGS a i
− C SGsur HC i i
∂t ∂z ∂z a
∂z r
∂r 2 r ∂r
SB,i
Hei SB,i
Hei
(B) Gas ∂εLB CLB ∂εLB uLB C LB ∂ ∂C LB ∂ 2 εLB C LB 1 ∂ εLB C LB C LB C LB + U LB − U SB

K
C SB − CLB
phase LB
i
= ± i
+ εLB DLB i
+ DLB i
+ i − KLa i
− C Li − kGS a i
− CSGsur HC i i
∂t ∂z ∂z a
∂z r
∂r 2 r ∂r
LB,i
Hei LB,i
Hei
Accumulation Convective Axial dispersion Radial dispersion G–L L–S
(C) Liquid ∂εL C Li ∂εL uL C Li ∂ ∂C L ∂ 2

εL C Li 1∂ εL C Li CSB i aS C i − C i
kLS L Ssur
phase = ± + εL DLa i + DLr + − KLa C Li − i

∂t ∂z ∂z ∂z ∂r 2 r ∂r
SB,i
Hei
volatile
C LB
− KLa LB,i CLi − i
Hei
(D) Liquid ∂εL C Li ∂εL uL C Li ∂ ∂C L ∂2 εL C Li 1 ∂ εL C Li i aS C i − C i
kLS L Ssur
phase non- = ± + εL DLa i + DLr +

∂t ∂z ∂z ∂z ∂r 2 r ∂r
volatile
(E) Solids ∂εSL C S ∂ ∂ ∂CS
= ± εSL uS − U SL CS + εSL DSa
dispersion ∂t ∂z ∂z ∂z
Accumulation Intraparticle diffusion G–S transfer L–S transfer Generation
(F) Liquid ∂εS C SL,isur + kLS

i aS CLi − CSL,isur N RL
phase in 1 − εB = + ρS ηLj νLij r jL

∂t j=1
solid surface
∗
(G) Gas ∂εS C SG,isur i aS C i − C G,i
+ k GS Ssur N RG
phase in 1 − εB = + ρS ηGj νGij r jG

∂t j=1
solid surface
(H) Liquid ∂εS CSL,iinn DSei ∂ ∂C inn S N RL
phase in = + ξ2 L,i + ρS νLij r jL

∂t ξ2 ∂ξ ∂ξ j=1
solid inner
(E) Gas ∂εS CG,i
Sinn
DSei ∂ ∂C inn S N RL
phase in = + ξ2 G,i + ρS νLij r jG

∂t ξ2 ∂ξ ∂ξ j=1
solid inner
Table 7.3 Generalized heat balance equations.
Accumulation Convective Axial dispersion Radial dispersion Conductive f–S transfer f–W transfer
(A) Gas phase ∂ εG T ∂ εG uG T ∂ ∂T 2

∂ εG T 1 ∂ εG T −ΔHvi − hGS aS T − T SS AW
CpG ρG = ± CpG ρG + εG λGa + λGr + − hGW T − TW
∂t ∂z ∂z ∂z ∂r 2 r ∂r V
(B) Liquid phase ∂ εL T ∂ εL uL T ∂ ∂T ∂ 2 εL T 1 ∂ εL T +ΔHvi − hLS aS T − T SS AW
CpL ρL = ± CpL ρL + εL λLa + λLr + − hLW T − TW
∂t ∂z ∂z ∂z ∂r 2 r ∂r V
Accumulation Convective Axial dispersion G–S transfer L–S transfer Generation
(C) Isothermal ∂ εS C S T S ∂ εS CS uSL T S ∂ ∂T S +hGSaS (T − TS) +hLSaS (T − TS) NRL NRG

CpS = ± CpS + εS λSa ρB ζ − ΔH Lrj ηLj r jL + − ΔH Grj ηGj r jG
solid phase ∂t ∂z ∂z ∂z j=1 j=1
Accumulation Axial dispersion f–W transfer
(D) Thermowell ∂T W 2
∂ TW AW AW
CpW ρW = + λW + hGW T − T W + hLW T − TW
∂t ∂z2 V V
Accumulation Intraparticle transfer Generation
(E) Non- ∂T S ∂2 T S 2 ∂T S NRL NRG

CpS ρS = + λSe + ρS − ΔH Lrj r jL + − ΔH Grj r jG
isothermal solid ∂t ∂ξ2 r ∂ξ j=1 j=1
phase
7.2.2 Solids Concentration

Eq. (E) in Table 7.4 describes the dynamic behavior of solid concentration in a solid–liquid mixture
along the reactor through the SDM. This model considers a solids axial dispersion flux and a solids
sedimentation flux superimposed on the average slurry convection flux. One of the main variables
in this equation is the particle settling velocity, which is the result between the liquid and particle
terminal settling velocities as shown in Figure 7.6. Terminal settling velocity utS is defined as the
highest velocity reach for particles immersed in a fluid; it occurs when the drag and gravity forces
are equal. For spherical shapes it typically has the form:
ρS − ρL gd2p
utS = 7 19
18μL
By definition, the terminal settling velocity could be positive or negative depending on the density
difference between solid and liquid phases and on the coordinate system adopted. On the other
hand, the prediction of the particles settling velocity is found in the literature given by different
correlations; for example, Parulekar and Shah (1980) and Kojima et al. (1989) consider the settling
velocity to be the same as the terminal settling velocity. Recently, Sehabiague et al. (2008) applied
the correlation proposed by Kato et al. (1972), where the particle settling velocity is proportional to
the superficial gas velocity and to the liquid fraction in the slurry phase:
0 25
UG
uS = 1 33utS ϕL2 5 7 20
utS
Most of the models that use the SDM consider a constant average value of the liquid and gas
superficial velocities, then an analytical solution of this equation is possible as a function of the
Peclet particle number. The mass balance on the external surface of catalyst particles is given in
Eqs. (F) and (G) in Table 7.4. The gradients are the product of an effective diffusion of liquid
and gas phases through the catalyst surface. It also has the generation terms, which include the
catalyst effectiveness factor associated to the reaction in the surface and at the inner of the catalyst
particle; however, for SPRs, the catalyst commonly used in FBR is crashed to generate small catalyst
diameters, which are known to have a high catalyst effectiveness (Macías and Ancheyta 2004), then
it could be assumed to be the unity.
7.2.3 Initial and Boundary Conditions

All transitory mass and energy balance equations presented in Tables 7.2 and 7.3 have initial (t = 0)
and boundary conditions (t > 0), which relate the surface properties to the bulk properties of the
reacting system. Considering a symmetric reactor of radius R in which gas, liquid, and solid phases
of radius ξ = dpe/2 enter at the bottom of the reactor at z = 0 and that the reactor outlet is located at
z = H, all coordinate system has been fixed. The initial conditions for the transitory state are typ-
ically set to the inlet properties of the system:
f f0
C i = Ci , C S = C0S , C Sfisur = C fiSsur 0 , C Sfiinn = C fiSinn 0 , T 0f = T 0f 7 21
These conditions could vary depending on the considerations made for the system under study;
for example, in some cases, it is considered that the liquid is saturated with the gaseous compound,
while in others there is no initial concentration of gas in the liquid phase. Also, the catalyst activity
Table 7.4 Boundary conditions of the generalized model.
Condition Gas phase Liquid phase Solid phase i at solid surface i at solid inner
z = 0, 0 ≤ r ≤ R dCGi dCLi dCS C Sfisur = CSfisur 0 CSfiinn = CSfiinn 0

DGa = uG CGi − CG0
i DLa = uL CLi − C L0
i
DSa = uSL CS − C0S
dz dz dz
dT G dT L dT S dT S dT S
λGa = uG ρG CpG T G − T 0G λLa = uL ρL CpL T L − T 0L λSa = uSL ρS CpS T S − T 0S λSa = uSL ρS CpS T S − T 0S λSa = uSL ρS CpS T S − T 0S
dz dz dz dz dz
z = H, 0 ≤ r ≤ R dCGi dCLi U L CS = C0S
=0 =0
dz dz
dT G dT L dT S — —
=0 =0 =0
dz dz dz
0 ≤ z ≤ H, r = R dT G dT L — —
− λGr = hGW T G − T W − λLr = hLW T L − T W
dr dr
ξ=0 — — — — dCSfiinn
=0
dξ
0 ≤ z ≤ H, 0 ≤ r ≤ R — — — — dT S
=0
dξ
dpe — — — — ∂CSfiinn
ξ= − DLei = kSi aS CSLiinn − CLi
2 ∂ξ
NRL
= ρB ζ νLij ηLj r jL
j=1
0 ≤ z ≤ H, 0 ≤ r ≤ R — — — — ∂CSfiinn
− DLei = kSi aS CSLiinn − CLi
∂ξ
NRL
= ρB ζ νLij ηLj r jL
j=1
uS
uL uL
utS
Figure 7.6 Schematic representation of movement of suspended particles.
and initial catalyst concentration may differ if recirculation is used. Moreover, since the system is
described by a set of PDE, four boundary conditions are needed in order to describe the longitudinal
and radial variation of the system properties (z = 0, 0 ≤ r ≤ R, z = H, 0 ≤ r ≤ R, 0 ≤ z ≤ H, r = 0, 0 ≤
z ≤ H, r = R), which typically have the form of Danckwerts’s conditions:
dΨ
at z = 0, = Ψ − Ψ0 7 22
dz
dΨ
at z = L, =0 7 23
dz
For the case of the reactor bottom at z = 0, 0 ≤ r ≤ R, there are different ways of setting the inlet
conditions; for example, some authors consider that the axial dispersion of mass and temperature is
too small compared with the convective term, therefore the gradients at the inlet could be omitted
and the conditions are the same as in Eq. (7.21). To describe the dynamics of solid concentration,
two conditions are needed because it is a second-order PDE, the fist condition is given at z = 0 for all
the radial coordinate 0 ≤ r ≤ R. Typically Danckwerts’s conditions are the most commonly used. In
this case, the inlet concentration of the catalyst is obtained as a function of the weight fraction of the
solid in the feed slurry through:
W S ρL
C 0S = 7 24
1 − ρS − ρL ρL W S
Instead of this equation, simple average concentrations could be also used (Sehabiague et al.
2008). The second condition is given at z = H for all the radial coordinate 0 ≤ r ≤ R, the condition
is given by a simple balance where the flux of solid entering the reactor must equal that leaving the
reactor:
U L CS = C0S 7 25
On the other hand, though considering internal and external particle diffusion effects in the mod-
eling could lead to a more realistic representation of real reactors, when modeling catalytic slurry-
phase reactors, due to reduced dimensions typically found in practice, these effects are commonly
ignored, and it is assumed that the reaction takes place on the liquid phase. However, the conditions
proposed in Table 7.4 for the concentration of the i component at the solid phase would be ignored.
7.2.4 Estimation of Model Parameters

Even in the most complex models, it is necessary to evaluate several parameters and chemical
properties of the system. An adequate description of the system dynamics needs reliable data on
parameters that are specific for the SPRs (kG, kL, and kS) and parameters that are not specific for
the type of reactor (k and DMi), as well as particle and fluid properties (μ and ρ). Those para-
meters can be estimated with existing correlations, whose accuracy is of great importance for
the whole robustness of the model. Density and viscosity of the liquid phase (heavy oil) play an
important role in system hydrodynamics. For slurry systems, these properties could be affected
by the presence of solid particles. High concentrations of particles in the feed oil could lead to
notorious changes in the viscosity of the liquid phase. There are existing correlations that esti-
mate the viscosity and density of the slurry phase as a function of the solids concentration (de
Swart and Krishna 2002a,b; Khadem-Hamedani et al. 2015). In general, all liquid and gas phase
properties at the operating conditions could be estimated through different correlations
reported in the literature as can be seen in the work of Mederos et al. (2009) for heavy oil
HCR. On the other hand, there are a few reviews in the literature dealing with SPRs and bubble
columns. Chaudhari and Ramachandran (1980), Shah et al. (1982), and Fan (1989) presented a
very complete review of previous works published until the 80 decade for parameter estimation
for slurry and bubble column reactors. Beenackers and Swaaij (1993) focused only on the mass
transfer phenomena in SPRs, and more recently Kantarci et al. (2005) reviewed bubble column
reactors with application to bioprocess. However, most of the parameters reported were
obtained for two-phase and low-viscosity systems, typically air–water or air–organic liquids.
In order to better reproduce the dynamics of the reactor, one should select the parameters that
were obtained under similar conditions to those used in the modeling system of interest. Leo-
nard et al. (2015) presented a review of bubble column reactors operating in high pressure and
temperature. Their work is focused on overhauling the effect of pressure and temperature on
hydrodynamic variables such as gas holdup, as well as mass transfer parameters and axial dis-
persion coefficients. On the other hand, Mederos et al. (2009) presented a summary of the typ-
ical correlations used in HCR systems for diffusion coefficients, viscosities, and densities, which
are needed in mass transfer calculations.
7.2.5 Gas Holdup

The gas holdup is one of the most important parameters characterizing the hydrodynamics of BCR
and SBCR, which is defined as the percentage of gas volume in the two- or three-phase mixture
inside the reactor. The gas holdup in conjunction with the knowledge of mean bubble diameter
allows for the determination of interfacial area, thus leading to the mass transfer rates between
phases. It depends mainly on the superficial gas velocity, but it is also sensitive to the physical prop-
erties of the liquid, gas and solid particles, column dimensions, operating temperature and pres-
sure, and gas distributor design. For practical simplicity, it is usually estimated using the
pressure profile method (Li and Prakash 2000), in which the gas holdup is obtained from the static
pressure drop as:
1 ΔP
εG = 1 − 7 26
g ρL ϕL + ρS ϕS ΔL
where ϕL and ϕS are the volume fractions of liquid and solid particles in the slurry phase,
respectively:
VL VS
ϕL = ,ϕ = 7 27
VS + VL S VS + VL
The term g(ρLϕL + ρSϕS)ΔL in Eq. (7.26) has been approximated to the static pressure difference
ΔP0, when no gas is flowing in the column (Tang and Heindel 2006). In these equations, the effect of
wall shear stress and liquid acceleration due to void changes are neglected. When more reliable
models are required, predictions of radial gas holdup profiles would lead to a better understanding
of system hydrodynamics. There are a large number of correlations proposed for the gas holdup,
especially for two-phase bubble columns (Kantarci et al. 2005; Shah et al. 1982), although there
is no general correlation applicable to any system mainly due to the extreme sensitivity of the
holdup to the materials properties and the trace impurities (Bach and Pilhofer 1978). It has been
pointed out that for some slurry systems, the presence of non-catalytic solids does not affect the gas
holdup significantly (Shah et al. 1982; Ying et al. 1980). However, recent experimental results
focused on catalytic systems show that increasing the solid concentration decreases the gas holdup
to a significant extent (Vandu and Krishna 2003).
There are several correlations of gas holdup for SPRs (Kito et al. 1976; Begovivh and Watson 1978;
Dudukovic and Mills 1984; Reilly et al. 1986; Sauer and Hempel 1987; Schumpe et al. 1987; Chen
et al. 1995; Luo et al. 1999; Krishna and Sie 2000; Kumar and Khanna 2014). Unfortunately, most of
these correlations have several limitations. On the one hand, these correlations were developed for
aqueous systems in small-diameter reactors operating under atmospheric pressure or ambient tem-
perature, and on the other hand, the majority of solids were non-catalytic particles. This leads some
authors to choose to solve the dynamic of the holdup along with the state variables of the system
introducing turbulence equations, despite the computational effort. Addressing this situation, the
correlations proposed by Behkish et al. (2006) predict the total holdup of single bubbles and LB in
pseudo-homogeneous and heterogeneous flows in BCR and SBCR, operating under a range of ele-
vated pressures (P = 0.1–15 MPa). In their study, several literature data were considered in order to
obtain the correlations. Afterward, the agreement between predicted and experimental values was
evaluated, presenting acceptable results with a standard deviation between 15% and 20%.
For HCR systems, most of the literature is focused on kinetic or catalyst studies at batch condi-
tions, thus validation of hydrodynamic parameters is complicated. At the pilot scale and based on
previous studies focused on the effect of superficial gas velocity, interfacial tension, and column
diameter, Parulekar and Shah (1980) proposed that the gas phase holdup in a slurry HCR system
operating at 10 MPa and 350 C was proportional to the variables according to Eq. (7.28).
εG = 0 0866σ L− 0 01 U 0G 5 DC− 0 08 7 28
While this correlation was implemented successfully in the model in order to obtain the
quantitative behavior of the system, the lack of experimental data does not allow to corroborate
its validity or even the model itself. Recently, the model for diesel hydrodesulfurization of
Khadem-Hamedani et al. (2015) includes hydrodynamic parameters obtained by Deckwer et al.
(1982), Krishna and Ellenberger (1996), Maretto and Krishna (1999), de Swart and Krishna
(2002a,b) for F–T synthesis. It should be noted that good results were obtained in the model
validations using these approximations.
Moreover, in EBR (dp ≥ 200 μm), wake models allow to reproduce the holdup of the three phases
inside the reactor. These models have been tested experimentally for ambient conditions as well as
for high temperature and pressure considering homogeneous and heterogeneous bubble phases
(Jiang et al. 1997; Ruiz et al. 2004). Although these models have been implemented efficiently to
EBR under HCR conditions, they have not been used for SPR modeling.
7.2.6 Gas–Liquid Mass Transfer Coefficients

The two-film theory is used to describe the mass transfer flux between the gas–liquid interphase:
1 RT G Z P,T G 1
= + 7 29
K Li k Gi H i kLi
where KLi, kLi, and kGi are the overall, liquid side, and gas side mass transfer coefficients for the i
component on each resistance. In slurry columns, the gas–liquid coefficients depend on the gas and
liquid velocity, the geometry and operating conditions of the reactor, physicochemical properties of
the system, and can be affected by the presence of solids. Gas side resistance to mass transfer
becomes important during the absorption of highly soluble gases. In two-phase bubble columns,
typically the gas-phase side resistance is negligible, then the overall mass transfer is governed
by the liquid side and only the mass transfer coefficient kLi is considered:
1 1
= 7 30
K Li kLi
In slurry reactors applied to HCR systems, the gas phase is mainly hydrogen (H2), which has high
solubility, then the gas-phase side resistance is also negligible. For a better comprehension of the
gas–liquid mass transfer mechanisms, the parameters kLi and a should be considered separately
(Kantarci et al. 2005). Nevertheless, only few literature data were found. On the other hand, there
are several correlations proposed in the literature when estimating the G–L mass transfer coeffi-
cient along with the interfacial area (kLa). For example, the review published by Beenackers
and Swaaij (1993) includes different correlations for two-phase systems (BCR) and the discussion
of the influence of solid concentration on the G–L mass transfer coefficient for three-phase systems
(SPR) showing different adjusting parameters, while Fukuma et al. (1987) and Kim and Kang
(1997) presented correlations applied specifically to SBCR. More recently, considering the effect
of operating conditions (T and P), the correlations presented in Table 7.5 were developed. However,
the only correlation for high operating conditions that considers the effect of reactor geometry is
that proposed by Lemoine et al. (2008), which is based on the considerations reported by Behkish
et al. (2006) for gas holdup in large-scale systems. These two correlations have been applied suc-
cessfully to an industrialized F–T reacting system by Sehabiague et al. (2008). Generally, the mass
transfer coefficient is obtained through correlations as a function of superficial gas velocity, liquid,
and gas phase properties, and in some cases solids concentration in the slurry. Other important
hydrodynamic parameters such as bubble size and gas holdup are not considered directly in the
correlations. This may be due to the gas–liquid interfacial area having a relation a = 6εG/db, thus
the competing effect of these variables becomes insignificant in mass transfer coefficient. However,
if calculations of the mean bubble diameter are required, there are correlations reported in the lit-
erature that depend on the sparger specifications, such as the one proposed by Lemoine et al. (2008),
or the method proposed by Kim and Kang (1997) in which the bubble diameter is a function of the
Reynolds number based on the bubble diameter and the Schmidt number. The gas phase is usually
assumed to move as plug flow along the reactor due to the characteristics of slurry-phase HCR, so
calculations of mean bubble diameter are not strictly necessary in the modeling.
Table 7.5 Recent G–L mass transfer correlations.
Reference Experimental conditions Correlation
0.17 ≤ P(MPa) ≤ 0.79 − 2 84

Behkish et al. (2006) ρL
k Li a = 0 18Sc − 0 6 × ρG U G 0 49 − 2 66C V
e
0.05 ≤ UG(m/s) ≤ 0.25 Ci
0 ≤ CV(%Vol) ≤ 36
Lemoine et al. (2008) 0.1 ≤ P(MPa) ≤ 19.8 ρL 0 26 μL 0 12 εG 1 21 DMi 0 5 DC 04

k Li a = 06 14 × 104 × 0 68 Γ0 11 e − 2 66CV
0.003 ≤ UG(m/s) ≤ 0.574 σ 0L 52 ρG 0 06 U 0G 12 dS 0 05 T DC + 1
275 ≤ T(K) ≤ 538 03
ρL 1 22 μL 0 08 ρG 0 02 T 1 66 0 14 DC
0.0382 ≤ DC(m) ≤ 7.72 dS = 37 19 × UG × × 1 − εG 1 56
Γ − 0 2 εG − 2 29X W + 2 81C V + 2 77ρp dp
MM 0G 12 ρL 1 52 DC + 1
0 ≤ CV(%Vol) ≤ 36
0.16 ≤ μL(N/m × 10−3) ≤ 75
Ferreira et al. (2010) 293 ≤ T(K) ≤ 308 DL UGg − 0 24 0 036 9 66

k Li a = 5 18 651U 0G 87 μeff × 1 − dp 1 − εS
m π μL
UG > 0 0072
s
1 ≤ P(MPa) ≤ 3 − 1 193 2 303
Jin et al. (2014) ρL CV
k Li a = 3 051Sc − 0 734 × ρG U G 0 524
1−
298 ≤ T(K) ≤ 423 Ci 0 85
DC(m) = 0.10
0 ≤ CV(%Vol) ≤ 20
7.2.7 Gas–Liquid Equilibrium

In the proposed generalized model, it is necessary to calculate the gradient for the i component
between the gas and liquid phases. For that matter, the equilibrium concentration of the i compo-
nent is needed. When modeling heavy oil HCR in presence of excess hydrogen, in the majority of
cases the gas phase is considered to be mainly composed by H2 and in a lower concentration of H2S,
therefore the volatility of other components of the oil is not considered. Thus, there are two ways for
performing equilibrium calculation, in one of them the equilibrium concentration is obtained
through Henry’s law employing solubility coefficients. The common procedure is the calculation
of solubility coefficients from Korsten and Hofffman’s correlations. Though the correlation for H2S
solubility is not of great accuracy for the high temperatures commonly used in HCR reactors.
The second method consists of obtaining the Henry’s coefficients. Either from assuming local
equilibrium at the gas–liquid interface or through fugacity coefficients as shown by Mederos
et al. 2009. In both cases, the use of an equation of state is necessary. This represents an advantage
because volatile compounds can be considered in the calculations. However, this method implies
the use of an equation of state that involves the calculation of thermodynamic parameters such as
binary interaction parameters, which may lead to considerable errors in the calculations due to the
experimental nature of these parameters, as well as increasing the computing time.
7.2.8 Liquid–Solid and Gas–Solid Mass Transfer Coefficients

Usually, the liquid–solid mass transfer coefficient for SBCR and STR has the following form:
Sh = 2 + αScβ Re γp 7 31
where Sh and Sc are the Sherwood and Schmidt numbers. The constant parameters α, β, and γ
depend mainly on the system hydrodynamics. Several considerations have been taken into account
for estimating the particle Reynolds number Rep. Usually, the particle slip velocity was used in the
computing; however, due to the complexity and uncertainty reported in some systems, the
approach based on Kolmogoroff’s isotropic turbulence theory has been preferred in SBCR (Shah
et al. 1982; Beenackers and Swaaij 1993). In the case of STR, Pangarkar et al. (2002) presented a
review on different theories including Kolmogoroff’s turbulence theory and correlations based
on a dimensional approach. In both cases, the authors concluded that the correlation depends
on the operating parameters like impeller speed, particle loading, and reactor geometry and the
selection of the correlation should be based on similar operation conditions.
Furthermore in order to describe all the mass transfer phenomena occurring from gas to solid
phases, two mechanisms are considered. On the one hand, the gas phase is absorbed into the liquid
wake with a velocity controlled by KLi, then the gas absorbed in the liquid reaches the solid surface
and the intraparticle diffusion process takes place. In SBCR only this mechanism is commonly used
in the modeling and is justified by the assumption of equilibrium conditions between gas and liquid
phases. On the other hand, the other mechanism considers the gas phase directly reach the solid
phase through the agglomeration of solid particles between the gas and liquid interfaces; this leads
to the conversion rate of a reaction limited by G–L mass transfer enhanced by the presence of small
catalyst particles adhering to gas bubbles (van der Zon et al. 1999, 2001). Ruthiya et al. (2004) com-
pared the classical G–L–S model with a modified G–L–S including G–S mass transfer in a STR.
Their results confirmed an increase in reaction rate when solid agglomeration was taken into
account; however, the estimation of the mass transfer coefficient kSG is obtained only through
experimental fitting. General models should contain a priori all the mass transfer resistances.
However, no correlation for the gas–solid mass transfer coefficient in SBCR or STR has been found
in the literature, therefore when this type of approximation is applied, the value of kSG should be
obtained through experimental data, or introducing models that include the enhanced factor due to
particle agglomeration-adhesion.
7.2.9 Dispersion Coefficients

The extent of liquid mixing is essential in the modeling, design, and optimization of BCR and SBCR.
The degree of mixing in these reactors is quantified by the axial dispersion coefficient Dfa. Most of
the models applied to SBCR have been formulated under the ADM whereby this parameter
becomes overriding. Despite ADM has shown good results simulating the behavior of different sys-
tems, there has been no systematic effort made for modeling effectively the combined effect of
superficial gas and liquid velocity and column diameter on liquid axial dispersion (Moustiri
et al. 2001). Also, there is an open question on how pressure affects the axial dispersion coefficient
(Tang and Heindel 2006); however, it has been found that at high pressure in the homogeneous
regime, there is no effect of pressure on the liquid axial dispersion coefficient (Yang and Fan
2003), which is generally the case for HCR systems. These results were obtained through comparing
experimental values with the correlation of Joshi (1998) formulated for two and three-phase
systems:
DLa = 0 29DC U L + U C 7 32
So far this is considered one of the most complete correlations because it considers liquid and gas
velocities as well as variables related to the gas bubble size and bubble rising velocity. Following
these considerations, Moustiri et al. (2001) proposed a general correlation for BCR supported by
experimental data for different sizes of columns:
U L DC 1 3 Re L0 1
Bo = = −2 3
7 33
1 − εG DLa 1 1 Re L
1 − εG + Ga0 6
1 − εG 235 Fr1 3
G
This correlation has shown good agreement with experimental data while the operating condi-
tions of both are similar. In HCR systems as well as for hydrogenation systems where the homo-
geneous bubble flow prevails, a common approximation for the liquid axial dispersion coefficient is
considered with the correlation proposed by Deckwer et al. (1974).
Recently Khadem-Hamedani et al. (2015) have shown, for an HDS system under sever conditions
of temperature and pressure, that the correlation proposed by Deckwer et al. (1982) formulated for a
F–T system can be satisfactorily included in the modeling regardless of the type of flow regime
(homogeneous or heterogeneous). Other correlations are presented extensively in different works
starting with Shah et al. (1982), Fan (1989), Degaleesan and Dudokivic (1998), and recently with
Sivaiah and Majumder (2013), who present a summary of the typical correlations for three-phase
systems and their respective range of parameters.
Less investigations have been found dealing with the radial dispersion phenomena. In general, it
has been found that the radial dispersion coefficient of the liquid phase in liquid fluidized beds
increases with the particle size and gas velocity. Radial dispersion coefficients are well represented
by the ratio of the fluid velocities and of particle-to-column diameters based on the concept of the
isotropic turbulence model (Han and Kim 1990). Some of the correlations found in the literature are
presented in Table 7.6. On the other hand, Hillmer et al. (1994) observed experimentally that the
Table 7.6 Correlations for the liquid radial dispersion coefficient.
Reference Correlation
1 16
dp UL
Kang and Kim (1986) Per = 28 3
DC UL + UG
0 85 − 0 16
dp UL
Han and Kim (1990) Per = 7 28 Re 0L 21
DC UL + UG
0 58 0 691
dp U L εG + εL
Kang et al. (2003) Per = 0 086 Re 0L 21
DC UL + UG
1 303 − 0 056 − 0 311
dp dp U L ρS GS
Cho et al. (2005) Per = 8 47
DC μL U L ρS
type of flow inside liquid fluidized bed bubble column reactors satisfies the condition that the
Schmidt number is close to unity. Therefore, the dispersion coefficient could be expressed by:
μf
Df = 7 34
ρf
where μf is the sum of turbulent and laminar viscosity. This consideration was applied successfully
by Matos and Guirardello (2002) and Matos et al. (2009) in a CFD model for petroleum
hydrotreating.
Moreover, the gas phase has been traditionally modeled as a single phase typically in plug flow or
in some cases considering axial dispersion. Nevertheless, when a heterogeneous flow regime is con-
sidered, usually the two bubble class models are introduced, then the SB are characterized by a
completely mixed phase while the LB are modeled either in plug flow or with an ADM (Rados
et al. 2005). Rarely both phases are modeled using axial dispersion. It is known that in the gas phase,
the axial dispersion is caused by the interactions among bubbles of the same class, therefore the
bubble rise velocity is the characteristic variable in the gas dispersion. Yet, the effect of solids
on gas dispersion in three-phase systems is not fully known, hence in most of the cases, correlations
for solid-free systems are used instead. However, even for G–L systems, these correlations are
scarce, some of which are presented by Fan (1989). In most cases, only approximations to the
gas Peclet number are used. For example, for the heterogeneous flow regime, de Swart and Krishna
(2002a,b) assigned the value of 100 to the gas Peclet number due to their plug-flow regime, and the
axial dispersion of SB was assumed to be equal to the axial dispersion coefficient of the liquid phase
according to the study of Baten and Krishna (2001). This last assumption has been successfully
applied in different heterogeneous models, in some cases only the value of the Peclet number
for LB is modified introducing correlations for the axial dispersion coefficient, as shown by Kha-
dem-Hamedani et al. (2015) who used the correlation of Wilkinson et al. (1992) for single- and two-
bubble class models.
As mentioned before, in slurry systems the suspension of solids in the liquid phase is usually
modeled as a single pseudo-homogeneous phase where the solids velocity is assumed to be equal
to the liquid velocity. Also, a uniform concentration of the solids throughout the reactor is sup-
posed. Given the small size of the solid particles that are typically used in SBC reactors (dp ≤
50 μm), the pseudo-slurry assumption is probably reasonable. However, some studies have shown
that despite the small size of the solids, the axial profile of solids concentration in the slurry may not
be flat (Rados et al. 2005). When the system properties are such that uniform concentration cannot
be an accurate consideration, the SDM is suitable to solve this issue. There are different formula-
tions of the SDM, but in all of them, the axial dispersion coefficient for the solid phase is needed.
Unfortunately, there is not much information on this issue. Some of the most common correlations
are also presented in the works of Kojima et al. (1989), Fan (1989), and Kato et al. (1985).
7.2.10 Heat Transfer Coefficients

The heat transfer coefficients in SPR are many times larger than that for single-phase flow or even
in liquid–solid fluidization. Due to the large contact area among phases, a common assumption in
the modeling is that the gas and liquid phases are in thermal equilibrium. This leads to several sim-
plifications; for example, a global heat transfer coefficient h between the continuous phase (G + L)
and the wall is considered instead of one for each resistance (G–W, L–W). Also, in most of the cases
due to the small particle diameter, intraparticle temperature gradients are negligible; however, in
some cases, the heat transfer coefficient between L–S and G–S is introduced.
The heat transfer coefficient h in three-phase fluidized beds has been found to be an increasing
function of the respective gas and liquid velocities, the size and density of the particles, column
diameter, and the thermal conductivity and heat capacity of the liquid phase, but a decreasing func-
tion of the dynamic viscosity of the liquid, and probably of the diameter of a heating cylinder
immersed inside the bed (Fan 1989). The heat transfer coefficient has been successfully correlated
and reviewed by several investigators in terms of experimental variables such as gas and liquid velo-
cities, particle size, and liquid viscosity (Shah et al. 1982; Kim and Kang 1997; Jamialahmadi and
Müller-Steinhagen 1999). The range of application of the correlations could vary depending on the
system under study, therefore the lecture on those works is recommended. Also, semi-theoretical
correlations have been proposed based on the energy dissipation rate; however, for the design and
scale-up of multi-phase flow reactors, it is desirable to predict h from dimensionless correlations.
Recently the work of Kim et al. (2014) proposed a correlation for viscous systems where the effects
of superficial gas velocity (0.002–0.164 m/s) and solid load (0–20%) were examined. It is concluded
that the heat transfer coefficient between the immersed heater and the bed increases with increas-
ing gas velocity. On the contrary, it decreases with viscosity of the slurry phase. These are important
results because they could be applied to the HCR of heavy oils. Different correlations can also be
found in the work of Jhawar and Prakash (2011).
7.2.11 Example of Simplification of the Generalized Model

As mentioned before, a model that considers all the dynamic equations presented in Tables 7.2 and
7.3 along with the initial and boundary conditions in Table 7.4 is way too difficult to solve, therefore
considerations are needed to simplify the model equations. However, these considerations would
depend on the feed oil properties as well as catalyst characteristics and operating conditions, there-
fore it would be difficult to establish a general criterion of simplifications. Although some of the
general considerations are as follows:
Homogeneous suspension: In order to consider a slurry-phase reactor, it should be considered a
good homogeneity between solid and liquid phases as well as small particle size, negligible effects of
the terminal and settling velocity and that the solids follow the liquid motion, therefore a constant
solids dispersion inside the reactor could be assumed and the phases liquid and solid could be con-
sidered as one phase (slurry). On the other hand, for larger particle diameters, this simplification
should be avoided, and the model would be approximate to a three-phase reactor.
Homogeneous bubble flow: As explained before, due to the operating conditions as well as the
system properties in HCR of heavy oils, the typical flow regime is the homogeneous bubble flow. In
that case, the N + 1 model equations simplify only to two general mass balance equations, one for
the gas phase and another for the slurry phase.
Gas phase: As seen in previous sections, due to hydrogen being found in excess inside the reactor,
most models consider the gas phase as a mixture only of H2 and H2S. Thus, only two equations for
the gas phase are needed, along with the corresponding equations for these components in the liq-
uid phase. Mass transfer between gas and liquid phases for the other lumps is not considered. Also
in HCR systems, because the gas phase is mainly hydrogen (H2) with high solubility, the gas-phase
side resistance is negligible. Another consequence of the large amount of hydrogen in the reactor is
avoiding coke formation.
Kinetic model: As it is known there are different types of kinetic models with different complex-
ity, and some of them are based either on lumps or continuous mixtures. In any case, both models
implied the aggrupation of components of the oil with similar characteristics. The number of lumps
is fundamental in the simplification of the model equations. For example, the kinetic model of
Krishna and Saxena (1989) includes six lumps; therefore, the dimension of the slurry-phase equa-
tions increases to six (in case volatility of any of these lumps is introduced in the modeling, it would
be necessary to include the corresponding gas-phase equations for each volatile lump).
Use of correlations: In the modeling of reactors, a common simplification is the use of correla-
tions for the hydrodynamic parameters. This would allow to solve the model equations, avoiding
the use of turbulence equations such as in CFD models. In that case, both variables stay out of the
differentials and only the variation of the concentration or temperature are computed. However,
care must be taken to select the parameters that were obtained under similar conditions to those
used in the modeling system of interest.
Isothermal conditions: So far, all the models reported in the literature concerning heavy oil HCR
assumed isothermal conditions because of the good mixture and uniform distribution of temper-
ature achieved with these reactors. However, it is not a general consideration.
7.3 Simplified Models
7.3.1 SPR 1D Model

The goal of this section is to formulate a mathematical model for an SPR that can be applicable at a
laboratory scale. Some of the main considerations have been reported in the literature (Bahroun
et al. 2010; Maretto and Krishna 1999; Rados et al. 2005; Khadem-Hamedani et al. 2015; Calderon
and Ancheyta 2016; de Swart and Krishna 2002a,b) and are shown below:
• Gas- and liquid-phase velocities are constant along the reactor because the variations are quite
small and do not affect the conversion of the reactor (Matos and Nunhez 2006). Therefore, only
mass and energy balances with constant holdup will be included in the model.
•• Due to the small size of the reactor, only axial effects will be considered.
Due to the high solubilities of hydrocarbons and high mixture between liquid and gas phases,
there won’t be mass transfer limitations and the hydrocarbon will be saturated for each of the
gas-phase components.
• The catalyst particles are completely pulverized, therefore no intraparticle diffusion will be
considered.
•• The catalyst and the liquid phase are considered to be a homogeneous phase (slurry phase).
Experimental conditions are isothermal.
Based on the above considerations, the reactor model is:

NC , NR
∂x i ∂x i ∂2 xi
= −u + Da 2 + r i,k 7 35
∂t ∂Z ∂Z i, k
With the boundary and initial conditions:
t = 0, 0 ≤ Z ≤ L, x i = x in
i 7 36
∂x i
0 ≤ t ≤ t f , Z = 0, Da = u x i − x 0i 7 37
∂Z
∂x i
0 ≤ t ≤ t f , Z = L, =0 7 38
∂Z
where xi stands for the mass fraction of component i in the hydrocarbon mixture (hydrocarbon +
gas) of feed and product, while x 0i represents the mass fraction of component i at the inlet. u is the
superficial gas velocity and Da the axial dispersion coefficient. t stands for the time of the run, and Z
is the axial coordinate with L being the reactor length.
7.3.2 SPR 2D Model

To describe the operation of an industrial reactor, most of the considerations presented in the pre-
vious section are considered; however, since this case is an industrial reactor, the effects of temper-
ature variation as well as radial dispersion are included. Because these reactors have a large
diameter and operate at low surface velocities, it is possible that gradients may occur in the radial
direction. Hence, the model becomes:
NC, NR
∂x i ∂x i ∂2 xi ∂2 xi 1 ∂x i
= −u + Da 2 + Dr + + r i,k 7 39
∂t ∂Z ∂Z ∂r 2 r ∂r i, k
∂TL ∂TL kL ∂2 TL kL ∂2 TL 1 ∂TL

= −u + + +
∂t ∂Z εL ρL CpL ∂Z 2 εL ρL CpL ∂r 2 r ∂r
NC, NR 7 40
1 ha TL − TG
+ ri, k − ΔHk −
εL VR ρL CpL i, k εL VR ρL CpL
∂T G ∂T G kG ∂2 T L kG ∂2 T G 1 ∂T G ha T L − T G
= −u + 2 + + + 7 41
∂t ∂Z εG ρG CpG ∂Z εG ρG CpG ∂r 2 r ∂r εG V R ρG CpG
With the boundary and initial conditions:
t = 0, 0 ≤ Z ≤ L, 0 ≤ r ≤ R, x i = x in
i 7 42
t = 0, 0 ≤ Z ≤ L, 0 ≤ r ≤ R, T Li = T in
L 7 43
t = 0, 0 ≤ Z ≤ L, 0 ≤ r ≤ R, T G = T in
G 7 44
∂x i
0 ≤ t ≤ t f , Z = 0, 0 ≤ r ≤ R, Da = u x i − x 0i 7 45
∂Z
∂T L
0 ≤ t ≤ t f , Z = 0, 0 ≤ r ≤ R, kL = u T L − T 0L 7 46
∂Z
∂T G
0≤t ≤ t f , Z = 0, 0 ≤ r ≤ R, kG = u T G − T 0G 7 47
∂Z
∂x i
0≤t ≤ t f , Z = L, 0 ≤ r ≤ R, =0 7 48
∂Z
∂T L
0≤t ≤ t f , Z = L, 0 ≤ r ≤ R, =0 7 49
∂Z
∂T G
0≤t ≤ t f , Z = L, 0 ≤ r ≤ R, =0 7 50
∂Z
∂x i
0≤t ≤ t f , 0 ≤ Z ≤ L, r = 0, =0 7 51
∂Z
∂T L
0≤t ≤ t f , 0 ≤ Z ≤ L, r = 0, =0 7 52
∂Z
∂T G
0≤t ≤ t f , 0 ≤ Z ≤ L, r = 0, =0 7 53
∂Z
∂x i
0≤t ≤ t f , 0 ≤ Z ≤ L, r = R, =0 7 54
∂Z
∂T L
0≤t ≤ t f , 0 ≤ Z ≤ L, r = R, =0 7 55
∂Z
∂T G
0≤t ≤ t f , 0 ≤ Z ≤ L, r = R, =0 7 56
∂Z
where TL and TG represent the liquid- and gas-phase temperatures, respectively; kf, εf, ρf, and Cpf
are the thermal conductivity, phase holdup, density, and heat capacity of f phase; and Dr is the
radial dispersion coefficient and ha the heat transfer coefficient.
7.3.3 Continous Stirred Tank Reactor Model

One of the key factors in reactor modeling is the reactions involved. In this case, the main reactions
that take place are HDC and HDT. Therefore, to study reaction kinetics is also necessary to provide
a continous stirred tank reactor (CSTR) model. The considerations for this model are similar to the
ones explained in previous sections, adding the perfect mixing assumption. Then the model sim-
plifies to:
NC , NR
∂x i
i − xi +
= − mT x in r i,k 7 57
∂t i, k
where the first term of the left hand represents the changing of mass fraction for component i in
time. The first term of the right hand stands for the convective flux gradients, while the last term of
the right hand states the reaction rates. The initial conditions have the form:
x i 0 = x in
i 7 58
7.3.4 Parameters
All models that describe physical and chemical processes, regardless of their complexity, require
the knowledge of properties and parameters that characterize the system under study. However,
for crude oil HCR systems, experimentally obtaining parameters is impractical and expensive due
to the complexity of the hydrocarbon mixture and the conditions under which these systems are
operated.
There are different experimental methods that allow to approximate these parameters by corre-
lations. Within this experimental framework, it is common to use tracers, which consist of using
fluids at ambient temperature with characteristics like the original fluids at the operating temper-
ature and pressure. These procedures have been effectively applied in Fischer–Tropsch reaction
systems, which is why most of the correlations reported in the literature are for these kinds of sys-
tems, which operate at temperatures ranging between 250 C and pressures of 5 MPa. On the other
hand, the lack of correlations is evident when it comes to crude oil HCR systems, which operate at
temperatures above 350 C and pressures between 5 and 14 MPa. Due to the difficulty of obtaining
experimental information on these systems, an alternative is to use correlations formulated for sim-
ilar fluids at similar operating conditions. This will be the case for the axial and radial dispersion
coefficients, molecular diffusivity, and heat of reaction, as well as for the density and viscosity of the
hydrocarbon at the operating conditions. Besides, an equation of state and correlations will also be
used to calculate the equilibrium concentrations for the volatile components.
In these types of systems, a temporary solution is to select the correlations that have been for-
mulated for similar fluids in conditions similar to those of the system studied. This will be the case
for the axial and radial dispersion coefficients, molecular diffusivity, and heat of reaction, as well as
for the density and viscosity of the hydrocarbon at the operating conditions.
Eqs. (7.59)–(7.65) show the correlations to calculate density at operating conditions (Chen et al.
1995; Hillmer et al. 1994):
ρL T, P = ρL,0 + ΔρP − ΔρT 7 59
2
P P
ΔρP = 0 167 + 16 181 × 10 − 0 0425
− 0 01 0 299 + 263 × 10 − 0 0603ρL,0
1000 1000
7 60
− 2 45
ΔρT = 0 0133 + 152 4 × ρL,0 + ΔρP T − 520
7 61
− 8 1 × 10 − 6 − 0 0622 × 10 − 0 764 ρL,0 + ΔρP
T − 520 2
Here, ρL(T, P) represents the liquid density as a function of temperature and pressure, ρL,0 is the
liquid density at standard conditions (15 C and 101.3 kPa), P is the pressure in psia, and T is the
temperature in R. On the other hand, the liquid viscosity at operating temperature as a function of
API is given by:
− 3 444 a
μL T = 3 141 × 1010 T − 460 log 10 API 7 62
a = 10 313 × log 10 T − 460 − 36 447 7 63
Finally, the heat capacities are calculated as:

CpL = 0 35 + 0 55K 0 6811 − 0 308SG + 0 000815 + 0 000306 T 7 64
1 3
T MeABP
K= 7 65
SG
where K is the Watson factor and TMeABP is the average boiling temperature. In addition, an
equation of state is used to improve the model and the accuracy of the Henry coefficients. The
Peng–Robinson equation has proved to be quite accurate for HCR systems (Mederos et al. 2009)
RT a
P= − 7 66
v−b v + δ1 b v + δ2 b
T is the temperature (K), P is the pressure (Pa), R is the ideal gas constant (J/mol K), v is the molar
volume (m3/mol), and δ1 and δ2 are universal constants for the Peng–Robinson equation. a and b
are temperature-dependent parameters that are calculated with:
NCL, NRL
a = am = x i x k aik 7 67
i, k
NCL, NRL
b = bm = x i bi 7 68
i
With
1 2
aik = aki = aii akk 1 − k ik 7 69
where kik is the binary interaction coefficient whose values are taken from the literature. Addition-
ally, the compressibility factors can be obtained with:
3 2
ZL − 1 − B ZL + A − 2B − 3B2 Z L = AB − B2 − B3 7 70
And the fugacity and Henry’s constants:
NCL
bi L A 2 x k aik bi Z L + δ2 B
ln ϕLi = Z − B − ln Z L − B − k
− ln
b B δ2 − δ1 a b Z L + δ1 B
7 71
aP
A= 7 72
RT 2
bP
B= 7 73
RT
The calculation procedure for the thermodynamic parameters is well documented (Mederos et al.
2009; Kesler 1976). Lastly, the critical properties Tc and Pc as well as the accentric factor ωi and
molecular weight MWi can be estimated based on the boiling point Tb and specific gravity SG with:
105
T c = 341 7 + 811SG + 0 4244 + 0 1174SG T b + 0 4669 − 3 2623 7 74
Tb
0 0566 0 2898 0 11857
ln Pc = 8 3634 − − 0 2444 + + 10 − 3 T b
SG SG SG2
3 648 0 47227 1 6977
+ 1 4685 + + 10 − 7 T b 2 − 0 42019 + 10 − 10 T b 3
SG SG2 SG2
7 75
1 408 − 0 0163K
ωi = − 7 904 + 0 1352K − 0 007465K 2 + 8 359T br + 7 76
T br
7.3.5 Reaction Kinetics VR

Heavy oils are complex mixtures of a large number of k4 k3 k2 k1
components; therefore, the study of all the reactions
involved in HCR and hydrotreating is quite complicated.
VGO
For the case of HCR, there are different types of kinetic
k6 k7
models that have been reported to approximate all the k5
reactions (Ancheyta et al. 2005a,b). Despite the depend-
ence of the kinetic parameters on the properties of the
MD
feed and the use of an invariant distillation range of pro- k9
ducts, a pseudo-component-based model will be used to k8
represent the HCR kinetic due to its simplicity and utility
in simulation and control applications. The kinetic model
N
is presented in Figure 7.7. This model was proposed by
Sanchez et al. (2005) and consists of five pseudo-compo- k10
nents: residue (538 C+), vacuum gas oil (343–538 C),
middle distillates (204–343 C), naphtha (IBP-204 C)
G
and gases, and 10 reaction paths for HCR of all lumps,
as well as 10 reaction rate coefficients to be calculated. Figure 7.7 Five-lump kinetic model for
Recently Martínez et al. (2010) conducted an experi- hydrocracking reported Adapted from
mental study based on the kinetic model presented in Sanchez, et. al. (2005).
Figure 7.7 to obtain the 10 intrinsic reaction rate coeffi-
cients in a CSTR. The kinetic expressions for HCR reac-
tions are presented in Eqs. (7.77)–(7.81). Although most literature reports propose that HCR
reactions of heavy oils follow a first-order reaction, in this work the HCR reaction of VR was of
second order while the rest of the HCR reactions were assumed to be of first order (Martínez
and Ancheyta 2012; Ellenberger and Krishna 1994). The second-order reaction is attributed to
the broad distribution of reaction rate values of compounds present in the residue fraction, which
generates that the reaction rate declines faster than that of a simple first-order reaction because the
most-reactive compounds disappear first and fast, leaving the most-refractory compounds in the
residue (Sánchez and Ancheyta 2007).
All the reaction rate expressions involve the calculation of effectiveness factors and catalyst deac-
tivation function, due to the large particle size and the deposition of coke on the catalyst surface.
r R = − φHDC η k 1 + k 2 + k3 + k4 y2R 7 77
r VGO = φHDC η k 1 y2R − k 5 + k6 + k7 yVGO 7 78
r D = φHDC η k 2 y2R + k5 yVGO − k 8 + k 9 yD 7 79
r N = φHDC η k 3 y2R + k 6 yVGO + k8 yD − k 10 yN 7 80
r G = φHDC η k 4 y2R + k 7 yVGO + k 9 yD + k10 yN 7 81
On the other hand, as hydrogen is present in the system, hydrotreating reactions also occur simul-
taneously, so kinetic expressions for all these reactions are needed. It is known that thermal and
catalytic reactions act sequentially via thermal scission and hydrogenation. Thus, in order to take
into account catalytic and thermal reactions, Martínez and Ancheyta (2014) proposed the expres-
sions presented in Eqs. (7.82)–(7.86). As well as for HCR reactions, all the expressions for hydro-
treating reactions require the knowledge of catalyst effectiveness factors and deactivation rates.
nC nC
x S S x H2H2 ηT
r HDS = − φS ηS k CS 2 + kTs x S S 7 82
1 + K H2 S x H2 S
n ηT
r HDAsph = − φAsph ηAsph k CAsph x Asph
Asph Asph
+ kTAsph x Asph 7 83
ηT
r HDCCR = − φCCR ηCCR k CCR x nCCR
CCR
+ kTCCR x CCR
CCR
7 84
nHDN ηT HDN
r HDNNBN = − φHDNNBN ηHDNNBN k HDNNBN x HDNNBN
NBN
+ kTHDNNBN x HDNNBN
NBN
7 85
n ηT
rHDNBN = − φHDNBN ηHDNBN kHDNBN xHDN
HDNBN
BN
HDNBN
+ kTHDNBN xHDN BN
nHDN ηTHDN 7 86
− φHDNNBN ηHDNNBN kHDNNBN xHDNNBN
NBN
− kTHDNNBN xHDNNBN
NBN
Catalyst deactivation is originated by coke deposition; however, the estimation of this phenom-
enon is based on deactivation functions. For HCR reactions, a time-dependent nonselective expres-
sion was used, which means that if the selectivity of the HCR reactions is not affected by coke
deposition, then the deactivation constant φ is the same for all the reactions. On the contrary,
for hydrotreating reactions, a time-dependent selective model was used. In this case, it is necessary
to take into account the effect of coke deposition on each hydrotreating reaction, and then there
would be a deactivation constant for each reaction, (k DS, kDAsph, k DCCR, kDNBN, and k DBN). In both cases,
the deactivation function is presented in Eq. (7.87), where mi is the deactivation order for reaction i
(HDC, HDS, HDAsph, HDCCR, HDNBN, HDBN).
1
φ= mi 7 87
1 + k Di t
In the case of the CSTR, Eq. (7.87) can be used directly in the reactor model because the catalyst
does not leave the reactor during the operation. However, for the SPR, it is considered that the cat-
alyst does not remain in the reactor; on the contrary, it enters and leaves the reactor along with the
feed, then the deactivation function changes to Eq. (7.88).
1
φ= mi 7 88
Z
1 + k Di
u
Z
where represents the variation of time that the catalyst remains in the reactor. Also, for the SPR,
u
because the catalyst particles are very small, the effectiveness factor is considered to be the unity for
all the reactions (ηi = 1).
7.3.6 Solution Method

The dynamic models presented in the previous sections for both the CSTR and the SPR are repre-
sented by a set of nonlinear differential equations. In the case of the pilot- and industrial-scale SPR,
a model is obtained in PDEs since both the composition and the temperature, which are the process
variables, change with the axial and radial positions as well as over the course of the process time.
These equation systems need to be solved by means of a numerical method. One of the most used
methods is the Runge–Kutta method. However, this method can only be applied to systems of ordi-
nary differential equations, so it is necessary to perform a reduction or discretization of the problem
in the axial and radial directions. To simulate the behavior of the pilot scale SPR, the orthogonal
collocation method will be used to solve the 1D model, which is based on the concept of interpo-
lation of unequally spaced points. Collocation methods have a different perspective from finite
difference methods; they belong to the group of so-called spectral methods and are widely used in
chemical engineering. One of the advantages of this method is that it does not depend on the geom-
etry of the system and does not need equidistant spacing, unlike finite difference methods. Orthog-
onal collocation consists of choosing an orthogonal polynomial function (typically Legendre
polynomials) to approximate the solution of a differential equation on a given interval. On the other
hand, for the 2D model, a finite difference method will be used.
7.4 Numerical Simulations
This section presents the models and simulation results for the different reactors presented in the
modeling section. First, the model of a CSTR is presented, which includes mathematical expres-
sions for the kinetics of the HCR and hydrotreating reactions. Subsequently, the case of a tubular
SPRs is analyzed for the same reactant system where the reactor model is presented, and the
dynamic behavior is analyzed through numerical simulations. Due to certain experimental condi-
tions, it has been determined that the CSTR and the SPR can behave in a similar way. Experimental
data obtained in the CSTR were used, and the results were compared with the dynamic behavior of
the SPR. Additionally, the dynamic model of an industrial-scale SPR is presented. This reactor has
far from ideal behavior, thus the numerical simulations allow to analyze the behavior of this tech-
nology at a commercial level. Finally, a parametric sensitivity analysis is presented that will allow
exploring the uncertainty of the model as well as possible applications for process control.
7.4.1 Experimental Reactors

To compare the performance of both reactors, the same operating conditions were assumed: An
isothermal operation, 891 std m3/m3 of H2–oil ratio, 100 ml of catalyst loaded in the reactors,
and a constant pressure of 9.8 MPa. The dimensions of the SPR are 38 cm in length and 7.7 cm
in diameter, while the CSTR total volume is 1.7 l. It is also assumed that in both reactors, the heavy
oil feed is preheated and mixed with hydrogen and then passed through the reactor.
7.4.1.1 Dynamic Simulations of CSTR and SPR

Martínez and Ancheyta (2012) performed an experimental study to obtain the intrinsic reaction
rate coefficients for HCR reactions of heavy oil involving catalyst deactivation and the catalyst effec-
tiveness factor in a continuous STR. In their research, the five-lump 11 kinetic model shown in
Figure 7.7 was used. All the reaction rate coefficients and feed properties were taken from that
investigation to simulate and compare the performance of both CSTR and SPR reactors.
A weight hourly space velocity (WHSV) of 0.98 hours−1 and a temperature of 380 C were consid-
ered. The rest of the data can be found in detail in the work of Martínez and Ancheyta (2012).
The dynamic behavior of the small-scale reactors for HCR reactions is presented in Figure 7.8.
The continuous line stands for the simulated composition obtained at the exit of the SPR, and the
dotted line is the simulated composition obtained in the CSTR. In both simulations, the same vol-
ume and weight hourly space velocity were considered. The symbols in each plot represent the
experimental data obtained in the CSTR. As can be seen, the simulation of the CSTR is in good
agreement with the experimental points. It is also observed that the catalyst starts to deactivate,
from the beginning of the reactions, so that the rate of the different HCR reactions reduces until
it reaches the steady state near the 150 hours. This situation does not occur in the SPR because the
catalyst leaves the reactor along with the hydrocarbon; therefore, a major residue conversion is
reached in the SPR in a shorter period, around 5 hours.
60
40
20
50
40
30
Composition (wt %)
20
15
10
5.0
2.5
0.0
10.0
8.0
6.0
0 50 100 150 200
Time (hr)
Figure 7.8 Comparison of reactor performance of heavy oil hydrocracking. (–) SPR, (- -) CSTR, (Symbols)
Experiment. (o) Residue, (∗) Vacuum gas oil, (+) Middle distillates, (Δ) Naphtha, (∇) Gases.
For thermal and catalytic hydrotreating reactions, the effects of temperature and space velocity
were reported in the literature (Martínez and Ancheyta 2012; Martínez and Ancheyta 2014). All
reaction rate coefficients were obtained in the kinetic regime; therefore, they are considered to be
intrinsic. Since the catalyst used in the experiments is of commercial size, the effectiveness factor
is included in the kinetic model for the CSTR. All the parameters needed for the reactor simula-
tions are reported by Martínez and Ancheyta (2012). The dynamic simulation of the small-scale
reactors for hydrotreating reactions is presented in Figure 7.9. Similar to HCR reactions, in cat-
alytic hydrotreating, the conversion is favored in the SPR due to the small catalyst size and that
the catalyst does not remain inside the reactor, which disfavors catalyst deactivation. On the other
hand, the effects of catalyst deactivation and internal diffusion are less severe in hydrotreating
reactions compared with HCR reactions, since HCR reactions are considered to occur only on
the catalyst surface, and hydrotreating involves thermal and catalytic conversions. In compari-
son, both reactors have a high conversion at the beginning of the operation because the catalyst
has its major activity, as seen in Figures 7.8 and 7.9, for HCR and hydrotreating reactions, respec-
tively. In the case of the CSTR, it is observed that after two hours of operation, the highest con-
version is reached, and then the conversion decreases due to catalyst deactivation. Catalyst
deactivation is also observed in Figure 7.10 for reactions of HDS and HCR of residue, in which
a major effect is presented in HCR reactions due to the heavy molecules present in the liquid
phase being easily clogged in the catalyst due to the presence of coke. Evidently, as the reaction
10
0
Concentration (wt %)
40
20
0
30
20
10
Concentration (ppm)
6000
4000
2000
0
0 50 100 150 200
Time (hr)
Figure 7.9 Comparison of reactor performance of heavy oil hydrotreating. (–) SPR, (- -) CSTR, (Symbols)
Experiment. (o) Sulfur, (∗) Asphaltenes, (+) CCR, (Δ) Basic nitrogen, (∇) Non-basic nitrogen.
time increases, the catalyst deactivation will also increase, generating a negative effect in reactions
conversion. In the SPR, the reacting system needs only around two hours to reach the steady state,
after that the conversion remains constant while fresh catalyst continuously enters the reactor as the
deactivated catalyst leaves it. This behavior is also observed in Figures 7.11–7.13, in which the
dynamic axial profile for HCR and hydrotreating reactions, as well as for hydrogen and hydrogen
sulfide, is presented. As can be seen, another major conclusion is that in the SPR, a constant axial
profile is presented, thus SPR can be modeled as a CSTR.
Figures 7.14 and 7.15 compare the yield of each lump for the conversion of feedstock as a function
of the yield of each lump vs the conversion of residue in the SPR and CSTR, respectively. As can be
observed in the SPR, a high feed conversion is obtained. The residue can reach a constant 60% of
conversion producing lighter fractions after 200 hours of operation, producing mainly vacuum gas
oil and distillates that change from 32 and 9 wt.% to 50 and 18 wt.% in the product, respectively.
A low production of naphtha and gases is also observed. On the other hand, the CSTR present a
similar selectivity; the main difference is observed after a long time of operation, at which the con-
version of residue decreases after it reaches its higher value, around 58% of residue conversion. This
phenomenon arises from catalyst deactivation, as the catalyst remains inside the CSTR during oper-
ation, leading to significant deactivation caused by coke deposition phenomenon is originated due
to catalyst deactivation since in the CSTR, the catalyst remains inside the reactor during the oper-
ation then there is a great deactivation due to coke deposition.
60
50
Sulfur Conversion (wt%)
40
30
20
10
60
Residue Conversion (wt%)
50
40
30
20
10
0
0 5 10 15 20
Time (hr)
Figure 7.10 Influence of catalyst deactivation in hydrodesulfurization and hydrocracking reactions.

(–) SPR, (- -) CSTR.
7.4.1.2 Steady-State Simulations of a SPR

Once the SPR model was validated with the CSTR experimental data, it was used to simulate all the
HCR and hydrotreating reactions simultaneously. The feed velocity (WHSV) and temperature vary
from 0.98 to 2.56 hours−1 and from 380 to 420 C, respectively, as shown in previous experimental
reports (Martínez and Ancheyta 2012; Martínez and Ancheyta 2014). Figures 7.16 and 7.17 show
the simulation results of hydrocarbon and impurities compositions at the exit of the reactor once
the steady state has been reached. Figure 7.12 shows the steady-state lumps composition. Since res-
idue is only present as a reactant in the kinetic model, a linear effect between the WHSV and tem-
perature variation is observed. In this case, increasing both the residence time of the feed and
temperature in the reactor increases the residue conversion; therefore, a low composition of residue
would be present in the hydrocarbon product. On the contrary, the vacuum gas oil appears in the
kinetic model as both reactant and product; therefore, VGO conversion depends on the selectivity
of residue and lighter fractions. For WHSV between 0.98 and 2.56 hours−1 and temperatures below
410 C, a linear function in the VGO composition is observed, which indicates that the rate of HCR
of residue is faster than that of the HCR of VGO. For temperatures higher than 410 C, low com-
position of VGO is observed, which is sign of a high HCR of this pseudo-component to produce
lighter fractions, such as middle distillates, naphtha, and gases. At a steady state, the light fractions
Residue (wt %)
50
25
50
VGO (wt %)
40
30
Distillates (wt %)
20
10
0
Naphtha (wt %)
10
Gases (wt %)
6
40
2
20 1
0 0
Z (cm) Time (hr)
Figure 7.11 Dynamic axial profile in the SPR for hydrocracking reactions.
such as distillates, naphtha, and gases exhibit a constant and linear behavior. In those cases, as well
as residence time and temperature increases, the conversion of residue and vacuum gas oil is
higher, and a major composition of these fractions is present in the product. A similar effect is
shown in Figure 7.17 for impurities contents, in which a high removal of impurities such as sulfur,
asphaltenes, Conradson carbon, and nitrogen is observed as temperature and residence time
increase.
6
Sulfur (wt %)
4
H2S (wt %)
12E3
Nitrogen (ppm)
8E3
4E3
Asphaltenes (wt %)
30
20
10
CCR (wt %)
20
15
10
40
2
20 1
0 0
Z (cm) Time (hr)
Figure 7.12 Dynamic axial profile in the SPR for hydrotreating reactions.
7.4.2 Industrial-Scale Reactor

Unfortunately, the literature only reports operating temperature and pressure ranges as well as
approximate conversions reached by the residue (538 C+) in a stable operation, as summarized
in Table 7.1 for the different technologies documented by Zhang et al. (2007) and Bellussi et al.
(2013). Consistent data of a dynamic operation do not exist as far as the knowledge of the authors;
thus, an alternative to validate the model is to compare the predictions of the two-directions model
7.4
Hydrogen (wt %)
7.2
6.8
6.6
4
Hydrogen sulfide (wt %)
0
40
30 2
20 1.5
1
10 0.5
Z (cm) 0 0 Time (hr)
Figure 7.13 Dynamic axial profile of gases in the SPR.
presented above with predictions of a previously validated model used to describe the behavior of
an experimental reactor. In such a case, the reactor presented in the previous section will be used as
a reference. It is important to mention that this experimental operation was performed isothermally
(the energy balance is not required) in an SPR of 38 cm in length and 7.7 cm in diameter at 891 std
m3/m3 of H2 − hydrocarbon ratio with 100 ml of catalyst loaded in the reactor, and a constant pres-
sure of 9.8 MPa. The heavy oil feed was preheated, mixed with hydrogen, and fed to the reactor with
an LHSV of 0.5. Although the model of the experimental reactor is based only on axial dispersion, it
is expected that the predictions of both models are similar, since in the experimental reactor the
diffusive effects in the radial direction are minimal due to the high axial velocity and the small
diameter size. Such results are shown in Figure 7.18, in which the error variable is defined as
the difference between the prediction with the ADM and the prediction with the two directions
model. As can be seen, if the order of magnitude of the error is near zero, then both models predict
the same response.
Once the model has been verified, what follows is to extrapolate the behavior of the reactor from
experimental to industrial scale. Figure 7.19 shows a comparison between the lump composition as
50
Residue (wt%)
35
20
50
VGO (wt%)
40
30
Distillates (wt%)
18
14
10
4
Naphtha (wt%)
0
8.5
Gases (wt%)
7.5
6.5
0 10 20 30 40 50 60
Figure 7.14 Yield of hydrocracked lumps as a function of residue conversion in the SPR.
a function of residue conversion (Figure 7.19a) and as a function of time (Figure 7.19b) for the
experimental and industrial reactors. It is expected that the commercial-size reactor exhibits better
conversions than the experimental-size reactor, even if they are simulated for the same type of cat-
alyst and operating conditions, since the industrial reactor operates adiabatically and the experi-
mental reactor isothermally. This was indeed corroborated since the industrial reactor reached
70% residue conversion compared with 60% achieved in the experimental reactor, as can be
observed in Figure 7.19a. This increase in conversion is mainly because in the industrial-scale reac-
tor, there is a better flow regime, which increases the dispersion of the components in the liquid
phase and directly promotes the HCR reactions. Figure 7.19b shows another interesting effect,
which is that since the industrial-scale reactor reaches higher residue conversion, the composition
of vacuum gas oil increases, which leads to a higher vacuum gas oil HCR that contributes to a
higher composition of middle distillates and naphtha during the operation.
50
Residue (wt%)
35
20
50
VGO (wt%)
40
30
Distillates (wt%)
18
14
10
4
Naphtha (wt%)
0
8.5
Gases (wt%)
7.5
6.5
0 10 20 30 40 50 60
Figure 7.15 Yield of hydrocracked lumps as a function of residue conversion in the CSTR.
7.4.2.1 Dynamic Simulations of the Industrial Slurry-Phase Reactor

In order to simulate the operation of a commercial-size reactor (20 m in length and 2 m in
radius), an H2-to-oil ratio of 2000 std m3/m3, 1% of catalyst content in the slurry phase, a con-
stant pressure of 13 MPa in the reactor, an inlet feed and initial temperature of 420 C, and an
LHSV of 0.4 hours−1 were assumed as operating conditions. Kinetic parameters for HCR and
hydrotreating reactions and feed properties used in the simulations were reported by Martínez
and Ancheyta (2012). As the model is proposed to describe a commercial unit, it would be nec-
essary to analyze the temperature profile distribution since this type of reactor operates adiabat-
ically. Figure 7.20 shows the dynamic behavior of reactor temperature in the axial and radial
directions. In the upper part of the figure, the temperature surface at the initial time is presented,
in which it is observed that the temperature remains constant throughout the reactor. The fol-
lowing surfaces show the temperature profile at 6, 12, 24, 30, and 42 minutes, while the surface
at the bottom shows the temperature distribution at steady state after two hours of operation.
40
Residue (wt %)
20
50
VGO (wt %)
48
46
25
Distillates (wt %)
20
15
10
Naphtha (wt %)
15
Gases (wt %)
10
5
4
420
2 410
400
0 390
380
WHSV (hr–1) Temperature (°C)
Figure 7.16 Steady-state simulations of hydrocracking reactions.
From this figure, it is possible to appreciate in a remarkable way that, contrary to what would be
expected in a large-diameter reactor, the temperature profile remains practically constant in
the radial direction during the whole time of operation. This effect is not presented in the axial
profile in which a temperature change can be observed from the inlet of the feed through the
output of the products. This temperature gradient is approximately 1–2 C. Nevertheless, for
285
7.4 Numerical Simulations
420
Temperature (°C)
Steady-state simulations of hydrotreating reactions.

400
380
0
WHSV (hr–1)
2
Figure 7.17
4
20
10
20
10
6000
4000
2000
4
Sulfur (wt %)
Asphaltenes (wt %) CCR (wt %)
Nitrogen (ppm)
× 10–7
5
4
Difference in yield composition
3
1D–2D models (wt %)
–1
–2
–3
–4
Time (hr)
Figure 7.18 Dynamic error of residue composition between the prediction with the axial dispersion model
and the axial and radial dispersion model.
simplification purposes, the operation can be considered as almost isothermal, which is one of
the main characteristics of SPRs.
Figures 7.21 and 7.22 show the dynamic profiles of HDC and HDT reactions, respectively, at 0, 6,
12, 24, 30, and 42 minutes and at steady state after two hours of operation. In the case of HDC reac-
tions, a decrease in the residue content in the feed of approximately 30% can be observed, while the
intermediate fractions of VGO, middle distillates, and naphtha present an increase in their com-
position until the system reaches the steady state. Naphtha is a product of the HCR of the heavier
components, with three direct pathways for its formation with reaction rates (k3, k6, and k8). While
its decomposition to light gases occurs with a reaction rate (k10), under the considered operating
conditions, the reaction rate of naphtha formation is greater than its decomposition, which is
the reason why there is a considerable increase in the composition of naphtha in the steady-state
operation. On the other hand, similar to the surface response for temperature, the composition of
each of the pseudo-components remains constant in the radial direction, while the axial composi-
tion presents variations of around 1 wt.% for each time of operation. This small change in the com-
position with respect to the entrance and the exit is due to the low feed velocities and high residence
times necessary to achieve high conversion rates. Similar behavior can be observed in Figure 7.22,
in which the dynamic variations of sulfur, asphaltenes, CCR, and nitrogen are presented. In such
cases, there is a decrease in the composition of around 60, 65, 43, and 40%. As well as for the HDC,
the composition of HDT components also remains constant in the radial position.
Figure 7.23 shows the surface of residue conversion as a function of pressure and temperature. As
expected, an increase in temperature would lead to an increase in residue conversion. The effect of
pressure is linear since in the model considerations hydrogen is fed in excess. According to the
information reported in Table 7.1 for industrial technologies, for temperatures between 460 and
480 C, residue conversions of more than 95% can be reached. However, although most of the res-
idue has been hydrocracked, the formation of light gases increases considerably as seen in
(a)
60
50
Yield composition (wt %)
40
30
20
10
0
0 10 20 30 40 50 60 70
Residue (538 °C+) conversion (wt%)
(b)
60
Yield composition (wt %)
40
20
0
0 1 2 3 4 5 6
Time (hr)
Figure 7.19 Yield composition as a function of time and residue (538 C +) conversion. (o) Residue, ( )
Vacuum gas oil, (∇) Middle distillates, (Δ) Naphtha, (∗) Gases. (−) Industrial-scale reactor, (−−) Experimental
scale reactor.
Figure 7.24 due to thermal effects. It can also be observed that an optimal operation that promotes
high composition of valuable products can be reached at temperatures between 400 and 440 C.
7.4.2.2 Sensitivity Analysis for the Industrial Slurry-Phase Reactor

Parametric sensitivity is a tool that allows to analyze the effect of the different process variables
on the dynamic behavior and steady state of a process and therefore the predictions of the
(a)
421
420
419
(b)
420.4
420.2
420
(c)
421
420.5
420
(d)
Temperature (°C)
422
421
420
(e)
422
421
420
(f)
422
421
420
(g)
422
421
420
1 1
0.5 0.5
0 0
r Z
Figure 7.20 Dynamic profile of reactor temperature in axial and radial positions. (a) 0, (b) 6, (c) 12, (d) 24, (e)
30, (f ) 42, and (g) 120 minutes.
mathematical model, whether they are process variables (space velocity, catalyst fraction, H2-to-
oil ratio, initial reactor temperature, feed inlet temperature, and reactor operating pressure)
or model parameters in which there is a certain degree of uncertainty, for example, the heat
of reaction, the axial dispersion coefficient, and the convective coefficient of heat transfer.
Figures 7.25–7.27 show the response behavior for the variables and parameters mentioned
above. The values that were used as a base case scenario are those proposed in previous
sections for the simulation of the commercial unit. From the base case solution, all the para-
meters of the model were varied by ±50% individually, thus the change percentage is
52 34 12 0.165 7
50 33 10 0.16 6.9 (a)
48 32 8 0.155 6.8
48.4 34.3 10.2 0.44 7.03

48.2 34.2 10.15 0.42 7.02 (b)
48 34.1 0.4 7.01
46.6 35.4 10.8 0.7 7.15
46.4 35.2 10.7 0.65 (c)

46.2 35 7.1
Middle distillates (wt %)
Naphtha (wt %)
Residue (wt %)
Gases (wt %)
VGO (wt %)
44 37 12 1 7.35
36.8
43.5 11.5 0.95 7.3 (d)
36.6
43 11 0.9 7.25
43 37.5 12 1.2 7.4

42.5 1.1 7.35 (e)
42 37 11.5 1 7.3
42 38.5 12.5 1.4 7.5

41 38 1.3 7.45 (f)
40 37.5 12 7.4
38 40.5 13.5 2 7.7

37 40 1.8 7.65 (g)
36 39.5 13 1.6 7.6
1 1 1 1 1 1 1 1 1 1
0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5
r 0 0 Z r 0 0 Z r 0 0 Z r 0 0 Z r 0 0 Z
Figure 7.21 Dynamic profile of pseudo-components for HDC reactions in axial and radial positions. (a) 0, (b) 6,
(c) 12, (d) 24, (e) 30, (f ) 42, and (g) 120 minutes.
defined as (ybase case − ychanged)/ybase case. In the case of temperatures and operating pressure,
simulations were carried out between 400 and 440 C and from 9 to 17 MPa.
Figure 7.25 shows the variation in the steady-state prediction for changes in the LHSV, catalyst
fraction, and H2-to-oil ratio, from which it was concluded that the HDC and HDT reactions are
more sensitive to changes in the solid fraction and H2-to-oil ratio than LHSV. On the one hand,
increasing the catalyst content increases the HCR mainly of the heavier fraction, which finally
decreases its composition and gives rise to the increase in the composition of the intermediate frac-
tions. On the other hand, increasing the H2-to-oil ratio favors the conversion of the intermediate
fractions. Another favorable effect on the conversion of residue is the increase in the residence time
of the oil in the reactor, which is achieved by decreasing the LHSV. In this case, decreasing LHSV
would increase the conversion up to 6% with respect to the previous results. In the case of HDT
reactions, a similar effect is observed for LHSV and H2-to-oil, however with a lower degree of var-
iation. Finally, it can be observed that the effect on the exit temperature is minimum for the three
process variables, where the percentages of change are less than 1%. Figure 7.26 shows the changes
in the process variables for variations in the temperature and operating pressure, as well as the feed
temperature, where it can be observed that these three process variables are the ones that most
affect the conversion of the HDC, HDT, and output reaction temperature, generating variations
between ±15%. Finally, in Figure 7.27, the variations in the conversion are presented in case of
uncertainty in any of the parameters of the model. In general, it can be observed that the errors
that occur for any of the parameters are minimal, which means that the accuracy of the predictions
of the model is not affected by using correlations that have been formulated for other systems or for
× 104
6.18 30.7 22 1.0901

6.17 30.6 20 1.09 (a)
6.16 30.5 18 1.0899
5.2 27.2 19.5 9800

5.15 27 19.4 9700 (b)
5.1 26.8 19.3 9600
5 24.5 18.4 9200

4.5 24 18.2 9000 (c)
4 23.5 18 8800
Nitrogen (ppm)
Sulfur (wt %)
Asph (wt %)
CCR (wt %)
4.4 20.5 16.5 8200
4.35 20 8000 (d)
4.3 19.5 16 7800
3.4 19 16 8000
3.2 18 15.5 7800 (e)
3 17 15 7600
3.5 17 15 7600
3 16 14.5 7400 (f)
2.5 15 14 7200
3 13 13 7000
2.5 12 12.5 6800 (g)
2 11 12 6600
1 1 1 1 1 1 1 1
0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5
r 0 0 r 0 0 r 0 0 r 0 0
Z Z Z Z
Figure 7.22 Dynamic profile of components for HDT reactions in axial and radial position. (a) 0, (b) 6, (c) 12, (d)
24, (e) 30, (f ) 42, and (g) 120 minutes.
Residue (538 °C+) conversion (%)
100
95
90
85
80
75
70
0
10
480
20 440 460
420
30 400
380
Pressure (MPa) Temperature (°C)
Figure 7.23 Pressure and temperature effects on residue conversion at steady state.
7.5 Conclusions 291
80
70
60
Composition (wt %)
50
40
30
20
10
0
380 400 420 440 460 480 500
Temperature (°C)
Figure 7.24 Temperature effect on lumps compositions. (o) Residue, ( ) vacuum gas oil, (∇) middle
distillates, (Δ) naphtha, (∗) gases.
utilizing constant values for these parameters. In a general way, the process variable most sensitive
to these changes is the residue fraction in the case of HDC and the composition of asphaltenes for
HDT. A particular case is the outlet temperature of the reactor, which is greatly affected when
changes occur in the initial temperature of the reactor or the inlet feed temperature.
7.5 Conclusions
Mathematical modeling of SPR is a difficult task, mainly due to the complexity of the hydrodynam-
ics involved. It has been found that the main variable is the gas velocity, which has an influence on
almost all the parameters involved in the model such as the flow regime, holdup, dispersion and
mass transfer coefficients, and even particle settling velocities. The usual modeling configuration is
in the form of SBCR in which ADMs are the most common formulation. Although there is much
uncertainty in the parameters used in these models, ADMs have been preferred and extensively
studied for different types of synthesis as well as for hydrogenation reactions. In addition, under
the conditions typically set for these reacting systems, the flow approximation commonly used
is the heterogeneous regime for the gas phase, divided as the SB phase moving with constant veloc-
ity along the reactor and the LB phase.
Moreover, only a few authors have dealt with HCR reactions in SPR, though in most cases using
the CFD modeling approach despite the computational effort. This may be induced by the uncer-
tainty of utilized correlations for hydrodynamic parameters that were obtained under ambient con-
ditions, at severe conditions of temperature and pressure. However, some of the correlations
presented in previous sections have been successfully implemented in industrial operations of
F–T synthesis as well as for HDS systems.
One of the main disadvantages of modeling HCR in continuous operation is the lack of exper-
imental information. Therefore, the existing models have been validated using experimental infor-
mation obtained for different systems. For example, Matonis et al. (2002) and Matos et al. (2009)
20 40 4
HDC Fraction Change (%)
10 20 2
0 0 0
–10 –20 –2
–20 –40 –4
6 0.4 0.2
HDT Wt% Change (%)
4 0.2 0.1
2 0 0
0 –0.2 –0.1
–2 –0.4 –0.2
× 10–3 × 10–3 × 10–3

Outlet Temperature Change (%)
4 2 2
2 1
1
0
0
–1 0
–2
–2
–1
–4 –3
–6 –4 –2
–50 0 50 –50 0 50 –50 0 50
LHSV Change (%) Catalyst Fraction Change (%) H2/Oil Change (%)
Figure 7.25 Sensitivity analysis of HDC; (o) residue, (Δ) vacuum gas oil, (×) middle distillates, (+) naphtha,
(∗) gases, HDT; (o) sulfur, (Δ) asphaltenes, (×) CCR and outlet temperature for LHSV, catalyst fraction and
H2-to-oil ratio changes.
compared their hydrodynamic models with the information previously obtained for air–water sys-
tems under different operating conditions, and after that they simulated the behavior including the
HCR kinetics. So far, only the ADM of Khadem-Hamedani et al. (2015) for HDS has been validated
with experimental data obtained previously by Deng et al. (2010) for this system. In this case, it
should be noted that favorable results were obtained with heterogeneous models.
Furthermore, only a few models include an equation for the solid dispersion inside the reactor for
HCR systems, mainly because slurry particles, as long as they are small (dp ≤ 60 μm), follow the liquid
motion except at very high slurry loadings exceeding 20–30%. Due to the operating conditions, all the
models developed for HCR in SPR are formulated under the homogeneous bubble flow.
HCR as well as catalytic and thermal hydrotreating were modeled in small-scale isothermal
CSTR and SPR based on intrinsic reaction rate coefficients involving catalyst deactivation and cat-
alyst effectiveness factors for each of the reactions. The dynamic simulations have shown similar
behavior in the CSTR and SPR. As expected, higher conversion of residue is reached in the SPR
(effectiveness factor equals one) and deactivation of catalyst is not taken into account. The results
7.5 Conclusions 293
–3
0.5 100 2 × 10
0 50 1
–0.5 0 0
–1 –50 –1
–1.5 –100 –2
3 60 0.1
HDT wt% Change (%)
2 40 0.05
1 20 0
0 0 –0.05
–1 –20 –0.1
× 10–5
0.1 10 5
0.05 5
0 0 0
–0.05 –5
–0.1 –10 –5
380 400 420 440 460 380 400 420 440 460 14 16 18 20
Initial Reactor Initial Feed Pressure (MPa)
Temperature Change (%) Temperature Change (%)
(∗) gases, HDT; (o) sulfur, (Δ) asphaltenes, (×) CCR and outlet temperature for inlet reactor temperature,
inlet feed temperature, and pressure changes.
of steady-state simulation have shown that it is possible to approximate an SPR for HCR with an
ADM with appropriate reaction rate coefficients.
As expected, for experimental purposes, the CSTR has some advantages compared with the SPR.
It promises lower cost in the experimental framework due to the amount of catalyst used. It was also
found that for modeling purposes, the experimental SPR can be conveniently approximated by
means of an ADM; however, to reduce its complexity, the SPR can be approximated to an ideal
CSTR model.
A commercial-size SPR for heavy oil HCR and hydrotreating was also modeled, considering axial
and radial dispersion effects for the state variables. The model equations were partially discretized
with finite central differences in the positional derivatives and adequately solved by a Runge–Kutta
method coded in MATLAB in subroutine ode32s and ode45 with a computing time of around 3 min-
utes for 10 and 5 axial and radial nodes. The operation was assumed to be non-isothermal; however,
simulations have shown that the reactor operates nearly isothermal, with a difference between inlet
and outlet temperature of 1–2 C. Also, the dynamic and steady-state simulations have shown that a
0.3 0.15 0.04

0.2 0.1 0.02
0.1 0.05 0
0 0 –0.02
–0.1 –0.05 –0.04
0.3 0.15 0.05

HDT Fraction Change (%)
0.2 0.1
0.1 0.05 0
0 0
–0.1 –0.05 –0.05

×10–3
0.01 2 0.1
0 0 0.05
–0.01 –2 0
–0.02 –4 –0.05
–0.03 –6 –0.1
–50 0 50 –50 0 50 –50 0 50
Da Change (%) ha Change (%) ΔH Change (%)
(∗) gases, HDT; (o) sulfur, (Δ) asphaltenes, (×) CCR and outlet temperature for axial dispersion
coefficient (Da), convective heat transfer coefficient (ha), and heat of reaction (ΔH) changes.
constant radial profile can be exhibited due to the large H2-to-oil ratio and the high residence time of
the feed within the reactor. Therefore, an ADM represents a good approximation to describe the sys-
tem for dynamics and steady-state performances. It was also shown by a sensitivity analysis that the
parameters the model requires can be well estimated through existing correlations since variations in
the values of these parameters do not produce a high variation in the model predictions, in the case of
axial dispersion coefficient, convective heat transfer coefficient, and heat of reaction.
Nomenclature
Symbols
A Bed-wall reactor heat transfer area, cm2
a Interfacial area per unit reactor volume, cm2 cm2r
f
CT Total concentration of f phase, g/cm3
C∗i Concentration of compound i in the liquid in equilibrium with the gas phase, mol i cm3L
Nomenclature 295
C fi Molar concentration of component i in the f phase, mol i cm3f

C Sf ,iinn Molar concentration of component i inside the solid filled with f phase, mol i cm3f
C Sf ,isur Molar concentration of component i at surface of the solid covered by f phase, mol i cm3f
CS Solids concentration, g cm3SL
CV Volumetric solids concentration, in the slurry cm3S cm3SL
Cpf Specific heat capacity of f phase, J/gf K
f
DM,i Molecular diffusion coefficient of component i in the f phase, cm3f cmS s
Dfa Mass axial dispersion coefficient of f phase, cm2r s
Dfr Mass radial dispersion coefficient of f phase, cm2r s
Dfei Effective Fickian diffusivity of component i inside a porous catalyst, cm3f cmS s
DC Column diameter, cm
dp Particle diameter, cm
dS Bubble mean diameter, cm
F Interfacial drag force, N/m3
g Gravity constant, cm/s2
He,i Henry’s law coefficient of component i, MPa cm3L mol i
hGS Gas-solid heat transfer coefficient, J cm2f s K
hGW Gas-wall heat transfer coefficient, J cm2f s K
hLW Liquid-wall heat transfer coefficient, J cm2f s K
K Overall LB–SB mass transfer coefficient, cm3LB cm2I s
k Intrinsic reaction rate constant, (for first-order reaction) 1/s
k GS
i
Gas-solid mass transfer coefficient for i component, cm2G cm2S s
k LS
i
Liquid–solid mass transfer coefficient for i component, cm2L cm2S s
kLi Mass transfer coefficient from gas–liquid interface to liquid phase, cm3L cm2I s
KL Overall gas–liquid mass transfer coefficient, cm3L cm3I s
L Equivalent longitude, cm
LB Large bubbles
MMf Molecular weight of f phase, g/mol
N Molar flux, mol/cm2
N fC Number of compounds in the f phase
NR, f Number of reactions in the f phase
P Pressure, MPa
PS Saturation pressure, MPa
PT Total pressure, MPa
R Gas law constant, J/mol K
r Referring to radial position, cm
r jf Rate of reaction j per unit of catalyst mass in the f phase, mol i/gS s
SB Small bubbles
t Time, s
Tf Temperature of f phase, C, K
Uf Superficial velocity of f phase, cm/s
uf Velocity of f phase, cm/s
utS Solid terminal settling velocity, cm/s
V Reactor volume, cm3r
Vf Volume of f phase, cm3f

W Weight fraction of solid in the feed slurry, gS/gSL
XW Weight fraction of the primary liquid in the mixture (1 ≤ XW ≤ 0.5), dimensionless
yf Dimensionless concentration of f phase, C fi Cfi 0
Z Compressibility factor, dimensionless
z Axial position, cm
Subscripts
a Referring to axial position
G Referring to gas phase
I Referring to interphase
i Referring to i component
L Referring to liquid phase
p Referring to solid particles
S Referring to solid phase
Ssup Referring to solid surface
SL Referring to slurry phase
W Referring to cooling fluid
Superscripts
∗ Equilibrium conditions
f Referring to f phase
Sinn Solid inner
Ssur Solid surface
0 Inlet conditions
eff Effective
Dimensionless groups
Bo Bodenstein number, U dpe/D
Da Damköler number, k εL H/UG0
Fr Froude number, U/(g DC)1/2
Ga Galilei number, g L3 ρ2/μ2
Pe Peclet number, U Le/DC
Re Reynolds number, U Le ρ/μ
Sc Schmith number, μ/ρ Df
Sh Sherwood number, kf dp/Df
St Stanton number, kLa H/UG
Greek letters
α Gas-phase contraction factor, dimensionless
αi, βm,i Matrix constant parameter
References 297
ΔHr, j Heat of reaction j, J/mol

ΔHv, j Heat of vaporization of i component, J/mol i
ΔH Height difference, cm
ΔP Pressure drop, MPa
ηfi Catalyst effectiveness factor of reaction j in the f phase, dimensionless
Γ Gas distributor parameter
λfa Axial thermal conductivity of f phase, J/cmr s K
λfe Catalyst thermal conductivity, J/cmS s K
λW Thermo-well thermal conductivity, J/cmW s K
μf Dynamic viscosity of f phase, cP
νfi,j Stoichiometric coefficient of component i in reaction j in the f phase, dimensionless
ϕ Weight fraction of f phase in the slurry, gf/gSL
Ψ Arbitrary system property
ρf Density of f phase, gf cm3f
σL Surface tension of liquid phase, dyn/cm
τ Dimensionless time
θ Dimensionless temperature, T/T0
εB Catalyst bed void fraction or catalyst porosity, cm3G + L cm3r
εf Holdup of f phase, cm3f cm3r
ξ Radial coordinate inside spherical catalyst particle, cmS
References
Ancheyta, J. (2013). Modeling of Processes and Reactors for Upgrading of Heavy Petroleum. CRC Press
Taylor and Francis Group.
Ancheyta, J., Sanchez, S., and Rodriguez, M. (2005a). Kinetic modeling of hydrocracking of heavy oil
fractions: a review. Catalysis Today 76–92.
Ancheyta, J., Rana, M.S., and Furimsky, E. (2005b). Hydroprocessing of heavy petroleum feeds: tutorial.
Angeles, M.J., Leyva, C., Ancheyta, J., and Ramírez, S. (2014). A review of experimental procedures for
heavy oil hydrocracking with dispersed catalyst. Catalysis Today 220–222: 274–294.
Bach, H. and Pilhofer, T. (1978). Variation of gas holdup in bubble columns with physical properties of
liquids and operating parameters of columns. German Chemical Engineering 1: 270–275.
Bahroun, S., Li, S., Jallut, C. et al. (2010). Control and optimization of a three-phase catalytic slurry
intensified continuous chemical reactor. Journal of Process Control 20: 664–675.
Bahroun, S., Li, S. Jallut, C., Valentin, C., Panthou, F. (2010) Modelling and control of a three-phase
catalytic slurry intensified continuous reactor. In European Symposium on Computer-Aided Process
Engineering, Ischia, Italy, June 6–9
Baten, J.M. and Krishna, R. (2001). Eulerian simulations for determination of the axial dispersion of
liquid and gas phases in bubble columns operating in the churn turbulent regime. Chemical
Beenackers, A.A.C. and Swaaij, W.P.M.V. (1993). Mass transfer in gas-liquid slurry reactors. Chemical
Begovivh, D. and Watson, J.S. (1978). Hydrodynamics characteristics of three phase fluidized beds.
Fluidization 190.
Behkish, A., Lemoine, R., Oukaci, R., and Morsi, B.I. (2006). Novel correlations for gas holdup in large
scale slurry bubble column reactors operating under elevated pressures and temperatures. Chemical
Engineering Journal 115: 157–171.
Bellussi, G., Rispoli, G., Landoni, A. et al. (2013). Hydroconversion of heavy residues in slurry reactors:
developments and perspectives. Journal of Catalysis 308: 189–200.
Bird, R.B., Stewart, W.E., and Lightfoot, E.N. (2002). Transport phenomena, Seconde. John Wiley.
Calderon, C. and Ancheyta, J. (2016). Modeling of slurry-phase reactors for hydrocracking of heavy oils.
Energy & Fuels 2525–2543.
Carbonell, M.M. and Guirardello, R. (1997). Modelling of a slurry bubble column reactor applied to the
hydroconversion of heavy oils. Chemical Engineering Science 52: 4179–4185.
Chaudhari, R.V. and Ramachandran, P.A. (1980). Three-phase slurry reactors. AIChE Journal 26: 177–201.
Chen, Z., Zheng, C., Feng, Y., and Hofmann, H. (1995). Modeling of three phase fluidized beds based on
local bubble characteristics measurements. Chemical Engineering Science 50: 231–236.
Chen, Z., Zhang, H., Ying, W., and Fang, D. (2010). Mathematical simulation of bubble column slurry
reactor for direct dimethyl ether synthesis process from syngas. World Academy of Science, Engineering
and Technology 4: 1443–1452.
Cho, Y.J., Song, P.S., Lee, C.G. et al. (2005). Liquid radial dispersion in liquid-solid circulating fluidized
beds with viscous liquid media. Chemical Engineering Communications 192: 257–271.
Costa, C.B., Rivera, E.A., Rezende, M.C.A.F. et al. (2007). Prior detection of genetic algorithm significant
parameters: coupling factorial design technique to genetic algorithm. Chemical Engineering Science 62:
4780–4801.
Deckwer, W.D., Burckart, R., and Zoll, G. (1974). Mixing and mass transfer in tall bubble columns.
Deckwer, W.D., Louisi, Y., Zaidi, A., and Ralek, M. (1980). Hydrodynamic properties of the fisher-tropsch
slurry process. Industrial & Engineering Chemistry Process Design and Development 19: 699–708.
Deckwer, W.D., Serpemen, Y., Ralek, M., and Schmidt, B. (1982). Modeling the Fischer-Tropsch synthesis
in the slurry-phase. Industrial & Engineering Chemistry, Process Design and Development 21: 231–241.
van Deemter, J. (1961). Mixing and contacting in gas-solid fluidized beds. Chemical Engineering Science
13: 143–154.
Degaleesan, S. and Dudokivic, M.P. (1998). Liquid backmixing in bubble columns and the axial
dispersion coefficient. AIChE Journal 44: 2369–2378.
Deng, Z., Wang, T., and Wang, Z. (2010). Hydrodesulfurization of diesel in a slurry reactor. Chemical
Domokos, L., Jongkind, H., Rigutto, M. S., van de Voort, E. H. C., Stork, W. H. K. (2006) Hydrocracking
catalyst composition. US Patent 0,207,917
Dudukovic, M.P. and Mills, P.L. (1984). Chemical and Catalytic Reactor Modeling. American Chemical
Society.
Ellenberger, J. and Krishna, R. (1994). A unified approach to the scale-up of gas-solid fluidized bed and
gas-liquid bubble column reactors. Chemical Engineering Science 49: 5391–5411.
Fan, L.S. (1989). Gas-liquid-solid fluidization. Butterworth’s Series in Chemical Engineering.
Ferreira, A., Ferreira, C., Teixeira, J.A., and Rocha, F. (2010). Temperature and solid properties effects on
gas-liquid mass transfer. Chemical Engineering Journal 162: 743–752.
Froment, G.F., Bischoff, K.B., and De Wilde, J. (2011). Chemical Reactor Analysis and Design, 3rde. Wiley.
Fukuma, M., Muroyama, K., and Yasunishi, A. (1987). Specific gas-liquid interfacial area and liquid
phase mass transfer coefficient in a slurry bubble column. Journal of Chemical Engineering of Japan 20:
321–324.
References 299
Han, J.H. and Kim, S.D. (1990). Radial dispersion and bubble characteristics in three-phase fluidized
beds. Chemical Engineering Communications 94: 9–26.
Hillmer, G., Weismantel, L., and Hofmann, H. (1994). Investigations and modelling of slurry bubble
columns. Chemical Engineering Science 49: 837–843.
Jakobsen, H.A., Lindborg, H., and Dorao, C.A. (2005). Modelling of bubble column reactors: progress and
limitations. Industrial & Engineering Chemistry Research 44: 5107–5151.
Jamialahmadi, M. and Müller-Steinhagen, H. (1999). Hydrodynamics and heat transfer of liquid fluidized
bed systems. Chemical Engineering Communications 179: 35–79.
Jhawar, A.K. and Prakash, A. (2011). Heat transfer in a slurry bubble column reactor: a critical overview.
Industrial & Engineering Chemistry Research 51: 1464–1473.
Jiang, P., Luo, X., Lin, T.J., and Fan, L.S. (1997). High temperature and high pressure three-phase
fluidization-bed expansion phenomena. Powder Technology 90: 103–113.
Jin, H., Yang, S., He, G. et al. (2014). Gas-liquid mass transfer characteristics in a gas-liquid-solid bubble
column under elevated pressure and temperature. Chinese Journal of Chemical Engineering 22:
955–961.
Joshi, J.B. (1998). Axial mixing in multiphase contactors: a unified correlation. Chemical Engineering
Research and Design 58: 155–165.
Jung, S., Becker, M., Agar, D., and Franke, R. (2010). One-dimensional modeling and simulation of
bubble column reactors. Chemical Engineering & Technology 33: 2037–2043.
Kang, Y. and Kim, S.D. (1986). Radial dispersion characteristics in three-phase fluidized beds.
Industrial & Engineering Chemistry, Process Design and Development 25: 717–722.
Kang, Y., Cho, Y.J., Lee, C.G. et al. (2003). Radial dispersion and bubble distribution in three-phase
circulating fluidized beds. The Canadian Journal of Chemical Engineering 81: 1–9.
Kantarci, N., Borak, F., and Ulgen, K. (2005). Bubble column reactors review. Process Biochemistry.
(Oxford, U.K.) 40: 2263–2283.
Kato, Y., Nishiwaki, A., Fukuda, T., and Tanaka, S. (1972). The behavior of suspended solid particles and
liquid in bubble columns. Journal of Chemical Engineering of Japan 5: 112–118.
Kato, Y., Morooka, S., Kago, T. et al. (1985). Axial holdup distributions of gas and solid particles in three-
phase fluidized bed for gas-liquid (slurry)-solid systems. Journal of Chemical Engineering of Japan 18:
308–313.
Kennepohl, D. and Sanford, E. (1996). Conversion of Athabasca bitumen with dispersed and supported
mo-based catalysts as a function of dispersed catalyst concentration. Energy & Fuels 10: 229–234.
Kesler, M. (1976). Improve prediction of enthalpy of fractions. Hydrocarb 153.
Khadem-Hamedani, B., Yaghmaei, B.K.-H., Fattahi, M. et al. (2015). Mathematical modeling of a slurry
bubble column reactor for hydrocracking of diesel fuel: single and two bubble configurations.
Chemical Engineering Research and Design 100: 362–376.
Khulbe, C. P., Belinko, K., Waugh, R. J., Perreault, M. (1989) Process for preparing an iron-coal slurry
catalyst for hydrocracking heavy oils. US Patent 4,923,838
Kim, S.D. and Kang, Y. (1997). Heat and mass transfer in three-phase fluidized bed reactor: an overview.
Kim, H.S., Kim, J.H., Kang, S.H. et al. (2014). Bubble and heat transfer phenomena in viscous slurry
bubble column. Advances in Chemical Engineering and Science 4: 417–429.
Kito, M., Shimida, M., Sakai, T. et al. (1976). Performance of tubular bed contractor, gas holdup and
interfacial area under liquid stagnant flow. Fluidization 411.
Kojima, H., Anjyo, H., and Mochizuki, Y. (1989). Axial mixing in bubble column with suspended solid
particles. Journal of Chemical Engineering of Japan 19: 232–234.
Krishna, R. and Ellenberger, J. (1996). Gas holdup in bubble column reactors operating in the churn
turbulent flow regime. AIChE Journal 42: 2627–2634.
Krishna, R. and Saxena, A.K. (1989). Use of an axial dispersion model for kinetic description of
hydrocracking. Chemical Engineering Science 44: 703–712.
Krishna, R. and Sie, S.T. (2000). Design and scale-up of the Fischer-Tropsch bubble column slurry reactor.
Fuel Processing Technology 64: 73–105.
Kriz, J. F., Ternan, M., Denis, J. M. (2009) Hydrocracking of bitumen and heavy oils. In 5th NCUT
Upgrading and refining conference, synthetic crude oil and heavy oil, Edmonton, Canada, Sept 14–16.
Kumar, S. and Khanna, A. (2014). Experimental analysis and development of correlations for gas holdup
in high pressure slurry co-current bubble columns. Korean Journal of Chemical Engineering 31:
1964–1972.
Lemoine, R., Behkish, A., Sehabiague, L. et al. (2008). An algorithm for predicting the hydrodynamic and
mass transfer parameters in bubble column and slurry bubble column reactors. Fuel Processing
Leonard, C., Ferrasse, J.-H., Boutin, O. et al. (2015). Bubble column reactors for high pressures and high
temperatures operation. Chemical Engineering Research and Design 100: 391–421.
Li, H. and Prakash, A. (2000). Influence of slurry concentrations on bubble population and their rise
velocities in three-phase slurry bubble column. Powder Technology 113: 158–167.
Li, Y., Wang, J., Jiang, L., Zhang, Z., Liu, J., Ren, S., Zhao, B., Jia, Y. (1999) Hydrocracking of heavy oil and
residuum with a dispersing-type catalyst. US Patent 6,004,454.
Li, S., Bahroun, S., Valentin, C. et al. (2010). Dynamic model based safety analysis of a three-phase
catalytic slurry intensified continuous reactor. Journal of Loss Prevention in the Process Industries 23:
437–445.
Liu, D., Kong, X., Li, M., and Que, G. (2009). Study on a water-soluble catalyst for slurry-phase
hydrocracking of an atmospheric residue. Energy & Fuels 23: 958–961.
Luo, X., Lee, D., Lau, R. et al. (1999). Maximum stable bubble size and gas holdup in high-pressure slurry
bubble columns. AIChE Journal 45: 665–680.
Macías, M.J. and Ancheyta, J. (2004). Simulation of an isothermal hydrodesulfurization small reactor
with different catalyst particle shapes. Catalysis Today 98: 243–252.
Maretto, C. and Krishna, R. (1999). Modelling of a bubble column slurry reactor for Fischer-Tropsch
synthesis. Catalysis Today 52: 279–289.
Mariano, A.P., Costa, C.B.B., deToledo, E.C.V. et al. (2011). Analysis of particle swarm algorithm in the
optimization of a three-phase slurry catalytic reactor. Computers & Chemical Engineering 35: 2741–2749.
short term catalyst deactivation. Fuel 193–199.
Martínez, J. and Ancheyta, J. (2014). Modeling the kinetics of parallel thermal and catalytic hydrotreating
of heavy oil. Fuel 27–36.
Martínez, J., Sánchez, J.L., Ancheyta, J., and Ruiz, R.S. (2010). A review of process aspects and modeling
of ebullated bed reactors for hydrocracking of heavy oils. Catalysis Reviews 52: 60–105.
Matonis, D., Gidaspow, D., and Bahary, M. (2002). CFD simulation of flow and turbulence in a slurry
bubble column. AIChE Journal 48: 1413–1429.
Matos, E. and Guirardello, R. (2002). Modeling and simulation of a pseudo-two-phase gas-liquid column
reactor for thermal hydrocracking of petroleum heavy fractions. Brazilian Journal of Chemical
Matos, E. and Nunhez, J. (2006). The effect of different feed flow patterns on the conversion of bubble
column reactors. Chemical Engineering Journal 116: 163–172.
Matos, E., Guirardello, R., Mori, M., and Nunhez, J. (2009). Modeling and simulation of a pseudo-three-
phase slurry bubble column reactor applied to the process of petroleum hydrodesulfurization.
Computers & Chemical Engineering 33: 1115–1122.
References 301
Mederos, F., Elizalde, I., and Ancheyta, J. (2009). Steady-state and dynamic reactor model for
hydrotreatment of oil fractions: a review. Catalysis Reviews – Science and Engineering 51: 485–607.
Melo, D.N., Toledo, E.C.V.D., Santos, M.M. et al. (2005). Off-line optimization and control for real time
integration of a three-phase hydrogenation catalytic reactor. Computers & Chemical Engineering 29:
2485–2493.
Mills, P.L., Turner, J.R., Ramachandran, P.A., and Dudokivic, M.P. (1996). The Fischer-Tropsch synthesis
in slurry bubble column reactors: analysis of reactor performance using the axial dispersion model.
Fuel Energy 38.
Moustiri, S., Hebrard, G., Thakre, S.S., and Roustan, M. (2001). A unified correlation for predicting liquid
axial dispersion coefficient in bubble columns. Chemical Engineering Science 56: 1041–1047.
Noguera, G., Araujo, S., Hernández, J. et al. (2012). A comparative activity study of a new ultra-dispersed
catalyst system for a hydrocracking/hydrotreating technology using vacuum residue oil: Merey/Mesa.
Chemical Engineering Research and Design 90: 1979–1988.
Ortiz-Moreno, H., Ramírez, J., Cuevas, R. et al. (2012). Heavy oil upgrading at moderate pressure using
dispersed catalyst: effects of temperature, pressure and catalytic precursor. Fuel 100: 186–192.
Panariti, N., del BIanco, A., del Piero, G. et al. (2000). Petroleum residue upgrading with dispersed
catalysts: part 2. effect of operating conditions. Applied Catalysis A: General 204: 215–222.
Pangarkar, V.G., Yawalkar, A.A., Sharma, M.M., and Beenackers, A.A.C. (2002). Particle-liquid mass
transfer coefficient in two/three-phase stirred tank reactors. Industrial & Engineering Chemistry
Research 41: 4141–4167.
Parulekar, S.J. and Shah, Y.T. (1980). Steady-state behavior of gas-liquid-solid fluidized-bed reactors.
Pfleger, D. and Becker, S. (2001). Modelling and simulation of the dynamic flow behaviour in a bubble
column. Chemical Engineering Science 56: 1737–1747.
Rados, N., Al-Dahhan, M., and Dudokivic, M.P. (2005). Dynamic modeling of slurry bubble column
reactors. Industrial & Engineering Chemistry Research 44: 6086–6094.
Reilly, I.G., Scott, D.S., Bruijin, T.J.W. et al. (1986). A correlations for gas holdup in turbulent coalescing
bubble columns. The Canadian Journal of Chemical Engineering 64: 705–717.
Rezair, H. and Smith, K.J. (2013). Catalyst deactivation in slurry-phase residue hydroconversion.
Energy & Fuels 27: 6087–6097.
Ruiz, R.S., Alonso, F., and Ancheyta, J. (2004). Effect of high-pressure operation on overall phase holdups
in ebullated-bed reactors. Catalysis Today 9: 265–271.
Ruthiya, K.C., van der Schaaf, J., Kuster, B.F.M., and Schouten, J.C. (2004). Modeling the effect of particle
to bubble adhesion on mass transport and reaction rate in a stirred slurry reactor: influence of catalyst
support. Chemical Engineering Science 59: 5551–5558.
Sahu, R., Song, B.J., Im, J.S. et al. (2015). A review of recent advances in catalytic hydrocracking of heavy
residues. Journal of Industrial and Engineering Chemistry 27: 12–24.
Sánchez, S. and Ancheyta, J. (2007). Effect of pressure on the kinetics of moderate hydrocracking of maya
crude oil. Energy & Fuels 653–661.
Sánchez, S., Rodríguez, M., and Ancheyta, J. (2005). Kinetic model for moderate hydrocracking of heavy
oils. Industrial & Engineering Chemistry Research 9409–9413.
Sauer, T. and Hempel, D.C. (1987). Fluid dynamics and mass transfer in a bubble column with suspended
particles. Chemical Engineering & Technology 10: 180–189.
Scheffer, B., van Koten, M.A., Röbschlager, K.W., and Boks, F.C. (1998). The shell residue
hydroconversion process: development and achievements. Catalysis Today 43: 217–224.
Schumpe, A., Saxena, A.K., and Fang, L.K. (1987). Gas-liquid mass transfer in a slurry bubble column.
Sehabiague, L., Lemoine, R., Behkish, A. et al. (2008). Modeling and optimization of a large-scale slurry
bubble column reactor for producing 10,000 bbl/day of Fischer-Tropsch liquid hydrocarbons. Journal
of the Chinese Institute of Chemical Engineers 39: 169–179.
Shah, Y.T., Kelkar, B.G., Godbole, S., and Deckwer, W.D. (1982). Design parameters estimations for
bubble column reactors. AIChE Journal 28: 353–379.
Shahrzad, H., Arturo, M., and Phillip, S. (2009). Dynamic simulation of gas hydrate formation in a three-
phase slurry reactor. Industrial & Engineering Chemistry Research 48: 6983–6991.
Sivaiah, M. and Majumder, S.K. (2013). Dispersion characteristics of liquid in a modified gas-liquid-solid
three-phase downflow bubble column. Particulate Science and Technology 31: 210–220.
Speight, J.G. (2004). New approaches to hydroprocessing. Catalysis Today 98: 55–60.
de Swart, J.W.A. and Krishna, R. (1997). Selection, design and scale-up of the fischer-tropsch slurry
reactor. Studies in Surface Science and Catalysis 107: 213–218.
de Swart, J. and Krishna, R. (2002a). Simulation of the transient and steady-state behaviour of a bubble
column slurry reactor for Fischer-Tropsch synthesis. Chemical Engineering and Processing 41: 35–47.
de Swart, J.K.A. and Krishna, R. (2002b). Simulation of the transient and steady-state behavior of a bubble
column slurry reactor for Fischer-Tropsch synthesis. Chemical Engineering and Processing 35–47.
Tang, C. and Heindel, T. (2006). Estimating gas holdup via pressure measurements in a cocurrent bubble
column. International Journal of Multiphase Flow 32: 850–863.
Todorova, T., Alexiev, V., and Weber, T. (2012). Energetics and electronic properties of defects at the (100)
mos2 surface studied by the perturbed cluster method. Reaction Kinetics, Mechanisms and Catalysis
105: 113–133.
de Toledo, E.C.V., de Santana, P.L., Maciel, M.R.W., and Filho, R.M. (2001). Dynamic modelling of a
three-phase catalytic slurry reactor. Chemical Engineering Science 56: 6055–6061.
Tomiyama, A. (1998). Struggle with computational bubble dynamics. Multiphase Science and Technology
10: 369–405.
Tomiyama, A. and Shimada, N. (2001). A numerical method for bubbly flow simulation based on a multi-
fluid model. Journal of Pressure Vessel Technology 123: 510–516.
Torvik, R. and Svendsen, H.F. (1990). Modeling of slurry reactors. a fundamental approach. Chemical
Vandu, C.O. and Krishna, R. (2003). Gas holdup and volumetric mass transfer coefficient in a slurry
bubble column. Chemical Engineering & Technology 26: 779–782.
Wilkinson, P.M., Spek, A.P., and van Dierendonck, L.L. (1992). Design parameters estimation for scale-
up of high-pressure bubble columns. AIChE Journal 38: 544–554.
Yang, G.Q. and Fan, L.S. (2003). Axial liquid mixing in high pressure bubble columns. AIChE Journal 49:
1995–2008.
Ying, D.H., Givens, E.N., and Weimer, R.F. (1980). Gas holdup in gas-liquid and gas-liquid solid flow
reactors. Industrial & Engineering Chemistry, Process Design and Development 19: 635–638.
Zehner, P. and Kraume, M. (2005). Bubble Columns: Encyclopedia of Industrial Chemistry. Wiley-VCH.
Zhang, S., Liu, D., Deng, W., and Que, G. (2007). A review of slurry-phase hydrocracking heavy oil
technology. Energy & Fuels 21: 3057–3062.
Zhu, Y.Q. and Zhou, D.L. (2011). Study on unsupported nano-MOS catalyst for clean fuel. Advanced
Materials Research 287–290: 1860–1865.
van der Zon, M., Hamersma, P.J., Poels, E.K., and Bliek, A. (1999). Gas-solid adhesion and solid-solid
agglomeration of carbon supported catalyst in three-phase slurry reactors. Catalysis Today 48: 131–138.
van der Zon, M., Thoolen, H., Hamersma, P.J. et al. (2001). Agglomeration and adhesion of catalyst
particles in gas-liquid reactors. Catalysis Today 66: 263–270.
303
Modeling of Fischer–Tropsch Synthesis Reactor

César I. Méndez1 and Jorge Ancheyta2
1
Centro de Investigación en Ciencia Aplicada y Tecnología Avanzada del Instituto Politécnico Nacional, Mexico City, Mexico
2
Instituto Mexicano del Petróleo, Mexico City, Mexico
8.1 Fundamentals of the Fischer–Tropsch Synthesis

to Produce Clean Fuels
Nowadays, the world faces the need to meet high energy demands, for which various fields of study
have been opened to explore useful alternatives that are economical and friendly to the environ-
ment. Fischer–Tropsch technology has become a viable option for the production of commercial
fuels (mainly gasoline, diesel, and airplane fuels), which has attracted the attention of the scientific
and industrial community (Jia-Yang et al. 2015). Different reactor technologies exist to carry out
the Fischer–Tropsch synthesis (FTS) process. These reactor technologies have been used for design,
modeling, and optimization studies and are applied according to the Fischer–Tropsch technology,
depending on the type of catalyst and operating temperature used: (i) high-temperature Fischer–
Tropsch (HTFT) at a temperature of 290–360 C with an iron catalyst; and (ii) low-temperature
Fischer–Tropsch (LTFT) at a temperature of 180–260 C with cobalt catalyst. The former used
fluidized-bed reactor and circulating fluidized-bed reactor, while the latter used fixed-bed
reactor, slurry bubble column, and microchannel reactor (Basha et al. 2015). Despite the advan-
tages of fluidized-bed reactor, slurry bubble column reactor, and the novel microchannel reactor
compared with those exhibited by the fixed-bed reactor, the latter has been preferably employed
in commercial applications (Chambrey et al. 2006; Maitlis and de Klerk 2013), because it shows
superiority over the others in both operational and economic matters, since it can be scaled from
a single tube to industrial plants (Saeidi et al. 2014), as can be seen in large commercial scale devel-
oped by Sasol (Dry 1996) and Shell (Sie 1998). The Fischer–Tropsch process is quite complex due to
the difficulty of predicting the products distribution generated by the synthesis. The use of different
catalysts has led to the proposal of various reaction mechanisms that attempt to accurately describe
the steps involved in this catalytic reaction (Dry 1996). On the other hand, there is a great diversity
of kinetic models that have been developed for the description of the FTS, which arise due to the
different reaction mechanisms that have been proposed, the operating conditions, and the type of
catalyst.
There is experimental evidence that a complex mixture of hydrocarbons ranging from methane
to wax is normally produced in a fixed-bed reactor (Caldwell and Van Vuuren 1986; Wang et al.
304 8 Modeling of Fischer–Tropsch Synthesis Reactor
2003a). Therefore, it is convenient to provide a comprehensive mechanism that contemplates all the
formation steps of the intermediate FTS species, initiators of the growth of the hydrocarbon chain,
and the product’s distribution, among other aspects related to the behavior of the catalyst, catalytic
support, and promoter (Wang et al. 2003a).
8.1.1 Fischer–Tropsch Synthesis Technology

The process of converting syngas into liquid hydrocarbons by using the X-to-liquid technology
(XTL) has proven to be an alternative to generate excellent clean-burning fuels, which has
increased the interest in studying the FTS to address the widespread demand for liquid fuels for
commercial use (Ahón et al. 2005). The XTL process is a viable option for the production of
improved clean fuels with a highly attractive business that can be generated from any material con-
taining carbon, natural gas, and biomass, as well as waste (e.g. carbon, natural gas, biomass or
waste) (Wilhelm et al. 2001; Mousavi et al. 2015).
The use of gas-to-liquid technology (GTL) for the syngas conversion to commercial value liquid
hydrocarbons by the FTS has received much attention due to the following advantages (Forghani
et al. 2009; Lee et al. 2015; Luque et al. 2012):
• Increased availability of natural gas located in remote areas and poor access for which there is no
nearby market.
•• Growing demand for middle distillate fuels used for transportation.

Improved cost-effectiveness of GTL technology resulting from the development of more active
catalysts and more advanced reactor design.
•• Involve low costs in transportation of liquid products compared with gaseous products.
Production of a high cetane number diesel, which is due to the low content of aromatics.
• Wide versatility in terms of different types of raw materials that can be used and different types
of products that are obtained, which are expected to be an important alternative for the energy
segment soon.
• Products are virtually free of sulfur, nitrogen, and aromatics, providing compatibility and good
blending with conventional fuels and the broad possibility to work efficiently with fuel infrastruc-
ture nowadays.
• Environment-friendly technology that produces almost free of impurities fuels.
Due to the demand for high-quality fuels to meet the energy needs of the industrial and trans-
port sectors, special attention has been paid to the development of more effective clean fuels pro-
duction and relatively environmentally friendly technologies. The FTS process, according to the
advantages already mentioned, has become a focus of attention around the technological, scien-
tific, and industrial sectors. It is considered a clean option for the environment and is also desig-
nated as the heart of the technological processes of gas conversion to liquid (Forghani et al. 2009),
which has been widely investigated to improve their aspects of design, implementation, start-up,
and optimization at various scales. This technology enables commercial liquid fuel production
with low aromaticity and impurities free, which reduces the environmental impact associated
with fuel from the conventional refining process of crude oil. Moreover, the effects of the FTS
process for fuel production do not cause a strong environmental impact in contrast to conven-
tional processes for fuel production and industrial chemicals. So, the FTS process is a viable tech-
nology for the production of synthetic fuels that are eco-friendly (Knottenbelt 2002; van Vliet
et al. 2009; Gill et al. 2011).
Syngas is produced with technologies already established in three ways: (i) partial oxidation,
(ii) steam reforming, and (iii) autothermal synthesis. Partial oxidation is based on the following
exothermic reaction (Vosloo 2001; Todic et al. 2015):
CH4 + 1 2O2 CO + 2H2 81
that can be generalized to any hydrocarbon:
Cn H m + 1 2nO2 nCO + 1 2mH 2 82
While in the steam reforming the following endothermic reaction is found:
CH4 + H2 O CO + 3H2 83
that can be generalized to any hydrocarbon:
Cn H m + nH 2 nCO + n + m 2 H 2 84
In autothermal synthesis, steam reforming and partial oxidation are combined. The heat pro-
duced by partial oxidation is used to provide energy for steam reforming, making a thermally
self-sufficient process. In the typical FTS process, the syngas catalytic conversion occurs mainly
in a predominant mixture of linear alkanes and alkenes whose product distribution exhibits a
recognizable pattern such as illustrated below:
Paraffins:
2n + 1 H 2 + nCO Cn H2n+2 + nH2 O 85
Olefins:
2nH2 + nCO Cn H2n + nH2 O 86
Alcohols:
nCO + 2nH2 Cn H2n+1 OH + n − 1 H2 O 87
Until then, a growing number of studies have been significantly reflected in the discussion on the
understanding of the integrated nature of surface species and the precise and detailed mechanical
sequence, the description by which the FTS reactions proceed over the catalyst. A wide variety of
FTS reaction mechanisms have been proposed whose common feature has been the consideration
of the polymerization reaction (Nakhaei et al. 2014a) and the stepwise chain growth hydrocarbons
process, which is supported by the fact that the carbon number distribution in the FTS products is
calculated based only on probabilities for the chain growth matching results observed experimen-
tally obtained from reactors of different sizes and configuration types over a wide range of operating
conditions and catalysts (Dry 1996).
According to Wojciechowski (1988), a reaction mechanism proposed for the description of the
FTS must comply with the following aspects:
• The complete adsorption of all species on the catalyst surface in a set of active sites resulting from
the decomposition of H2 and CO to hydrogen atoms and the respective adsorption of carbon and
oxygen atoms causes interaction between the species located on the surface, leading to the for-
mation of CHx, OH compounds, etc.
• The CH2 should be the monomer species to carry out the oligomerization. The formation rate
arises from the adsorbed species C and H, and this becomes the rate-determining step for CO
hydrogenation kinetics.
• The radical growth on the catalytic surface is immobile except for species C1 − C4, and the chain
grows only through the participation of a nearby monomer. The growing chain may be formed or
migrated through the diffusion between the set of appropriate active sites.
• The chain growth surface can spontaneously generate 1–2 accompanying displacements that can
result in the formation of branched hydrocarbons.
• The termination stage and therefore the product type are determined by the type of component
that occupies the vacant site adjacent to the growth radical.
• Classical distributions involve primary species produced, each having its own chain length
distribution of the Anderson–Schulz–Flory (ASF) type. This distribution is typical of a proper
collocation grouping of growth, monomers, and the termination stage sites that constitute
the growth place location for the molecular species. Such locations have a stable composition
and only continue to produce one type of molecular species to a certain set of given reaction
conditions.
• Operating parameters, such as temperature, system total pressure, and the H2/CO ratio, are
extremely important factors affecting both the kinetics and the product’s distribution.
According to Adesina (1996) and Förtsch et al. (2015), both the rate of the catalytic reaction of
Fischer–Tropsch that normally occurs in the catalyst pores and the product’s distribution are
affected by various physical and chemical phenomena that include the following: (i) the reactants
diffusion into a porous particle to active sites, (ii) reactants adsorption on catalytic active sites,
(iii) chain initiation, (iv) chain growth (or propagation), (v) chain termination, (vi) product desorp-
tion process, (vii) re-adsorption and further reaction, and (viii) products diffusion toward the par-
ticle outside. Förtsch et al. (2015) summarized the FTS reaction system’s basic stages in simplified
schematic reactions involved in the following way:
∗ R1 ∗
i chain initiation + CO + 2H 2 CH 2 + H2O 88
propagation
∗ Rn ∗
ii chain propagation CH2 n + CO + 2H2 CH2 n+1 + H2 O 89
Rtermination
∗ n,paraffins
iii termination to paraffins CH2 n + H2 Cn H2n+2 8 10
Rtermination
∗ n,olefins
iv termination to olefins CH2 n Cn H2n 8 11
re−adsorption
Rn,olefins
∗
v re − adsorption Cn H2n CH2 n 8 12
hydrogenation
Rn
vi secondary hydrogenation Cn H2n + H2 Cn H2n+2 8 13
It is important to note that for decades, some assumptions have been made about the forms
in which the FTS monomer and initiator and its chemical formation occur. The most important
monomers identified through the FTS process history are methylene (−CH2−), hydroxyl
carbine (CHOH), and CO, which lead to the following three important mechanisms:
(i) carbide, (ii) the enol mechanism, and (iii) the CO insertion mechanism, respectively
(Mousavi et al. 2015).
Other reaction mechanisms have been proposed that differentiate in the assumptions that arise
for monomer, initiator, and their chemical formation such as the alkyl, alkenyl and alkylidene
hydride-methylidene, and formate mechanisms (Basha et al. 2015; Frennet and Hubert 2000; Brady
and Pretit 1981; Vannice 1976; Satterfielf et al. 1982).
8.1.2 Fischer–Tropsch Synthesis Catalysts

At present, there are some challenges in terms of product selectivity that are of particular interest
(such as commercial gasoline and diesel fuels) and improve the FTS reactor performance used under
severe operating conditions, for which research on the design of novel catalysts has been an issue of
interest. Moreover, it is known that the spectrum of products obtained from the FTS technology is
broad and requires a second refining step to improve quality, thus achieving commercial fuel pro-
ducts that are usually of higher quality compared with the fuels generated by conventional refining
processes (Zhang et al. 2013). Based on this, for the option of liquid fuels of high quality from syngas
by a single-stage FTS without the use of a subsequent refining step, it is necessary to design a novel
catalyst with high selectivity toward the desired range of hydrocarbons (Kang et al. 2011; Sun et al.
2012; Valero-Romero et al. 2016). In addition, the FTS process can be very successful by developing
novel active catalysts having high selectivity toward generating wax; in other words, the design of a
good active catalyst is the key element for excellent operation (Storsaeter et al. 2005).
As for the catalyst types employed in the FTS process, they are based on transition metals such as
cobalt, iron, nickel, and ruthenium. It is relevant to mention that one of the biggest problems pre-
sented in the design of an FTS process is the apparent complexity exhibited by the reaction mech-
anism and the large number of species involved (Glasser et al. 2012). Moreover, two main types of
products are generated: commercial hydrocarbons (gasoline, diesel, and olefins of low molecular
weight) and organic oxygenates (methanol, ethanol, and mixtures of alcohols). That is, the forma-
tion of a particular product type is based on the catalyst type employed (Zhang et al. 2013) among
other features. The choice of a particular type of catalyst lies mainly in the operating mode (high or
low temperature) used and the type of raw material employed (Luque et al. 2012). The FTS process
can proceed at low and high temperatures, distinguishing between LTFT synthesis and HTFT syn-
thesis, respectively, and each of these two classifications will typically correspond to two specific
catalysts, cobalt-based catalyst and iron-based catalyst respectively.
8.1.2.1 Cobalt-Based Catalysts

The cobalt catalyst has been one of the most used in LTFT technology due to the following
advantages:
• Exhibits high activity and selectivity toward the production of long-chain paraffins (Storsaeter
et al. 2005; Hinchiranan et al. 2008; Trépanier et al. 20009).
• Provides satisfactory stability in the production of linear hydrocarbons from mixtures of CO/H2
(Singh and Gu 2010).
• Shows low selectivity toward the production of oxygenates (byproducts) (Hinchiranan

et al. 2008).
• The gas–water shift (WGS) reaction has low activity that can be discarded in kinetic models pro-
posed to describe the FTS process (Hinchiranan et al. 2008; Trépanier et al. 2009).
• The water concentration increases with time-on-stream (TOS) in FTS for cocatalysts because of
enhanced catalyst deactivation, which reduces the selectivity to water–gas shift (Fazlollahi
et al. 2012).
• High resistance to wear catalyst when employed in suspension phases, e.g. slurry reactor opera-
tions (Luque et al. 2012).
•• Provides higher yields and longer lifetime compared with iron catalysts (Chaumette et al. 1995).
Not inhibited by water, thus exhibiting higher productivity compared with catalysts involving
reaction water–gas shift and higher syngas conversion (van Berge and Everson 1997).
The most important disadvantages of cobalt catalysts are higher prices compared with iron and
low performance due to excessive methane production, which can occur when the proper support
and promoter are not used and/or when it is not manufactured under the required quality stan-
dards. However, despite these disadvantages, it remains the most preferred catalyst (Trépanier
et al. 2009).
In addition, the lifetime of the catalysts is influenced by at least two factors: (i) the physical prop-
erties such as attrition of the catalyst, wax buildup within the catalyst pellet, pressure drop through
the catalytic bed, and sudden variations in temperature and pressure, and (ii) loss of catalyst active
sites through poisoning and/or contamination (Luque et al. 2012). Some other causes that lead to
catalyst deactivation are (Tsakoumis et al. 2010): (i) poisoning, (ii) re-oxidation of cobalt active sites,
(iii) formation of carbon species on the catalyst surface, (iv) carbidization, surface reconstruction,
(v) sintering cobalt crystallites, (vi) wear, and (vii) solid-state reactions between metal and support.
A mechanism of cobalt catalyst regeneration has been also proposed, which consists of three steps:
(i) dewaxing, (ii) oxidation, and (iii) reduction. Luque et al. (2012) recently reported classification
of the different types of deactivation mechanisms that affect the cobalt-based catalysts used in
the LTFT system: (i) poisoning by sulfur compounds and/or nitrogen in the syngas feed,
(ii) oxidation of cobalt metal to active inactive cobalt oxide, (iii) formation of support-cobalt com-
pounds, (iv) sintering cobalt small crystallites, (v) structural reconstruction, and (vi) carbon
formation.
8.1.2.2 Iron-Based Catalysts

The main FTS products of iron catalysts are gasoline and diesel components with a relatively higher
fraction of olefins (Yan et al. 2014). The iron catalyst can be used in both configurations of FTS at
high and low temperatures. The following are the main advantages of iron catalysts:
•• Much lower cost compared with cobalt catalysts (Rao et al. 1992; Jager and Spinoza 1995).
Exhibit high activity in the water–gas shift reaction and high selectivity toward olefins that
appear to be stable when converting syngas with high H2/CO ratio (Rao et al. 1992; Jager and
Spinoza 1995).
• Can operate under a wide range of temperatures and ratios of H2/CO showing low selectivity to
methane (Fu and Li 2015).
• Improve the syngas conversion from raw materials such as carbon and biomass with low
H2/CO ratios due to the high activity of the water–gas shift reaction (Fun and Li 2015;
Mabry 2014).
• Quite small iron catalyst particles (diameter approximately 100 μm) such that carbon deposition
on the catalyst surface does not adversely affect reactor performance (Mabry 2014).
• Heavy paraffinic waxes as the main product that can be converted into usable products such as
commercial waxes or high-quality diesel (Mabry 2014).
• Production of hydrocarbons such as linear alkanes, alkenes, and oxygenates (Fu and Li 2015).
Some disadvantages of iron catalysts reported in the literature are as follows: (i) due to the high
activity shown by the iron catalysts toward the gas–water shift reaction, hydrothermal sintering and
oxidation can occur if the partial pressure of H2O increases progressively as the FTS reaction pro-
ceeds through the catalyst bed in fixed-bed reactors (Schanke et al. 1995); (ii) for the commercial
FTS fixed-bed reactors, the iron-catalyst bed becomes progressively deactivated toward the reactor
exit as a result of the oxidation of iron when considerable amounts of H2O and CO2 are produced by
the water–gas shift reaction (Dry 1991), which depends on the operating conditions; and (iii) it is
susceptible to produce aromatics when operating at high temperatures in the FTS, which would
have a negative effect on the environment (van Dijk 2001).
8.1.2.3 Catalyst Support

The design of catalytic supports is extremely essential to significantly increase the FTS catalyst
activity, and product’s selectivity of major interest, among others. The most common supports
are metal oxides such as aluminum, silicon, and titanium (Storsaeter 2005; Hinchiranan et al.
2008; Girardon et al. 2005; Fischer et al. 2013). The following are some advantages of the supports
that significantly improve the performance of FTS heterogeneous catalysts (Fu and Li 2015;
Surisetty et al. 2010; Gardezi et al. 2012):
• Positive effect of the FTS process performance due to interactions that arise between metal-
support, the acidity of the support, the porosity, and the mass transfer limitations.
• Dispersion of particles of cobalt on the catalyst surface after the reduction process and then to
maintain stable cobalt particles during the reaction.
• Porosity of the support, i.e. the average pore diameter, volume, and surface area positively influ-
ence the dispersion and reducibility property, as well as activity–selectivity of the FTS.
• Improved diffusion properties of both reactants and products within the granules of the catalyst
support.
• Dissipation of heat reaction and reduction of temperature gradients in a fixed-bed reactor.
Catalyst activity increases according to the following order of supports: MgO > MnO > C > SiO2 >
Al2O3 (Reuel and Bartholomev 1984).
Some catalysts supported on carbon have been widely suggested by the high potential shown to
increase the catalytic performance of the FTS process, within which the nanofibers, nanotubes,
spheres, and mesoporous carbon exhibit unique properties such as high mechanical strength, good
electrical conductivity, high thermal stability, and large surface area (De Jong and Geus 2000; Tava-
soli et al. 2010). It has been indicated that silica is one of the catalytic supports suitable for the design
of cobalt Fischer–Tropsch catalysts for its implementation in fixed-bed reactors due to its greater
surface area, porosity, stability, and weak metal support interaction (Oukaci et al. 1999).
8.1.3 Fischer–Tropsch Synthesis Kinetic Models

One of the high difficulties in developing accurate kinetic models for describing the behavior of FTS
over a range of carbon numbers obtained experimentally has been the detailed and improved con-
struction of FTS mechanisms (Glasser et al. 2012; Dry 2001). The adequate kinetic formulation that
describes the reactions involved in the FTS system has been considered a task of high complexity for
use in industrial practice, becoming a prerequisite for the design of industrial-scale processes, opti-
mization, and numerical simulation (Atashi et al. 2010; Visconti et al. 2011a). Different approaches
for kinetic modeling have been reported to describe the reactions occurring during the FTS, such as
kinetic rate designed from a semi-empirical and/or empirical mode, models derived from the use
of certain catalytic reaction mechanisms, as well as kinetic expressions represented by a type of
power-law equation (Zimmerman and Bukur 1990; van der Laan and Beenackers 1999). The kinetic
models depend on the type of catalyst and operating conditions used in the FTS process (Basha
et al. 2015).
Kinetic models are to be developed on the basis of a fairly large set of experimental data distribu-
tions under certain process conditions that are varied enough and independently, such as space
velocity, temperature, and total and partial pressures of CO and H2, with heat transfer and mass
gradients almost negligible (Van der Laan and Beenackers 1999). The design of kinetic models is
quite broad, which arises from the wide variety of catalysts used.
The reported detailed kinetic models have differences due to the assumptions regarding the rate
of determining steps (Gideon Botes et al. 2009). Some disadvantages of these models are as follows
(van Berge and Everson 1997): high complexity of the catalytic mechanism, statistical uncertain-
ties for many kinetic constants to be estimated, and questionable assumptions to generalize the
models developed. Furthermore, the cases of FTS kinetic models that use groups based on the
number of carbon atoms in the molecules, as well as those that are simply treated as a group
or division set of paraffins and olefins, can give oversimplified information of syngas total con-
version, so that they are not able to provide detailed information on the product’s distribution
(Wang et al. 2003b).
Kinetic models for the FTS process using catalysts based on cobalt have been organized into the
following three different classes: (i) kinetic models based on the occupation of a single active site by
CO or carbon surface, (ii) kinetic models based on occupancy of two active sites by the CO or a
reaction intermediate, and (iii) kinetic models based on nontraditional reaction orders for hydrogen
(Gideon Botes et al. 2009). Recently 14 kinetic models were investigated for the FTS process with
cobalt catalyst (Keyvanloo et al. 2016). Some correspond to models of the Langmuir–Hinshelwood
(LH) type, Eley–Rideal type, and power-law.
The FTS kinetic modeling is a catalytic process that involves a comprehensive approach of the
reaction steps involved during the initiation, propagation, and termination processes of the hydro-
carbon chain, which is a highly complex task that recognizes the scientific community (Hall et al.
1952; Brötz 1949). There are several studies cited in the literature about various kinetic modeling
schemes of the FTS process that derive from a catalytic reaction mechanism developed under cer-
tain considerations and assumptions that may be valid for a set of operating conditions that define a
specific operational scenario. However, it is important to mention that there is no generalized
model that accurately and uniquely describes syngas consumption and the hydrocarbon chain for-
mation and distribution process because it is a highly catalytic reaction.
There is a great diversity of FTS kinetic models for both iron-based and cobalt catalysts and
kinetic models for the WGS reaction that is involved in the water formation during the reaction
process. In addition, the main difference between the kinetics of cobalt-based catalysts and the iron
catalysts lies in the former’s inactivity toward the WGS reaction since H2O is not adsorbed on the
catalyst active sites.
8.1.3.1 Kinetic Models Developed with Iron Catalyst

Table 8.1 reports the different kinetic models developed from experiments carried out in fixed-bed
reactors with iron catalysts. Table 8.2 shows some reaction rates for water–gas shift with catalysts
based on iron. Also, Table 8.3 shows the bases for the kinetic model development and the specifics
of the reaches of each kinetic model for the FTS catalyzed with iron.
8.1.3.2 Kinetic Models Developed with Cobalt Catalyst

Table 8.4 shows the different kinetic models applied in fixed-bed reactors with cobalt catalysts, and
Table 8.5 shows the bases for the kinetic model development and the specifics of the reaches of each
kinetic model for the FTS catalyzed with cobalt.
One of the differences between cobalt catalysts and iron catalysts is that for kinetic modeling, the
WGS reaction is not considered, which is due to the very small amount of water produced that can
be neglected (Basha et al. 2015). However, there are cases of FTS fixed-bed reactor modeling that
Table 8.1 FTS kinetic models based on iron catalyst.
Operation conditions
Type of catalyst/
promoter P (MPa) T ( C) H2 Kinetic model Reference
CO
Fe p2H2 Hall et al.

— — — − RFT = k FT (1952)
pCO
Reduced fused 2.2.4.2 250–320 2.0 RFT = kFTPT Brötz and
Systematik
(1949)
Fe 2.2 225–240 0.25–2.0 − RFT = ap0H62 p0CO4 − f p0H52 O R0FT5 Anderson
and Karn
(1960)
Fe/K2O/Al2O3/SiO2 1.0–2.8 225–265 1.2–7.2 RFT = k FT pH2 Dry (1972)
pH2 pCO Dry (1976)
RFT = k FT
Fe 0.5–4.0 200–340 1.0–7.3 pCO + ac pH2 O
ac = K H2 O K CO
Reduced fused 2.0 250–315 2.0 pH2 pCO Atwood and

RFT = k FT
Fe/K2O/Al2O3/SiO2 pCO + bc pH2 O Bennet
(1979)
Precipitated copper Feimer et al.
pH 2
promoted by 1.0–2.0 220–270 1.0–6.0 RFT = k FT (1981)
potassium/iron p0CO25
Reduced precipitated 1.0–3.2 220–300 1.1–2.8 pH2 pCO Liu et al.
RFT = k FT
Fe/Cu/K pCO + Cc pH2 O (1995)
Cc ratio of the adsorption equilibrium constants
of the two kinetic models
k 1 pCO Lox and

Reduced precipitated pH 2 αn − 1 Foment
k 1 pCO + k 5 pH2
Fe/CuO/K2O/Na2O/ 0.6–2.1 250–350 3.0–6.0 RCn H2n+2 = k5 (1993a)
k 1 pCO 1
SiO2 1+
k 1 pCO + k 5 pH2 α − 1
(Continued)
Table 8.1 (Continued)
Type of catalyst/
CO
k 1 pCO
pH 2 αn − 1
k 1 pCO + k 5 pH2
RCn H2n = k 6
k 1 pCO 1
1+
k 1 pCO + k 5 pH2 α − 1
k 1 pCO
α=
k 1 pCO + k 5 pH2 + k 6
1 Jess et al.
RFT = k FT pH2 p
Fe 2.4 220–260 1.87–2.0 1 + 1 6 H2 O (1999)
pCO
Fe/Cu/K 1.1–3.1 220–269 1.0–3.0 k 5M pH2 α1 Wang et al.
RCH4 = , n=1 (2001)
1 p 1 1 j
1+ 1+ + H22 O + + N
i=1 j = 1 αj
k2 k3 k4 pH 2 k 2 k 3 pH 2 k4
j
k 5 pH 2 j = 1 αj
RCn H2n+2 = , (n ≥ 2)
1 p 1 1 j
1+ 1+ + H22 O + + N
i=1 j = 1 αj
k2 k3 k4 pH 2 k 2 k 3 pH2 k4
j
k 6 1 − βn j = 1 αj
RCn H2n = , n≥2
pH 2 O 1 1 N j
j = 1 αj
1
1+ 1+ k2 k3 k 4 + + + i=1
p2H2 k 2 k 3 pH 2 k4
05
pCO pH2 O pCO2 pH 2
kV
p0H52 KP
RCO2 = pCO pH2 O
1 + kV 05
pH 2
k 1 pCO
α1 = ;n = 1
k 1 pCO + k 5M pH2
k 1 pCO
αn = ; n≥2
k 1 pCO + k 5 pH2 + k 6
k−6
p
k 6 Cn H2n
βn = ; n≥2
k 1 pCO k 1 pCO n j−1
αnA − 1 + α pC n − i
i=2 A H n−1 + 2
k 1 pCO + k 5 pH2 k 1 pCO + k 5 pH2 + k 6 2
k 1 pCO
αn = ; n≥2
k 1 pCO + k 5 pH2 + k 6
k1, k2, k3, k4, and k5M: constants given by the Arrhenius equation
KP: equilibrium constant that is determined by a polynomial.
kV: rate constant of CO2 formation
Reduced Fe/Mn 1.0–3.0 267–327 1.0–3.0 Ak Yang et al.
RCH4 =
Bk + C k + Dk 2 (2003)
k7
RCn H2n+2 = 2
k 7M Bk + Ck + Dk
k 8+ 1 − βn K 3 pCO j
α
j=1 j
pH 2 O
RCn H2n = 2
Ak + Bk + C k
k 7M K 4 K 6 K 3 pCO p3H2
Ak =
pH 2 O
K 3 pCO p2H2
Bk = 1 + K 4 pH2 + K 1 pCO +
pH 2 O
K 6 K 04 5 K 3 pCO p0H52
Ck = K 1 K 2 pCO pH2 +
pH 2 O
N i
K 3 pCO p2H2
Dk = 1 + K6 K 4 pH 2 αj
pH 2 O i=1 j=1
whose kinetic rate constants involved in the model are established in the form given by Arrhenius
(Continued)
Type of catalyst/
CO
Fe/Co/Ni 0.1–0.7 250–270 1.0–2.5 k FT bCO pCO bH2 pH2 Mirzaei et al.
− RFT = 2 (2015)
1 + bCO pCO + bH2 pH2
bCO and bH2 : adsorption coefficients of CO and H2, respectively
Fe 1.1–2.61 260–300 0.67–2.05 k HC1 1 pCO Zhou et al.
αC1 = pH 2 O (2016)
k HC1 2 pCO + k HC5 l pH2 + k HC7 1
K HC3 K HC4
k HC1 n pCO
αCn = pH2 O ; (n ≥ 2)
k HC1 n + 1 pCO + k HC5 n pH2 + k HC6 n + k HC7 n
K HC3 K HC4
k HC6re n
βCn = pH2 O ; (n ≥ 2)
k HC1 n + 1 pCO + k HC5 n pH2 + k HC6 n + k HC7 n
K HC3 K HC4
n n
φ= 1 + αC 1 + αC j pCi H 2i + βCn pCn H 2n
i=2 j=i+1
k HC5 1 αC1 pH2

RCH4 =
φ
n n n
k HC5 n i = 1 αC i + i=2 j = i + 1 αC j βC i pC i H 2i + βCn + pCn H 2n PH 2
RCn H2n+2 = ; (n ≥ 2)
φ
n n n
k HC6 n i = 1 αC i + i=2 j = i + 1 αC j βC i pC i H 2i + βCn + pCn H 2n − k HC6re n pCn H 2n
RCH 3 H 2n = ;
φ
(n ≥ 2)
k HC7 1 αC1 pH 2 O
RCH 3 OH =
K HC3 K HC4 φ
n n n
k HC7 n i = i αC i + i=2 j = i + 1 αC j βC i pC i H 2i + βCn pCn H 2n pH 2 O
RCn H 2n+1 OH = ; (n ≥ 2)
K HC3 K HC4 φ
Table 8.2 Kinetics models for water–gas shift based on iron catalyst.
Type of catalyst/
CO
Fe 0.5–4.0 200–340 1.0–7.3 RWGS = RWGSpCO Dry (1976)

Reduced 1.0–2.0 220–270 1.0–6.0 RWGS = kWGSpCO Feimer et al.
precipitated (1981)
Fe/Cu/K2O
Reduced 0.6–2.1 250–350 3.0–6.0 pCO2 p0H52 Lox and
precipitated pCO pH2 O − Froment
K eq
Fe/CuO/K2O/ RWGS = k WGS 2 (1993b)
Na2O/SiO2 pH 2 O
1+a 05
pH 2
Fe 2.4 220–260 1.87–2.0 RWGS = k WGS pH2 O Wang et al.

(2001)
Reduced 0.6–2.6 210–250 1.6–4.1 pCO2 pH2 Kayser et al.
precipitated RWGS = k WGS pf pCO − (2000), Yang
K eq pH2 O
Co/MnO et al. (1979)
Reduced
Fe/Cu/K
Fe/Cu/K 1.1–3.1 220–269 1.0–3.0 pCO pH2 O pCO2 p0H52
− Wang et al.
p0H52 KP
RWGS = k V (2003b), Yang
pCO pH2 O et al. (2003)
1 + kV 05
pH 2
Reduced Fe/Mn 1.0–3.0 267–327 1.0–3.0 pCO pH2 O pCO2 p0H52 Yang et al.
− (2003)
p0H52 K eq
RWGS = ac
pCO pH2 O
1 + bc 05
pH 2
consider the WGS reaction, given that the amount of water experimentally produced is consider-
able (Dai et al. 2014; Moazami et al. 2015a).
8.1.4 General Aspects of Fischer–Tropsch Catalytic Mechanisms

Generally, not only taken as a valid assumption but also accepted as necessary is the fact that there
are different ways or reaction systems that depend on the type of catalyst and operating conditions
used, among other features that directly affect the approach of a single generalized reaction surface
mechanism during the FTS, thus giving rise to the existence of several alternate paths parallel oper-
ation also becoming an object of scientific research.
Currently, the scientific literature has reported a wide variety of studies on the reaction mechan-
isms involved in the FTS process, which today has resulted in the existence of a dispute about which
chemical reactions occur in this process. It is also important to keep in mind that for decades var-
ious chemical formation pathways of the monomer and the initiator of hydrocarbon chain growth
have been proposed. However, the most important monomers identified throughout the history of
Table 8.3 Bases for the kinetic model development and reaches of the FTS process description based on iron
catalyst.
Reference Bases for the kinetic model development Scope
Hall et al. (1952) The differential method was used for the The model describes the synthesis gas
development of the kinetic model, which is consumption well, but for a narrow range
a function of the partial pressures of the of operating conditions. Because the
syngas (CO and H2). kinetics was not based on a reaction
mechanism, i.e., the phenomena of
formation of the hydrocarbon chain, the
olefins readsorption, and others are not
contemplated, their use is limited for
cases where operating conditions of
industrial interest are explored.
Brötz and The kinetic model of the FTS reaction is The kinetic model was not based on any
Systematik directly proportional to the total system reaction mechanism, so it is limited for
(1949) pressure. use in case studies under industrially
relevant conditions.
Anderson and Use of the fractional coverages of the Methane production and the effects of
Karn (1960) surface by reactants and products. These CO2 inhibition are ignored. The postulate
surface coverages are approximated by of methylene formation as the complex
Freundlich isotherms since the coverage of species involved in the hydrocarbon
each component is proportional to its growth chain requires a broad
partial pressure in a positive exponent reconsideration of recent concepts on the
smaller than one. synthesis mechanism development,
especially in the formation of oxygenated
molecules.
Dry (1972) The Fischer–Tropsch synthesis mechanism At different low values of the H2O/CO2
involves the reaction of the adsorbed ratios, the kinetic model can predict that
hydrogen with the adsorbed CO, such that H2O is the predominant primary product
the first-order dependence of the hydrogen rather than CO2 and that at a fixed carbon
pressure means that the adsorption of monoxide partial pressure the activity
hydrogen or the reaction of the hydrogen increases with increased hydrogen partial
molecules adsorbed with the CO is the rate- pressure.
controlling step.
Dry (1976) The CO2 formation occurs easily under the At higher temperatures, the H2/CO ratio
process conditions used so that the water– does not manifest negative consequences
gas shift reaction is considered in the in the description of hydrocarbon
kinetics development. selectivity, and the selectivity seems to be
Langmuir’s adsorption theory was used to predominantly controlled by the CO2
represent a system in which the CO partial pressure.
molecule competes with the H2O, CO2, and Because Langmuir’s adsorption theory was
H2 molecules to occupy the adsorption simplified according to the operating
sites. scenario used, the model is limited to
certain operating conditions and therefore
to a good prediction of a broad spectrum of
hydrocarbons.
Atwood et al. The kinetic equation is based on the CO The kinetic model is limited to the
(1979) consumption rate. The diffusional effects description of a short range of liquid
on the catalyst particle were neglected, so hydrocarbons. However, it can predict the
the derived kinetics is intrinsic. H2O and CO2 production.
Feimer et al. The model was derived from the power law The model can predict the formation of
(1981) to represent the formation of C1 to C5 hydrocarbons with a low number of
hydrocarbons and alcohols based on the carbon atoms and alcohol production.
partial pressures of CO and H2. The reaction mechanism does not
contemplate the CO readsorption or the
olefins readsorption/hydrogenation, so the

reaction kinetics is limited to the
description of long-chain hydrocarbons.
Liu et al. 1995 The intrinsic kinetic model was developed Intrinsic kinetics predicts the inhibition of
based on the concept of passivation and the FTS reaction by the effect of water
using the lumping method. produced but not its inhibition by CO2.
Lox and The reaction kinetics was developed using The initial rates of carbon dioxide
Froment the LHHW theory under the assumption formation and various hydrocarbons
(1993a) that the elementary step that gives depend on the partial pressures of CO
formation to the monomer is the rate- and H2.
determining step. The model can predict the influence of CO
and H2 partial pressures on the formation
of olefins and n-paraffins with more than
two carbon atoms.
The model is limited to the use of an ideal
hydrocarbon growth probability model.
Lox and The kinetic model assumes that different The hydrocarbon distribution is limited to
Froment active sites are arranged on one side for the the ideal description given by the
(1993b), Jess gas–water shift reaction to occur and other Anderson–Schulz–Flory model.
et al. (1999) sites on the other side for the formation of The model only contemplates the
n-paraffins and 1-olefins to occur. elementary reactions of CO adsorption and
The rate kinetic expressions for the hydrocarbon desorption as those of greater
hydrocarbon formation were developed relevance in the kinetic model
from the elementary reactions dictated in development, leaving aside the
the carbide mechanism that are based on re-absorption of short-chain olefins
CO adsorption and products hydrocarbon truncating the long-chain paraffins
desorption, which are not in equilibrium. formation.
For purposes of simplifying the kinetic All the elementary reactions leading to
model, it is assumed that the active sites are hydrocarbon products formation are
occupied by surface hydrocarbon single-site reactions.
intermediates.
The kinetics of the gas–water shift reaction
implies a surface intermediate formate on
elementary reactions. The intermediate
formate formation is the rate-determining
step.
Wang et al. The kinetic expression is based on the The kinetic model can accurately describe
(2001) overall consumption rate of CO, the the paraffins and olefins formation.
hydrocarbon distribution model, and However, it is limited to the description of
assuming the olefin readsorption process. oxygenated compounds.
Yang et al. The alkylidene mechanism was used in The model good fits with the
(2003) which readsorption and secondary experimental data on the hydrocarbon
reactions of olefins are considered and distribution under the use of industrially
deviations of hydrocarbon distribution relevant conditions.
from the use of the conventional ASF The olefin-to-paraffin ratio was not
model are described. intrinsically described.
Both the product desorption and the
insertion of methylene into the metal-
alkylidene bond are considered the rate-
determining steps.
Only five surface steps are contemplated
for the derivation of the kinetic model:
(i) adsorption of reactants (CO and H2);
(Continued)
(ii) chain initiation; (iii) chain propagation;

(iv) chain termination and product
desorption; (v) readsorption and secondary
reaction of olefins.
The dissociative adsorption of hydrogen is
considered to occur both in the molecular
state and in the dissociative state.
Mirzaei et al. The LHHW theory was used and The diffusion limitations in the catalyst
(2015) associative adsorption of CO and H2 were pore play an important role in the
considered. catalytic syngas conversion that must be
considered for the derivation of a more
precise kinetic model.
Zhou et al. The elementary steps are based on the The model shows a deviation from the
(2016) carbide mechanism with hydrogen- linear distribution of ASF, which is
assisted CO insertion and CH2 as the because the structure of the reactants and
growth monomer. the transition states intermediates were
The readsorption of olefins was considered, not considered.
observing a negligible effect on the
hydrocarbon product distribution.
The oxygenation reation is considered and
the reaction scheme used was expressed in
terms of a mechanistic carbide model with
hydrogen-assisted CO insertion.
the FTS process are methylene (─CH2─), hydroxyl carbine (CHOH), and CO, which lead to
approach the following three important mechanisms FTS: (i) carbide, (ii) the enol mechanism,
and (iii) CO insertion mechanism, respectively. However, today different reaction mechanisms
have been developed with the aim of describing in a precise way the FTS process, which have been
proposed based on differences between the proposed mechanisms because of the assumptions that
arise for monomer, initiator, and their chemical formation. Such mechanisms are the alkyl, alkenyl
and alkylidene hydride-methylidene, and more recently the formate mechanism (Frennet and
Hubert 2000).
In the case of cobalt-based catalysts, there are two main mechanisms that describe the overall
scheme of FTS reactions: (a) the direct CO dissociation (carbide mechanism) and (b) hydrogen-
assisted CO dissociation. Although there are several studies that indicate that the mechanism of
hydrogen-assisted CO dissociation is the one that describes the FTS product’s distribution more
precisely, it is not able to predict the presence of atomic carbon on the catalytic surface (Hall
et al. 1952). Recently, a mechanism based on a parallel hydrogen-assisted mechanistic pathway
was used to develop an LH-type kinetic model to predict the presence of atomic carbon on the cat-
alyst surface. This mechanism is characterized by including the direct dissociation of CO, hydro-
gen-assisted CO dissociation, and parallel hydrogen-assisted CO dissociation (Keyvanloo et al.
2016). Recently, a work proposed a comprehensive mechanism-derived FTS kinetic model whose
results were compared with the most recent findings reported in the literature (eight kinetic mod-
els). It was shown that the kinetic model developed based on a combination of the alkyl/alkenyl
mechanisms for FTS reactions (to produce the n-paraffins and α-olefins) together with the
formate mechanism for the WGS reaction can provide more accurate predictions of the syngas
Table 8.4 FTS kinetic models based on cobalt catalyst.
Type of Operation Conditions

catalyst/
CO
Co/MgO/ 0.1 185–200 2.0 p2H2 Hall et al.

ThO2/ k FT = (1952)
pCO
Kieselguhr
Co/CuO/ 1.7–5.5 235–270 1.0–3.0 05 Yang et al.

RFT = k FT pH2 pCO
Al2O3 (1979)
Reduced 0.015–0.1 250 0.25–5.0 pH2 p0CO5 Outi et al.

precipitated RFT = k FT 3 (1981)
1+ K CO pCO + K H 2 pH 2
K: pseudo-adsorption equilibrium
constant for CO and H2 species
Co/ 0.2–1.5 190 0.5–8.3 Model 1:

Kieselguhr
p0H52 p0CO5
RFT = k FT 2;
1 + ac p0CO5 + bc p0H52
ac = 0.446, bc = 0.914
Model 2:
p0H52 pCO
RFT = k FT 2;
1 + ac pCO + bc p0H52
ac = 0.0116, bc = 0.0354
Co/TiO2 2.0265 180–240 1.0–3.5 − 0 24 0 74 Zennaro

RFT = k FT pCO pH 2
et al.
ac pCO p0H74
2 (2000)
RFT = k FT 2
1 + bc pCO
Co − Mn/ 0.1–1.0 190–280 1.0–3.0 k FT K CO pCO pH2 Visconti

TiO2 RFT = et al.
1 + bc pCO
(2011a)
k FT K CO pCO p2H2
RFT = 2
1 + 2 bc pCO
KCO: equilibrium constant of CO
adsorption step
Co − Ru/ 0.1 186–207 0.8–3.5 ac p0H52 p0CO5 Irankhah

Al2O3 RFT = 2 et al.
1 + bc pH2 + cc p0CO5 + dc pCO
(2007)
Co − Ce/ At 200–300 1.0–3.0 pH 2 Mansouri

RFT =
SiO2 atmospheric 1 + ac pCO et al.
pressure ac: constant lumped (2013)
Co/Al2O3 1.0–2.5 190–220 1.5–3.0 pCO pH2 Kaiser

RFT = k FT
1 + ac pCO et al.
ac: determined by the Arrhenius equation (2014)
Table 8.5 Bases for the kinetic model development and reaches of the FTS process description based on
cobalt catalyst.
Visconti et al. The two kinetic models were based on the Both models predict the synthesis of
(2011a) Langmuir–Hinshelwood theory. synthesis gas (CO and H2) and the
selectivity of hydrocarbon products.
However, it was only validated for a
narrow range of operating conditions.
A rigorous kinetic mechanism for the
development of kinetics is not
contemplated.
Yang et al. Kinetics is based on the power rate law. Kinetics indicated that hydrocarbon
(1979) The conventional ASF model was used to formation increases linearly with syngas
describe the hydrocarbon chain growth. conversion, limiting its use with a reactor
model where high linearities are present.
Outi et al. FTS kinetics consider that the process The kinetics can explain the partial
(1981) initiation is carried out by the CO pressure effect of the syngas on the olefin/
dissociation and the formation of the CH2 paraffin ratio and on the methane
surface intermediate, where its formation is selectivity. The kinetics illustrates that the
a rate-determining step. It is assumed that chain growth probability constant is a
the hydrocarbon chain growth proceeds complex function of the CO and H2 partial
through the addition of the same CH2 pressures.
groups to the growing molecule.
Sarup and Two kinetic models are of LHHW type in The step of CH formation as a building
Wojciechowski which CO conversion to hydrocarbons and block of the hydrocarbon chain is
(1989) methane rate formation are used. The first assumed in the reaction mechanism
model is based on the hydrogenation of which allows the kinetics to predict CO
surface carbon and the second on a conversion and methane formation good.
hydrogenation-assisted dissociation of CO Both models fit satisfactorily to CO
both considered as the rate-determining conversion and methane formation data
steps of the FTS process. for a limited range of operating
conditions.
Zennaro et al. Two kinetics (i) of the power law type and Both kinetics are limited to their
(2000) (ii) a simplified version of Langmuir– application to a wide range of pressures,
Hinshelwood are developed. gas space velocities, and feed syngas ratio.
They do not integrate a hydrocarbon
chain distribution model.
Irankhah et al. LHHW model was used to obtain kinetics The kinetic modeling is validated through
(2007) and parameters to describe CO conversion a linear relationship between the reaction
to CH4 formation under supercritical rate of the CO consumption measured
conditions. Kinetics formulation was based with respect to that estimated in
on the following steps: (i) adsorption and supercritical conditions. However, it lacks
dissociation of CO and H2 occur at the validation against the description of
active sites of the cocatalyst; (ii) removal of hydrocarbon selectivity and estimation of
oxygen from the catalytic surface through the hydrocarbon chain. The range of
water and CO2; (iii) hydrogenation of operating conditions was narrow.
adsorbed carbon, oligomer production and
formation of monomer (CH2∗S) and alkyl
species (CH3∗S); (iv) methane formation.
Mansouri et al. Kinetics is based on the LHHW approach. The activation energy value is too low
(2013) Kinetics assumes the following: (i) intrinsic compared to the values reported in the
reaction rates are proportional to the literature for a set of similar operating
surface coverage of reactants; (ii) CO conditions. The kinetics well describes the
consumption and CH2 intermediary hydrogenation of CO toward light olefins.

formation are the irreversible rate- The LHHW approach is relatively simple.
determining step while the other reactions The application of kinetics is limited to the
are in quasi-equilibrium; (iii) there is no use of operating conditions of industrial
accumulation in the catalytic surface at relevance.
steady-state conditions; (iv) operation is
considered under isothermal regime;
(v) the catalyst is fixed and uniformly
distributed; (vi) elementary CO and H2
adsorption are in quasi-equilibrium within
the concentration of gas phase; (vii) water
is removed after irreversible CO
decomposition; (viii) CO and carbon
concentrations are assumed to be the
dominant concentrations.
Kaiser et al. Intrinsic kinetics for CO consumption and CO2 and H2O formation show negligible
(2014) its parameters were evaluated using the effect on FTS reaction rate. Water
Langmuir–Hinshelwood (LH) approach. produced has minor effect on product’s
The influence of pores diffusion was distribution as predicted by kinetics.
assumed. Simplified LH approach despises the
adsorption of H2. The model predicts well
FTS and CH4 reactions for medium range
of commercially relevant operating
conditions.
consumption, hydrocarbons formation, and a better description of the FTS reaction network
(Moazami et al. 2017a). An important finding of this study given by Moazami et al. (2017a) was
that the combination of the alkyl and alkenyl mechanisms for the formulation of a comprehensive
kinetic model that allows to describe well the initiation, propagation, and termination steps of the
hydrocarbon chain, since it is a fact that the alkenyl mechanism largely represents the formation of
α-olefins, while the alkyl mechanism allows to describe well the formation of n-paraffins. In addi-
tion, it was found the alkyl mechanism favors the paraffins formation rather than olefins; hence it is
a better choice for this case compared to other mechanisms, e.g. alkenyl, CO insertion, and/or enol
mechanisms. However, care should be taken only for the use of the alkyl mechanism since erro-
neous underestimations of olefin selectivities can be obtained. On the other hand, it was observed
that considering the WGS reaction kinetics, the formate mechanism, in which the formate species
that were formed from the reaction between the adsorbed CO intermediate and a hydroxyl surface
species (─OH), was considered the most relevant route for the construction of a kinetic model, since
it is able to describe in a precise way the CO consumption and the higher hydrocarbon distribution
(Moazami et al. 2017a).
8.1.5 The Fischer–Tropsch Synthesis Product Distribution Models

The FTS product distribution is usually predicted by the classical model of ASF, which is repre-
sented by the following equation:
f C = n 1 − α 2 αn−1 8 14
where fC is the fraction of carbon atoms that make up the length of the hydrocarbon chain of size
n and α is the chain growth probability factor that is determined by the following expression
(Todic et al. 2015):
rp
α= 8 15
rp + rt
where rp and rt are the propagation and chain termination velocities, respectively. On the other
hand, van Dijk (2001) reported the following equation to determine the probability of chain
termination:
1−α
αt = 8 16
α
The growth of hydrocarbons considering a polymerization FTS reaction scheme is illustrated in

Figure 8.1, in which the probability of growth is characterized by C that depends on the type of
catalyst and operating conditions. The chain termination leads to the formation of products (olefins
and paraffins), where the hydrogenation reaction of paraffins is irreversible whereas for olefins it is
reversible, which implies readsorption of these in catalytic surface and again the hydrogenation
reaction to continue with the formation of paraffins.
However, it is noteworthy that the idealized ASF model has certain deviations from the precise
description of the hydrocarbon spectrum generated in the FTS process. It is known that the selec-
tivity of the carbon number of the FTS products does not fully obey the classical ASF model due to
three types of deviations that have been observed and have been reported previously: a production
Probability
CH4 1–α
C2H6 α(1 – α)
C3H8 α2(1 – α)
C4H10 α3(1 – α)
⋮ ⋮
Cn–1H2n αn–2(1 – α)
CnH2n+2 αn–1(1 – α) C5H12
⋮ C5H10
αt = 1 – α α
C5* C2H4
CnH2n+2 C2H6
CH4
C*n
C*2 C3H6
C*
CnH2n C*3
C3H8
C4H10 C6H12
C4H8 C*6
C*4
Cn–1H2(n–1)
C6H14
Cn–1H2n
C*n–1
Figure 8.1 FTS hydrocarbons growth simplified scheme.

of methane higher than expected; a lower production of ethene than expected; and higher selec-
tivity to heavy products, which is due to increase in the chain growth probability with the number
of carbon (Lozano-Blanco et al. 2009; Todic et al. 2014; Novak et al. 1982). Some authors identified
several factors that may be responsible for such deviations, for example, Masuku et al. (2011) and
van der Laan and Beenackers (1999) attribute these deviations to phenomena such as diffusion lim-
itations, vapor–liquid equilibrium (VLE), and readsorption. It is also believed that deviations can
come from analytical difficulties (Puskas et al. 1993) and/or due to handling and the reactor under
unsteady-state conditions (Huff and Saterfield 1985).
However, others suggest that deviations from the ideal ASF model may be due to higher produc-
tion of methane than expected, low selectivity to ethane and heavier products, as well as a lot of FTS
products showing a strong deviation, which may be due to mainly secondary reactions of the pri-
mary hydrocarbon product, such as reinsertion into the chain growth process, hydrogenation, and
hydrogenolysis (Kuipers et al. 1996; Schulz and Claeys 1999). The secondary reactions exhibit a
strong influence on the FTS products spectrum, which is critical to optimizing the selectivity to
the desired product range (Kuipers et al. 1996).
Other cases of the current study aim at improving the ASF model by describing in more detail the
FTS product’s distribution and may be able to explain the deviations that may happen with the
classical ASF model (Todic et al. 2015).
Nowadays, the interest in having more accurate predictive models for a better description of the
product’s distribution has increased due to the improvements in the analysis of all isomers and pro-
ducts involved in the FTS, which cannot be calculated using the classical ASF model distribution
(van der Laan and Beenackers 1999).
It is worth noting that product selectivity in the FTS process can not only be affected by the cat-
alyst but also by certain operating conditions such as temperature, syngas feed composition, TOS,
and catalyst reduction (Moutsoglou and Sunkara 2011). van der Laan and Beenackers (1999) pre-
sented a more detailed discussion about the way the after-mentioned operating conditions affect the
performance and overall efficiency of the FTS process.
It has been experimentally found that the water production during the FTS reaction with cobalt
catalyst favors the higher hydrocarbons selectivity (C5+). In addition, according to the study con-
ducted by Rytter et al. (2016) for the FTS process with cocatalyst supported on Al2O3, SiO2, and
TiO2, it was found that the selectivity to C5+ increases and the selectivity to CH4 decreases with
increasing water partial pressure.
Table 8.6 reports some of the most important products derived from the FTS technology and the
corresponding carbon number (Mabry 2014).
Table 8.6 Products of interest in the FTS process and its carbon number range.
Name Carbon number range Name Carbon number range
Fuel gas C1 − C2 Diesel C13 − C17

LPG C3 − C4 Middle distillates C10 − C20
Gasoline C5 − C12 Soft wax C19 − C23
Naphtha C8 − C12 Medium wax C24 − C35
Kerosene/jet fuel C11 − C13 Hard wax C35+
8.1.6 Final Remarks

The development of a specific scheme of reactions should involve the full range of FTS products;
however, such an approach of a reaction mechanism to accurately describe the synthesis process is
a rather complex task. There is a great diversity of works in the literature for the development of FTS
kinetic models, which consider all the aspects involved in the catalytic steps that give rise to the
formation of intermediary species and initiators of the hydrocarbon chain growth. However,
despite the great effort of many researchers, there is still a need for a complete model that describes
the range of FTS products under any scheme of operating conditions, types of catalyst, and raw
material used.
8.2 Modeling of Catalytic Fixed-Bed Reactors for Fuels Production

by Fischer–Tropsch Synthesis
8.2.1 Introduction
A comprehensive review of the various studies reported in the literature up to date on the math-
ematical modeling of fixed-bed reactors to produce fuels by the FTS was carried out. It is quite clear
that most of the proposed models are based on a set of assumptions that allow their wide simpli-
fication by reducing the models into forms of low complexity, since in most cases the effects of
phase equilibrium are neglected and relatively simple Fischer–Tropsch kinetics of the power-
law type are used. In addition, most of the proposed modeling schemes neglect the effects of resis-
tances to gas–liquid and liquid–solid mass transfer. On the other hand, few reports consider the
energy effects under the consideration of a non-isothermal operation assuming a plug-flow behav-
ior and a gas–liquid system. A generalized model of a fixed-bed FTS reactor is proposed, which takes
into account all the mass and heat transfer phenomena, as well as hydrodynamics and VLE, based
on the information given in the literature. It is evident that for fixed-bed reactors for fuel production
using Fischer–Tropsch technology, there is little experimental information for validation and a
need to explore different types of reactor models, such as reactor models under a trickle-flow regime
considering the effects of phase distribution and dispersion under transient state conditions.
8.2.2 Modeling of Fixed-Bed Fischer–Tropsch Reactors

Although different types of reactors have been reported to carry out the FT synthesis for fuels pro-
duction (diesel and gasoline), the FBR has increased the interest of the scientific and industrial
community for the following reasons (Hooshyar et al. 2012; Kaskes et al. 2016):
1) Strong potential over slurry bubble column because it operates with a concentration gradient of
plug-flow type.
2) High catalyst holdup compared with other reactor configurations.
3) No requirement for a separation step and recovery of the catalyst, which is a major advantage
from the point of view of time and reduction of operating costs.
4) Easy of scale-up.
In addition, it is possible to produce different types of hydrocarbons due to the following reasons:
1) The wide range of catalysts that can operate at low and high temperatures.
2) Different operating conditions.
3) Type of material and geometry employed in the reactor design.
8.2 Modeling of Catalytic Fixed-Bed Reactors for Fuels Production by Fischer–Tropsch Synthesis 325
Mathematical modeling is a tool with high potential for an accurate description of the different
types of reactors that are used by FTS technology (Moazami et al. 21015b). The importance of devel-
oping models is to predict the effects of operating conditions on the thermal behavior of the reactor,
especially in industrial application reactors or so-called large-scale reactors (Rados et l. 2013).
In the case of fixed-bed reactors, there are some factors that complicate the modeling (Wang et al.
2001; Wang et al. 2003a):
1) Multidimensional distributed configuration.
2) The nature of the two phases (gas and liquid) involved in the reactor bed.
3) Nonlinear dependence of the reaction rate on the temperature.
4) Uncertainties of both the packaging and the flow through this.
5) The uncertainties of the parameters for heat and mass transfer.
6) Parameters that vary depending on the place within the catalyst bed.
7) The complex nature of the reaction kinetics involving many chemical species.
8) The time-varying reaction rate due to catalyst deactivation.
For modeling of FTS catalytic fixed-bed reactor, the transport parameters of heat and mass, the
hydrodynamics, effects of pressure drop, holdups of both phases (gas and liquid), phase equilib-
rium, chemical kinetics, selectivity, and product distribution must be taken into account to obtain
a generalized mathematical model to predict the dynamic behavior under a wide range of operating
conditions. In addition, the model must be fitted to the data obtained in experimental runs for val-
idation purposes and scaling-up. To give robustness to the model, further verification is required,
which means that the model must accurately describe other experimental data.
8.2.2.1 Classification of Fixed-Bed Fischer–Tropsch Reactor Models

It is well-known that a multi-tubular fixed-bed FTS reactor is like a conventional shell and heat
exchanger tube reactor with the FTS reaction that occurs in the tube side. Water is generally used
as a coolant, which circulates through the shell side to keep the process under an isothermal regime.
The proposed mathematical model for such a system must take into accounted the interactions that
arise between the side shell and the catalyst tube where highly exothermic reaction occurs. There-
fore, the phenomenon that occurs in the catalytic fixed-bed FTS reactor can be characterized
according to the following (Iordanidis 2002):
•• Effects of heat and intraparticle mass diffusion.

Heat and mass convective transport in the fluid phase.
•• Heat and mass effects occurring between the fluid and solid phases.
Effect of heat exchange between the side of the exothermic FTS reaction and tube wall.
• The process of thermal conduction and mass diffusion in the solid phase.
The typical modeling approaches widely used for accurate description of fixed-bed reactors are
the continuum models, which are classified according to Figure 8.2 (Jakobsen 2014; Froment
et al. 2011):
8.2.2.2 One- and Two-Dimensional Pseudohomogeneous Model

In the pseudohomogeneous model, the presence of the solid is not considered, i.e. the resistances to
heat and mass transfer occurring between the solid phase and the fluid phase are negligible, assum-
ing that the catalyst surface is exposed to bulk fluid conditions, and intraparticle diffusion effects
are ignored. Such a model is classified into one-dimensional plug-flow, one-dimensional plug-flow
with axial dispersion, two-dimensional plug-flow with radial dispersion, and two-dimensional with
axial and radial dispersion models (Froment et al. 2011).
Continuous
Models
Pseudohomogeneous Heterogeneous
1-D plug flow 2-D plug flow 1-D plug flow 1-D axial
with radial dispersion
1-D with axial dispersion 2-D axial
dispersion and and radial
2-D axial and dispersion
radial dispersion Integration Interparticle
effectiveness effects
factor
Figure 8.2 Schematic representation of the models commonly used for the description of catalytic fixed-bed
reactors (Jakobsen 2014; Froment et al. 2011).
8.2.2.3 One- and Two-Dimensional Heterogeneous Model

In both one-dimensional (axial dispersion) and two-dimensional (axial and radial dispersion) het-
erogeneous models, the intraparticle diffusion limitations are considered (Froment et al. 2011). In
other words, when heat and mass transport within the catalyst particle offer considerable resist-
ance, the reaction rate through the catalyst particle is not uniform (Jakobsen 2014). On the other
hand, in the FTS process, the catalysts are completely filled with liquid, for which it is important to
take into account the limitations of intraparticle mass transfer for model effects. In addition, the
FTS reaction is strongly influenced by the diffusion effects on the catalyst pore and the effective
rate (Jess and Kern 2012a). Thus, the modeling of an FTS reactor under a heterogeneous regime
is necessary for a more accurate prediction of the FTS process involved in the fixed-bed reactor.
8.2.3 Development of a Generalized Fixed-Bed Fischer–Tropsch Reactor Model

From the literature review, it must be highlighted that it is strictly necessary to develop a general
model for FTS fixed-bed reactors to propose a robust mathematical model that can accurately
describe certain phenomena naturally occurring in FTS processes for the gasoline or diesel produc-
tion according to either at high- or low-temperature operation. It is also important to have well-
analyzed heat and mass transport phenomena to safely develop a mathematical model that can
predict what really happens in commercial fixed-bed reactors.
The formulation of the mathematical model depends on the mass and heat balances for the var-
ious phases involved (gas, liquid, and solid), as well as on the balance of momentum and pressure
drop through the catalyst bed. It is important to note that during the development of a generalized
reactor model, all the terms related to resistances to mass and energy transfer, among other terms,
must be incorporated into the mass and energy balance. It is recognized that the solution of the
resulting model can be extremely difficult and requires considerable computational time for its
numerical solution.
8.2.3.1 General Equations of the Model

To perform proper balances of mass and energy, it is necessary to define the types of phases involved
in the FTS process carried out in the fixed-bed system. The species in the gas phase are the syngas
mixture (CO and H2), CO2 and H2O (water steam), and light hydrocarbons formed during the FTS
process: CH4, C2 − C4 (C2H4, C2H6, C3H6, C3H8), i − C4H10, n − C4H10, while the liquid phase is
constituted by volatile species such as short-chain hydrocarbons (volatile liquid hydrocarbons such
as some olefins and paraffins in the gasoline and diesel, respectively, and oxygenated compounds
such as alcohols and ketones) and nonvolatile species such as long-chain hydrocarbons (heavy liq-
uid hydrocarbons such as waxes); and solid is the catalyst. Then, the terms involved in the general-
ized model for the fixed-bed catalytic FTS reactor must be as follows: (i) for mass balance, the terms
of accumulation, convective transport, and axial and radial mass dispersion are to be considered, as
well as the gas–liquid, liquid–solid, and gas–solid mass transfer effects due to flow regimes; (ii) for
the energy balance, the terms of accumulation, convective transport, and axial and radial heat dis-
persion, as well as heat due to mass transfer effects between the different phases, are to be consid-
ered. The mass and energy balances of the generalized model for fixed-bed FTS reactor are
summarized in Tables 8.7–8.10. The following assumptions are considered:
• Properties of gas and liquid (mass and heat dispersion coefficients, specific heats, holdups, den-
sities, and viscosities), of the catalyst particle (porosity, size, activity, effectiveness, etc.), and bed
void fraction are assumed to be constant throughout the entire catalyst bed.
• Regarding the behavior inside the catalytic particle, the coefficients of mass- and heat-effective
diffusivity are considered as constants.
•• The gas and liquid superficial velocities are variable.

For the non-isothermal case, both physicochemical and thermodynamic properties must be eval-
uated at each point of discretization during the numerical simulation.
• The reactor parameters that are assumed to be constant can be put out of the partial derivatives
with respect to the axial and radial spatial coordinates involving them.
Although in the literature some correlations have been reported to predict the variation of some
parameters with respect to spatial coordinates, their use in conventional continuum models is quite
complex (Mederos et al. 2009).
The proposed generalized model for fixed-bed FTS reactors is based on a set of mass, energy, and
momentum balance equations developed for each component in each phase f, which are derived
from the fundamental form of the multifluid continuity equation considering the terms of reaction
rate and convective mass transfer (Jakobsen 2014; Froment et al. 2011; Bird et al. 2001).
Mass Balance The equations involving the mass balances in Table 8.7 assume that each phase in
the fixed-bed system is a continuum able to be represented by an Eulerian–Eulerian framework
model. Figure 8.3 shows schematically the concentration profiles in a fixed-bed FTS reactor,
and the mass transport in different phases is identified.
It is important to note that in the generalized model, the fixed-bed FTS reactor was considered to
operate in a trickle-flow regime. According to Caldwell and Van Vuuren (1986), if the operating
conditions and the catalyst formulation suitably enhance the FTS process, it is possible to obtain
a liquid phase inside the reactor operating in a trickle-flow regime where the syngas and liquid
product flow downstream in a concurrent way. On the other hand, it is reasonable to assume that
there is a distribution of the gas and liquid phases within the reactor, which varies in time along the
axial and radial directions. Therefore, in the proposed generalized reactor model, the behavior of
the gas–liquid distribution phases through the bed using dynamic holdups of both phases involved
is taken into account. In addition, it is convenient to take into consideration the partial wetting of
the catalyst since this effect is common in a trickle-bed reactor. Relatively high surface velocities of
the gas and liquid along the reactor allow simplifying the proposed generalized model by consid-
ering an axisymmetric flow field of the gas and liquid phases in the whole bed, and dynamic
Table 8.7 Generalized mass balance equations for gas and liquid phases.
Axial
Accumulation Convective dispersion Radial dispersion G–L transfer G–S transfer L–S-transfer
Gas phase ∂CGi ∂ uGs C Gi ∂ C Gi

2
1 ∂ ∂C Gi − k GL GL
i aI CGi − CLi − 1 − f w k GS GS
i aI + f w k LS
i aI
LS
ϵG = − + ϵG DGax + ϵG DGrad r
∂t ∂z ∂z2 r ∂r ∂r CGi − CSi sur CLi − CSi sur
Liquid phase: volatile ∂CLi ∂ uLs CLi ∂ 2 C Li 1 ∂ ∂CL − k GL GL

i aI CGi − CLi — + f w k LS
i aI
LS
species 1 − f St ϵL = − + ϵL DLax + ϵL DLrad r i
∂t ∂z ∂z2 r ∂r ∂r CLi − CSi sur
Liquid phase: nonvolatile ∂CLi ∂ uLs CLi ∂ 2 C Li 1 ∂ ∂CL — — + f w k LS

i aI
LS
species 1 − f St ϵL = − + ϵL DLax + ϵL DLrad r i

∂t ∂z ∂z2 r ∂r ∂r CLi − CSi sur
Stagnant liquid ∂C Li — — — − k GL GL
i aI CGi − CSt — + f w k iLSt S aLS
f St ϵL = L,i I
∂t L,i − C i
CSt Ssur
Table 8.8 Generalized mass balance equations for gas–solid phase and liquid–solid phase.
Accumulation Intraparticle diffusion G–S transfer L–S transfer Generation
p ∂C
Solid phase/ Ssur
wet surface 1 − εB ϵL L,i = — — I C St,i − C St,i

+ kiLS aLS L Ssur
—
∂t
Solid phase/ ∂C Ssur NS
dry surface 1 − εB ϵpG G,i = — + 1 − f w k GS GS
i aI CGi − CSG,isur — + ρB vij ηj RFT,j CSi sur , T Ssur
∂t j=1
Solid phase/ ∂C SG,iins Spherical catalyst — — NS

inner
p
ϵG = + ρS vij ηj RFT,j C Si ins , T Ssur
∂t DGeff ,i ∂ ∂CSins j=1
+ 2 ξ2 G,i
ξ ∂ξ ∂ξ
Cylindrical catalyst
DGeff ,i ∂ ∂C Sins
+ ξ G,i
ξ ∂ξ ∂ξ
Hollow cylinder catalyst
G
Deff ,i ∂ ∂C SG,iins
+ ξδcyl + r inn
ξδcyl + r inn ∂ξ ∂ξ
Table 8.9 Generalized heat balance equations for fluid phases (gas and liquid).
Axial Fluid-interface
Accumulation Convective dispersion Radial dispersion transfer Conductive
Gas ∂ ρ G
CGP,i T G ∂ uGs ρG C GP,i T G 2
∂ T G 2
∂ T 1 ∂T
G G
− hGL GL
c aI T −T
G I N CG pGi
phase ϵG = − + ϵG λGax + ϵG λGrad + + K GL GL
L,i aI − CLi C GP,i T I − T G − ΔH vi
∂t ∂z ∂z2 ∂r 2 r ∂r i=1 Hei
Liquid ∂ ρL C LP,i T L ∂ uLs ρL C LP,i T L ∂2 T L ∂2 T L 1 ∂T L N CL pGi

ϵL = − + ϵL λLax + ϵL λLrad + + K GL GL
L,i aI − CLi C LP,i T I − T L + ΔH vi
phase ∂t ∂z ∂z2 ∂r 2 r ∂r − hGL GL
c aI TI − TL i=1 Hei
Fluid–solid transfer Fluid wall transfer
Gas phase − 1 − f w hGS GS

c aI T G − T Ssur − 1 − f w hGW
ABW G
T − TW
c
Vr
Liquid phase − f w hLS
c aI T − T
LS L Ssur
− f w hLW
ABW L
T − TW
c
Vr
Accumulation Axial dispersion Generation
Solid phase-non-isothermal ∂T Sins

∂ T2 Sins
2 ∂T Sins N CG
ρS CSP = + λSeff + ρS − ΔH FTR,j RFT , j CSi ins , T Sins
∂t ∂ξ2 ξ ∂ξ
j=1
Table 8.10 Generalized heat balance equations for the solid phase (catalyst).
Axial
Accumulation dispersion Radial dispersion G–S transfer L–S-transfer Generation
Solid phase- ∂T Ssur ∂ 2 T Ssur ∂ 2 T Ssur 1 ∂T Ssur − 1 − f w hGS GS

c aI − f w hLS LS
c aI T G − T Ssur NS
isothermal ϵS ρS C SP = + ϵS λSax + ϵS λSrad + ρB − ΔH FTR,j vij ηj RFT,j CSi sur , T Ssur
∂t ∂z2 ∂r 2 r ∂r T G − T Ssur
j=1
KiGS holdups can be neglected; in consequence, the holdups

dynamics in mass balances of gas and liquid phases,
Liquid i the volatile liquid species, and zones of the stagnant liq-
I C L uid are equal to 1. Accordingly, the concept of catalyst
Ci i
Gas i Solid wetting factor for modeling purposes can be neglected,
since it is very common that in the FTS process the thick-
CiG CSsur Pore ness of liquid film on the catalyst surface is negligible, as
i
Ci* reported by Jess and Kern (2012a), by which external dif-
CSin fusion limitations can be ignored at operation tempera-
i
kiL
G
ture below 400 C. The proposed generalized model for a
ki
k LS fixed-bed FTS reactor takes into account the wetting
i
effect of the catalyst particle proposed by Mederos
et al. (2009) for a trickle-bed reactor applied to the hydro-
treating of petroleum fractions, as well as a similar anal-
KiLS ysis for the mass balance equations of the liquid phase in
KiGL
stagnant zones, on the surface and inside of the catalyst
particle.
Figure 8.3 Concentration profiles in the
fixed-bed FTS reactor. As for the gaseous species (reactants and volatile
liquids produced) present in the fixed-bed FTS reactor,
the convective flow term is considered to be plug-flow type, which implies that the concentration
and temperature gradients reached only occur in the axial direction, i.e. along the catalytic bed. The
mass balance for the gas phase describes the dynamic behavior of the reactor for the generalized
non-steady-state model (represented by the accumulation term) and its steady-state behavior (with-
out considering the transient effect).
The effect of axial dispersion of fluid flow in fixed-bed reactors (on a commercial scale) can be
neglected in cases where gas and liquid velocities are high enough (Mederos et al. 2009). On the
other hand, when the gas-flow pattern exhibits a plug-flow type behavior, the axial dispersion
effects can be discarded because the axial dispersion coefficients tend to zero (Salmi et al. 2000).
As in the case of the mass radial dispersion term in a fixed-bed system, this can be neglected when
the ratio of diameters dt/dpe is higher than 25, which means that the variation of the radial porosity
within the reactor can be completely omitted (Mederos et al. 2009). In addition, it has been reported
that the dispersion coefficient in the axial direction is greater than the radial dispersion coefficient
by a factor of 5 when working with Reynolds numbers greater than 10, whereby the radial disper-
sion of mass in a packed bed may be ignored when operating at high Reynolds number values
(Delgado 2006).
The resistance to mass transfer is explained in detail by the theory of the two films as schematized
in Figure 8.4. The mass transfer from gas to liquid is determined using a fictitious liquid-phase con-
centration, which is the concentration that could be in equilibrium with the corresponding bulk
partial pressure in the case of light hydrocarbons. Then the following equation is used for its
determination:
pG,i
C∗i = 8 17
Hei
where Hei is Henry’s constant taken as an equilibrium ratio representing the solubility of the i com-
ponents in the gas state in the fraction of liquid product. In addition, it is important to considerer
that when the liquid phase produced is slightly volatile under the operating conditions of the FTS
Reactants diffusing
into the pores Profile of concentration during the
Fischer-Tropsch reaction
Resistance to internal mass

CI CL transfer
i i
CS sur
i
Gas i Pore
Liquid filled catalyst CS in

i
pores
Liquid i
Solid
Resistance to gas-liquid external

mass transfer (usually negligible)
Figure 8.4 Schematic diagram of the Fischer–Tropsch reaction in the fixed-bed system on the catalyst
particle.
process, the gas–liquid mass transfer effect on the liquid side can be ignored. However, when the
wetting efficiency of the catalytic particle is assumed to be 1 ( fw = 1), the direct contact between the
gas and the solid catalyst particle is lost, resulting in a null conversion of the reactants CO and H2,
and the reaction finished since it takes place directly into the solid, which is an extreme case to
consider.
The dynamic holdup of the liquid phase must be considered similarly to that of the gas phase,
because as the FTS reaction occurs along the catalytic bed the gas and liquid fractions inside
the reactor change, which is mainly due to the consumption of the syngas (convection) and pro-
duction of liquid that are a function of time and of the axial and radial positions. In addition, if
the liquid produced has a partial volatilization effect, this will reflect a variation of the dynamic
holdup along the catalytic bed (Mederos et al. 2009).
It is well-known that inside fixed-bed reactors involving a liquid phase, there may be some areas
of liquid stagnation, the so-called dead zones. Many regions of stagnant liquid regularly occur
mainly in several regions of the catalytic bed affecting the holdup dynamics of both the liquid
and gas. For this reason, in this work, an equation for stagnant liquid is proposed taking into
account the gas–liquid transfer effects on the liquid side and mass transfer from stagnant liquid
to solid catalyst.
To develop the mass and energy balance in solid catalysts, the following assumptions were pro-
posed based on the schematic diagram depicted in Figure 8.4 (Jess and Kern 2012a, b; Pöhlmann
and Jess 2016a):
• The Fischer–Tropsch reaction is carried out inside pores as commonly a heterogeneous catalytic
reaction occurs.
•• The catalyst pores are considered to be fully filled with liquid.

Since inner catalyst pores are filled with waxy liquid, the liquid film thickness is commonly
neglected.
• Due to the formation of liquid within the catalyst, concentration profile agrees with the two-film
theory.
• The wax formed inside the pores is liquid by which solid waxes in pores are discarded.
For the performance of mass balances of the solid catalyst particle, two cases were considered:
(i) a dry surface of the catalyst and (ii) a partially wetting. In both cases, the internal mass gradients
in the solid phase must be evaluated using the effectiveness factor (Mederos et al. 2009), which is
defined as the ratio of the overall reaction rate to the reaction rate that arises when the whole inner
surface is exposed to the external condition (Lee and Chung 2012).
The mass gradient that occurs within the catalytic particle results from effective diffusion, which
depends mainly on two important factors: (i) the porosity of the catalyst and (ii) the size of the mol-
ecule diffusing through the catalytic pores. To achieve maximum FT catalyst effectiveness, it is
recommended to operate under conditions where the mass transfer limitations at the liquid–solid
interface are neglected (Mederos et al. 2009). There is evidence that reducing the size of the catalyst
particle allows for an increase in the effectiveness factor (ηi) by reducing the path lengths inside the
particle (Wang et al. 2001). However, the catalyst pellet size plays an important role in reactivity and
selectivity during the FTS process because they are size-dependent (Jess and Kern 2012a, b).
According to the reactivity and selectivity analysis provided by Wang et al. (2001), a large size
of catalyst has a negative effect on the reactivity and selectivity of desired products because strong
diffusion effects occur and small-size particles are preferred instead. In industrial-scale applica-
tions, catalyst pellets 2–4 mm in diameter are commonly used to keep the pressure drop as low
as possible and to remove the generated heat efficiently (Sie and Krishna 1999). However, according
to Jess and Kern (2012a) and Jess and Kern (2012b), a particle diameter >3 mm is suitable for
obtaining a high production rate of higher hydrocarbons. Moreover, under effective conditions
(i.e. with particles of 1 mm as used in fixed-bed reactors), the FTS is affected by internal mass trans-
port limitations, which lead to an increased H2/CO ratio inside the particle, which has an impact on
the local reaction rate and selectivity (Jess and Kern 2012b).
On the other hand, the effect of the particle size of the catalyst on the reaction rate can also be
quantified with the effectiveness factor, as well as with the Thiele modulus (Macías and Ancheyta
2004). A simple expression for the effectiveness factor and the Thiele modulus valid for spherical
particles and first-order kinetics is (Post et al. 1989):
1 1 1
η= − 8 18
ΦS tan h 3ΦS 3ΦS
05
k in
ΦS = 8 19
Deff
In industrial operation, it is impossible to reduce the catalyst pellets to small sizes without having
higher pressure drops; however, high yields of products are required (Sie 1998). Furthermore,
according to Jess and Kern (2012a) for typical FTS conditions, the particle size suitable to avoid
an excessive pressure drop in the catalyst bed must be higher than 1 mm. It is therefore imperative
that research on the development of novel FT catalysts strives to design materials that consider the
advantages and disadvantages that may arise in their direct application (Macías and Ancheyta
2004). However, it is noteworthy that pore diffusion greatly affects the effective rate constant
within a typical temperature range of 200–250 C for particle diameters above 0.5 mm (Jess and
Kern 2012a).
Post et al. (1989) reported significant external diffusion limitations caused by the resistance of the
film of gas or liquid around the catalyst particle, which was confirmed by observing no changes in
synthesis gas conversion when changing the bed length at the same space-time values.
Top view of the catalytic bed
Central region
Near-wall region
Figure 8.5 Top view of the split in the center and near the wall regions catalyst bed.
To develop the mass balance within the catalyst particle, pores having uniform properties and
being completely filled with liquid are to be assumed. A close-up of the catalyst particle is shown
in Figure 8.5, in which the concentration profile occurring within the catalyst particle with mass
transfer resistance effects can be seen. The equations that model the behavior of the FTS process
inside the solid catalyst particle arise from the mass balance on an infinitesimal volume element of
the catalyst particle considering different geometric forms (Table 8.8).
Energy Balance It is well-known that the temperature control of a fixed-bed reactor deserves
special attention for modeling and simulation. There are several studies of fixed-bed reactor mod-
eling in which an isothermal regime is assumed. However, this assumption could not be reliable
since the model could generate erroneous predictions of the actual thermal behavior in a reactor,
especially at industrial scale. It may be noted that the FTS reaction is highly exothermic, with
a ΔHFT = −165 KJ/molCO for the paraffin formation and a ΔHFT = −204 KJ/molCO for the olefin
formation (Kölbel and Ralek 1980). Then, due to the high exothermicity of the FTS reaction, it is
necessary to have an effective temperature control of the fixed-bed reactor, since according to Jager
and Espinoza (1995), the following circumstances may occur:
•• Generation of hot spots (temperature peaks) along the catalyst bed is common.
Axial and radial temperature profiles arise inside the tubes.
• To maximize the synthesis gas conversion, the maximum average temperature must not exceed
the maximum permissible temperature peak (hot spot) to prevent carbon deposition on the cat-
alyst surface.
• Carbon deposition deactivates the catalyst, which results in activity loss and catalyst replacement.
It is also important to evaluate the temperature profiles as well as the heat transfer rates to ensure
control and performance of the reactor (Guardo et al. 2006). As aforementioned, the FTS reaction is
highly exothermic and good control of the heat released during the process is required. In addition,
the formation of undesirable methane as well as the deactivation of the catalyst usually occurs
(Schulz 1999). It is also possible that the handling of high temperatures leads to a considerable loss
of selectivity and thermal runaways. Therefore, an efficient alternative is the primary mechanism
for removing the generated heat by radial heat transfer, for which it is necessary to match the radial
heat transfer rate to the reaction rate of the catalyst to ensure that the catalyst bed temperature can
be controlled, according to Zhu et al. (2010).
It has been determined by Pölhmann and Jess (2016b) that, in exothermic reactions such as that
of the FTS process, the generation of temperature gradients can occur within the particle, which
may lead to an overheating of the particle, which would not be beneficial to reactor performance.
The heat balance equations for the gas and liquid phases considering the isothermal and non-
isothermal cases are reported in Table 8.9. In the heat generated by axial and radial dispersion,
the fluxes of fluid-interface convection heat are observed where the driving force is produced by
a temperature gradient generated among the temperature in the bulk fluid phase (gas and liquid)
and temperature at the interface. The transport of energy by conduction of the gas-film side at the
gas–liquid interface occurred through the transfer of interfacial mass. The conductive heat flux
term for both phases considers the heat generated by evaporation/condensation that occurs
between the gas and liquid phases. Convective heat transfers of both the gas and liquid phases
on the external catalyst surface are also involved.
In the case of heat balance in the liquid phase, heat evolved by vaporization/condensation is con-
sidered. Here ΔHvi represents the latent heat corresponding to the heat that is consumed by vapor-
ization effects only for the products of the FTS reaction.
Wang et al. (2001) investigated the transfer and reaction phenomena in a catalyst pellet for FTS,
the interaction between diffusion and reaction in the pellet, and its effects on product selectivity.
The results of simulations showed that the temperature differences between the external surface
and the center of the pellet were less than 0.02 K due to the excellent thermal conductivity of
the Fe–Cu–K catalyst. Accordingly, it is possible to assume, in certain cases of modeling of
fixed-bed FTS reactor, an isothermal behavior of the catalyst particle.
In Table 8.10, the heat balance equation for the catalyst is established considering the isothermal
case in which concentrations and temperatures on the solid surface must be used. It should be
noted that the sign of the heat of reaction (ΔHFTR, j) is negative because the FTS reactions are highly
exothermic.
Furthermore, the resistances that exhibit the films (gas and liquid) greatly influence the heat
transfer, whereas heat transport within the catalyst particle is normally fast. Thus, the heat transfer
phenomena occurring within the particle pores can be described with the Fourier law in terms of
the partial differential equation (PDE) with respect to the temperature for the non-isothermal case,
as illustrated in Table 8.10.
8.2.3.2 Boundary Conditions of the Proposed Generalized Model

The generalized model for the fixed-bed FTS reactor consists of the mass balances of each species
and the heat balance, which imply a system of PDEs, that require a set of initial and boundary con-
ditions to be z = 0 and z = LB at the reactor inlet and outlet, respectively. As indicated by Wärna and
Salmi (1996), in a fixed-bed reactor model by an axial dispersion model, appropriate boundary con-
ditions to describe steady-state fluxes are needed (Wang et al. 2001; Danckwerts 1953). It is impor-
tant to indicate that boundary conditions at the reactor outlet for the case of a dispersion model are
needed because back-mixing takes place. The set of initial and boundary conditions that define the
proposed generalized model for the FTS reactor is given in Tables 8.11 and 8.12.
At z = 0, the concentration of the gas phase is in physical equilibrium with the inlet temperature
and pressure, and often the gas concentration at the reactor inlet is expressed as a function of con-
version considering the following ideal gas ratio (Dai et al. 2014; Moazami et al. 2015a; Park et al.
2014; Moazami et al. 2015c)
C Gi = pGi Ru T 8 20
Table 8.11 Initial conditions (t = 0) of generalized mass and heat balance equations of the model.
Gas phase: i = H2,

CO, CO2, CH4 and Liquid phase: i =
light light and heavy
hydrocarbons: hydrocarbons:
paraffins and paraffins and Solid phase: Solid phase:
olefins (C2 − C4 olefins (C2 − C4 surface (all inside (all Stagnant
Condition and C5+) and C5+) components) components) zone
z=0 CGi = C Gi,0 C Li = CLi,0 CSf ,isur = 0 CSf ,iins = 0 CSf ,isur = 0
0≤r≤R T G = T G0 = T 0 T L = T L0 = T 0 = 0 T Ssur = T S0 sur = T 0 T Sins = T S0 ins = T 0
0 ≤ z ≤ LB CGi =0 C Li =0 CSf ,isur =0 CSf ,iins = 0 CSSt,i = 0
0≤r≤R T G = T G0 = T 0 T L = T L0 = T 0 = 0 T Ssur = T S0 sur T Sins = T S0 ins = T 0
0≤ξ≤R CSf ,iins = 0
0 ≤ z ≤ LB
T Sins = T S0 ins
0≤r≤R
pGi = PT yGi 8 21
Effects of axial and radial dispersion of mass and heat in the dispersion model give second-order
differential equations so that the two boundary conditions necessary according to Danckwerts
(1953) are as follows:
The Danclwerts’ boundary condition at the reactor inlet at z = 0 is:
f
∂Ci f f
− εf Dfax = ufs Ci − Ci 8 22
∂z 0 z = 0+
z = 0+
∂T f f
− εf λfax = ufs ρf C P T f − Tf 8 23
∂z z=0 +
0 z = 0+
which is simplified as
C if = T f Cif ; Tf = Tf 0
8 24
•
0
Such boundary conditions are acceptable with reliability since the axial dispersion of mass and
heat are relatively small and concentration and temperature gradients at the reactor inlet are
usually fairly flat (Chen et al. 2001).
The boundary condition at the exit of the reactor at z = LB is:

f
∂C i ∂T f
= =0 8 25
∂z ∂z
In addition, Pearson (1959) showed a way of manipulating the boundary conditions of
Danckwerts to impose the continuity of the reactant concentration in continuous-flow reactors that
suggests different boundary conditions can be applied to the inlet and outlet in packed-bed systems,
as well as in more complex situations involving problems of diffusion of multicomponent systems
with variation in temperature.
Table 8.12 Boundary conditions of generalized mass and heat balance equations of model.
Condition Gas phase Liquid phase Solid phase (surface) Solid phase (interior)
z=0 CGi = CGi,0 CLi = 0

0≤r≤R T G = T G0 = T 0 TL = 0
z = LB ∂C Gi ∂C Li
=0 =0
∂z ∂z
0≤r≤R ∂T G ∂T L
=0 =0
∂z ∂z
r=0 ∂C Gi ∂C Li
=0 =0
∂r ∂r
0 ≤ z ≤ LB ∂T G
∂T L
=0 =0
∂r ∂r
r=R ∂C Gi ∂C Gi ∂T Ssur
=0 =0 − λSrad = hGS T Ssur − T W
∂r ∂r ∂r c
0 ≤ z ≤ LB ∂T G ∂T L
− λGrad = hGW TG − TW + λLrad = hLW TS − TL
∂r c
∂r c
ξ=0 — — ∂C Sf ,iins ∂T Sins

= =0
0 ≤ z ≤ LB ∂ξ ∂ξ
dpe — — ∂C SL,iins
ξ= − DLeff = f w k LS LS Ssur
− CLi
2 i aI C i
∂ξ
0 ≤ z ≤ LB
+ k iLSt −S C Si sur − CSt
L,i
∂C SG,iins
− DGeff = 1 − f w k GS GS
I aI CSi sur − CGi
∂ξ
N NS
= ρB ε B vij ηj RFTR C Si sur , T Ssur
j=1
∂T S
− λSeff = 1 − f w hGS GS
c aI T Ssur − T G
∂ξ
+ f w h c aI T − T L
LS LS Ssur
8.2.3.3 Pressure Drop

In designing and modeling of a fixed-bed FTS reactor, it is mandatory to predict the pressure drop
through the catalyst bed. In the case of the LTFT process to produce clean fuels, the main products
are heavy oil and waxy materials whose molecular weights are relatively high, and in consequence,
the effect of flow velocities occurring through the packed bed could cause severe attrition of
catalysts.
Among various mathematical models presented in previous sections, there are not many of them
that consider the effect of the pressure drop through the bed and many of them make use of the
classical Ergun correlation for its prediction. However, the classical model of Ergun to estimate
the pressure drop in packed-bed reactors is inefficient when various geometries of catalyst are used.
In different fields of engineering, the semiempirical Ergun equation has been employed for predict-
ing pressure drop across packed beds with particles made of various materials (Allen et al. 2013). In
the case of FTS, there are several models for fixed-bed FTS reactors that use such a correlation to
describe the effect of pressure drop through the packed bed. However, various research works have
been conducted in which a comparison between modified correlations of Ergun and contempora-
neous correlation has been made, and as a result, a deficiency is obtained in predicting pressure
drop by the classical model of Ergun compared to the modified correlations (Allen et al. 2013; Ozahi
et al. 2008; Vollmari et al. 2015). It has been established in the literature that deviations from the
predicted pressure drop across packed beds that exhibit the classical correlation of Ergun may be
due to the high sensitivity of the equation to variations in the bed porosity by which it is necessary to
estimate accurately the packed-bed porosity, especially when dealing with non-spherical geometry
(Koekemoer and Luckos 2015). It has been found that several factors influence the bed porosity,
such as the particle diameter, the particle size distribution, surface roughness, Reynolds number,
and others (Vollmari et al. 2015; Bai et al. 2011). In summary, the literature reports give both the-
oretical and experimental evidence that pressure drop depends on flow velocity and physical prop-
erties (viscosity and density), the average bed porosity, shape and surface of the particles packaging,
bed height, and contribution of the ratio of particle to the container diameters (Eisfeld and Schnit-
zlein 2001; Rangel et al. 2001; Montillet et al. 2007).
Flow regimes in catalytic fixed-bed reactors are quite important as they are directly related to the
pressure drop. Following four flow regimes in fixed-bed reactors are reported in the literature
(Dixon et al. 2006):
• At the flow rate of Darcy ore creeping-flow at Rei < 1, the pressure drop varies linearly with flow
rate, and interstitial number range is controlled by viscous forces.
• In a laminar flow regime in a stationary Reynolds interstitial number range of 10 < Rei < 150,
pressure drop shows a nonlinear relationship with interstitial flow rate.
• In a laminar flow regime in the nonstationary Reynolds interstitial number range of 150 < Rei <
300, oscillations of laminar wake are present in the pores and vortices form and the number of
Rei = 250.
• In a flow regime considered highly nonstationary at Reynolds interstitial number above 300,
eddies arise like turbulent flow in pipes.
The effects of pressure drop involving both the gas phase and liquid phase must strictly be con-
sidered separately to get better accuracy and approach to the real pressure drop across the catalyst
bed. Bai et al. (2011) proposed an empirical correlation to predict the pressure drop of a two-phase
flow through pebble beds in a gas–water system by considering the use of the relative permeability
in the gas phase while the bed is subdivided into two regions: (i) one near walls and (ii) another in
the central zone of the bed to consider the effects of wall and improvements to low ratios of tube
diameter to the particle. The correlation used to calculate the pressure drop in two-phase flow was
established in dimensionless terms to cover the whole range of Reynolds and Galileo numbers,
assuming that the relative permeability is dependent on the void fraction bed and the subdivision
of the bed into two regions to consider the effects of wall. Bai et al. (2011), based on previous studies
reported by Eisfeld and Schnitzlein (2001) and Reichelt (1972), proposed the following equations:
2
ΔPf dpe Re f Re f
Ψf = 2 = κ1 β1w + κ 2 β 2w 8 26
ρf ufs
L Gaf Gaf
where
f 2
us ρf dpe ρf dpe εB
Re f = ; Gaf = gc 8 27
μ 1 − εB
f μf 1 − εB
6V p
dpe = 8 28
Ap
κ1 and κ2 are Ergun-type constants, and β1w and β2w are the coefficients that consider the effect of
wall under the following restrictions:
i) When small packed beds are used (Dhyd/dpe < 10), the central region of the bed is negligible and
the following wall effect coefficients must be considered:
d∗pe
β1w = a + b 8 29
1 − εw
−2
β2w = cd∗pe 8 30
dpe
d∗pe = 8 31
dh
ii) When large packed beds are used (Dhyd/dpe <10), the near-wall region is neglected and the coef-
ficients of wall effect will be equal to 1. In Eqs. (8.26)–(8.31), κ1 = 180, κ 2 = 1.8, a = 0.8, b = 2,
c = 3, d = 1, d∗pe is the dimensionless particle diameter and dh is the hydraulic diameter of the
packed beds.
Figure 8.5 shows a top view of the catalyst bed to specify in detail the effects of pressure drop in
the central and near-to-the-wall zones when the ratio of the hydraulic diameter to the equivalent
particle diameter is lower than 10.
The proposed generalized model for predicting the pressure drop considers two general Ergun-
type approaches: (i) a first Ergun model in which a bed porosity is assumed to be uniform (i.e. a
constant bed void fraction); (ii) a second Ergun model in which an oscillatory bed porosity behavior
(i.e. bed porosity as a function of radial position) is assumed. Therefore, considering the second
Ergun model approach, the variable porosity bed relative to the radial direction (oscillatory behav-
ior of the bed porosity) results in the loss of viscous energy, which is proportional to the linear veloc-
ity of the fluid. The second approach for pressure drop is an adapted version from Bey and
Eigenberger (1997) and considers the effects of shear stress that are negligible for most cases.
For the proposed model, the pressure drop is calculated by considering the equation of Ergun and
the general balance of momentum (Tables 8.13 and 8.14, respectively) recently reported in the
literature (Rangel et al. 2001), where the coefficients that determine the viscous effects and
kinetics (inertial) correspond to general constants κ 1 and κ 2, respectively. The model incorporates
Table 8.13 General Ergun equation for pressure drop.
General Ergun equation for FTS reactor
Viscous energy loss term Viscous energy loss term
f
∂PT ∂usf ∂ usf
2
= κ 1 β1w κ 2 β2w
∂z ∂z ∂z
Model 1 (not oscillatory porosity behavior):

2
1 − εB ρ f usf
f 2
dPT 1 − εB μ f usf
= κ1 β1w + κ 2 β2w ; (Kürten et al. 1966; Hicks 1970)
dz ε3B ψdp
2
ε3B ψdp
Model 2 (with oscillatory porosity behavior):

f
f
dPT f 1 − εB r
2
f 1 − εB r 2 μeffd du f r
= κ1 β1w μeff 2 us r + κ 2 β2w ρ
f
usf r + εB r r s ;
dz εB r dp
3 εB r dp εB r dr dr
(Kürten et al. 1966; Hicks 1970; Çarpinlioğlu and Özahi 2008)
Boundary conditions
f f
P = P 1 z = 0 , P f = P 2 z = LB
dusf r = 0
= 0, usf r = R = 0
dr
Table 8.14 Generalized momentum equations for FTS reactor.
Generalized momentum equation for FTS reactor
Momentum in ∂ ρ f usf z ∂PT 1 ∂ ∂ f ∂us z

f
axial direction = − εB − r εB ρ f usf r usf z − εB μeff
∂t ∂z r ∂r ∂r ∂r
∂ ∂ f ∂us z
f
− εB ρ f usf z usf z − εB μeff
∂z ∂z ∂z
1∂ f ∂u f r ∂ f ∂us z
f
+ εB μeff r s + εB μeff −fz
r ∂r ∂r ∂z ∂z
Momentum in ∂ ρ f usf r ∂PT 1 ∂ 1∂ f ∂u r
f
radial direction = − εB − r εB ρ f usf r usf r − εB μeff s
∂t ∂z r ∂r r ∂r ∂r
∂ ∂ f ∂us r
f
1∂ f ∂u f r
− εB ρ f usf r usf z − εB μeff + εB μeff r s
∂z ∂z ∂z r ∂r ∂r
∂ f ∂us z
f
f usf r
+ εB μeff − εB μeff −fr
∂r ∂z r2
the parameters β1w and β2w, which consider the effects of the reactor wall on pressure drop across
the bed (Bai et al. 2011).
Table 8.13 displays the widespread Ergun equation corresponding to the prediction of pressure
drop caused by both fluids (gas and liquid) involved in the FTS fixed-bed system, while Table 8.14
shows the mathematical terms of the generalized model of momentum.
f f
The effective viscosity (μeff = ηdyn ) has been used in the generalized model for pressure drop
because this parameter considers the effects of turbulent-flow fluctuations within the package.
Therefore, some correlations reported in the literature to estimate the effective viscosity for
spherical and cylindrical catalyst particles, respectively, need to be used (Bey 1998; Bey and
Eigenberger 1997):
f 2
μeff −6 dt
f
ρ f u s dp
=1+ 7 × 10 + 2 × 10 − 5 8 32
μf dp μf
f 2
μeff −5 dt
f
ρ f u s dp
=1+ 1 7 × 10 + 2 × 10 − 5 8 33
μf dp μf
8.2.4 Model Parameters

To have a robust mathematical model, it is necessary to have enough information about all para-
meters involved in the mass and energy balances. Several empirical correlations for estimating the
following parameters have been reported: (i) mass transport between phases (G–L, G–S, and L–S);
(ii) scattering coefficients for axial and radial mass transfer; (iii) heat transfer such as heat conduc-
tivities in the axial and radial directions; (iv) heat transfer in the solid phase; (v) holdups of the gas
and liquid phases; (vi) coefficients of convective heat transfer; (vii) wetting factor of the catalytic
surface; and (viii) void fraction of the bed, etc. For simulations using the generalized model or sim-
plified forms, it is necessary to evaluate distinct physical and thermodynamic parameters involved
to solve the set of PDEs. Consequently, the set of parameters can be estimated by means of existing
correlations.
8.2.4.1 Mass Transfer Parameters

Phenomena representing mass transport and dispersion coefficients are two aspects to be consid-
ered in the development of mass balances of the FTS reactor model. The mass transfer phenomenon
in fixed-bed reactors occurs according to the diffusion of reactants through the liquid that is pro-
duced to reach the active sites before the reaction takes place in the solid catalyst. Once the catalyst
surface is reached, the reactants diffuse through the inner particle pores from mouth to active sites,
and then the chemical reaction occurs under the kinetic conditions.
The mass transfer flux at the gas–liquid interface is described according to the two-film theory
valid for fixed-bed configurations:
1 Ru T G Z P T , T G 1
= G
+ L 8 34
K L,i k I,i Hei k I,i
where the overall external resistance to mass transfer is formed by the mass transfer resistances
offered by the gas (k GI,i ) and liquid (kLI,i ) films.
Delgado (2005) developed a rigorous investigation of variables that directly affect the Newtonian
and non-Newtonian fluids axial and radial dispersion coefficients involving fixed-bed packed con-
figurations, as well as the correlations for prediction. The variables that have significant effects on
dispersion coefficients were found to be (i) length of the bed, (ii) ratio of bed diameter to particle,
(iii) ratio of the bed length to particle diameter, (iv) particle size distribution, (v) particle shape, and
(vi) fluid properties (viscosity, density, velocity, and temperature). The results allowed one to estab-
lish two expressions for predicting axial and radial Peclet numbers exhibiting higher accuracy com-
pared with other correlations, which are valid for all values of Peclet and Schmidt numbers (Sc).
Carbonell and Whitaker (1983) proposed a criterion to evaluate the effects of axial dispersion in
fixed-bed systems that consists of neglecting the dispersion in the axial direction if the following
expression is satisfied:
t
1 − εB Dij
1 8 35
εB d2p
On the other hand, Han et al. (1985) showed that the values of the coefficient of axial dispersion
when packed beds of uniform size are used and measured at different positions in the bed are a
function of the position unless it approximates the following criterion:
2
LB 1 1 − εB Dij t
≥ 0 3; or ≥01 8 36
dp Pe εB d2p
8.2.4.2 Heat Transfer Parameters

The heat transfer mechanism is an important aspect used in the design, modeling, simulation, and
optimization of catalytic fixed-bed reactors implemented in various chemical processes. According
to Dixon (2012), the heat transfer phenomenon that arises in a fixed bed of catalyst particles may be
due to different factors such as (i) heat flow by conduction through the fluid phase, (ii) conduction
along the particle, (iii) conduction caused by the contact points in existing particle–particle inter-
actions and because of the amounts of gas stagnant around the contact points between the particles,
(iv) transfer mechanism radiative heat from particle to particle and intraparticle void fraction to
intraparticle void fraction, (v) convective transport of energy due to fluid travels around the catalyst
particles along the bed, (vi) heat transfer of solid-particle fluid through the surrounding film,
(vii) heat transfer of wall-particle contact, and (viii) heat transport of fluid-wall contact (Barker
1965). Moreover, an important aspect in the modeling of the temperature profile in fixed-bed
systems refers to the heat transfer originated near the reactor wall region because in this region
a greater amount of catalyst particles is concentrated when a low ratio of the reactor diameter
to the catalytic particle diameter is present (Dixon 2011). Therefore, in the region near the wall,
a significant increase in resistance to heat transport is found due to some factors (Yagi and Wakao
1959; Tobis and Ziókowski 1988): (i) increasing the void fraction implies a decrease of the thermal
conductivity of the bed, (ii) a viscous layer (laminar) that serves as a boundary, (iii) damping of the
convective heat transfer, which is mainly due to a decrease of the lateral displacement of the fluid.
Furthermore, the oscillatory behavior of the bed porosity along the radial position makes velocity
perform an oscillatory behavior directly affecting the heat transfer in packed beds (Mueller 1992;
Béttega et al. 2011).
Dixon (2011) reported various factors affecting the effective radial heat transfer solid coefficients.
The coefficients of heat transfer by convection of the fluid phase to the wall, as well as several
empirical correlations to estimate these two heat transfer coefficients. It is concluded that there
are disagreements in estimation of λeff,rad and hw values; however, the modeling approaches men-
tioned can be used as complementary tools in studies of the phenomena of fixed-bed packed.
Larachi et al. (2003) collected extensive information on parameters of heat and mass transfer in
packed-bed systems, such as parameters for radial effective thermal conductivity and heat transfer
on the wall and particle fluid influenced by packing particle size and particle materials. Based on
these data, these authors proposed new correlations for calculating these heat transport parameters
using artificial neural network, which is a powerful tool to develop more effective correlations in
transport equations.
It is known that an important parameter to characterize the thermal resistance between the fluids
inside the reactor (syngas) and the cooling shell is the overall heat transfer coefficient. According to
Reid et al. (1977), the overall heat transfer coefficient between the cooling medium (commonly boil-
ing water is used in fixed-bed FTS reactors) of the shell side and the bulk of the gas phase inside the
reactor is determined by means of the following correlation,
dout
Ainn In
1 1 dinn Ainn 1
= GS + + 8 37
Ur hc 2πLr k w Aout hLS
c
where hw,inn is the convection heat transfer coefficient between the gas phase and the reactor wall,
and it is calculated as follows (Smith 1981):
2 3 − 4 407
hLW
c C Pmix,G μmix,G 0 458 ρGmix uGs dp
= 8 38
CPmix,G ρGmix μmix,G k GT εB μmix,G
Moreover, the following Leva correlation is used to calculate the heat transfer coefficient of the
cooling medium (boiling water) on the shell side at high pressure:
4 3
hLW
c = 282 2 PT ΔT; 0 7 < PT < 14 Mpa 8 39
The Chilton–Colburn analogy (Chilton and Colburn 1934) has proved to be appropriate whose
j-factor for the transfer of mass and heat are, respectively, given by
Sh
jD = =f 8 40
Re f Sc1 3
Nu
jH = =g 8 41
Re f Pr1 3
The literature is scarce in showing correlations to estimate the heat transfer coefficients in gas or
liquid film interacting with the interfaces (gas–liquid and liquid–solid interfaces). It is often con-
venient to make use of the Chilton–Colburn analogy (jD = jH) for calculating these coefficients.
The determination of the heat of reaction is another important issue for modeling the tempera-
ture profiles, but not all products could be known in the FTS process. Therefore, it is preferred to
use lumping techniques to encompass the FTS products of interest. Furthermore, there are several
modeling works in which the formation of CO2, CH4, C2H4, C2H6, C3H8, i − C4H10, and n − C4H10 is
considered as a net scheme of the FTS reactions involved in a given process (see reaction scheme
elsewhere [Bayat et al. 2014; Lozano-Blanco et al. 2009; Park et al. 2015]).
The following expression given by Sandler (2006) can be used to calculate the enthalpy change
of any state (T1, P1) to (T2, P2):
T2
H T, P − H IG T, P
T 2 , P2 − H T 1 , P1 = C P dT + T c
T1 Tc T r2 ,Pr2
8 42
H T, P − H IG
T, P
−
Tc T r1 ,Pr1
In addition, the database for thermodynamic properties given by Burcat and Ruscic (2005) is used
in order to obtain the ideal gas enthalpy of the FTS species as a function of temperature:
T T2 T3 T4 A6
H IG = Ru T A1 + A2 + A3 + A4 + A5 + 8 43
2 3 4 5 T
8.2.4.3 Phase Equilibrium

In most literature reports related to mathematical models of the FTS process in fixed-bed reactors
including other reactor technologies, authors have assumed that both the reactants and products
remain only in a single phase, either vapor or liquid. Although many of these modeling schemes
have served to analyze FTS technology in terms of selectivity, reactivity, and reactor performance,
the lack of robustness is common by ignoring the VLE. However, a more complete modeling of the
FTS reactor needs to consider the effect of the VLE, because it is necessary to know the gas solubility
(reactant and some light hydrocarbons) through the liquid wax produced. In addition, the contri-
bution of VLE in reactor modeling is also because the waxy liquid is partially volatilized at oper-
ation conditions of the FTS process (Guardo et al. 2006), whose phase distribution as well as the
thermal behavior of the reactor is directly affected. However, according to Pöhlmann et al.
(2016) concerning the catalyst pore filling, vaporization of volatile hydrocarbons in the bulk phase
of the gas can occur. In summary, The VLE data play an important role in the design, operation, and
development of the process (Caldwell and Van Vuuren 1986).
There is evidence in the literature that the cubic state equations widely used to describe the VLE
of the FTS system are Soave–Redlich–Kwong (SRK) (Kim et al. 2009; Huang et al. 1988; Wang et al.
1999) and Peng–Robinson (PR) (Marano and Holder 1997; Visconti and Mascellaro 2013; Irani
2014). Wang et al. (1999) developed a modified version of the SRK equation of state to represent
the gas–liquid equilibrium (solubility in heavy waxes) in the FTS, considering a wide range of
solutes such as CO, H2, CO2, CH4, C2H4, C2H6, and heavy wax solvents from C20 to C61 using a
wide range of temperatures and pressures. Such an SRK equation has been preferably used
(Huang et al. 1988; Wang et al. 1999; Mikhailova et al. 2003; Karimi et al. 2012) because of its high
predictive and descriptive ability of the gas–liquid equilibrium behavior in the FTS. Karimi et al.
(2012) showed a comparison between the results obtained by the SRK state equation, the PR equa-
tion as well as its modified versions, and Henry’s empirical equation for the determination of phase
equilibrium in the process FTS, where it is strongly shown that the modified SRK equation agrees
with the experimental data collected from the literature. Other methods employed including
the modified PR equation show high deviations. It should be noted that the modification in the
SRK and PR state equations was based on the replacement of the use of the acentric factor by
the molecular weight of the FTS products as a characteristic parameter.
The modified SRK state equation for the prediction of VLE equilibrium behavior is:
RT αSRK a
P= − 8 44
P−b V V + b
where P is the pressure, T the absolute temperature, and V the molar volume. The constants a and b
take the following values:
R2u T 2c
a = 0 42747 8 45
Pc
Ru T c
b = 0 08664 8 46
Pc
This equation can be rewritten as a function of the compressibility factor as follows:
Z 3 − Z 2 + Z A − B − B2 − AB = 0 8 47
aPT
A= 8 48
R2u T 2
bPT
B= 8 49
Ru T
The definition of the α parameter by Soave is given as follows:
2
αSRK = 1 − m 1 − T r 8 50
The parameter m is obtained on the basis that the equation reproduces the vapor pressure of non-
polar compounds and is correlated as a function of the acentric factor (ω), whose expression is
adopted as follows:
m = f ω = 0 48508 + 1 55171ω − 0 15613ω2 8 51
The acentric factors of heavy n-paraffins that can be used in the SRK equation of state can be
obtained by the following correlation proposed by Lee-Kesler (1975):
− In Pr − 5 97214 + 6 09648θ − 1 + 1 28862In θ − 0 169347θ6 Tb

ω= ; θ= 8 52
15 2518 − 15 687θ − 1 − 13 472In θ + 0 43577θ6 Tc
where Tb is the temperature at the normal boiling point and Tc is the critical temperature. However,
the parameter α proposed by Soave for the prediction of hydrocarbon vapor pressures implies cer-
tain errors, since this form of α increases rapidly at relatively low reduced temperatures. In addition,
for a given finite temperature beyond the critical temperature, the parameter α becomes zero and
then is raised again with increasing temperature, whose behavior results in obtaining binary inter-
action parameters quite large and dependent on the temperature in hydrogen–hydrocarbon sys-
tems. Such drawbacks clearly indicate that the original version of the SRK equation of state is
inadequate for the direct correlation of solubility data of the FTS process (Wang et al. 1999). Wang
et al. (1999) improved the term α by replacing the acentric factor with the molecular weight as a
characteristic parameter, since the acentric factor of the waxy product is usually inaccurate,
whereas the molecular weight of the wax is a property easily calculated. Then the term m was pro-
posed as follows:
n
m = f Mm = A0 M im− 1
I=i
= 0 45445 + 0 62675 × 10 − 2 M m − 0 93697 × 10 − 5 M 2m + 0 37387 × 10 − 8 M 3m

8 53
The equilibrium between the gases (reactant and products) and the Fischer–Tropsch liquid wax is
established as a function of the fugacities of each component in its respective gas or liquid phase as
follows:
f Vi = f Li 8 54
f fi = Pyfi ϕfi 8 55
where ϕfi is the fugacity coefficient for component i in f phase, which can be estimated with the
SRK equation of state as follows:
f bi A bi 2 B
In ϕi = Z − 1 − In Z − B + − yi αSRK a ij In 1 + 8 56
bm B bm αSRK,i ai j
Z
where am and bm are the parameters for mixtures, which are estimated by the following mixing rule:
am = yi yj ai aj 1 − k ij 8 57
i j
bm = y i bi 8 58
i
In addition, the solubility of the gases in the Fischer–Tropsch waxy product can be determined by
means of:
xi
Si = 8 59
1− xi M wax
m
i
where M wax
m is the molecular weight of the wax and xi is the molar fraction of each component in
the wax.
It should be noted that the critical properties (temperature and critical pressure) with respect to
the components of the FTS must be determined for the VLE calculations; however, the information
for heavy hydrocarbons is scarce since there are only data reported in the literature for hydrocar-
bons up to C20. Thus, the following Gasem and Robinson (1985) correlation including data of nor-
mal boiling points and number of carbon atoms (NC) is used:
1
C1 C1 1 − C2
Y= − − C 13 − C2 exp − C 4 N C − 1 1 − C 2 8 60
C4 C4
where Y represents either Pc or Tb/Tc, and the constants C1 − C4 are given in the work by Gasem and
Robinson (1985).
Mikhailova (2003) modified the SRK equation, and in this case, the molecular weight distribution
of the FTS products was estimated by the ideal ASF polymerization model:
n−1
Mm = 1 − α 8 61
However, it should be noted that the molecular weight distribution can be estimated by noniden-
tical ASF models (Förtsch et al. 2015; Kuipers et al. 1996; Donnelly et al. 1988).
8.2.4.4 Catalyst Particles Parameters

In the case of the catalyst particles, it is necessary to determine the different properties involved as
well. A review of the different correlations published in the literature was done for the parameter
estimation of the catalyst particle used in the FTS process.
Due to the formation of a film or thin layer of waxy liquid product on the catalytic particle surface,
the external mass transport effects may be negligible, but the influence of pore diffusion is impor-
tant (Kaiser and Jess 2014). Thus, the Thiele modulus is an important dimensionless parameter that
allows determining the presence of diffusion limitations within the particle, which for the case of a
pseudohomogeneous first-order reaction is defined as follows (Jess and Kern 2012a; Thiele 1939):
05 05
kA CGCO 1
ΦS = Lp = Lp k int = Lp k int 8 62
Deff ,L Deff ,L CLCO Deff ,L Ru HeTCO
where Lp is the characteristic length of the reaction system also known as the form factor, kA is the
kinetic rate constant for the homogeneous first-order reaction, and Deff,L is the effective diffusivity
for CO at the liquid phase.
One approach given in the literature to determine the mass transfer limitations of the catalyst
consists of solving the mass balance of a differential shell taken into the catalyst particle by con-
sidering a spherical geometry of the pellet or catalyst particle, in which a reaction of first order
occurs as follows (Iglesia et al. 1993):
2r
sinh 3ΦS
C fi dpe dpe
Ssur
= 8 63
Cf ,i 2r sinh 3ΦS
whose modified form for a reaction of order n was established as (Bischoff 1965; Petersen 1965):
n−1
k int C Sf ,isur
ΦS = Lp n+1 f
8 64
2Deff
The shape factor Lp was used to modify the solution of different geometries, which in the case of a
cylinder pellet turns into dp/4 and dp/6 for a spherical pellet. kint is denoted by the intrinsic reaction
f
rate coefficient per volume unit for the nth-order reaction considered, and Deff is the effective
diffusivity given as follows:
εp
Dfeff = Dcomb 8 65
τ
V pore ρB
εp = 8 66
1 − εB
On the other hand, for the determination of the effective diffusivity, the following correlations are
used in modeling of fixed-bed configuration in which the FTS occurs (Cybulski and Moulijn 1994;
Bosanquet 1944):
f 1
Deff ,ax,i = DT 1 + Pe2 8 67
192
where
dt uGs Ru
Pe = and Dij =
Dij 6πμGmix dp
f εS 1
Deff = F λG 8 68
τ 1
f + 1
f
DM DK,i
i
Whose restrictive factor F(λG) is given by Iliuta et al. (2006), in which rsolute is the hydrodynamic
molecular radius of the solute, rpm is the mean radius of the pore, and Z is a dimensionless constant
used in the Bosanquet’s formula:
Z r solute
F λG = 1 − λ G ; λG = 8 69
r pm
The effectiveness factor of the catalyst can be estimated by the correlations reported in Table 8.15.
Mederos et al. (2009) presented some other formulas for the determination of the effectiveness
factor and the Thiele modulus for different shapes of catalysts.
It is possible to obtain close approximations if the Thiele modulus is used by making the neces-
sary adjustment for an n-order reaction (Eq. 8.64). However, it is important to mention that there
are no analytical methods to determine the effectiveness factor for cases involving nonlinear reac-
tion rates and arbitrary catalyst particle shapes. Thus, its approximation based on the consideration
of nonlinear reaction rates and catalysts of an arbitrary shape can be calculated by the equations
shown in Table 8.15 using the generalized Thiele modulus (Iordanidis et al. 2002):
−1 2
RFT CSf ,isur , T Ssur C Sf ,isur
ΦS = L p RFT CSf ,isur , T Ssur dC Sf ,isur 8 70
2dpe 0
In the case where η tends to 1, the shape adopted by the catalytic particle becomes an important
factor that influences the estimated value of the effectiveness factor. Wijngaarden et al. (1998) pro-
posed a novel approach to calculate the effectiveness factor in the case where chemical reactions
whose rate is nonlinear and catalytic particles with an arbitrary geometry are contemplated that
involved two dimensionless groups:
Zero Aris number:
RFT C Sf ,isur , T Ssur CSf ,isur

An0 = L2p RFT C Sf ,isur , T Ssur dC Sf ,isur 8 71
2dp 0
First Aris number:
Γ ∂RFT Cf ,i , T
Ssur Ssur
An1 = Lp 8 72
dpe ∂CSf ,isur
CSf ,isur
Table 8.15 Effectiveness factors for conventional geometry of catalyst particles and their concentration
profiles according to analytical solution.
Geometry of
the system Lp η Observations
Rectangular LB tan h ΦS —
ΦS
Cylindrical dp 2I 1 ΦS For pellets of cylindrical catalyst with a particle diameter
2 ΦS I 0 ΦS of less than 200 μm (dp < 200 μm). I0 and I1 are the typical
Bessel functions of first and second order, respectively.
Spherical dp 3 1 1 Spherical particles with a diameter greater than 200 μm
− (dp > 200 μm).
6 ΦS tan h ΦS ΦS
where Γ is the geometric factor that depends only on the particle shape. Γ is 2/3, 1, and 6/5 for a
slab, an infinite cylinder, and a sphere, respectively. The zero Aris number was introduced to cal-
culate the effectiveness factor in a low region of η, whereas the first Aris number determines η when
its value approaches 1. The following expression is a generalized form of the effectiveness factor
that implies the two numbers of Aris already defined:
1
η= 8 73
1 + 1 − η An0 + ηAn1
which can be solved by an iterative procedure.

When kint is unknown, it is not possible to use Eq. (8.64) for the Thiele modulus already
described, so if only information on the observed experimental reaction rate is available, the
Wagner–Weisz–Wheeler modulus (Mw) must be used (Levenspiel 1999):
L2p r v ρS n + 1
Mw = S
= Φ2S η 8 74
Dfeff Cf ,isur 2
where ρS is the density of the pellet and rv is the observed rate of reaction of species i, which also
translates as the CO depletion rate in the FTS reaction per unit mass of the catalyst.
In Eq. (8.75), the diffusivity Dcomb represents the combination between the bulk diffusivity Dij
and Knudsen diffusivity DK that can be estimated from the correlations reported by Pollard and
Present (1949):
1 1 1
= + 8 75
Dcomb Dij DK
In addition, the diffusivity of each component involved in the gas mixture is given by the follow-
ing expression (Wilke 1949):
1 − yi
DGi,mix = yi 8 76
D
i j ij
where Dij (where i diffuses through the liquid wax j) was estimated using the following correlations,
which are valid for a wide range of temperatures and pressures (Erkey et al. 1990a; Erkey et al.
1990b; Fuller et al. 1966):
0 945 T
Dij = 1 134 8 77
M 0m239
i
0 781
Mm j
σi σj
1 1
10 − 7 T 1 75 +
M mi M mj
Dij = 2 8 78
1 3 1 3
ρGmix νci + νvj
where M mi and M mj are the molecular weight of the solute (either CO or H2) and the molecular
weight of the solvent (waxy liquid product), respectively, both in g/mol; σ i and σ j are the hard-
sphere diameter for the solute (either CO or H2) and hard-sphere diameter for the solvent (waxy
liquid product), respectively, both in Å; νci and νcj are the critical volumes of components i and j,
respectively; and V is the specific volume of the different species in m3/mol. Meanwhile the
diffusivity of Knudsen DK is calculated by the following correlation obtained from the kinetic
theory of the gases (Welty et al. 2007):
T
DK = 4850dp 8 79
M mi
Welty et al. (2007) indicate that when DK exceeds Dij by more than two orders of magnitude under
the set of operating conditions used, the effects of diffusion resistance can be neglected.
It should be noted that the diffusivity of CO in the produced wax is usually lower than that of H2,
and it is considered as the limiting diffusion rate. On the other hand, when the synthesis gas feed
stream contains low concentrations of H2, it is convenient to consider that the diffusivity of the H2
in the waxy product is the limiting diffusivity (Hallac et al. 2015).
In the case of using catalyst pellets with irregular geometry, the equivalent particle diameter (dpe)
must be used to utilize the previous correlations.
The equivalent particle diameter is referred to as the diameter of a sphere whose surface area
(or volume) is equal to that of the catalyst particle, being a characteristic property of the particle
that depends on both its size and shape. One way to determine the equivalent particle diameter
for fixed-bed systems is provided by Cooper:
Vp
6 Sp
dpe = 8 80
ϕS
The geometric forms of the catalyst are ideally spherical, but it would be possible to use any cat-
alyst shape; however, correlations for the effectiveness factor are not possible in all cases mainly for
irregular geometry. Therefore, there are other parameters that are involved in calculating the effec-
tiveness factor that can be estimated for other types of catalyst particles such as the particle size also
known in the literature as the form factor Lp. It should be noted that the construction of the model
for an irregular geometry catalyst particle in the FTS process, would be more complicated.
Nowadays, the design of novel FTS catalysts is of utmost importance for the purposes of improv-
ing the synthesis gas conversions, product selectivities, and overall performance of the reactor
(Khodakov et al. 2007; Zhang et al. 2010). In the research of catalyst design applied not only to
the FTS process but also to the hydrotreatment processes, it is very important to improve the geo-
metries of the catalyst, mainly to diminish the mass and heat transfer intraparticle limitations. The
form factor (Lp) defined as the volume ratio to the catalyst particle surface (Vp/Sp) is frequently
employed to correlate certain characteristics such as the size and shape of the catalyst with the cat-
alytic activities, as well as to perform studies on the characteristics of the catalyst, packed bed, intra-
particle diffusion limitations, catalytic pore size, pressure drop restrictions, etc. Ancheyta et al.
(2005) proposed some equations for the determination of the area and the external volume for some
porous and nonporous particles of different shapes and sizes. This study is mainly oriented to cal-
culations of Vp and Sp for different types of hydrotreating catalysts; however, its application can be
extended to catalysts used in the FTS process. Calculation of Vp and Sp of lobe-shaped catalyst
particles are as follows:
V p = n πr 2c L − A1 L 8 81
Sp = n 2πr 2 L + 2πr c L ± 2A1 − nA2 8 82
where n is the number of lobes, A1 is the lateral area of the geometric shape between the lobes,
which represents a rhombus for a bi-lobular geometry, a triangle for a tri-lobular shape, and a frame
Table 8.16 Definition of parameters to determine the volume and surface of particles of different geometries.
Shape n θ rc A1 A2
Cylinder and pellet 1 — dp/2 0

dp d2p π
sin θ − sin θ 2
r cyl L
Bi-lobular 2 45
3 4142 = 8 88348 × 10 − 3 d2p 2
8 1 + sin θ
dp d2p 2sin θ − 1 2 π
r cyl L
Tri-lobular 3 60
4 = 3 86751 × 10 − 2 d2p 3
8 tan θ
2 π
dp 2 cos θ − 1 r cyl L
Tetra-lobular 4 45 d2p = 2 94373 × 10 − 2 d2p 2
4 8284 1 + 2 cos θ
for tetra-lobular geometry, while A2 is the common area between each cylinder and between each
side of the shape between the lobes. In Eq. (8.82), the term after parenthesis is negative for bi-
lobular forms and positive for tri- and tetra-lobular shapes. Table 8.16 shows the parameters
involved in Eqs. (8.81) and (8.82) for cylinders, spheres, pellets, bi-lobular, tri-lobular, and tetra-
lobular shapes.
An important aspect in the design of catalysts, simulation, and modeling of a fixed-bed reactor is
the analysis of transport limitations at the particle level. Mears (1971a) indicated that concentration
and temperature gradients can exist in three domains: (i) intraparticle, referring to the gradients
that occur within the individual catalytic particles; (ii) interface, wherein the gradients are gener-
ated between the external surface of the catalytic particles and the adjacent layer of the fluid; and
(iii) interparticle, with the axial and radial gradients within the reactor being considered.
It is of interest to have precise criteria that allow predicting the existence of intraparticle mass
transports, since it is of importance in determining suitable pathways for model simplification
of the catalyst particle, such as neglecting or not the effects of internal diffusion in the pore. On
the other hand, it is stated that it is only possible to obtain information related to the intrinsic kinet-
ics in the absence of effects caused by resistances to mass and heat transport in the catalytic particle
(Ibrahim and Idem 2006). There is a criterion given in the literature to determine if the presence of
intraparticle diffusion limitations exists; this is the Weisz–Prater criterion given by the dimension-
less parameter.
− RFT,obs ρS L2p
CWP = f
8 83
Deff C Gi
where RFT,obs is the observed reaction rate (mol/(Kgcat s)), ρS is the density of the solid catalyst
particle, and CGi is the limiting reactant concentration in the syngas mixture. This criterion estab-
lishes that if the value of CWP becomes much larger than 1, the effects of limiting the internal pore
diffusion are significant, i.e. there is a strong resistance to intraparticle diffusion resulting in a con-
centration gradient from the catalyst surface to its pores (Ibrahim and Idem 2006). Akpan et al.
(2007) showed experimentally that the diffusion resistance is null when the particle size (form fac-
tor) is less than 0.3 mm. This criterion of Mears (1971a) is often considered for the determination of
the onset of external mass transport limitation that is given by
r obs ρB Lp n
< 0 15 8 84
k FT
To determine that external mass transfer resistance has no significant effect on the reaction rate,
the following criterion is used, relating the observed reaction rate to the reaction rate if the resist-
ance to the film controls (Ibrahim and Idem 2006):
observed rate − RFT,obs
= ; i = reactants CO and H2 8 85
rate if film resistance controls k c C fi
However, Levenspiel (1999) points out that the film resistance to mass transfer should not influ-
ence the rate of reaction.
Ibrahim and Idem (2006) established the following two criteria: (1) the heat transfer resistance
within the pore and (2) the limitations of heat transfer across the gas film, to be:
Dfeff C Si sur − Ccp

i − ΔH RFT
ΔT max,particle = 8 86
λeff
L − RFT,obs − ΔH RFT
ΔT max,film = 8 87
h
where ΔTmax,particle and ΔTmax,film are the upper limit to temperature variation between the pellet
center and its surface and the upper limit to temperature difference between the gas bulk and the
cp
pellet surface, respectively; ΔHRFT is the heat of reaction; C Si sur and Ci are the concentrations at the
pellet surface and center, respectively; Dfeff is the effective mass diffusivity; and L is the characteristic
length.
One way of assessing the thermal severity resulting from the FTS process in fixed-bed systems is
by using the Mears (1971b) rigorous criterion, which allows one to compare the heat generated by
the reaction with the capacity of the catalytic bed to transfer heat from the reaction zone toward the
wall of the tube. This criterion was used in a pseudohomogeneous modeling of a fixed-bed FTS
reactor proposed by Philippe et al. (2009):
dp T a d2t 8λrad
1 − εB ΔH RFT,j 1+ <04 8 88
2 4λeff,rad T 2w U r dt
When this criterion is fulfilled, it can be assumed that the radial temperature profile is almost flat,
which means that the reaction heat is eliminated in the radial direction. However, this does not
ensure that there is no generation of hot spots along the catalytic bed (Philippe et al. 2009). The
following criteria can be applied in cases where the resistance to heat transfer at the wall is signif-
icant, as well as its impact on catalyst particle:
ΔH RFT,j RFT,B d2t Ea 1 0 4Ru T w

< 8 89
4λeff,rad T w Ru T w dp Ea
1+8 Biw
dt
where RFT, B is the reaction rate per unit bed volume and Biw is the dimensionless number of Biot
defined as hwdt/λeff,rad. It should be noted that the term λeff,rad/8hwdt represents the heat transfer
resistance in the wall that cannot be ignored when low ratios of the diameter of the tubes to the
diameter of the particle are used (dt/dp ≈ 10) (Mears 1971a; Mears et al. 1971b). Otherwise, Mears
(1971a) pointed out that the magnitudes of heat transfer resistance in experimental reactors
behave as follows: interparticle> interphase> intraparticle.
8.2.4.5 Catalytic Bed Parameters

An extremely important parameter in the modeling of fixed-bed systems involves the determination
of the bed void fraction (or bed porosity), as this indicates the space within the fixed bed that is filled
with catalyst particles, i.e. it states the fraction of the catalyst bed volume that is available for the
phases involved in a multiphase multicomponent system. Alternatively, having a reliable estima-
tion of the bed void fraction is important for minimizing the pressure drop across the catalyst bed.
Table 8.17 reports some correlations for calculating the void bed fraction applicable to many pro-
blems involving fixed-bed configurations.
Antwerpen et al. (2010) found a wide range of correlations to estimate the porosity of the bed of
structured packages that can be used in fixed-bed reactor modeling.
Once the bed void fraction is determined, the density of the solid catalyst particle can be
calculated according to the following equation:
ρB
ρS = 8 90
1 − εB
Based on the consideration of volume conservation, the bed void fraction is the sum of the gas
and liquid holdups (Whitaker 1973):
εB = ϵL + ϵG 8 91
ϵL + ϵG + ϵS = 1 8 92
Relationships between the holdups within the solid catalyst particle are given by
εS = ϵpL + ϵpG 8 93
p p p
ϵL + ϵG + ϵS =1 8 94
Table 8.17 Correlations for the calculation of the void fraction bed.
Correlation Reference
0 2024 1 0814 Benyahia and O’Neill (2005)

εB = 0 1504 + +
ϕp dt 2
+ 0 1226
dpe
−1
1 − εB r − r inn White and Tien (1987)
(1) εB r = 1 + 1 − exp − 2
εB dp
r out − r inn
for r inn ≤ r ≤
2
−1
1 − εB r out − r
(2) εB r = 1 + 1 − exp − 2
εB dp
r out − r inn
for ≤ r ≤ r out
2
dt Froment et al. (2011), Haughey and

−2 Beveridge (1969)
dpe
εB = 0 38 + 0 073 + 1 + 2
dt
dpe
The external surface area of the catalyst particles per unit volume of the fixed-bed reactor can be
calculated as
6 1 − εB
aS = 8 95
dpe
It is also convenient to estimate the pore size to determine the pore filling of the catalyst particle
with water and liquid hydrocarbons using the Kelvin equation (Ermolaev et al. 2015):
2σ i
dpore = 8 96
ρSi Ru T pi
pi − psat − ln
M mi psat,i
Because it is possible that the formation of waxy liquid is carried out inside the catalytic reactor,
and that the velocities reached by the liquid flow are relatively low in the range of a trickle regime, it
is necessary to determine the liquid holdup for generalized modeling of the fixed-bed FTS reactor.
This parameter is useful in design and operation of a fixed-bed reactor under a trickle-flow regime
(Larachi et al. 1991). Liquid holdup has been shown to greatly affect the pressure drop across the
bed, the amount of catalyst wetting, and the thickness of the liquid film surrounding the catalyst
particle (Ellman et al. 1990). In addition, the holdup knowledge is essential in the case of high
exothermicity of the reactions such as the FTS, since it allows avoiding the formation of hot spots
in the catalytic bed and the prevention of reactor runaways. On the other hand, it is a parameter
that affects the wetting efficiency of the catalyst and hence the selectivity, which depends on
whether the reaction is carried out in the wet catalyst area or in the wet and dry catalyst regions
alike (Al-Dahhan et al. 1997).
Bazmi et al. (2013) proposed a correlation for the prediction of liquid holdup where the effects of
the catalyst shape (extruded trilobe particle) on the pressure drop are considered, as well as the
behavior of the liquid holdup to the different methods of loading the catalyst into trickle-bed
systems:
35 2 0 17 35 2
We0L 5 ε3B Re L WeL0 5 ε3B Re L
ϵL,d = 0 07 exp 8 97
XL 1 − εB Re G XL 1 − εB Re G
where
05
uLs ρL
XL = 05 8 98
uGs ρG
uLs dpe ρL
WeL = 8 99
σL
The equation is valid within the range of gas superficial velocities from 0.01 to 0.1 m/s and the
liquid superficial velocities from 0.001 to 0.01 m/s included in the range of the trickle-flow
regime. Such a correlation was compared with the correlation proposed by Larachi et al.
(1991) as well as with experimental data and showed acceptable improvements in predicting
liquid holdups.
Al-Dahhan et al. (1997) also reported a comprehensive review on various correlations given in the
open literature for the prediction of the transition from trickle-flow regimes to pulse, for liquid
holdups, two-phase pressure drop, and catalyst wetting efficiency (liquid–solid contacting effi-
ciency), among others in high-pressure trickle-bed reactors.
8.2.5 Final Remarks

The search for new and better alternatives to produce improved clean fuels that meet more strin-
gent environmental constraints is a worldwide aim. The scientific and industrial community is
increasingly studying the implementation of the FTS as it has been considered as one of the most
viable options to replace conventional technologies for producing fuels of commercial value. In
addition, the design of more efficient catalysts and novel reactor technologies (microchannel reac-
tor and monolith reactor) has been a field of research currently active to greatly improve the overall
FTS process. However, the fixed-bed catalytic reactor has been preferably used because of its
marked advantages over other reactor technologies, which include the relatively easy scaling up
from a single tube. On the other hand, the FTS process is considerably complex since the distribu-
tion of FTS products has not been accurately known because the products are quite broad and dif-
ficult to accurately predict. FTS behavior is attributed to the effect of (i) the catalyst type and/or
support promoter and (ii) the type of reactor used, among others.
Despite the large number of mathematical models published in literature until now, it is nec-
essary to propose a generalized modeling scheme of the fixed-bed reactor for the FTS obtained
from simplified versions reported in the literature. However, solving the generalized model is a
rather complex task, since it requires an in-depth study of hydrodynamics, the analysis of the
phases distribution in the catalytic bed (vapor−liquid equilibrium), mass and heat transport phe-
nomena in the bulk and particle phases, a robust model for pressure drop, and the search for pre-
cise correlations that estimate the several parameters of the model. Therefore, it becomes evident
that there is a need to study different modeling schemes that can be derived from this proposed
generalized model and that must be validated through comparisons between simulated and exper-
imental data.
8.3 Importance of Proper Hydrodynamics Modeling in Fixed-Bed

Fischer–Tropsch Synthesis Reactor
8.3.1 Introduction
The importance of hydrodynamics, particularly gas density, superficial gas velocity, and total pres-
sure in axial and radial directions, was analyzed for the modeling of catalytic reactors using a non-
isothermal pseudohomogeneous approach. The modeling of a fixed-bed reactor in one and two
stages for the CO conversion by FTS was taken as a study case. For the validation of the proposed
model, the results of the simulations for the CO conversion and temperature profiles were com-
pared with experimental data reported in the literature. Simulations for CO conversion and reactor
temperature profiles confirmed the model’s ability to predict the selectivity of the liquid products in
the FTS reactor in one and two stages. The proposed model predicts more suitable profiles of CO
conversion and temperature along the reactor, which makes it a more robust and efficient tool for
design, optimization, and control purposes.
8.3.2 Mathematical Modeling of the Fixed-Bed Fischer–Tropsch Synthesis Reactor

To predict the syngas conversion and reactor temperature profiles of a fixed-bed multitubular FTS
reactor, a two-dimensional pseudohomogeneous model is proposed, which is formulated using
8.3 Importance of Proper Hydrodynamics Modeling in Fixed-Bed Fischer–Tropsch Synthesis Reactor 355
cylindrical coordinates. The radial dispersive plug flow is described by the mass and heat balances,
and the following assumptions were considered:
•• Steady-state operation
Plug-flow regime with no channeling effects along the catalyst bed
• Negligible axial dispersion effects of mass and heat, as well as concentration gradients in the
radial direction
•• Variable density, total pressure, and gas velocity along the axial direction
Radial heat transfer within the catalytic bed as a function of the constant radial effective thermal
conductivity
•• Constant temperature of the cooling medium

Negligible formation of a wax layer on the catalyst surface of mass and heat external transfer
limitations
• The liquid hydrocarbon diffusion out of the catalyst pores occurs fast enough to neglect any
(internal and external) mass transfer resistances.
It is worth mentioning that the liquid hydrocarbon accumulation (mainly wax) on the catalyst
particle surface is important for diffusional effects and external mass transport (Pöhlmann et al.
2016). It has been shown experimentally and by means of numerical simulations how the filling
of the pores catalyst with liquid hydrocarbons changes as a function of time. Also, the effectiveness
factor varies with the degree of catalytic pores filling (Pöhlmann and Jess 2016b). On the other
hand, for low-temperature FTS (<250 C), the catalyst pores are filled with liquid hydrocarbons
(waxes), which leads, inevitably, to pore diffusion effects within the catalyst particles and a decrease
in optimum catalyst performance and selectivity (Jess et al. 2013; Calderone et al. 2013). Further-
more, it has been shown that the diameter of the reactor and the catalyst particle play an important
role in the diffusional and external mass transfer effects (Jess and Kern 2012a; Jess and Kern 2012b).
However, for diameters of catalyst particles of 3–4 mm (Pöhlmann et al. 2016; Jess and Kern 2012b),
with a reactor diameter to catalyst particle diameter ratio of dt/dp = 7.6, the external diffusional
effects are not significant. In addition, the risk of the thermal runaway phenomenon occurring
becomes negligible (Pöhlmann et al. 2016). Based on these literature reports, the assumption that
the external mass transfer effects that are due to the wax layer or liquid hydrocarbons formed on the
catalytic surface are negligible can be justified.
8.3.2.1 Reactor Model

For the two-dimensional axisymmetric plug flow inside the fixed-bed FTS reactor, when the effect
of axial mass dispersion is neglected and at steady state, the mass balance equation results in the
following expressions:
∂x CO 1 − x CO ∂us Der ∂ 2 x CO 1 ∂x CO ρB RFTS

= + + + 8 100
∂z us ∂z us ∂r 2 r ∂r us C 0,CO
∂x H2 1 − x H2 ∂us Der ∂ 2 x H2 1 ∂x H2 ρB υj RFTS
= + + + 8 101
∂z us ∂z us ∂r 2 r ∂r us C0,H2
While the heat balance equation in the fixed-bed FTS reactor is:
∂T λer ∂2 T 1 ∂T ρ RFTS
= f + − f B − ΔH FTS 8 102
∂z ρ us C Pmix,G ∂r 2 r ∂r ρ us CPmix,G
Eqs. (8.100)–(8.102) represent the syngas conversion (CO and H2) and the temperature variation
in the axial and radial directions, which are typical equations used in two-dimensional pseudoho-
mogeneous models (Philippe et al. 2009; Sharma et al. 2011; Rafiq et al. 2011; Dai et al. 2014).
The equations related to hydrodynamics in the axial variation (density, surface velocity, and gas
pressure) are as follows:
∂ρf M mmix,G 1 ∂PT PT ∂T

= − 2 8 103
∂z Rc T ∂z T ∂z
∂us us ∂ρf
= − f 8 104
∂z ρ ∂z
∂PT u2 ρf
= −fp s 8 105
∂z dp
The boundary conditions used for solving the proposed model are as follows:
f
at z = 0 and 0 ≤ r ≤ R0 ; x CO = x CO,0 , T = T 0 , us = us,0 , ρf = ρ0 , PT = PT,0 8 106
∂x CO ∂T
at r = 0 and 0 ≤ z ≤ LB ; = =0 8 107
∂r ∂r
∂x CO ∂T
at r = R0 and 0 ≤ z ≤ LB ; = 0, λer = − hw T − T w 8 108
∂r ∂r
It is important to mention that in the proposed FTS reactor model, the intraparticle effects were
neglected for simplicity (two-dimensional pseudohomogeneous model). This was done to compare
it with the same model given in the literature that lacks robustness in terms of hydrodynamics, i.e.
the variation in the superficial velocity and the density of the gas mixture is not taken into consid-
eration. Another difference is the use of a novel correlation for the friction factor used for the cal-
culation of the total pressure.
8.3.2.2 Kinetics
For the development of kinetic models that describe the FTS reactions, the main approach reported in
the literature is based on the power-law rate model, in which the kinetics for reactants conversion is
described (syngas: CO and H2) (Pinna et al. 2002). This model has been successfully used and validated
in a series of case studies (Van der Laan and Beenackers 1999; Das et al. 2005). The use of this type
of kinetic model that represents a simple description of a reaction system in the FTS process is valid
(from a theoretical approach) if the products do not affect or participate in the formation mechanism
of initiating monomers of the Fischer–Trospch hydrocarbon chain growth (Visconti et al. 2007).
The kinetic equation of FTS is given by a power-law model as a function of partial pressures of CO
and H2, which represents the global consumption of CO during the synthesis:
RFTS = Ap e −
Ea R T
c pm n
H2 pCO 8 109
The following kinetic parameters were taken from the literature (Dai et al. 2014): Ap = 2.45 ×
106 mol/(Kgcat/(MPa0.3 s)), Ea = 73 KJ/mol, m = 1.3, and n = − 1. This CO consumption kinetics
equation has been previously used by others (Visconti et al. 2011b; Zhen et al. 2009; Zennaro et al.
2000; Yates and Saterfield 1991; Outi et al. 1981).
The molar product selectivity was calculated by the following:
moles of CO2 produced
SCO2 = × 100 8 110
moles of inlet CO − moles of outlet CO
moles of CH4 produced

SCH4 = × 100 8 111
moles of C2 − C 4 produced
SC2 − C4 = × 100 8 112
SC 5 + = 1 − SCO2 − SCH 4 − SC2 − C4 × 100 8 113
In order to guarantee the proposed model’s validity in terms of syngas conversion and product
selectivity, it is convenient and necessary to respect the same kinetic modeling scheme considered
in this case study (Dai et al. 2014). In addition, it is important to mention that the kinetic parameters
of a certain FTS kinetic model depend considerably on the operating conditions used (Gideon Botes
et al. 2009).
8.3.2.3 Other Parameters and Correlations

Table 8.18 shows the reactor characteristics and fluid properties used for the simulations,[15]
while Table 8.19 presents the correlations used to estimate the parameters involved in the proposed
model.
8.3.2.4 Numerical Method

The simulations were performed in MATLAB software, in which the finite difference method was
applied for the solution of the PDEs system representing the proposed model. The number of nodes
used for the discretization of the first partial derivatives in the axial direction was 1200 with a step
size of 0.01 m, while for the first and second partial derivatives in the radial direction, 25 nodes with
a step size of 0.0005 m were used. The numerical method was presented in a matrix form in which
the set of matrices that arise from the discretization of the system of PDEs exhibits a stable value of
the matrix norm of the maximum absolute value of the eigenvalues of the matrix, indicating that
the matrices are well conditioned, i.e. the numerical method is stable.
Table 8.18 Characteristics of the FTS fixed-bed reactor and fluid properties (Dai, et. al. (2014) / Elsevier).
Parameter Value Parameter Value
Inner tube radius 13 mm Heat capacity of wáter CPW = 4220 J Kg K

Total catalyst 30 l Initial pressure PT = 1.5 MPa
loading
Catalyst particle dp = 3.20 mm Inlet temperature T0 = 493 K
diameter
Bulk density of ρB = 630 Kg/m3 Cooling temperature TW = 486 K
catalyst bed
Bed void fraction εB = 0.42 Synthesis gas flow Fmix,G = 0.205 Kg/m2/s
−5
Gas viscosity μmix,G = 4.54 × 10 Kg/(m s) Inlet syngas 63.11 H2/30.76 CO/
composition (mol %) 2.37CH4/0.62 CO2/3.14 N2
Heat capacity of CPmix,G = 2700 J Kg K Cooling water flow FW = 780 Kg/h
gas mixture
Table 8.19 Correlations used for estimation of parameters involved in the proposed model.
Parameter Correlation Reference
Radial dispersion dp εB Der 1 0 38 Rase (1990)

For >01 = +
coefficient dt us dp m Re f
dp εB Der 1 0 38
For <01 = +
dt dp m Re f
us dp 1 + 19 4
dt
where
Re f > 400 11
m=
20 < Re f < 400 57 85 − 35 36 log Re f + 6 68 log Re f
Radial effective dt 0 331 Wilhelm
λer = 0 199 + 0 015 + 2
thermal conductivity dp dt (1962)
1 + 8 54
dp
Convective heat hw = 1000 W/m2K Haarlemmer
transfer coefficient and
on the side wall Bensabath
(2016)
T
Reaction enthalpy ΔH FTS = ΔH 0FTS + 298 K C P,FTS dT
Sandler
(2006)
Friction factor 172 4 36 d p ρf u s Allen et al.
fp = + 0 12 ; Re f = (2013)
Re f Re f μmix,G 1 − εB

The results of the simulations with the proposed model were compared with those reported by Dai
et al. (2014). They used a one-stage fixed-bed reactor with 12 m of catalytic bed and a two-stage FBR
with 6 m of catalytic bed for each stage. The internal diameter of the reactor for both configurations
was 13 mm. The experiments were carried out with a CoMo/γ − Al2O3 catalyst at the following
reaction conditions: initial temperature (T0) of 493 K, initial pressure of 1.5 MPa, gas synthesis flow
(Fmix,G) of 0.205 Kg/m2/s, and H2/CO molar ratio of 2.05. The model proposed by Dai et al. (2014)
was developed based on the following assumptions: (i) steady state; (ii) plug-flow regime assumed
with no channeling along the bed; (iii) the axial dispersion of mass and heat as well as radial con-
centration gradients are ignored; (iv) the build-up of wax layer on the catalyst surface is negligible
for mass and heat transfer; (v) described by the mass balance of CO and heat balance; (vi) constant
surface velocity and gas density; and (vii) the total gas pressure was estimated by the total sum of the
partial pressures of each component of the gas mixture.
8.3.3.1 Simulations for the One-Stage Reactor

Figure 8.6 illustrates the results of the simulations carried out with the proposed model and those
reported by Dai et al. (2014) for the CO conversion in the one-stage reactor.
It can be seen that both models quite accurately predict the CO conversion at the reactor outlet.
The absolute error between experimental and predicted values of CO conversion, calculated
yexp − ycal
as × 100, was 2.7% for Dai et al. (2014) model and 1.8% for the proposed model.
yexp
90
80
70
CO conversion XCO(%)
60
50
40
30
20
10
0
0 2 4 6 8 10 12
Catalytic bed length, LB(m)
Figure 8.6 CO conversion profile in one-stage fixed-bed FTS reactor. Proposed model (solid line), literature
model (Moazami et al. 2015a) (dotted line), experimental data (■) (Dai et al. 2014).
Although the prediction of CO conversion at the reactor outlet is similar, the trend of the sim-
ulated CO conversion profiles of both models is different. In majority of the reactors, the proposed
model predicts higher values of CO conversion than the literature model. The highest difference
is found in the middle of the reactor (~10% CO conversion of difference). It is then anticipated
that these dissimilar profiles may cause the prediction of the temperature profile along the axial
direction to be different with both reactor models.
In the last third part of the reactor, the proposed model predicts small variations in CO conver-
sion (quasi steady state), while CO conversion with the model presented in the literature
continuously increases.
The profiles of CO conversion calculated with the proposed model correlate with those reported
by Rafiq et al. (2011), as can be seen in Figure 8.6, who used a similar model with the following
assumptions: (i) variable mass and heat radial dispersion; (ii) variable gas density and surface veloc-
ity in axial direction; and (iii) variable axial total pressure. It should be noted that the simulations of
Rafiq et al. (2011) were carried out in a 2-m long reactor with the use of a spherical unpromoted
cobalt catalyst and the following reaction conditions: a wall temperature of 473 K; an initial total
pressure of 2 MPa; and a gas hourly space velocity (GHSV) of 37 NmL/(gcat h).
On the other hand, it is important to mention that the global conversion of hydrogen at the outlet
of the reactor in one stage is 89.1% (Figure 8.7). However, neither experimental data nor the sim-
ulated value was provided by Dai et al. (2014) for comparison. The hydrogen conversion profile
during the FTS reaction has a different tendency than the CO conversion, i.e. higher conversions
are achieved at the exit of the reactor (89.1% versus 77.5%). The H2 conversion is greater than the CO
conversion precisely because there is an excessive consumption of hydrogen until it almost reaches
its exhaustion, which is due, in principle, to a considerable production of methane and light gases.
When the reaction between H2 and CO occurs on the surface of the cobalt-based catalyst, it forms
methylene groups, (-CH2-) which, by successively joining these groups, produce long hydrocarbon
chains (mainly paraffins). However, it should be noted that each −CH2 formed during the FTS
reaction releases a higher average heat than in typical catalytic reactions that are carried out in
90
80
70
H2 Conversion, XH2(%)
60
50
40
30
20
10
0
0 2 4 6 8 10 12
Figure 8.7 H2 conversion profile in one-stage fixed-bed FTS reactor.
conventional petroleum refining processes (Rafiq et al. 2011). Therefore, it is extremely important
to have a safe control reactor temperature due to the high exothermicity of the Fischer–Tropsch
reaction, to avoid the runaway phenomenon, and to guarantee a high selectivity and productivity
(Rafiq et al. 2011). Thus, the reactor must be modeled in non-isothermal conditions to obtain infor-
mation about its behavior under this regime and to identify the parameters that have a direct influ-
ence on the thermal behavior of the reactor, guaranteeing the control of the temperature.
The literature reports two-dimensional pseudohomogeneous models in which the suitable tem-
perature profile of fixed-bed FTS reactors has been verified; however, that is not the case for uni-
dimensional pseudohomogeneous models (Marvast et al. 2005; Rafiq et al. 20011; Marvast et al.
2005; Lee and Chung 2012; Chabot et al. 2015). It is worth noting that the former model did not
take into consideration the variation in the velocity and density of the gas along the axial direction.
Assuming they are constant or not, these two hydrodynamic variables may cause differences in the
conversion profiles and, hence, in the temperature profile along the reactor.
Figure 8.8 shows the results of the temperature profiles with the proposed model, the model
presented in the literature, and experimental data provided by Dai et al. (2014) for one-stage
multitubular fixed-bed FTS reactor. The absolute average differences between the experimental
n exp
i = 1 Ti − Ti
cal
and predicted values with the proposed model and the model presented in
the literature (Dai et al. 2014) were 4.8% and 4.3%, respectively.
Both models predict a temperature jump (runaway) at the beginning of the reactor, which is
slightly lower than the proposed model. This temperature profile behavior is due, in principle,
to the gradual increase of the gas mixture velocity in the axial direction of the reactor. This result
correlates with Chabot et al. (2015), who studied the effect of the inner diameter of the reactor tube
on the performance of the reactor. They confirmed that an increase in the space gas velocity in the
axial direction causes an increase in the heat released inside the bed. They also observed that as the
diameter decreases, the maximum temperature induced by the heat released by the FTS decreases
toward the reference value of the cooling temperature. The temperature profile predicted with the
proposed model is like those presented by others with the same type of pseudohomogeneous model
in two dimensions (Jess and Kern 2009; Marvast et al. 2005; Rafiq et al. 2011).
504
502
500
Temperature (K)
498
496
Initial temperature, T0
494
492
490
0 2 4 6 8 10 12
Figure 8.8 Temperature profile in one-stage fixed-bed reactor. Proposed model (solid line), literature model
(dotted line) (Moazami et al. 2015a), experimental data (■) (Dai et al. 2014).
It should be noted that the differences in CO conversion profiles shown in Figure 8.6 are due to
the high reaction temperatures that are reached at the beginning of the reactor. It should be noted
that the model presented in the literature performed simulations with an initial operating temper-
ature of 498 K, 5 degrees higher than that used in our simulations (493 K), which is the correct inlet
temperature as indicated by Dai et al. (2014).
Both models predict that the hot spot is slightly lower than the experimental value (1.5-2 K). It has
been reported that the presence of a hot spot near the reactor entrance zone is due to two important
causes: (i) the composition of syngas in the feed, and (ii) the cooling temperatures that are used
(Marvast et al. 2005). The displacement of the hot spot beyond the reactor inlet can be achieved
by handling higher cooling temperatures; however, this results in the potential for the reactor
to encounter a thermal leakage, leading to the irreversible deactivation of the catalyst (Rafiq
et al. 2011).
Figure 8.9 shows the behavior of the total gas pressure in the axial direction of the reactor. The
absolute average differences between the experimental values for the proposed model and the lit-
erature model Pexp T − P T were 0.01 MPa and 0.04 MPa, respectively. Similar to the CO conversion
cal
profiles, the total pressure at the reactor outlet is predicted to be more or less equal with both mod-
els; however, the variation in the total pressure along the reactor is indeed different since the values
calculated with the proposed model were always higher than those found with the model proposed
in the literature. This difference is due to the correlation that is used in the proposed model (Allen
et al. 2013), which considers the effects of the bed porosity and the Reynolds number of the catalyst
particle. Dai et al. (2014) model, however, estimates the pressure drop in the catalyst bed with the
sum of the partial pressures of the components of the mixture gas, which may be erroneous since
the characteristics of the catalytic bed are neglected.
It is relevant to mention that the correlation proposed by Allen et al. (2013) for the prediction of the
pressure drop through the catalytic bed is presumed to be superior to the classical Ergun equation,
which is due, in principle, to the fact that the latter tends to over-predict the pressure drop through
1.6
1.4
1.2
Total pressure (MPa)
0.8
0.6
0.4
0.2
0 2 4 6 8 10 12
Figure 8.9 Total gas pressure in one-stage fixed-bed reactor. Proposed model (solid line), literature model
(dotted line) (Moazami et al. 2015a), experimental data (■) (Dai et al. 2014).
randomly packed or structured beds with smooth spheres at a number >700. In addition, the con-
stant of 1.75 established in the Ergun equation can vary by a factor of up to almost 5, which depends
on the alignment and the shape of the obstacle offered by the packing arrangement (Allen et al.
2013). On the other hand, the correlation given by Allen et al. (2013) was derived from an analysis
that considers the friction and dragging generated from the tortuosity of a packed bed with smooth
or rough spheres and any other geometry such as cylindrical packing of irregular shapes.
Figure 8.10 depicts the gas density profile predicted by the proposed model, in which it can be
appreciated that density decreases proportionally to the total gas pressure (Figure 8.9). The
4.5
3.5
Density of gas, ρG(Kg/m3)
2.5
1.5
0.5
0 2 4 6 8 10 12
Figure 8.10 Gas density in one-stage fixed-bed reactor according to the proposed model.
reduction in gas density is a result of a decrease in the concentration of the syngas as the FTS reac-
tion proceeds along the axial direction. On the other hand, the calculation of the gas mixture density
depends, to a large extent, on the amount of syngas moles and of the light products. As can be seen
in Figure 8.10, the gas mixture density tends to decrease because as the syngas is exhausted when
the FTS reaction is carried out, a small amount of light gases is generated (CO2, CH4, and C2 − C4),
which is denser than the syngas (CO and H2).
Figure 8.11 shows the gas mixture velocity profile determined using the proposed model. It can be
seen that the superficial velocity increases as the FTS reaction occurs. This is because the syngas
concentration decreases. On the other hand, it is true that light gases are generated; however,
the quantity produced is relatively low, and their appearance is not reflected in a decrease
in the superficial velocity of the gas mixture. Another reason for the increase in gas velocity is
the decrease in total gas pressure (Rafiq et al. 2011).
Figure 8.12 shows the comparison between the simulation and the experimental value of the
product selectivity at the outlet of the one-stage fixed-bed FTS reactor. The light gases produced
(CO2, CH4, and C2 − C4) and the generation of liquids with a number of carbon atoms greater than
5 (C5+) are compared. The absolute error between the experimental value given by Dai et al. (2014)
and the predicted value acquired using the proposed model of the product selectivity of CH4, CO2,
yexp − ycal
C2 − C4, and C5+, calculated as yexp × 100, was 12.1, 12.1, 20.1, and 0.8 %, respectively. Even
though the calculated error for the selectivity of the light gases (C2 − C4) is ~20%, it can be consid-
ered negligible compared with the selectivity of the other products.
It should be noted that a rise in the reaction temperature generates the undesirable production of
methane (Eidt et al. 1994). It is known that methane production is undesirable in the FTS process,
and a means of inhibiting its production needs to be determined. According to Ranhimpour and
Elekaei (2009), if a sufficient amount of hydrogen is added to the FTS process, this can increase
the conversion of H2O to CO2 by the gas–water shift reaction, which in turn would lead to a
decrease in methane production. However, according to Ranhimpour and Elekaei (2009), the
decrease in methane production can also be carried out by the two-stage effect. This is because
the separation and condensation of water and liquid hydrocarbons that are produced between
0.509
0.508
0.507
Superficial velocity (m/s)
0.506
0.505
0.504
0.503
0.502
0.501
0.5
0 2 4 6 8 10 12
Figure 8.11 Superficial gas velocity in one one-stage fixed-bed reactor according to the proposed model.
(a) (b)
8 3.5
CO2 Selectivity (%)

CH4 Selectivity (%)
7 3
6 2.5
5 2
4 1.5
3 1
2 0.5
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Catalytic bed length, LB(m) Catalytic bed length, LB(m)
(c) (d)
3.5 98
C2 – C4(%) Selectivity
C5+ Selectivity (%)

96
2.5
94
2
92
1.5
90
1
0.5 88
0 86
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Catalytic bed length, LB(m) Catalytic bed length, LB(m)
Figure 8.12 Profiles of products selectivities in one stage FTS reactor: (a) CH4; (b) CO2; (c) C2 − C4; (d) C5+.
(−) predicted value by proposed model; (□) experimental value by Adapted from Dai et al. 2014.
the two stages of the FTS reactor, and by controlling its temperature, can reduce the generation of
methane and, therefore, increase the production of C5+ hydrocarbons (Figures 8.12a and 8.12d).
It can be seen in Figure 8.12b that the production of CO2 is relatively low according to the exper-
imental value given by Dai et al. (2014). However, the value predicted by the proposed model is not
far from the experimental value reported, which indicates the efficacy of the model. In addition, low
CO2 production is desirable since it is not a product of interest in the FTS process.
Figure 8.12c shows the light gases selectivity, which is relatively low. This is because a low pro-
duction of light gases is desired in the use of a multistage reactor, which positively contributes to the
production of a significant number of liquid hydrocarbons. However, the low generation of light
hydrocarbons also means that the gas mixture density does not increase during the FTS reaction.
It can also be seen that the proposed model agrees with the experimental value reported by Dai
et al. (2014).
8.3.3.2 Simulations for the Two-Stage Reactor

The predicted and experimental values of CO conversion are presented in Figure 8.13. For the first
stage, the absolute error between experimental and predicted values is 7.5% for the proposed model,
while it is 6.3% for the model proposed in the literature. For the second stage, the absolute errors are
1.3% and 2.0%, respectively. The proposed model predicts the CO conversion more accurately at the
exit of the second stage compared to the model proposed by Dai et al. (2014). In general, the pro-
posed model tends to overestimate the experimental values, while the model from the literature
tends to underestimate them.
The H2 conversion in the two-stage reactor was 71.2% for the first stage and 96.5% for the
second stage, and the axial profile is illustrated in Figure 8.14. It can be seen that the achieved
100
90
80
CO conversion, XCO(%)
70
60
50
40
30
20
10
0
0 2 4 6 8 10 12
Catalytic length bed, LB(m)
Figure 8.13 CO conversion profile in two-stage fixed-bed FTS reactor. Proposed model (solid line), literature
model (dotted line) (Moazami et al. 2015a), experimental data (■) (Dai et al. 2014).
100
90
80
H2 Conversion, xH2(%)
70
60
50
40
30
20
10
0
0 2 4 6 8 10 12
Catalytic bed length, (m)
Figure 8.14 H2 conversion profile for the fixed-bed FTS reactor in two stages according to the
proposed model.
H2 conversions are higher than the CO conversions. This may be because when the surface gas
velocity is increased, it produces a low CO conversion but a high H2 conversion as the residence
time is reduced (Rafiq et al. 2011).
Figure 8.15 shows the temperature profile along the axial direction for both stages of the reactor.
The absolute error between the experimental values and those predicted by the proposed model
were 2.8 K and 2.4 K for the first and second stages, respectively, while for the Dai et al. (2014)
model, the errors were 2.0 K and 1.9 K for the first and second stages, respectively. In general,
504
502
500
Temperature (K)
498
496
Initial temperature, T0
494
492
0 2 4 6 8 10 12
Figure 8.15 Temperature profile in the fixed-bed FTS reactor in two stages. Proposed model (solid line),
literature model (dotted line) (Moazami et al. 2015a), experimental data (▲) (Dai et al. 2014).
the error in the prediction of the temperature is negligible, although the Dai et al. (2014) model
provides a slightly more accurate prediction than the experimental data. This is because the tem-
perature curve for the two reactor stages shifted to an initial temperature of 498 K. It is important to
mention that the proposed model and the Dai et al. (2014) model predict the experimental hot spot
with a difference of 2 K and 1 K, respectively.
In recent literature reports (Park et al. 2014; Marvast et al. 2005; Sharma et al. 2011; Rafiq et al.
2011; Chabot et al. 2015), the influence of the variation in the reactor tube diameter on the reactor
performance has been analyzed with an aim to reduce the possibility of the formation of high hot
spots and temperature runaway. Thus, it is important to optimize the diameter of the reactor tube in
order to minimize the heat released by the FTS reaction to eliminate the possibility of a thermal
runaway.
It has also been found that the catalytic activity decreases considerably because of the generation
of hot spots in the reactor (Chabot et al. 2015). Therefore, temperature control in the reactor is
advised to avoid this type of problem; otherwise, the possibility of achieving high yields of liquid
products and conversion of syngas is reduced. The proposed model tends to approximate the tem-
perature of the refrigerant (boiling water at high pressure), thus showing its reliability, potential for
the prediction of the thermal runaway phenomenon, and application in the design of commercial
FTS reactors.
Figure 8.16 shows the results of the comparison of simulation and experimental data for the pres-
sure drop across the catalytic bed for the two-stage reactor. It can be seen that the total pressure at
the beginning of the second stage increases from 0.81 to 1.2 MPa with the proposed model, while it
increases from 0.6 to 1.2 MPa with the Dai et al. (2014) model. According to Dai et al. (2014), this
increase is due to the liquid products and water that are produced and separated by condensation at
the end of the first stage until the initial syngas pressure in the second stage of the reactor reaches
1.2 MPa. The increase in the partial pressure of the syngas in the second stage is due to the increase
in the syngas concentration. In another report, it was observed that the increase in the syngas
1.6
1.4
1.2
Total pressure (MPa)
0.8
0.6
0.4
0.2
0 2 4 6 8 10 12
Figure 8.16 Pressure drop profile in the fixed-bed FTS reactor in two stages. Proposed model (solid line),
literature model (dotted line) (Moazami et al. 2015a), experimental data (■) (Dai et al. 2014).
pressure contributes favorably to the formation of high-molecular-weight hydrocarbons and the

improvement of the FTS reaction rate when using an alumina-supported cobalt-based catalyst
(Gideon Botes 2009).
It should be noted that the increase in the operating gas pressure allows higher conversions of CO
and H2 to be obtained. In addition, it has been confirmed in the literature that high pressure and
temperature enhance the amount of CO conversion (Rohde et al. 2008; Rahimpour et al. 2011a;
Rahimpour et al. 2011b; Wang et al. 2003a). This is the reason why multi-stage fixed-bed config-
urations improve syngas conversion and the selectivity and productivity of the liquid products.
There is a clear difference in the pressure drop predicted by both models for the two-stage reactor,
although the two models accurately predict the experimental value reported at the exit of the reac-
tor. It should be noted that the pressure drop is estimated by Dai et al. (2014) with the sum of the
partial pressures of the syngas (CO and H2), CO2, and light gases (CH4 and C2 − C4) in the axial
direction, while in the proposed model it is done by using the correlation recently presented by
Allen et al. (2013). This correlation takes into account the packing form of the catalyst particles
(including the packing method), since it has been demonstrated that this has a high influence
on the estimation of delta-P. This correlation shows the similarity between the calculated pressure
drop and the experimental value given by Dai et al. (2014) in both reactors.
Figure 8.17 presents the results of the simulation of the superficial gas velocity according to the
proposed model for the two-stage reactor. It is observed that the velocity increases gradually as the
FTS reaction occurs along the axial direction in both stages. However, the difference in gas velocity
change is not significant in the first stage after 6 m compared with the second stage of the reactor. In
general, the increase in the gas velocity observed in the case of both reactor stages allows the reactor
temperature to drop so rapidly to approximate the cooling water temperature due to the gradual
decrease in the retention time of the reactants (CO and H2). Such a behavior has been observed
by Rafiq et al. (2011), who pointed out that a considerable increase in the synthesis gas velocity
would lead to a significant decrease in the conversion of the reactants. The proposed model predicts
small velocity gradient.
0.51
0.509
0.508
Superficial velocity (m/s)
0.507
0.506
0.505
0.504
0.503
0.502
0.501
0.5
0 2 4 6 8 10 12
Figure 8.17 Superficial velocity profile in the fixed-bed FTS reactor in two stages according to the
proposed model.
On the other hand, Sharma et al. (2011) presented an analysis of the effects of gas superficial
velocity on the effective radial thermal conductivity and on the global heat transfer coefficient.
In this study, it was observed that an increase in gas velocity produces a significant change in these
two thermal parameters, which increases both the released heat of reaction and the temperature of
operation, which are beneficial due to the high conversions of CO and H2 can be achieved.
All these observed behaviors show the importance of considering the effects of the superficial gas
velocity for modeling purposes in a single and multi-stage fixed-bed FTS reactor.
In addition, due to the considerable decrease in the pressure drop, the density of the gaseous
mixture diminishes proportionally, as can be seen in Figure 8.18. This is due to the decrease in
4.5
4
Density of gas, ρG(Kg/m3)
3.5
2.5
1.5
1
0 2 4 6 8 10 12
Figure 8.18 Density profile in the fixed-bed FTS two-stage reactor according to the proposed model.
(a) (b)
14 7
12 6
CH4 selectivity, (%)
CO2 selectivity, (%)

5
10
4
8
3
6
2
4 1
2 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Catalytic bed length Catalytic bed length
(c) (d)
7 100
C2–C4 selectivity, (%)
6
95
C5+selectivity, (%)
5
4 90
3 85
2
80
1
0 75
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Catalytic bed length Catalytic bed length
Figure 8.19 Profiles of product selectivities in the two-stage fixed-bed FTS reactor according to the proposed
model. (−) predicted value by proposed model; (□) experimental value by Dai et al. (2014). a) CH4 selectivity,
b) CO2 selectivity, c) C2 − C4 selectivity, d) C5+ selectivity.
the amount of the denser gases (C2 − C4) as the production of liquid hydrocarbons increases, as
reported by Dai et al. (2014).
Figure 8.19 shows the comparison between the simulations and experimental values for the selec-
tivity toward light gases (CO2, CH4, and C2 − C4) and the generation of liquids with a number of
carbon atoms greater than 5 (C5+) at the outlet of the two-stage fixed-bed FTS reactor. In general, it
can be seen that the production of methane, carbon dioxide, and light gases increases during the
operation of the second stage of the FTS reactor. However, the production of C5+ hydrocarbons
decreases as a result of the effect in multistage operation. The increase observed in gaseous products
is also a result of the increase in the pressure system and hence the concentration of syngas. This is
because of the condensation of water and the liquid hydrocarbons produced between the two stages
of the fixed-bed FTS reactor.
Figures 8.20a and 8.20b show the variations of CO (a) and H2 (b) conversion, respectively, as
a function of the axial and radial directions of the reactor. It can be observed that there is a slight
change in the syngas conversion in the radial direction, which could be negligible. The same can be
stated for the radial variation of the gas mixture surface velocity, pressure drop, and gas mixture
density. However, in the case of radial variation of the reactor temperature, the changes are more
significant, which is of vital importance in modeling studies of fixed-bed reactors for FTS.
Figure 8.21 shows profiles of some process variables in axial and radial positions. For instance,
from Figure 8.21d, it can be seen that as the tube diameter increases, the temperature of the reactor
decreases until the inlet temperature is reached. This phenomenon has already been observed by
others who studied the influence of the variation in the reactor tube diameter on the reactor per-
formance (Marvast et al. 2005; Park et al. 2014; Sharma et al. 2011; Rafiq et al. 2011; Chabot
et al. 2015).
(a)
CO Conversion, XCO(%)
100
50
0
12 10 8 12 14
6 8 10
4 2 4 6
0 0 2
Catalytic bed length, LB(m) Tube radius (mm)
(b)
H2 Conversion, XH2(%)
100
50
0
12 10 8 12 14
6 8 10
4 2 4 6
0 0 2
Catalytic bed length, LB(m) Tube radius (mm)
Figure 8.20 Syngas conversion profile in fixed-bed two-stage reactor: (a) CO, (b) H2.
(a) (b)
Sperficial velocity (m/s)
1.5 0.51
Syngas pressure,
PT (Mpa)
1
0.505
0.5
0 0.5
15 15
10 15 10 15
10 10
5 5 5 5
0 0 0 0
Catalytic bed length, LB(m) Tube radius (mm) Catalytic bed length, LB(m) Tube radius (mm)
(c) (d)
Gas density, ρ (Kg/m3)
Temperature (°K)
6 505
G
4 500
2 495
0 490
15 15
10 15 10 15
10 10
5 5 5 5
0 0 0 0
Tube radius (mm) Catalytic bed length, LB(m) Tube radius (mm)
Figure 8.21 (a) Syngas pressure; (b) superficial velocity; (c) gas density; (d) temperature profiles in the fixed-
bed FTS two-stage reactor according to the proposed model.
8.4 Dynamic One-Dimensional Pseudohomogeneous Model for Fischer–Tropsch Reactors 371
8.3.4 Final Remarks

A more robust model was developed for the simulation of a fixed-bed reactor in one and two stages
for the conversion of CO by the FTS. The FTS was used as a study case to demonstrate the impor-
tance of considering hydrodynamics in the model formulation that aims for the design, simulation,
and optimization of commercial reactors. The developed two-dimensional pseudohomogeneous
model was more accurate than the models proposed in the literature for describing the behavior
of syngas conversion, product selectivity, and temperature profile in the axial and radial directions.
More precise correlations for the diffusivity and thermal conductivity, as well as an equation for the
friction factor that allows for a more accurate prediction of the gas pressure in the system, were
incorporated in the model. It was proven that the variation of superficial velocity, density, and total
pressure of gas allows for a better understanding of the effects of these variables on reactor perfor-
mance in terms of syngas conversion and temperature profiles. By model simulations, it was con-
firmed that the FTS can be improved by using multistage fixed-bed reactors since they enhance
syngas conversion, product selectivity, and temperature profile.
8.4 Dynamic One-Dimensional Pseudohomogeneous Model for

Fischer–Tropsch Reactors
8.4.1 Introduction
A dynamic one-dimensional pseudohomogeneous fixed-bed FTS reactor model in a non-isothermal
regime is developed. The proposed dynamic model is validated with experimental data obtained
from the literature and is compared with simulations of a steady-state one-dimensional isothermal
fixed-bed FTS reactor model recently reported. The best set of initial operating conditions showing
highest syngas conversions (CO and H2) and high heavy hydrocarbons (C5+) selectivities corre-
spond to temperature, pressure, and GHSV of 503 and 518 K, 1.0 and 1.5 MPa, and 1800 (NmL/
(gcat h)), respectively. The proposed model can accurately predict the CO conversion experimental
data and product’s selectivity. Furthermore, it allows us to evaluate the thermal behavior of
the reactor, as well as to analyze the hot spot generation typically observed in highly exothermic
reactions such as in the case of the FTS reaction.
8.4.2 Formulation of the Model

8.4.2.1 Model Equations and Solution
The proposed model to describe the dynamic behavior of the fixed-bed FTS reactor was implemen-
ted under plug-flow regime at non-isothermal conditions, taking into account the variations of
the superficial gas velocity, the gas mixture density, and total pressure of the system in the axial
direction, which have been demonstrated to be important parameters to improve predictions of
syngas conversion, product’s selectivity, and temperature profiles (Méndez et al. 2019).
The proposed dynamic model consists of the following set of equations that describe the
behavior of the fixed-bed FTS reactor:
RN
∂Ci ∂Ci
= −us + ρB β vij Rj 8 114
∂t ∂z j=1
RN
∂T ∂T ρB β 4U r
= −us + −ΔH j vij Rj − T − Tw 8 115
∂t ∂z ρmix,G CPmix,G j = 1 dt ρmix,G CPmix,G
∂us us ∂ρmix,G
= − 8 116
∂z ρmix,G ∂z
∂ρmix,G M mix,G 1 ∂PT PT ∂T
= − 2 8 117
∂z Rc T ∂z T ∂z
∂PT u2 ρ
= − f p s mix,G 8 118
∂z dp
subject to the boundary conditions:

at t = 0, z Ci = C i,0 , T = T 0 , us = us,0 , ρmix,G = ρmix,G,0 , PT = PT,0 8 119
at z = 0, t Ci = C i,0 , T = T 0 , us = us,0 , ρmix,G = ρmix,G,0 , PT = PT,0 8 120
The proposed model was solved using the finite difference and the Runge–Kutta method using
the ode23 routine of MATLAB software. Twenty-five nodes were employed to approximate the first
derivatives in the axial direction, and an appropriate step size was also employed to reduce the com-
putational time and cost derived from the dynamic model simulations.
8.4.2.2 Model Parameters, Correlations, and Kinetics

The parameters and correlations used in the proposed model are given in Table 8.20. The global
heat transfer coefficient was estimated using the correlation given by Haarlemmer and Bensabath
(2016). This parameter is given as a function of the convective heat transfer coefficient on the inter-
nal side reactor (hint), the convective heat transfer coefficient of the external side reactor (hext), and
the reactor thermal conductivity (λw), which in this case is of stainless-steel material. The pressure
drop along the catalytic bed was estimated using a correlation recently reported in the literature for
the friction factor. Such a correlation was validated by comparing it with many correlations
reported in the literature showing superior accuracy (Allen et al. 2013). The heat conductivity of
the steel wall reactor was assumed to be 60 Wm−1h−1 (Sharma et al. 2011), and the value of the
heat transfer coefficient (film) of the external side corresponding to boiling water coolant was
1000 Wm−1h−1 (Haarlemmer and Bensabath 2016).
Table 8.20 Parameters and correlations of the dynamic one-dimensional pseudohomogeneous fixed-bed
FTS reactor model.
Parameter
n 172 4 36
μmix,G = yi μi fp = +
i=1 Re Re 0 12
μi = a + bT + cT2 + dT3 εB = 0.40; ϕ = 4/3
n −1
M mix,G = yi μi 1 δw 1
Ur = + +
i=1 hext λw hint
−4 407
−4 074 dp 0 458
β = 0.255 hint = ρmix,G CPmix,G μmix,G
μmix,G εB
dp ρmix,G us CPi = a + bT + cT 2 + dT 3
Re =
μmix,G SN
CPmix,G = yi C P i
i=1
Other parameters were considered as constants, such as the hollow fraction of the bed and
the active side volume fraction of the catalyst. There are several studies on the kinetics of the
FTS reactions focused on the formation of hydrocarbons. Most of the developed kinetic models
are of the Langmuir–Hinshelwood–Hougen–Watson (LHHW) type; however, some authors
have developed simpler kinetic models based on the power-law type (Marvast et al. 2005; Yang
et al. 2003; Wang et al. 2003b), with which good predictions and fit to experimental data
have been reported. The kinetic model used in the proposed reactor model is of the following
power-law type:
m n Ea,j mj nj
Rj = kj pCOj pHj 2 = Aj exp − p p 8 121
RT CO H 2
The set of FTS reactions considered was:
R1 CO + 3H 2 CH4 + H 2 O 8 122
R2 2CO + 4H 2 C 2 H 4 + 2H 2 O 8 123
R3 2CO + 5H 2 C 2 H 6 + 2H 2 O 8 124
R4 3CO + 7H 2 C 3 H 8 + 3H 2 O 8 125
R5 4CO + 9H 2 n − C4 H 10 + 4H 2 O 8 126
R6 4CO + 9H 2 i − C 4 H 10 + 4H 2 O 8 127
R7 6 05CO + 12 23H 2 C6 05 H 12 36 C5+ + 6 05H 2 O 8 128
R8 CO + H 2 O CO2 + H 2 8 129
The product’s selectivity was calculated using the Eqs. (8.110)–(8.113).
The values of the pre-exponential factors, activation energies, and reaction orders for carbon
monoxide and hydrogen were reported by Moazami et al. (2015a) and are summarized in
Table 8.21.

8.4.3.1 Experimental Data
The simulations with the proposed model are compared with experimental data reported by
Moazami et al. (2015a). They used a bench-scale fixed-bed reactor with a length of 52.83 cm, a cat-
alytic bed of 11 cm length, and an internal diameter of 15.7 mm. The experiments were carried out
using a Co/SiO2 catalyst, which was diluted with 2 g of silicon carbide at a mass ratio of 1:12 (Co/
SiO2/SiC). A simulated nitrogen-rich syngas was used with a concentration of 17, 33, and 50 vol % of
CO, H2, and N2 respectively.
8.4.3.2 Conversion of CO and H2

Table 8.22 reports the operating conditions of 6 experimental cases used for the simulations of
the proposed reactor model. Two temperatures (503 and 528 K), four pressures (1.0, 1.5, 2.0,
−1 −1
and 2.5 MPa), and four GHSVs (1800, 2400, 3000, and 3600 NmLgcat h ) were used for the vali-
dation of the model. Figures 8.22 and 8.23 show the simulated CO conversion profiles for 6 cases as
a function of time at the exit of the reactor and along the catalytic fixed-bed at the steady-state con-
ditions, respectively. Figure 8.22 shows the dynamic profile CO conversion for the 6 experimental
cases considered. For cases I, II, V, and VI, the conversions are practically 100%. For cases V and VI,
the CO conversion reaches the steady state about half the time used for the simulations (180 seg).
Table 8.21 Experimental data of the kinetic parameters of the FTS reaction reported in the literature (Adapted from Moazami, et. al. (2015a)).
Reaction R1 R2 R3 R4 R5 R6 R7 R8
Kinetic
parameters A1 Ea1 A2 Ea2 A3 Ea3 A4 Ea4 A5 Ea5 A6 Ea6 A7 Ea7 A8 Ea8
65.56165 83523.9 0.045438 65017 0.001144 49782 1.19E−06 34885.5 1.84E−10 27728.9 6.59E−09 25730.1 1.98E−08 23564.3 3.78E−05 58826.3
Reaction m1 n1 m2 n2 m3 n3 m4 n4 m5 n5 m6 n6 m7 n7 m8 n8
order
Case 1 0.6095 −0.4679 0.3733 −0.2613 −0.0283 0.1913 0.4354 0.0876 0.0353 1.1163 −0.2515 0.0092 0.8187 0.0342 −0.362 1.2639
Case 2 1.295 −1.0527 −0.4637 0.4007 1.2866 −1.0054 0.6653 −0.0195 0.1816 0.8522 0.1657 0.7574 1.3976 −0.5635 1.4825 −0.6426
Case 3 0.6142 −0.3101 −0.6284 0.2391 1.2625 −0.9602 0.9859 −0.3488 0.2189 0.8523 0.336 0.5542 1.2153 −0.3777 1.4034 0.6075
Case 4 −0.4755 0.8201 0.014 0.0746 −0.0594 0.4128 −0.7131 1.4167 0.1037 −0.4789 −0.2626 1.1713 1.2895 0.4571 0.472 0.7706
Case 5 0.4831 −0.2516 −1.2039 1.395 −0.1382 0.4795 0.1607 0.5001 −0.0166 1.0945 0.3423 0.5573 −0.5105 1.4353 −0.4506 1.4841
Case 6 0.9771 −0.8641 −0.1176 −1.1396 −0.2278 0.5233 1.1815 −0.6838 0.278 0.8884 −0.9539 0.1122 1.0146 −0.1848 −0.4465 1.465
Table 8.22 Experimental conditions for fixed-bed FTS reactor (Adapted from Moazami, et. al. 2015a).
Case Temperature (K) Pressure (Mpa) GHSV (NmL/gcat/h)
I 503 1.0 1800

II 503 1.5 2400
III 503 2.0 3000
IV 503 2.5 3600
V 518 1.0 2400
VI 518 1.5 1800
(a) (b)
100 100
CO conversion, xCO(%)
50 50
T = 503 K T = 503 K
P = 1.0 MPa P = 1.5 MPa
GHSV = 1800 NmL/(gcath) GHSV = 2400 NmL/(gcath)
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Time, t(s) Time, t(s)
(c) (d)
100 100
50 50
T = 503 K T = 503 K
P = 2.0 MPa P = 2.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(e) (f)
100 100
50 50
T = 518 K T = 518 K
P = 1.0 MPa P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Figure 8.22 Simulated dynamic CO conversion profiles at the reactor outlet. (a) case I; (b) case II; (c) case III;
(d) case IV; (e) case V; (f ) case VI.
(a) (b)
100 100
CO conversion, (%)
CO conversion, (%)
50 T = 503 K 50 T = 503 K
P = 1.0 MPa P = 1.5 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Catalytic bed length, z(m) Catalytic bed length, z(m)
(c) (d)
100 60
CO conversion, (%)
CO conversion, (%)
40
50 T = 503 K
T = 503 K
20 P = 2.5 MPa
P = 2.0 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(e) (f)
100 100
CO conversion, (%)
CO conversion, (%)
50 50
T = 518 K T = 518 K
P = 1.0 MPa P = 1.5 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Figure 8.23 Simulated CO conversion profiles in axial position. (○) experimental data; (-) proposed model; (--)
literature model [15]; (a) case I; (b) case II; (c) case III; (d) case IV; (e) case V; (f ) case VI.
For cases III and IV, the CO conversion is close to 95%. In general, the results of the CO conversion
dynamic profile indicate that for relatively short times the conversions are high. The differences in
CO conversion shown in all the experimental cases under study are mainly due to changes in total
pressure and GHSV. The difference in CO conversion is mostly seen in cases III and IV. In case III, a
decrease of about 98% is observed in comparison to the first two experimental cases. This is due to
the effect of increasing the gas space velocity. This is expected because an increase in the gas space
velocity translates into a decrease in residence time, which results in a decrease in CO conversion
(Kwack et al. 2011a). However, the CO conversion in cases V and VI increases considerably until
reaching a maximum conversion as the simulation time is extended. Moreover, it is important to
note that an increase in total pressure corresponds to a slight increase in CO conversion as reported
in the literature (van Berge and Everson 1997; Yan et al. 2009). In addition, the effect of a temper-
ature increase of 503 to 518 K also leads to an improvement in the synthesis gas conversion in terms
of an appreciable increase (Dry 1981).
Figure 8.23 shows the simulated CO conversion profiles along the catalytic bed within the reactor
at steady-state conditions, their comparison with experimental data, and the results obtained with
another model reported in the literature (Moazami et al. 2015a). It is important to mention that the
experimental data given for the syngas conversion (CO and H2) are only reported at the reactor
outlet.
The simulated CO conversion profiles with the proposed model and with the literature
model (Moazami et al. 2015a) are quite similar for the first 4 cases (Figures 8.23a-8.23d). How-
ever, for cases V and VI (Figures 8.23e and 8.23f ), the profiles are different, particularly for
case VI.
The lower CO conversions predicted by the literature model (Moazami et al. 2015a) are due to
the isothermal conditions used by the authors, while in the proposed model the non-isothermal
conditions are considered.
The highest difference in predictions with both models is found at the reactor inlet, where the
catalytic activity is greater than in any other part of the reactor (van Berge and Everson 1997). On
the other hand, the CO conversion is higher for case II; however, for cases III and IV, it is even
lower. This is due to the combined effect of the simultaneous changes in the total pressure and
GHSV. It has been confirmed experimentally that the increase in total pressure generates an
increase in the syngas conversion (van Berge and Everson 1997; Yan et al. 2009), but if the
gas space velocity is increased, this causes a decrease in the residence time and decrease in
the CO conversion (Kwack et al. 2011a). However, it is important to note that care should be
taken with the simultaneous changes in total pressure and gas space velocity because high pres-
sures generate a shift of the FTS process toward the production of heavy hydrocarbons (Dry 1981).
Increasing the gas space velocity can cause a slight decrease in CO conversion, but if adequate
values are established, this allows for high production rates without altering the quality of the
product (Yan et al. 2009).
The errors between the experimental and predicted values at the reactor outlet were calculated
with Eq. (8.130). Table 8.23 summarizes the calculated errors for the proposed model and the lit-
erature model (Moazami et al. 2015a) for the conversion of CO, H2, and product’s selectivity (CO2,
CH4, light gases: C2–C4, and heavy hydrocarbons: C5+) for the 6 cases studied. In general, lower
errors are obtained with the proposed model.
yi,exp − yi,calc
error = × 100; where i = specie 8 130
yi,exp
Figures 8.24 and 8.25 illustrate the hydrogen conversion profiles as a function of time at the reac-
tor outlet and along the catalytic bed length at steady-state conditions for the 6 cases given in
Table 8.22. Both figures present the comparison between the simulations with the proposed and
literature models and the experimental data given only at the reactor outlet.
Only for case V (Figure 8.24e), the hydrogen conversion reaches the steady state at 180
seconds. For the rest of the experimental cases, longer time is required, but at 180 seconds
the hydrogen conversions are relatively high (above 90%), which is beneficial for the FTS
process since the formation of long-chain liquid hydrocarbons is enhanced (Dry 1981). The
hydrogen conversion profile along the catalytic bed length of the FTS reactor (axial direction)
is illustrated in Figure 8.25. In general, the proposed dynamic model predicts a profile similar to
the literature steady-state model.
For case VI (Figure 8.25f ), the literature model (Moazami et al. 2015a) tends to predict higher
H2 conversion. It should be noted that no experimental data on hydrogen conversion are
available at the reactor outlet, so the accuracy with respect to hydrogen conversion cannot
be verified.
Table 8.23 Errors between experimental and predicted values with the proposed model (PM) and the
literature model (LM) (Adapted from Moazami, et. al. 2015a).
Case I Case II Case III
Species % error (PM) % error (LM) % error (PM) % error (LM) % error (PM) % error (LM)
CO 1.9 0.87 1.2 0.92 1.6 0.83

CO2 0.09 0.11 0.42 0.20 0.21 0.15
CH4 0.62 0.59 0.56 0.48 0.33 0.47
C2 0.88 0.69 0.57 0.48 0.35 0.33
C 3H 8 0.23 0.40 0.17 0.21 0.11 0.07
C4 0.41 0.33 0.09 0.03 0.08 0.04
C5+ 0.08 0.06 0.17 0.12 0.09 0.12
Case IV Case V Case VI
Species % error (PM) % error (LM) % error (PM) % error (LM) % error (PM) % error (LM)
CO 1.9 0.5 0.05 0.01 0.30 6.0

CO2 0.05 0 0.04 0 0.23 0.07
CH4 0.09 0 0.06 0 0.09 0
C2 1.2 0 0.12 0.04 0.04 0.09
C 3H 8 1.0 0 0.10 0.06 0.03 0
C4 1.4 0 0.09 0.11 026 0.16
C5+ 0.08 0.01 0.21 0.01 0.07 0.02
8.4.3.3 Temperature Profiles

Figures 8.26 and 8.27 illustrate the dynamic and axial temperature profiles at the reactor outlet as a
function of time and along the axial direction of the reactor, respectively. No comparison is made with
the literature model (Moazami et al. 2015a) since it simulates the FTS reactor in isothermal mode. It is
relevant to note that important variations in temperature profiles are observed, thus considering the
reactor as isothermal is totally inadequate, which is due to the highly exothermic reactions that occur
during FTS. In addition, using fixed-bed FTS reactor under isothermal conditions is: not feasible to be
implemented in scales of greater extension (semi-industrial and industrial).
Figures 8.26a, 8.26e, 8.26f and 8.27a, 8.27e, 8.27f show the temperature reactor profiles for experi-
ments I, V, and VI. In these cases, the temperature quickly drops from its initial value to the cooling
jacket temperature value. It should be noted that the boiling point of water (cooling agent) was
established at a constant value of 498 K for the 6 cases. The temperature was established at this
value since the range suggested experimentally is between 5 and 10 K below the initial temperature
(Dai et al. 2014) to avoid as much as possible the hot spots formation (Rahimpour and Elekaei 2009).
In addition, the use of high temperatures in fixed-bed reactors where a highly exothermic reaction
is carried out, such as the FTS, is highly prone to fall into a thermal runaway problem (Kaskes et al.
2016; Yan et al. 2009; Rafiq et al. 2011). Due to this, it is advisable to have good temperature control
of the FTS reactor, since poor control of it can cause serious problems in the FTS process, such as
catalytic deactivation and poor production of desired products (Lee and Chung 2012). It has been
demonstrated that the temperature directly affects the syngas conversion and, therefore, the pro-
duct’s distribution (effects on the hydrocarbon chain length) (Park et al. 2014).
(a) (b)
100 100
H2 conversion, xH (%)
2
2
50 T = 503 K 50 T = 503 K
P = 1.0 MPa P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(c) (d)
100 100
2
2
50 T = 503 K 50 T = 503 K
P = 2.0 MPa P = 2.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(e) (f)
100 100
2
50 T = 518 K 50 T = 518 K
P = 1.0 MPa P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Figure 8.24 Simulated dynamic H2 conversion profiles at the reactor outlet. (a) case I; (b) case II; (c) case III;
(d) case IV; (e) case V; (f ) case VI.
On the other hand, if the heat released in a chemical FTS reaction can be adequately removed,
then the temperature runaway can be avoided so that a higher hydrogen-to-carbon monoxide feed
ratio can be operated safely for a higher conversion. However, too much heat removal makes the
reactor stay at relatively lower temperatures, in which the FTS reaction also cannot be activated
(Park et al. 2014).
The temperature profile observed in Figures 8.26d and 8.27d, corresponding to case IV, shows a
slight hot spot near the reactor inlet. This temperature profile is a good result of the operating con-
ditions combination used for this case and the temperature of the cooling jacket used because it
almost eliminates the presence of the hot spot.
However, in Figures 8.26b, 8.26c and 8.27b, 8.27c, corresponding to cases II and III, respectively,
the presence of a hot spot near the reactor inlet is quite appreciated. This may be due to the negative
effect of the simultaneous increase in the total pressure and gas hourly space velocity. The hot spot
increases by 4 and 2 degrees for the operating conditions used in cases II and III, respectively.
(a) (b)
100 100
2
2
50 50 T = 503 K
T = 503 K
P = 1.0 MPa P = 1.5 MPa

0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(c) (d)
100 100
T = 503 K
2
2
P = 2.5 MPa
GHSV = 3600 NmL/(gcath)
50 50
T = 503 K
P = 2.0 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(e) (f)
100 100
2
50 50
T = 518 K T = 518 K
P = 1.0 MPa P = 1.5 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Figure 8.25 Dynamic-state H2 conversion profile in axial position. (-) proposed model; (--) literature
model literature model (dotted line) (Moazami et al. 2015a). (a) case I; (b) case II; (c) case III; (d) case IV;
(e) case V; (f ) case VI.
However, this does not happen in the case of Figure 8.26d since apparently the considerable
increase in gas space velocity appreciably reduces the residence time within the reactor (Kwack
et al. 2011a), which favors the FTS reaction in terms of practically negligible hot spot formation.
In fact, the presence of hot spots in the vicinity of the reactor inlet can cause the thermal runaway
phenomenon to occur. It is important to mention that the location (distance) of the hot spot with
respect to the reactor inlet depends on the feed and cooling temperatures (Rahimpour and Elekaei
2009; Dimitriadis and Pistikopoulos 1995).
Therefore, thermal management is important in an exothermic catalytic reaction to avoid unde-
sirable thermal effects such as the presence of high hot spots and the thermal runaway problem
(Lee and Chung 2012).
A solution that is mentioned in the literature to attenuate the presence of hot spots is by elim-
inating excess heat that is released from reactors in which highly exothermic reactions occur by
increasing the heat transfer coefficient or increasing the temperature gradient using lower cooling
temperatures (Lee and Chung 2012).
(a) (b)
520 520
T = 503 K T = 503 K
Temperature, (K)
Temperature, (K)
515 P = 1.5 MPa 515 P = 1.5 MPa
510
510
505
505
500
500
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(c) (d)
520 520
T = 503 K T = 503 K
Temperature, (K)
Temperature, (K)
P = 2.0 MPa 515 P = 2.5 MPa
515
510
510
505
505
500
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(e) (f)
520 520
T = 518 K T = 518 K
Temperature, (K)
Temperature, (K)
P = 1.0 MPa P = 1.5 MPa

510 GHSV = 2400 NmL/(gcath) 510 GHSV = 1800 NmL/(gcath)
500 500
490 490
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Figure 8.26 Dynamic-state temperature profile, (a) case I; (b) case II; (c) case III; (d) case IV; (e) case
V; (f ) case VI.
It is important to highlight that the thermal behavior of fixed-bed FTS reactors is only possible to
visualize when using a dynamic model. In addition, it is extremely important to improve the design
and optimization aspects of fixed-bed reactors in order to verify the feasibility and flexibility of the
process under a broad set of operating conditions and uncertainties in the reactor parameters, thus
guaranteeing the thermal stability of the process (Dimitriadis and Pistikopoulos 1995; Soroush and
Kravaris 1993a; Soroush and Kravaris 1993b).
8.4.4 Product’s selectivity

Figure 8.28 shows the CO2 selectivity profile as a function of time at the reactor outlet, and
Figure 8.29 shows the CO2 selectivity profile along the catalytic bed at unsteady-state conditions.
It can be seen that for experimental cases I–IV, corresponding to Figures 8.28a–8.28d and
Figure 8.29a–8.29d, the CO2 selectivity is relatively low, between 4% and 5%.
(a) (b)
520 520
T = 503 K T = 503 K
Temperature, (K)
Temperature, (K)
P = 1.0 MPa 515 P = 1.5 MPa
510
510
505
500
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(c) (d)
520 520
T = 503 K T = 503 K
Temperature, (K)
Temperature, (K)
515 P = 2.0 MPa 515 P = 2.5 MPa
510 510
505 505
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(e) (f)
520 520
T = 518 K T = 518 K
Temperature, (K)
Temperature, (K)
P = 1.0 MPa P = 1.5 MPa

510 510
500 500
490 490
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Figure 8.27 Dynamic temperature profile in axial position. (a) case I; (b) case II; (c) case III; (d) case IV;
However, for cases V and VI (Figures 8.28e, 8.29e and 8.28f, 8.29f, respectively), the CO2 selec-
tivity variation ranges between 14 and 15%, which is relatively high compared with the results
given in cases I–IV. This is due to the simultaneous changes in initial temperature, total pressure,
and gas space velocity. The initial temperatures for cases V and VI are high, which favors a higher
CO2 selectivity. It is worth mentioning that CO2 is an undesirable product for the FTS process so
high selectivities should be avoided, which leads to the proper use of feed inlet temperatures.
Therefore, the set of operating conditions given in cases I to IV provides an ideal amount of
CO2 produced.
According to the calculated errors given in Table 8.23, a good fit of the proposed model with the
experimental data is observed. The error for CO2 selectivity with the literature model cannot be
calculated because the authors reported only the selectivity data of FTS products at the reactor out-
let. That is why the simulations of the FTS product selectivity profiles with the literature model are
not shown.
(a) (b)
20 20
T = 503 K T = 503 K
CO2 selectivity (%)
CO2 selectivity (%)

15 P = 1.0 MPa 15 P = 1.5 MPa
10 10
5 5
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(c) (d)
20 20
T = 503 K T = 503 K
CO2 selectivity (%)
CO2 selectivity (%)

15 P = 2.0 MPa 15 P = 2.5 MPa
10 10
5 5
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(e) (f)
20 20
CO2 selectivity (%)
CO2 selectivity (%)
15 15
10 10
T = 518 K T = 518 K
5 P = 1.0 MPa 5 P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Figure 8.28 Dynamic-state CO2 selectivity. (a) case I; (b) case II; (c) case III; (d) case IV; (e) case V; (f ) case VI.
Figures 8.30 and 8.31 present the results of the proposed model simulations for the methane selec-
tivity profile as a time function and bed length, respectively. For the dynamic case of methane selec-
tivity, Figures 8.30b–8.30e (cases II–V, respectively) show that methane selectivity is high with
values of 26, 34, 30, and 31%, respectively. Similar to CO2, care should be taken in the FTS process
with high methane productivity since it is also considered an undesirable product.
The high values of CH4 selectivity are due to the increase in total pressure and gas space velocity
since the latter translates into a decrease in the residence time in the reactor.
The use of high values of gas space velocity leads to an increase in the production of light gases,
especially methane, since there is a short time to favor the re-adsorption of olefins for the growth of
the hydrocarbon chain, thus giving rise to the formation of high-molecular-weight liquid hydrocar-
bons (Bechara et al. 2001).
It is observed that an increase in gas space velocity induces a considerable decrease in the pro-
duction of liquid hydrocarbons. However, the increased total gas pressure in steps of 0.5 MPa does
not allow the reduction of the liquid hydrocarbons selectivity to be so drastic. In addition, the light
(a) (b)
20 20
T = 503 K T = 503 K
CO2 selectivity (%)

CO2 selectivity (%)
15 P = 1.0 MPa 15 P = 1.5 MPa

10 10
5 5
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(c) (d)
20 20
T = 503 K T = 503 K
CO2 selectivity (%)
CO2 selectivity (%)

15 P = 2.0 MPa 15 P = 2.5 MPa
10 10
5 5
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(e) (f)
20 20
CO2 selectivity (%)
CO2 selectivity (%)
15 15
10 10
T = 518 K T = 518 K
5 P = 1.0 MPa 5 P = 1.5 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Figure 8.29 Dynamic-state CO2 selectivity profile in axial position. (○) experimental data; (-) proposed
model; (a) case I; (b) case II; (c) case III; (d) case IV; (e) case V; (f ) case VI.
gases selectivity does not undergo considerable changes, which is beneficial when an FTS process
with high liquid hydrocarbon productivities is desired. Kwack et al. (2011a) corroborated that an
increase in total pressure leads to a slight improvement in the production of heavy hydrocarbons.
Also, recently Moazami et al. (2017b)found that high pressures favor an increment in the liquid
hydrocarbons selectivity, which was not observed in the first four study cases due to the simulta-
neous increase in gas space velocity (i.e., a decrease in residence time), and there is a low diminish
in the concentration of reactants (CO and H2), which in turn results in a decrease in the long-chain
liquid hydrocarbons selectivity.
Under typical reaction conditions, a thin layer of liquid hydrocarbon covers the catalyst particles.
As a result, before the reactants reach the catalyst surface, they have to diffuse inside this layer,
while reaction products have to do the same in the opposite direction before being desorbed. It
(a) (b)
40 40
T = 503 K
CH4 selectivity
CH4 selectivity
30 P = 1.0 MPa 30

20 20
T = 503 K
10 10 P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(c) (d)
40 40
CH4 selectivity
CH4 selectivity
30 30
20 20
T = 503 K T = 503 K
10 P = 2.0 MPa 10 P = 2.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(e) (f)
40 40
T = 518 K
CH4 selectivity
CH4 selectivity
30 30 P = 1.5 MPa
20 20 GHSV = 1800 NmL/(gcath)

T = 518 K
P = 1.0 MPa
10 10
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Figure 8.30 Dynamic-state CH4 selectivity profile. (a) experimental case I; (b) experimental case II;
(c) experimental case III; (d) experimental case IV; (e) experimental case V; (f ) experimental case VI.
is known that olefins, in contrast to paraffins, can be readsorbed on the active sites, reinserted in the
chain growth process, or can be hydrogenated to the corresponding paraffins (Moazami et al. 2017b;
Kuipers et al. 1996; van der Laan and Beenackers 1999). However, in experimental case I -
(Figure 8.30a), the methane selectivity profile over time is favorable since the selectivity achieved
is low, while in experimental case VI (Figure 8.30f ), methane selectivity at the reactor outlet is mod-
erate since intermediate selectivity is layered, but even so, it is always desirable to decrease the
methane production by establishing an adequate set of operating conditions. In Figure 8.31, the
methane selectivity profiles are illustrated along the catalytic bed length. In general, the selectivities
are not so high. Cases II, III, IV, and V report selectivities close to 20% (see Figure 8.31b–8.31e).
According to Table 8.23, the error between the predictions with the proposed model and the exper-
imental data is relatively small. However, the steady-state literature model presents an error a little
bit lower compared with the proposed model. In general, the light gases selectivities (C2, C3, and C4)
are relatively low for all the cases, as can be seen in Figures 8.32–8.37.
Kwack et al. (2011a) reported that the variation in the total pressure and gas space velocity con-
siderably affects the light gases selectivity and the product’s total distribution. They varied the gas
space velocity in an order of 600 units as compared with the experiments carried out by Moazami
et al. (2015a). Such a variation in said parameter does not affect the product’s selectivity in a
(a) (b)
40 40
T = 503 K T = 503 K
CH4 selectivity (%)
CH4 selectivity (%)

30 P = 1.0 MPa 30 P = 1.5 MPa
20 20
10 10
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(c) (d)
40 40
T = 503 K T = 503 K
CH4 selectivity (%)
CH4 selectivity (%)

30 P = 2.0 MPa 30 P = 2.5 MPa
20 20
10 10
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(e) (f)
40 100
T = 518 K T = 518 K
CH4 selectivity (%)
CH4 selectivity (%)
30 P = 1.0 MPa P = 1.0 MPa

20 50
10
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Figure 8.31 Dynamic-state CH4 selectivity profile in axial position. (○) experimental data; (-) proposed
relevant manner. For example, the variation of light gases (C3–C4), except for C1 and C2 (for which
it is about 1.3%), represents a change of approximately 2% from one case to another.
The proposed model predicts fairly accurately the experimental data given at the reactor outlet.
In the FTS process, it is important to reduce, as much as possible, the light gases production and
increase the selectivity of heavy hydrocarbons.
Kwack et al. (2011a) reported that the variation in the total pressure and gas space velocity con-
siderably affects the light gases selectivity and the product’s total distribution. They varied the gas
space velocity in an order of 600 units as compared with the experiments carried out by Moazami
et al. (2015a). Such a variation in said parameter does not affect the product’s selectivity in a rel-
evant manner. For example, the variation of light gases (C3-C4), except for C1 and C2 (for which it is
about 1.3%), represents a change of approximately 2% from one case to another.
On the other hand, the liquid hydrocarbons selectivity (C5+) presents a considerable variation
due to the simultaneous effects of temperature, total pressure, and the gas space velocity.
Figures 8.38 and 8.39 show the product’s selectivity as a function of time and along the axial
direction. The lowest productivity is presented in cases II, III, IV, and V (Figures 8.38b–8.38e
(a) (b)
4 4
T = 503 K
P = 1.0 MPa
C2 selectivity (%)
3 3
C2 selectivity (%)

2 2
T = 503 K
1 1 P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(c) (d)
4 4
C2 selectivity (%)
3 3
C2 selectivity (%)
2 2
T = 503 K T = 503 K
1 P = 2.0 MPa 1 P = 2.5 MPa

0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(e) (f)
4 4
T = 518 K
P = 1.5 MPa
C2 selectivity (%)
3 3
C2 selectivity (%)

2 2
T = 518 K
1 P = 1.0 MPa 1
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Figure 8.32 Dynamic-state C2 species selectivity profile. (a) case I; (b) case II; (c) case III; (d) case IV;
and 8.39b–8.39e), in which the selectivity toward C5+ hydrocarbons is around 60–64%. This
decrease is mainly because of gas space velocity. The use of high gas space velocities helps maintain
an adequate heat transfer due to the short residence time of the syngas inside the reactor, which
agrees with the results reported by Rahimpour and Elekaei (2009).
But for the purpose of increasing the CO conversion and improving the product’s selectivity, the
variations in the gas space velocity do not have a positive impact on the productivity of the FTS
process.
In summary, the best set of operating conditions that allows high conversions of the syngas (CO
and H2) and improvements in the product’s selectivity in terms of low methane and other light
gases (C2–C4) production and high production of C5+ hydrocarbons are cases I and VI. In these
cases, the inlet temperature of the reactor quickly approaches the temperature of the cooling water
without giving rise to the formation of hot spots. The syngas conversion is high, around 80 and 100%
(a) (b)
3 3
T = 503 K T = 503 K
C2 selectivity (%)
C2 selectivity (%)
P = 1.0 MPa P = 1.0 MPa
2 2
1 1
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(c) (d)
3 3
T = 503 K T = 503 K
P = 2.0 MPa P = 2.5 MPa
C2 selectivity (%)
C2 selectivity (%)
2 2
1 1
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(e) (f)
3 3
T = 518 K
P = 1.5 MPa
C2 selectivity (%)
C2 selectivity (%)
2 2
T = 518 K
1 1
P = 1.0 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Figure 8.33 Dynamic-state C2 species selectivity profile in axial position. (○) experimental data; (-) proposed
for the CO, and around 85 and 90% for the H2, respectively. In addition, the methane and light gases
selectivities are relatively low, while the heavy hydrocarbons selectivity (C5+) is approximately
82 and 80%, respectively, which is favorable.
8.4.5 Final Remarks

The proposed dynamic model for the FTS reactor has been successfully validated with six experi-
mental cases reported in the literature, which considered simultaneous variations in temperature,
total pressure, and syngas space velocity at the inlet of the reactor. The reactor dynamics indicates
that for most of the operating conditions in each case, the reactor reaches the steady-state condition,
high syngas conversions, and improvements in the product’s selectivity. The thermal behavior of
(a) (b)
6 6
T = 503 K
C3H8 selectivity (%)

P = 1.0 MPa
4 GHSV = 1800 NmL/(gcath) 4
2 2 T = 503 K
P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(c) (d)
6 6

4 4
2 T = 503 K 2 T = 503 K
P = 2.0 MPa P = 2.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(e) (f)
6 6
T = 518 K
P = 1.5 MPa
4 4 GHSV = 1800 NmL/(gcath)
2 T = 518 K 2
P = 1.0 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Figure 8.34 Dynamic-state C3H8 selectivity profile. (a) case I; (b) case II; (c) case III; (d) case IV; (e) case V;
(f ) case VI.
the reactor is properly evaluated from the consideration of a non-isothermal regime, whereby the
generation of hot spots is analyzed, to prevent a negative effect in terms of the thermal runaway
formation or low catalytic activity. With the proposed model, it was confirmed that the best set
of operating conditions that allow high syngas conversions (CO and H2), high C5+ selectivity,
low production of CH4, and the absence of hot-spot formation (improvements in the thermal behav-
ior of the FTS reactor) correspond to 503 and 518 K, 1.0 and 1.5 MPa, and 1800 (NmL/(gcat h)) of
temperature, pressure, and gas space velocity, respectively. The proposed dynamic model guaran-
tees the viability of its application for scaling up purposes (semi-industrial and industrial scales),
design of control schemes, and dynamic optimization of the FTS process.
(a) (b)
4 4
T = 503 K

3 P = 1.0 MPa 3
2 2
T = 503 K
1 1 P = 1.5 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(c) (d)
4 4

3 3
2 2
T = 503 K T = 503 K
1 P = 2.0 MPa 1 P = 2.5 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(e) (f)
4 4
T = 518 K
3 3 P = 1.5 MPa
2 2
T = 518 K
1 P = 1.0 MPa 1

0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Figure 8.35 Dynamic-state C3H8 selectivity profile in axial position. (○) experimental data; (-) proposed
8.5 Modeling and Control of a Fischer–Tropsch Synthesis Reactor

with a Novel Mechanistic Kinetic Approach
8.5.1 Introduction
A modeling and dynamic control study for an FTS fixed-bed reactor under the one-dimensional
pseudohomogeneous scheme representation with integration of the cooling jacket dynamic
approach has been developed. The proposed reactor model contemplates the use of a novel
8.5 Modeling and Control of a Fischer–Tropsch Synthesis Reactor with a Novel Mechanistic Kinetic Approach 391
(a) (b)
4 4
T = 503 K
P = 1.0 MPa
C4 selectivity (%)
C4 selectivity (%)
3 3
2 2
T = 503 K
1 1 P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(c) (d)
4 4
C4 selectivity (%)
C4 selectivity (%)
3 3
2 2
T = 503 K T = 503 K
1 P = 2.0 MPa 1 P = 2.5 MPa

0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(e) (f)
4 4
T = 518 K
P = 1.5 MPa
C4 selectivity (%)
C4 selectivity (%)
3 3
2 2
T = 518 K
1 P = 1.0 MPa 1
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Figure 8.36 Dynamic-state C4 species selectivity profile. (a) case I; (b) case II; (c) case III; (d) case IV;
mechanistic kinetic model based on the Langmuir theory recently reported in the literature. To
ensure good thermal behavior and improved reactor performance, a proportional–integral control-
ler (PIC) is implemented in the cooling jacket. The robustness of the controller, the dynamic model,
and the mechanistic kinetic approach are evaluated and validated against the use of four different
sets of initial operating conditions of temperature, pressure, and gas hourly space velocity. The
results indicate the reactor model potential under the PIC action since it allows to completely atten-
uate the typical hot-spot formation in fixed-bed reactors in which highly exothermic reactions are
carried out. In addition, the mechanistic kinetic model represents to a great extent the FTS
reaction network in terms of the syngas conversion and the light gases and heavy liquid hydrocar-
bons selectivity.
(a) (b)
3 3
T = 503 K
P = 1.0 MPa
C4 selectivity (%)
C4 selectivity (%)
2 GHSV = 1800 NmL/(gcath) 2
T = 503 K
1 1
P = 1.5 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(c) (d)
3 3
C4 selectivity (%)
C4 selectivity (%)
2 2
T = 503 K T = 503 K
1 1
P = 2.0 MPa P = 2.5 MPa

0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(e) (f)
3 3
T = 518 K
C4 selectivity (%)
C4 selectivity (%)
P = 1.5 MPa
2 2
T = 518 K
1 1
P = 1.0 MPa
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Figure 8.37 Dynamic-state C4 species selectivity profile in axial position. (○) experimental data; (-) proposed
8.5.2 Formulation of the Model

8.5.2.1 Model Equations and Solution
The model to describe the dynamic behavior of the fixed-bed FTS reactor was implemented under
plug-flow regime at not-isothermal conditions, taking into account the variations of the superficial
gas velocity, the gas mixture density, and the total pressure of the system in the axial direction,
which have been demonstrated to be important parameters to improve predictions of syngas con-
version and temperature profiles. It also incorporates the dynamic model (energy balance) of the
cooling jacket (Méndez et al. 2022).
(a) (b)
100 100
C5+ selectivity (%)
C5+ selectivity (%)

50 50
T = 503 K T = 503 K
P = 1.0 MPa P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(c) (d)
100 100
C5+ selectivity (%)
C5+ selectivity (%)
50 50
T = 503 K T = 503 K
P = 2.0 MPa P = 2.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
(e) (f)
100 100
C5+ selectivity (%)
C5+ selectivity (%)
50 50
T = 518 K T = 518 K
P = 1.0 MPa P = 1.5 MPa
0 0
0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 120 140 160 180
Figure 8.38 Dynamic-state C5+ species selectivity profile. (a) case I; (b) case II; (c) case III; (d) case IV;
The dynamic model is given by the set of Eqs. (8.114)–(8.118) that describe the behavior of the
fixed-bed FTS reactor with the boundary conditions given by Eqs. (8.119) and (8.120). The dynamic
cooling jacket model is represented by the following equation:
∂T w F w,0 ∂T w
C P w ρw V c = − C P w ρw V c + F w,0 C w ρw T − T w + U r Ac T − T w 8 131
∂t Ac ∂z
with the boundary condition:
T w 0, z = T w t, 0 = T w,0 8 132
(a) (b)
100 100
C5+ selectivity (%)
C5+ selectivity (%)

50 50
T = 503 K T = 503 K
P = 1.0 MPa P = 1.5 MPa

0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(c) (d)
100 100
C5+ selectivity (%)
C5+ selectivity (%)

50 50
T = 503 K T = 503 K
P = 2.0 MPa P = 2.5 MPa

0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(e) (f)
100 100
C5+ selectivity (%)
C5+ selectivity (%)
50 50
T = 518 K T = 518 K
P = 1.0 MPa P = 1.5 MPa

0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Figure 8.39 Dynamic-state C5+ species selectivity profile in axial position. (○) experimental data; (-) proposed
CPw , ρw is the heat capacity of water whose value is 4.6 KJ/(KgK) and the ρw is the water density.
The equations of the proposed model are solved by using the orthogonal collocation numerical
method. For approaching to the solution of the model, 7 collocation points were used. The
Legendre orthogonal polynomials were used in the estimation of each collocation point
(Constantinides and Mostoufi 1999). Python software was used for the development of numerical
simulations.
8.5.2.2 Model Parameters and Correlations

The parameters and correlations used in the reactor model are given in Table 8.19. The
global heat transfer coefficient was estimated with the correlation proposed by Haarlemmer
and Bensabath (2016). This parameter is given as a function of the convective heat transfer
coefficient on the internal side of the reactor (hint), the convective heat transfer coefficient of
the external side (hext), and the reactor thermal conductivity (λw), which in this case is stain-
less steel.
The pressure drop along the catalytic bed was estimated using a correlation recently reported in
the literature for the friction factor. Such a correlation was validated by comparing it with a large
number of correlations showing superior accuracy (Allen et al. 2013).
The heat conductivity of the steel wall reactor was assumed to be 60 Wm−1h−1 (Sharma et al.
2011), and the value of the heat transfer coefficient (film) of the external side corresponding to
boiling water coolant 1000 Wm−1h−1 (Haarlemmer and Bensabath 2016). Other parameters
were considered as constants, such as the hollow fraction of the bed and the active side volume
fraction of the catalyst. The other parameters of the fixed-bed FTS reactor are given in
Table 8.20.
8.5.2.3 The Mechanistic FTS Kinetic Model

The FTS reaction network is classified as a number of lumped reactions by means of the kinetic
characteristics of reaction molecules, consisting of 11 reacting components (i.e. CO, H2, CO2,
H2O, CH4, C2H4, C2H6, C3H8, i-C4H10, n-C4H10, and C5+). Based on this consideration, the reaction
scheme used in the present study is given by Eqs. (8.122)–(8.129).
The kinetic model given by Moazami et al. (2017a) was derived from the LHHW rate theory based
on a combination of alkyl/alkenyl mechanism for FTS reactions (for production of n-paraffins and
n-olefins), which consist of various elementary reaction step routes and carbon chain distribution
pathways (i.e. adsorption, initiation, propagation, and termination steps). The surface elementary
reaction steps are given by four categories (van der Laan and Beenackers 199; Moazami et al.
2017a): (i) adsorption of the reactants (molecular CO and H2 species) on the catalyst surface,
(ii) chain initiation step, (iii) chain growth (propagation) step, and (iv) chain termination and
desorption of the products step. The following expressions represent the Fischer–Trospch synthesis
and water-gas shift (WGS) reaction rates, respectively:
k 8 K 7 K 6 K 5 K 1 K 22 5 K 3 K 4 P0CO5 PH
1 25
2
RFT = 2
1 + K 1 PCO + K 2 P0H52 + K 1 K 2 K 3 PCO PH
05
2
+ K 1 K 2 K 3 K 4 PCO PH 2 + η1 + η2
where
K 1 K 21 5 K 3 K 4 K 5 K 6 K 1 K 21 5 K 3 K 4 K 5 K 6 K 7 0 5 0 75
η1 = P0CO5 PH
0 25
; η2 = PCO PH 2 8 133
K 1 K 22 5 K 3 K 4 K 5 K 6 K 7 k 8
2
k8
PCO PH 2 O PCO2
k WGS4 K W 1 K W 2 K W 3 K W 6 − k WGS − 4
PH 2 KW5
RWGS = 2
PH 2 PH O 0 5 PH 2 O PCO2
1 + K W 1 PCO + + KW2 KW3 KW6 2 + KW2 KW +
KW6 PH 2 6
PH 2 KW5
8 134
The reaction rate constants, adsorption coefficients, equilibrium constant, and Gibbs free
energy are calculated by using the following equations:
Ea,j
kj = k j,0 exp − 8 135
RT
ΔCads,i
K i = K i,0 exp − 8 136
RT
PCO2 PH 2
KP = 8 137
PCO PH 2 O
ΔG0R
ln K P = − 8 138
RT
products reactants
1
ΔG0R,298 15 = αi ΔG0f ,298 15 − αi ΔG0f ,298 15 8 139
αk i i
4861 49
ln K P = − 3 72 + − 6 90 × 10 − 3 T + 1 33 × 10 − 5 T 2 − 8 38 × 10 − 9 T 3
T 8 140
+ 1 25 × 10 − 12 T 4
The product selectivities were calculated by Eqs. (8.110)–(8.113).
8.5.3 Implementation of the PI Controller

To guarantee a viable operation of the reactor in terms of a safe thermal operation, it is necessary to
implement a control scheme in the cooling jacket. It is well-known that the FTS reaction is highly
exothermic, which requires good temperature control in any reactor configuration used (Kaskes
et al. 2016; Kwack et al. 2011a). The high temperatures of the FTS reaction give rise to the hot-spot
formation, which can lead to the presence of the runaway phenomenon (Kaskes et al. 2016a; Rafiq
et al. 2011).
The literature indicates that a PI controller is suitable in those cases that involve a highly non-
linear process and where it is intended to eliminate the static error (off-set) (Stephanopoulos 1985).
On the other hand, it is stated that a PI control is a form of continuous adjustment that originated in
the process industries and usually considered part of (automatic) engineering process control (Box
and Luceño 1997).
To ensure good temperature control of the fixed-bed FTS reactor, a PIC scheme is implemented in
the cooling jacket where the reactor temperature is controlled by adjusting the temperature of the
refrigerant inside the cooling jacket by manipulating the flow velocity of the latter. The PI controller
is given by the following expression:
t
KC
ζ t = KCe + edt 8 141
τI 0
where ζ(t) is the variable control (flow velocity of refrigerant), KC is the controller gain, τI is the reset
time, and e is the error given by the temperature difference of the set-point (system temperature;
513 K for the first three experimentation cases and 523 K for the fourth experimentation case) and
the exit temperature (system response: outlet temperature) as follows:
e = T sp − T 8 142
whose error is a constant.

The set-point established in the studied control problem becomes the reactor temperature of 503
K, and to achieve a negligible deviation in the process response, the flow velocity of the refrigerant is
manipulated. The controller parameters (controller gain KC and reset time τI) are determined by the
adjustment rules given by Tyreus and Luyben (1992) whose values were KC = 2.6 and τI = 0.25 min.
To verify the robustness of the PI controller, changes in total system pressure and gas space velocity
were considered as disturbances.

The model simulations are compared with experimental data reported by Moazami et al. (2017a).
The experiments were carried out using a Co/SiO2 catalyst, which was diluted with 2 g of silicon
carbide at a mass ratio of 1:12 (Co/SiO2/SiC). A simulated nitrogen-rich syngas was used with con-
centrations of 17, 33, and 50 vol% of CO, H2, and N2, respectively. The dimensions of the bench-
scale fixed-bed reactor used in this case study are a catalytic bed of 6 m in length and an internal
diameter of 13 mm (Dai et al. 2014). An important aspect in the modeling is the extrapolation of
laboratory-scale reactor dimensions given by Moazami et al. (2017a) for one-stage reactor dimen-
sions given by Dai et al. (2014), in order to verify the model robustness subject to changes made in
the reactor dimensions and in the initial operating conditions. In addition, it is always important to
try to scale up the FTS reactor size in order to verify its performance at higher scales (pilot, semi-
industrial, and industrial) and its economic feasibility.
8.5.4.2 Simulations of the Syngas Conversion, Light Gases, and Heavy Liquid Selectivity
Although it is known, it is advisable to use kinetic models of reduced complexity to develop math-
ematical models of FTS reactors (fixed-bed, fluidized-bed, and slurry bubble column) for design and
scaling-up purposes (Kwack et al. 2011b). However, it is also essential to have an integrated com-
prehensive kinetic model that allows us to accurately describe the FTS product’s distribution
(Kwack et al. 2011b; Mosayebi et al. 2016; Quian et al, 2013).
To validate the proposed model of the fixed-bed FTS reactor with the integration of a novel com-
prehensive LHHW kinetics, four sets of initial operating conditions of temperature, pressure, and
GHSV were used (Table 8.24). From these experimental data, the effect of the main variables (tem-
perature, pressure, and GHSV) can be studied and used for validation of the model.
Figure 8.40 shows the CO conversion profile as a function of time and along the catalytic bed. It is
important to point out that the changes in the initial operating conditions were made every 10 min
(stage), considering that the syngas initial concentration is the same in each stage.
For the four experimental cases, the simulations show that the maximum CO conversion is
reached after 1.3 min. That is, after this period, the reactor reaches a stable steady state. In the first
experimental case (Figure 8.40a), it can be seen that the CO conversion is around 99%. In the rest of
the cases (II, III, and IV), the CO conversion is close to 90%. The stable state shown for the CO
conversion allows to verify that the controller action causes the inside temperature of the FTS reac-
tor to be maintained in almost isothermal conditions. On the other hand, Figure 8.40b shows the
Table 8.24 Experimental cases used in the proposed model simulations.
Case Temperature (K) Pressure (Mpa) GHSV (NmL/gcat/h)
I 518 1.0 2400

II 518 1.5 1800
III 518 2.0 3600
IV 528 1.5 3600
(a)
100
80
I II III IV
60
40
20
0
0 500 1000 1500 2000
Time, t(seg)
(b)
100
I
80
IV
60 II
40
III
20
0
0 1 2 3 4 5 6
Catalytic bed length, z(m)
Figure 8.40 CO conversion profile. Proposed model (-), experimental data: case I (○), case II (◊), case III
(Δ), and case IV ( ). (a) CO conversion profile as a function of time, (b) CO conversion profile along
the catalytic bed.
CO conversion along the catalytic bed (axial direction) for the four case studies. In all the cases, the
model manages to predict the experimental data at the exit of the reactor, with a negligible devi-
ation. The maximum error was observed for cases I and III, which is not of considerable attention.
Figure 8.41 illustrates the H2 conversion profile as a function of time along the catalytic bed length
for the four experimental cases. For this profile, no experimental data were reported at the output of
the reactor, so there is no comparison with the values predicted by the proposed model.
Figure 8.41a shows the maximum H2 conversion that is reached at the output of the FTS reactor.
In the first three cases, it is slightly above 80%, while for case IV, the H2 conversion falls below 70%.
Figure 8.41b shows how the H2 conversion profile along the catalytic bed differs for the four cases
under study.
The high syngas conversions in terms of CO and H2 amounts consumed indicate the little influ-
ence of the water produced during the FTS reaction. This is because cobalt-based FTS catalysts tend
to have little affinity for water production, which in turn results in a non-negative effect on the
syngas conversion (CO and H2), since there is not enough water that reacts with the CO
(Alihellal and Chibane 2016). The water–gas shift reaction is spontaneous toward the side in which
the hydrogen reacts with the CO2.
Figure 8.42 shows the CO2 selectivity for the four experimental cases. All the cases show a low
CO2 production (selectivity < 15% for cases I and II, around 10% for case III, < 20% for case IV). This
low CO2 productivity is mainly due to the little effect of the water–gas shift reaction. That is, the
small amount of produced water slightly reacts with the CO to give CO2. This low CO2 selectivity is
beneficial for the FTS process because it is a greenhouse gas whose production is sought to reduce as
much as possible.
(a)
100
IV
80
H2 conversion, xH2 (%)
I II III
60
40
20
0
0 500 1000 1500 2000
(b) Time, t(seg)

100
II I
80
H2 conversion, xH2 (%)
60
IV
III
40
20
0
0 1 2 3 4 5 6
Figure 8.41 H2 conversion profile according to the proposed model. (a) H2 conversion profile as a function of
time, (b) H2 conversion profile along the catalytic bed.
The errors between the experimental and predicted values at the output of the reactor were cal-
culated with Eq. (8.130). Table 8.25 summarizes the calculated errors for the proposed model versus
the literature model (Moazami et al. 2017a), for the CO conversion and product selectivity (CO2,
CH4, light gases: C2, C4, and hydrocarbons: C5+) for the 4 cases.
The errors in the prediction of CO2 selectivity in the four cases are relatively low (1.2–5.1%), with
case II having the highest error.
It is well known that methane is an undesirable product in the low- or high-temperature FTS
process (Gideon Botes 2009; Rahimpour and Elekaei 2009). On the other hand, the FTS reaction
is a surface reaction in which the catalytic characteristics are the main cause of rising to high
amounts of methane (Nakhaei Pour and Modaresi 2016). This is because the initiating block of
the hydrocarbon chain growth process is the methyl building block or monomer (CH2), which,
depending on the catalytic and operating conditions, can give rise to methane production as the
main paraffin of the FTS process or chemically re-adsorbed on the catalyst surface to continue with
the process of forming short or long-chain olefins or paraffins (Nakhaei et al. 2013; Nakhaei et al.
2014a, 2014b).
Figure 8.43 shows the CH4 selectivities profile for the four experimental cases. In experimental
case II, methane production in terms of selectivity is around 10%, which is considered relatively
low. For cases I and III, the selectivity of methane does not account to more than 20%, which is
also not of considerable importance. However, for experimental case IV, methane production rises
(a)
20
15
CO2 selectivity (%)
10
I II III IV
5
0
0 5 10 15 20 25 30 35 40
Time, t(time)
(b)
20
IV
II
15
CO2 selectivity (%)
I
10
III
5
0
0 1 2 3 4 5 6
Figure 8.42 CO2 selectivity profile. Proposed model (-), experimental data: case I (○), case II (◊),
case III (Δ) and case IV ( ). (a) CO2 selectivity profile as a function of time, (b) CO2 selectivity profile along
the catalytic bed.
Table 8.25 Error between model predictions and experimental data given by Adapted from Moazami,
et. al. (2015a).
Error (%)
Component I II III IV
CO 7.25 0.50 7.58 2.92

H2 — — — —
CO2 1.22 5.12 2.21 1.87
CH4 1.38 10.0 9.10 3.92
C 2H 6 1.21 23.65 7.24 9.34
C 3H 8 10.90 34.30 0.11 6.19
C4H10 0.27 35.0 5.65 0.59
C5+ 0.0003 1.34 3.31 4.45
(a)
40
I II III IV
CH4 selectivity (%)
30
20
10
0
0 5 10 15 20 25 30 35 40
Time, t(min)
(b)
40
IV
CH4 selectivity (%)
30
I
20
III
10
II
0
0 1 2 3 4 5 6
Figure 8.43 CH4 selectivity profile. Proposed model (-), experimental data: case I (○), case II (◊), case III (Δ),
and case IV ( ). (a) CH4 conversion profile as a function of time, (b) CO conversion profile along the
catalytic bed.
to 40%. This is because the reactor operating temperature increases near 10 K, as reported in the
literature (Nakhaei Pour and Housaindokht 2014; Rahimpour et al. 2011b).
Figure 8.43b depicts the CH4 selectivity profile along the catalytic bed, which fits well with the
experimental data given at the reactor outlet. The highest error between the calculated and
the experimental values is observed for study cases II and III (~10%).
The dynamic profiles of light gases selectivity are shown in Figures 8.44a, 8.45a, and 8.46a, while
the profiles of the same gases are illustrated along the axial direction of the FTS reactor in
Figures 8.44b, 8.45b, and 8.46b, respectively.
It can be observed that the amount of light gases produced is relatively low (Figures 8.44a and 8.44b)
since in the worst case the selectivity reaches at most 5% (case IV). From a point of view of the FTS
process, it is advisable to have simultaneously a low selectivity toward light gases and methane, and at
the same time a high selectivity toward liquid hydrocarbons. It should be mentioned that case IV
reflects the worst scenario in terms of a high selectivity toward light gases (especially CO2 and
CH4) and a low selectivity toward liquid hydrocarbons (C5+). This effect is due to two main causes:
(i) effect of high temperatures: the CO conversion is sometimes favored at the expense of high pro-
duction of light gases, particularly methane, and more water generation (Dai et al. 2014; Rafiq et al.
2011; Bechara et al. 2001); (ii) effect of space gas velocity: a high value of the space gas velocity trans-
lates into a low residence time, which causes the gas to have little reaction time and polymerization of
long hydrocarbon (liquid hydrocarbon) chains (Rafiq et al. 2011; Dry 1983; Kuipers et al. 1996).
(a)
6
5 I II III IV
C2 selectivity (%)
0
0 5 10 15 20 25 30 35 40
Time, t(min)
(b)
5
IV
4
C2 selectivity (%)
I
3
2 III
1
II
0
0 1 2 3 4 5 6
Figure 8.44 C2H4–C2H6 selectivity profile. Proposed model (-), experimental data: case I (○), case II (◊),
case III (Δ), and case IV ( ). (a) C2H6 selectivity profile as a function of time, (b) C2H6 selectivity profile
along the catalytic bed.
Visconti et al. (2016) established that the light gaseous hydrocarbons selectivity decreases due to
the variation exerted in the gas pressure, the H2/CO feed ratio, and the gas space velocity. They
emphasized the GHSV effect on the FTS product’s selectivity and CO conversion, in which it is
verified that at high gas space velocity, the olefin selectivity increases because of the short residence
time in the reactor to give time for secondary re-adsorption of olefins to long-chain paraffins (liquid
hydrocarbons).
It can be seen in Figures 8.45b, 8.46b, and 8.47b that the light gases selectivity is well adjusted to
the experimental data reported at the FTS reactor outlet. The worst experimental case in which a
greater error is observed is case II. Such an error may be due to the thermal effect that is considered
in the reactor model (non-isothermal regime), the process dynamics, and especially in the reactor
parameters uncertainty (global heat transfer coefficient and catalyst active fraction) and kinetic
parameters.
Figures 8.47a and 8.47b show the selectivity profiles toward the liquid hydrocarbons (C5+) as a
function of the process dynamics and along the axial direction of the FTS reactor, respectively.
High temperatures are found to favor the formation of light hydrocarbons, while low tempera-
tures tend to promote the production of heavy hydrocarbons. That is, the C5+ selectivity tends to
decrease due to the increase in the inlet temperature. For case IV, the light gases selectivity
increases (Figures 8.43–8.46) which is negatively reflected in the liquid hydrocarbons selectivity
since it tends to decrease considerably.
(a)
6
5 I II III
4
IV
3
0
0 5 10 15 20 25 30 35 40
Time, t(min)
(b)
5
4 III
IV
I
3
1 II
0
0 1 2 3 4 5 6
Figure 8.45 C3H8 selectivity profile. Proposed model (-), experimental data: case I (○), case II (◊), case III (Δ),
and case IV ( ). (a) C3H8 selectivity profile as a function of time, (b) C3H8 selectivity profile along the
catalytic bed.
On the other hand, as mentioned above, the gas space velocity also plays an important role in the
FTS hydrocarbons selectivity. For case IV, the reactor temperature and the highest GHSV of the set
of initial operating conditions considered results in a low liquid gases productivity that should be
avoided in an FTS process when it is sought to make it more efficient.
In cases I and III, the liquid hydrocarbons selectivity is around 70%, where the change in
the initial pressure and GHSV conditions cannot be seen to a large extent. Meanwhile, in case
II, the highest C5+ selectivity is verified, and the CO and H2 conversions are 93 and 82%,
respectively.
For case II, a low light gases productivity is verified, which results in a high C5+ selectivity of
around 89%. It is important to mention that for this case, a temperature of 518 K favors the FTS
process with low productivity in light gases. The kinetic model used in this work shows a satisfac-
tory prediction of the FTS product’s distribution (light gases and C5+). The latter is sustained by the
fact that the error between the predicted and experimental values is relatively low.
The trend shown in the C5+ selectivity profile is according to the predictions made by a two-
dimensional steady-state model for a fixed-bed FTS reactor (Gideon Botes 2009) whose data of
the dimensions were taken for the simulations development of the model proposed. Therefore,
the proposed model predicts with sufficient accuracy the conversion profiles of the syngas and
the C5+ selectivity, considering the valuable aspect in the extrapolation of the reactor dimensions
(from laboratory scale to pilot scale).
(a)
6
5 I II III IV
C4 selectivity (%)
0
0 5 10 15 20 25 30 35 40
Time, t(min)
(b)
5
(b) IV
4 III
C4 selectivity (%)
I
2
1 II
0
0 1 2 3 4 5 6
Figure 8.46 n-C4H10, i-C4H10 selectivity profile. Proposed model (-), experimental data: case I (○), case II (◊),
case III (Δ), and case IV ( ). (a) n-C4H10-i-C4H10 selectivity profile as a function of time, (b) n-C4H10-i-C4H10
selectivity profile along the catalytic bed.
8.5.4.3 Simulations of the Fischer–Tropsch Fixed-Bed Reactor and the Cooling Jacket
Thermal Behavior
Figures 8.48a and 8.48b show the temperature profile for each of the experimental cases under
study over time and along the catalytic bed length (axial direction), respectively. It can be appre-
ciated that the PI controller action is fast since it manages to make the FTS temperature reactor be
established in the value of the temperature that is set as the set-point. A fast and satisfactory PI
controller action is due to a good adjustment of the controller parameters (controller gain and reset
time) determined by the adjustment rules given by Tyreus and Luyben (1992).
It is important to point out that to ensure good control of the FTS reactor temperature and to
distance itself from the presence of the thermal runaway phenomenon, it is recommended to
use a coolant temperature of 5 K below the feed inlet temperature (initial operating temperature)
(Jess and Kern 2012a). Therefore, the refrigerant temperature is set at 513 K for the first three cases
and at 523 K for case IV.
However, for diameters of catalyst particles of 3 mm (Pöhlmann et al. 2016) to 4 mm (Jess and
Kern 2012a), with a diameter ratio of the tube to the catalyst of dt/dp = 7.6, the external diffusional
effects are not present in a considerable manner. In addition, the risk of the thermal runaway
phenomenon occurring becomes negligible (Pöhlmann et al. 2016).
(a)
100
80 I II III IV
C5+ selectivity (%)
60
40
20
0
0 5 10 15 20 25 30 35 40
(b) Time, t(min)
100
80
C5+ selectivity (%)
60 II I
III
40
IV
20
0
0 1 2 3 4 5 6
Figure 8.47 C5+ selectivity profile. Proposed model (-), experimental data: case I (○), case II (◊),
case III (Δ), and case IV ( ). (a) C5+ selectivity profile as a function of time, (b) C5+ selectivity profile along
the catalytic bed.
Figures 8.49a and 8.49b show the temperature profile of the coolant as a function of time and
along with the cooling jacket length. It is seen how the refrigerant temperature increases at
the jacket inlet because at the reactor inlet the maximum temperature is reached due to the high
catalytic activity exhibited by the FTS catalyst. However, the cooling jacket temperature quickly
stabilizes at its initial value, which indicates a rapid response from the PI controller.
8.5.4.4 Surfaces of the Syngas Conversion and the Heavy Liquids Selectivity as a Function
of the FTS Reactor Temperature
Figures 8.50a, 8.50b, 8.50c, and 8.50d show the surfaces of the change in CO conversion and the
liquid hydrocarbons selectivity (C5+) as a function of the FTS reactor temperature for the 4 cases.
The objective of the development of these surfaces is to illustrate how the CO conversion and the
liquid hydrocarbons selectivity (C5+) are improved simultaneously during the course of the FTS
reaction and, at the same time, to show how the C5+ hydrocarbons selectivity is favored with
the effective CO conversion. It can be seen that case II is the best in terms of the CO conversion
and liquid hydrocarbons selectivity.
(a)
528
526 I II III IV
Temperature profile (K)
524
522
520
518
516
514
0 5 10 15 20 25 30 35 40
Time, t(min)
(b)
530
(b)
Temperature profile (K)
IV
525
520
III II
515
I
510
0 1 2 3 4 5 6
Figure 8.48 Temperature profile of the fixed-bed FTS reactor: (a) temperature profile as a function of time,
(b) temperature profile along the catalytic bed length.
(a)
526
524
Temperature, (K)
522
I II III
520 IV
518
516
514
512
0 5 10 15 20 25 30 35 40
Time, t(min)
(b)
526
IV
524
Temperature, (K)
522
520
518
III
516 II
514
I
512
0 1 2 3 4 5 6
Cooling jacket length, z(m)
Figure 8.49 Temperature profile of the cooling jacket: (a) temperature profile as a function of time,
(b) temperature profile along the catalytic bed length.
(a) (b)
100 100
C5+ selectivity (%)
C5+ selectivity (%)
50 50
0 0
150 150
100 518 100 518
517 517
50 516 50 516
514 515 514 515
0 513 0 513
Temperature profile (K) Temperature profile (K)
CO conversion, xCO (%) CO conversion, xCO (%)
(c) (d)
100 60
C5+ selectivity (%)
C5+ selectivity (%)
40
50
20
0 0
150 150
100 518 100 528
517 527
50 516 50 526
514 515 525
0 513 524
0 523
Temperature profile (K) Temperature profile (K)
CO conversion, xCO (%) CO conversion, xCO (%)
Figure 8.50 Surfaces of the CO conversion and C5+ selectivity along the temperature of the fixed-bed FTS reactor. Experimental cases: (a) case I; (b) case II;
(c) case III; (d) case IV.
8.5.5 Final Remarks

It can be appreciated that the robustness of the proposed dynamic model, considering a novel
kinetic Fischer–Trospch synthesis approach and a simple control scheme, is quite satisfactory.
To verify and guarantee the proposed dynamic model with comprehensive FTS kinetics under
the PI controller action that was applied to a fixed-bed FTS reactor with pilot-scale dimensions,
four sets of operating conditions were used with variable temperature, pressure, and GHSV at
the reactor inlet. Case II shows the best result in terms of high CO conversion and liquid hydro-
carbons selectivity (C5+), as well as a low amount of unwanted methane production. The PI con-
troller action ensures good control of the reactor temperature allowing the CO conversion and
product’s selectivity to be favored by the FTS reaction temperature. It is important to highlight that
the implementation of a control scheme in the cooling jacket of the fixed-bed FTS reactor is the key
element to ensure good thermal behavior and elimination of the runaway phenomenon. This new
approach shows great potential to be extended to scales of larger dimensions.
8.6 On the use of Steady-State Optimal Initial Operating

Conditions for the Control Scheme Implementation of a Fixed-Bed
Fischer–Tropsch Reactor
8.6.1 Introduction
A control scheme implementation for a fixed-bed FTS reactor using the optimal initial operating
conditions at steady state is developed. The proposed approach consists of solving a nonlinear con-
strained optimization problem for a fixed-bed FTS reactor with incorporation of a control scheme in
the cooling jacket to find the optimal values of the initial operating conditions of temperature,
GHSV, pressure, and coolant flow velocity under a steady-state regime that maximizes CO conver-
sion and C5+ product selectivities. Such optimal operating conditions at steady state are used to
evaluate the dynamic behavior of the reactor under the PIC scheme action that is implemented
in the cooling jacket. An extended dynamic one-dimensional pseudohomogeneous model of a
fixed-bed FTS reactor recently reported in the literature is used. The obtained results show the
importance of the implementation of dynamic control strategies for the FTS carried out in a
fixed-bed reactor, since they allow for ensuring the thermal control of such highly exothermic reac-
tions and assure better reactor performance in terms of high CO conversion and liquid product’s
selectivity (C5+).
8.6.2 Methodology
In this section, the set of dynamic equations of the reactor model based on mass and energy bal-
ances is stated, including the model that describes the dynamic behavior of the cooling jacket. The
correlations are cited to calculate the overall heat transfer coefficient and the pressure drop across
the catalyst bed. In addition, the optimization problem with non-linear constraints is described.
8.6.2.1 Model Equations and Numerical Solution

The model describing the dynamic behavior of the fixed-bed FTS reactor was implemented under
plug-flow regime at non-isothermal conditions, considering the variations of the GHSV, the gas
mixture density, and total pressure of the system in the axial direction, which have demonstrated
to be important parameters to improve predictions of syngas conversion and temperature profiles
8.6 On the use of Steady-State Optimal Initial Operating Conditions 409
(Méndez et al. 2019). It also incorporates the cooling jacket dynamic equation. The equations of the
dynamic reactor model are given by (8.114)–(8.120) and (8.131)–(8.132). The equations are solved
by using the orthogonal collocation numerical method. For approaching to the solution of the
model, five collocation points were used. The Legendre orthogonal polynomials were used for
the estimation of each collocation point (Constantinides and Mostoufi 1999).
8.6.2.2 Model Parameters, Correlations, and Kinetics

The parameters and correlations used in the reactor model are given in Table 8.20. The global heat
transfer coefficient was estimated with the correlation proposed by Haarlemmer and Bensabath
(2016), which is given as a function of the convective heat transfer coefficient on the internal side
of the reactor (hint), the convective heat transfer coefficient of the external side (hext), and the reac-
tor thermal conductivity (λw), being the reactor made of stainless steel for this case. The pressure
drop along the catalytic bed was estimated using a correlation recently stated in the literature
for the friction factor in which superior accuracy was observed by comparison with several reports
(Allen et al. 2013). The heat conductivity of the steel wall reactor was assumed to be 60 Wm−1h−1
(Sharma et al. 2011), and the value of the heat transfer coefficient (film) of the external side cor-
responding to boiling water coolant was 1000 Wm−1h−1 (Haarlemmer and Bensabath 2016).
Other parameters were considered as constants, such as the hollow fraction of the bed and the
active side volume fraction of the catalyst. The kinetic model, reaction scheme, and the correspond-
ing product’s selectivity used in the proposed reactor model are given by Eqs. (8.121)–(8.129) and
(8.100)–(8.113), respectively. There are several studies on the kinetics of the FTS reactions focused
on the formation of hydrocarbons based on the LHHW model; however, some authors have devel-
oped simpler kinetic models based on power-law type (Marvast et al. 2005; Yang et al. 2003;
Wang et al. 2003) fitting well to experimental data. Table 8.26 details the kinetic parameters
(pre-exponential factor, reaction orders for the partial pressures of H2 and CO, and the activation
energy) used in this work and previously reported by Moazami et al. (2015a).
8.6.2.3 Steady-State Nonlinear Constrained Optimization Problem

Satisfactory performance of the FTS process is defined in terms of high syngas conversion and high
liquid hydrocarbon productivity (SC5+). However, it is important to ensure good thermal behavior
of the process by having good control over the FTS reactor, which is temperature-restricted to
provide high syngas conversions and liquid hydrocarbons productiveness. For this purpose, the
temperature interval allowing a syngas conversion and liquid hydrocarbons productivity above
or equal to 80% is 493 K ≤ T ≤ 513 K.
Then, in order to establish an optimal zone of operation at steady state that faces process distur-
bances and a fixed value in the uncertain parameters, the following nonlinear constrained optimi-
zation problem is defined:
max x CO 8 143
η, d , ξ
subject to
h x, η, d, ξ = 0 8 144
T ≤T≤T
l u
8 145
x CO ≥ x lCO 8 146
SC5+ ≥ SlC5+ 8 147
Table 8.26 Kinetic parameters of the FTS reaction (Adapted from Moazami, et. al. 2015a).
Reaction R1 R2 R3 R4 R5 R6 R7 R8
Kinetics
parameters A1 Ea1 A2 Ea2 A3 Ea3 A4 Ea4 A5 Ea5 A6 Ea6 A7 Ea7 A8 Ea8
65.56165 83523.9 0.045438 65017 0.001144 49782 1.19E-06 34885.5 1.84E-10 27728.9 6.59E-09 25730.1 1.98E-08 23564.3 3.78E-05 58826.3
Reaction
order m1 n1 m2 n2 m3 n3 m4 n4 m5 n5 m6 n6 m7 n7 m8 n8
−0.4755 0.8201 0.014 0.0746 −0.0594 0.4128 −0.7131 1.4167 0.1037 −0.4789 −0.2626 1.1713 1.2895 0.4571 0.472 0.7706
where h(x, η, d, ξ) is the proposed model given by Eqs. (8.114)–(8.120) and as a function of
state variables (x), process disturbances (d), and parameter uncertainties (ξ). Tu and Tl are the upper
and lower bounds of the reactor temperature, and x lCO and SlC5+ are the lower bounds for the CO
conversion and liquid hydrocarbons selectivity (SC5+ ), respectively.
The optimal operating conditions for a steady-state process are often not optimal for a process at
unsteady state because the optimality region for a steady-state process differs considerably from the
optimal operating region for a dynamic process. Figure 8.51 shows schematically how the optimal-
ity region of a process in a steady state tends to move due to process dynamics. In addition, it is
important in any dynamic problem to consider the process disturbances effects and the parameters
uncertainty to determine a robust solution under different restrictions. On the other hand, the phe-
nomenon of “temperature runaway” is very sensitive to the dynamics of the inlet temperature, the
hot spot can appear immediately or later at an initial temperature value. For this reason, the pre-
sented optimization method is applied to find the best fixed-bed reactor inlet temperature for
steady-state situations.
Figure 8.51a highlights the optimality of the steady-state process considering a fixed value in
the uncertain parameters and the disturbances effect. It is seen that certain process constrained
separates the unfeasible region. The point located in the constrained line indicates the steady-
state optimal design point, and the surrounding region becomes the set of points in an optimal
neighborhood in which the process can feasibly operate. However, this point might violate con-
straints when transient changes are accounted for in the analysis, as schematically illustrated by a
black dotted region in Figure 8.51a (Mehta and Ricardez-Sandoval 2016; Rafiei-Shishavan et al.
2017). Nevertheless, for this case study, the effect of the process dynamics does not greatly alter
the feasibility.
Therefore, in order to find the optimal values of the inlet operating conditions (temperature,
pressure, and coolant flow rate) that determine the optimality region in steady state, the nonlinear
constrained optimization problem to solve can be written in the following general form:
max f x, η, d, ξ
η, d , ξ
(a) (b)
x1 x1 g(x1,x2) ≤ 0 Dynamic region
Dynamic region The optimal point displacement
Optimal steady-state point
g(x1,x2) ≤ 0
Unfeasible region x2 Unfeasible region x2
Figure 8.51 Schematic diagram of optimality (Mehta and Ricardez-Sandoval 2016; Rafiei-Shishavan et al.
2017): (a) steady state, (b) dynamic state.
subject to
gj x ≥ bj, j = 1, …, m 8 148
where x = (x1, …, xn) is the vector of decision variables, and each bj is a constant. The objective f and
constraint functions gj are assumed to have continuous first partial derivatives. Corresponding to
each constraint gj = bj, a Lagrange multiplier λj is defined, and λj = (λ1, …, λm) is the vector of these
multipliers (Tosun 2007). The Lagrangian function for the problem is:
m
L x, λ = f x + λj gj x − bj 8 149
j=1
and the first-order necessary conditions for optimality according to Kuhn–Tucker are:
∂L ∂f m
∂hj
= + λj = 0, i = 1, …, n 8 150
∂x i ∂x i j=1
∂x i
gj x ≥ bj, j = 1, …, m 8 151
where m are the set of constraints and n are the process variables. In addition, it is essential that the
necessary and sufficient second-order Kuhn–Tucker conditions are met to ensure the optimization
of the nonlinear constraint optimization problem. The second-order optimality conditions involve
the matrix of second partial derivatives with respect to x (the Hessian matrix of the Lagrangian func-
tion) and may be written as follows (Biegler 2010; Edgar and Himmelblau 2001):
yT ∇2x L x ∗ , λ∗ y > 0 8 152
for all nonzero vectors y such that:

J x∗ y = 0 8 153
where J(x∗) is the Jacobian matrix, whose rows are the gradients of the constraints that are active
at x∗. The abovementioned equation determines a set of vectors that are orthogonal to the active
constraints (Edgar and Himmelblau 2001).
The steady-state nonlinear constrained optimization problem was initialized using an initials
temperature, pressure, and flow rate (at the reactor inlet) of 505 K, 3.0 MPa, and 15.6 L/min, respec-
tively, as well as an initial coolant temperature (boiling water) of 498 K. The uncertain parameters
considered in the optimization problem were the active catalyst fraction (β) and the overall heat
transfer coefficient (Ur). The disturbance that greatly affects the reactor behavior at steady state
−1 −1
is the GHSV. The values used for the GHSV were 1800, 2400, 3000, and 3600 NmLgcat h .
The results of the optimization problem solution indicate optimum initial conditions of temper-
ature, pressure, and flow rate of 503 K, 2.5 MPa, and 13.5 L/min, respectively. On the other hand,
the value of the Lagrange multiplier that allows obtaining the best optimal operating conditions
was 0.71. Such a set of optimal values defines the solution of the steady-state nonlinear constrained
optimization problem accomplishing the established inequality constraints. These optimum values
were used to solve the closed-loop reactor model by implementing a PI control scheme in the cool-
ing jacket.
The steady-state nonlinear constrained optimization problem for the FTS fixed-bed reactor was
solved using Python software. In Figure 8.52, the algorithm of the nonlinear optimization problem
with implementation of the control scheme is shown.
PI controller
FTS model t
KC
h(x, η, d, 𝜉) = 0 –
𝜁 (t) = KCe + 𝜏
I
∫0e– dt
If error
–
e ≠ Tsp – T
max f (x, 𝜂, d, 𝜉)
𝜂,d,𝜉
subject to
gj(x) ≥ bj, j = 1,…, m
j = 1,…, n
Lagrangian function
m
L(x, 𝜆) = f (x) + Σ
j=1
𝜆j [gj (x) – bj ]
First-order condition
Second-order condition
m
𝜕hj yT∇x2L(x*, 𝜆*)y > 0
𝜕L
𝜕xi
=
𝜕f
𝜕xi
+ Σ
j=1
𝜆j
𝜕xi
= 0, i = 1,…, n
J(x*)y = 0
Solution
Figure 8.52 Algorithm of the nonlinear optimization problem with implementation of the control scheme.
8.6.2.4 Implementation of the Control Scheme

To ensure a viable operation of the reactor in terms of a safe thermal operation, it is necessary to
implement a control scheme in the cooling jacket. It is well known that the FTS reaction is highly
exothermic, which requires good temperature control in any reactor configuration used (Kaskes
et al. 2016; Kwack et al. 2011a). The high temperatures of the FTS reaction give rise to the hot-spot
formation, which can lead to the presence of the runaway phenomenon (Kwack et al. 2011a; Rafiq
et al. 2011).
The literature indicates that a PI controller is suitable in those cases that involve a highly non-
linear process and where it is intended to eliminate the static error (off-set) (Stephanopoulos 1985).
On the other hand, it is stated that PI control is a form of continuous adjustment originating in
the process industries and is usually considered part of engineering process control (Box and
Luceño 1997).
To ensure adequate temperature control of the fixed-bed FTS reactor, a PIC scheme is implemen-
ted in the cooling jacket where the reactor temperature is controlled by adjusting the temperature of
the refrigerant inside the cooling jacket by manipulating the flow velocity of the coolant. The PI
controller is given by Eq. (8.141), and e is the error given by Eq. (8.142).
The set-point established in the control problem becomes the reactor temperature of 503 K, and
in order to achieve a negligible deviation in the process response, the flow velocity of the refrigerant
is manipulated. The controller parameters (controller gain KC and reset time τI) are determined
by the adjustment rules given by Tyreus and Luyben (1992) whose values were KC = 2.6 and
τI = 0.25 min.
Figure 8.53 illustrates the line diagram of the PI controller implemented in the cooling jacket.
8.6.3 Results and discussion

In this section, the results of simulations are discussed. The conversion of CO and selectivity of
liquid hydrocarbons are obtained. The temperature profiles of the FTS reactor and cooling jacket
are analyzed. CO conversion and selectivity of liquid hydrocarbons as a function of time under the
PI controller action are also shown.

The model simulations are compared with experimental data reported by Moazami et al. (2015a).
The experiments were carried out using a Co/SiO2 catalyst, which was diluted with 2 g of silicon
carbide at a mass ratio of 1:12 (Co/SiO2/SiC). A simulated nitrogen-rich syngas was used with a
concentration of 17, 33, and 50 vol.% of CO, H2, and N2, respectively.
The dimensions of the bench-scale fixed-bed reactor used in this case study are 6 m in length as a
catalytic bed with an internal radius (Dai et al. 2014).
+
T −
FTS plant model ∑
–
u(t) e(t)
+ t
KC
∑ 𝜏I ∫0 edt–
+
–
KCe
Figure 8.53 Line diagram of the PI controller implemented in the cooling jacket.
8.6.3.2 Simulations of the Steady-State Nonlinear Constrained Optimization Problem: CO

Conversion, SC5+ Selectivity, and Temperature Profiles
As previously mentioned, the optimal operating conditions at the inlet of the FTS reactor for the non-
linear constrained optimization problem were: Tin = 503 K, PT,in = 2.50 MPa, and Fw,in = 15.6 L/min.
These optimal conditions values were used for the development of the steady-state one-dimensional
pseudohomogeneous model simulations.
The CO conversion and selectivity of liquid hydrocarbons are shown in Figures 8.54a and 8.54 b,
respectively. The CO conversion reached is close to 100% due to the reactor temperature effect. The
optimal initial operating conditions found according to the steady-state nonlinear constrained opti-
mization problem do not affect the restrictions imposed in the CO conversion. On the other hand,
the liquid hydrocarbons selectivity is above 81%. It can be verified that the selectivity toward pro-
ducts is enhanced by the effect of the temperature along the catalytic bed. In addition, the low pro-
duction of light gases also allows the liquid hydrocarbons selectivity to increase because of the effect
of GHSV. In such a way that the optimal values for temperature, GHSV, and coolant flow velocity at
the reactor inlet do not infringe the constraints imposed in the liquid hydrocarbon products accord-
ing to the nonlinear problem defined.
Figures 8.55a and 8.55b illustrate the FTS reactor and the cooling jacket temperature profiles
along the catalytic bed at steady-state conditions, respectively. It can be clearly seen that the hot
spot along the catalytic bed is 3 K, almost formed at the reactor inlet. However, the temperature
(a)
100
80
60
40
20
0
0 1 2 3 4 5 6
(b)
100
80
C5+ selectivity (%)
60
40
20
0
0 1 2 3 4 5 6
Figure 8.54 CO conversion and liquid hydrocarbons profiles evaluated with the optimal inlet condition
values: (a) CO conversion, (b) hydrocarbons liquid (SC5+) selectivity.
(a)
506
505
Temperature, (K)
504
503
502
501
0 1 2 3 4 5 6

(b)
502
501.5
Temperature, (K)
501
500.5
500
0 1 2 3 4 5 6
Cooling jacket length, z(m)
Figure 8.55 Temperature profiles evaluated with the optimal inlet condition values: (a) FTS fixed-bed
reactor, (b) cooling jacket.
peak is relatively low such that the runaway phenomenon is not present. The temperature of the
cooling jacket rises by about 1.7 K and does not become stable, which is due to the process operating
in open loop (without control action on the cooling jacket). It is important to note that the temper-
ature along the catalytic bed of a fixed-bed reactor, in which a highly exothermic reaction is carried
out such as in the case of an FTS reaction, can be triggered to cause the presence of the runaway
causing damages in the reactor. However, a good control of the FTS reactor temperature is not only
guaranteed by the integration of a cooling jacket in the reactor as can be seen in Figure 8.55b. There-
fore, it is interesting to verify the integration of a dynamic control scheme in a jacketed fixed-bed
reactor involving highly exothermic reaction takes place to ensure good control of temperature, as
will be demonstrated below.
8.6.3.3 Simulations of the Control Scheme Implementation: CO Conversion, SC5+ Selectivity,

and Temperature Profiles
The closed-loop system (FTS reactor with control in the cooling jacket) is studied to evaluate the
dynamic behavior of the process under the action of the PI control scheme. To verify the robustness
of the PI control implemented in the FTS reactor, variations in the GHSV are made by setting a
(a)
100
80
60 I II III IV
40
20
0
0 500 1000 1500 2000
Time, t(seg)
(b)
100
80
60
IV
40 III
II
20
I
0
0 1 2 3 4 5 6
Figure 8.56 Simulated CO conversion profile in the fixed-bed FTS reactor under PI action control
(I: GHSV = 1800, II: GHSV = 2400, III: GHSV = 3000, IV: GHSV = 3600 NmL/gcat/h): (a) Dynamic CO conversion,
(b) CO conversion along the catalytic bed, (○) Experimental data (Adapted from Kuipers, et. al. (1996)).
value in the uncertain parameters. It should be noted that the continuous processes have been con-
tinuously improved to achieve the process optimality around a nominal steady-state operating
point (Bansal et al. 2000). In addition, it is important to determine those conditions in which a con-
tinuous process can operate in a feasibility region (Pistikopoulos and Mazzuchi 1990; Dimitriadis
and Pistikopoulos 1995).
Figure 8.56a shows the CO conversion as a function of time under the PI controller action, for
four changes in steps made in the considered disturbance (GHSV). It can be clearly seen that the CO
conversion is close to 99% for each experimental run lasting 10 min. On the other hand, the CO
conversion along the catalytic bed (Figure 8.56b) is about 88%. The high conversions are due to
the effect of the FTS reactor temperature. The higher the pressure, the higher the syngas conversion
(van Berge and Everson 1997; Yan et al. 2009), but if the syngas space velocity is increased, a
decrease in the residence time is observed, thus diminishing the CO conversion (Lee and Chung
2012). In addition, the error observed between the experimental data reported by Moazami
et al. (2015a) at the reactor outlet and the simulated value for the CO conversion is low. However,
it is important to note that care should be taken with the simultaneous changes of temperature and
space velocity, because high temperatures shift the FTS process toward the production of heavy
hydrocarbons (Dry 1981), while increasing the space velocity and decreasing slightly the CO con-
version, but if adequate values are established, then high production rates without altering the qual-
ity of the product can be obtained (Lee and Chung 2012). The optimal operating conditions
(a)
100
80 I II III IV
C5+ selectivity (%)
60
40
20
0
0 5 10 15 20 25 30 35 40
Time, t(min)
(b)
100
80
C5+ selectivity (%)
60 I
II
III
40
I
II
IV
20 III
IV
0
0 1 2 3 4 5 6
Figure 8.57 Simulated SC5+ liquid hydrocarbons selectivity profile in the fixed-bed FTS reactor under PI action
control (I: GHSV = 1800, II: GHSV = 2400, III: GHSV = 3000, IV: GHSV = 3600 NmL/gcat/h): (a) Dynamic CO
conversion, (b) CO conversion along the catalytic bed; (○) Experimental data (Adapted from Kuipers,
et. al. (1996)).
determined for the steady-state nonlinear constrained optimization problem accomplish the con-
straints imposed on the CO conversion for the dynamic problem under the PI controller action.
Figure 8.57a shows the liquid hydrocarbons selectivity dynamic profile under the PI control
action scheme implemented. For the fourth step changes made in the gas space velocity, the liquid
product’s selectivity is around 44–45%. In this case, the constraints imposed on the liquid hydro-
carbons selectivity are not accomplished, which is largely due to the process dynamics (Mehta
and Ricardez-Sandoval 2016; Rafiei-Shishavan 2017). However, a value of between 44 and 45%
in the liquid product’s selectivity is acceptable. Nevertheless, the SC5+ selectivity along the axial
direction of the reactor remains above 90% (Figure 8.57b). It is important to note that the selectivity
of SC5+ does not show a significant variation (Figure 8.57b inset), which is caused by the simulta-
neous effect of temperature, pressure, and GHSV (Kuipers et al. 1996). The error observed between
the experimental data reported by Moazami et al. (2015a) at the reactor outlet and the simulated
value for SC5+ selectivity is practically negligible. Figure 8.58a shows the dynamic temperature pro-
file of the FTS reactor under the PI controller implementation. As GHSV increases, the formed hot
spot also increases. However, the action of the PI controller is quite satisfactory since the reactor
temperature does not rise considerably and it stabilizes around the established set point. On the
(a)
506
505.5
I II III IV
Temperature (K)
505
504.5
504
503.5
503
0 5 10 15 20 25 30 35 40
Time, t(min)
(b)
506
505.5 IV III
Temperature (K)
505
II
504.5
I
504
503.5
503
0 1 2 3 4 5 6
Figure 8.58 Simulated temperature profile in the fixed-bed FTS reactor under PI action control
(I: GHSV = 1800, II: GHSV = 2400, III: GHSV = 3000, IV: GHSV = 3600 NmL/gcat/h): (a) Dynamic temperature,
(b) temperature along the catalytic bed.
other hand, Figure 8.58b illustrates the temperature profile of the FTS reactor along the axial direc-
tion. As the disturbances (GHSV) value increases in each pitch step applied every 10 min, the reac-
tor temperature rises until a hot spot of 3 K is formed above the inlet temperature of the FTS reactor.
Thus, the PI controller shows satisfactory robustness because it can stabilize the reactor tempera-
ture near the set point with negligible error.
A very important aspect that is verified in the thermal behavior of the FTS reactor is that the hot
spot formed undergoes a delay due to the control action. That is, the hot spot that is usually formed
has been shown to originate near the reactor inlet. However, the PI controller allows stabilizing the
reactor temperature near the set point and achieves the hot spot formed beyond the reactor inlet.
Finally, the temperature profiles along the cooling jacket length and time are shown in Fig-
ures 8.59a and 8.59b. Figure 8.59a illustrates the dynamic behavior of the refrigerant temperature
(boiling water) under the control action produced by the PI controller. The changes are made in the
GHSV (disturbances) with the same initial conditions of concentration and pressure of the syngas.
For the temperature at the reactor outlet (which is remarkably close to the input temperature of 503
K) the initial temperature for the next experimental run (simulation) is used.
According to the temperature profile of the coolant, hot spots are formed beyond the cooling
jacket inlet, which is neglected (Figure 8.59b). However, in terms of the process response to the
controller action, it is verified that the traditional PI control scheme can keep the reactor thermal
(a)
505
I II III IV
504.5
Temperature (K)
504
503.5
503
0 5 10 15 20 25 30 35 40
Time, t(min)
(b)
505
IV
504.5
Temperature (K)
III
504 II
I
503.5
503
0 1 2 3 4 5 6
Cooling jacket length, z (m)
Figure 8.59 Simulated temperature profile in the cooling jacket under PI action control (I: GHSV = 1800,
II: GHSV = 2400, III: GHSV = 3000, IV: GHSV = 3600 NmL
gcat h): (a) Dynamic cooling jacket temperature,
(b) cooling jacket temperature along the catalytic bed.
operation within an operational feasibility region. On the other hand, the behavior of the controller
against the disturbances made in the process (GHSV) and the uncertainty parameters (the active
side catalyst and overall heat transfer coefficient) under the optimal operating conditions (accord-
ing to the solution of the steady-state nonlinear constrained optimization problem) are completely
satisfactory.
8.6.4 Final Remarks

The solution of a nonlinear constrained optimization problem at steady-state conditions is: a crucial
step in determining the operational feasibility region. In this case study, the process optimality in
terms of the inlet operating optimal conditions values is used to evaluate the process dynamics
under the traditional PI control scheme implementation. It was possible to demonstrate that the
PI controller is robust against disturbances of the GHSV for a fixed value in the uncertain para-
meters considered. The controller allows keeping the FTS reactor temperature around the estab-
lished set-point with an insignificant error. In addition, the CO conversion and the liquid
hydrocarbons selectivity are satisfactory, verifying a negligible error between experimental and
simulated values. On the other hand, the optimal values of the inlet operation conditions (temper-
ature, pressure, and refrigerant flow velocity), although when used in the evaluation of the closed-
References 421
loop process (on control action), do not accomplish the constraint imposed on the liquid product’s
selectivity (SC5+), the dynamic process behavior is not of considerable attention, since the selectivity
reached at the reactor outlet is high. In summary, the implementation of a dynamic control scheme
in fixed-bed FTS reactors is extremely important, as demonstrated in this study.
References
Adesina, A.A. (1996). Hydrocarbon synthesis via Fischer-Tropsch reaction: travails and triumphs.
Ahón, V.R., Costa, E.F., Monteagudo, J.E.P. et al. (2005). A comprehensive mathematical model for
the Fischer–Tropsch synthesis in well-mixed slurry reactors. Chemical Engineering Science 60:
677–694.
Akpan, E., Sun, Y., Kumar, P. et al. (2007). Kinetics, experimental and reactor modeling studies of the
carbon dioxide reforming of methane (CDRM) over a new Ni/CeO2-ZrO2 catalyst in a packed bed
tubular reactor. Chemical Engineering Science 62: 4012–4024.
Al-Dahhan, M.H., Larachi, F., Dudukovic, M.P., and Laurent, A. (1997). High-pressure trickle-bed
reactors: a review. Industrial and Engineering Chemistry Research 36: 3292–3314.
Alihellal, D. and Chibane, L. (2016). Comparative study performance of Fischer-Tropsch synthesis in
conventional packed bed and in membrane reactor over iron-and cobalt-based catalysts. Arabian
Journal for Science and Engineering 41: 357–369.
Allen, K., von Backström, T., and Kröger, D. (2013). Packed bed pressure drop dependence on particle
shape, size distribution, packing arrangement and roughness. Powder Technology 246: 590–600.
Ancheyta, J., Muñoz, J., and Macías, M. (2005). Experimental and theoretical determination of the
particle size of hydrotreating catalysts of different shapes. Catalysis Today 109: 120–127.
Anderson, R.B. and Karn, F.S. (1960). Kinetics and reaction mechanism of the Fischer-Tropsch synthesis.
Journal of Physical Chemistry 64: 805–808.
Atashi, H., Siami, F., Mirzaei, A., and Sarkari, M. (2010). Kinetic study of Fischer–Tropsch process on
titania-supported cobalt–manganese catalyst. Journal of Industrial and Engineering Chemistry 16:
952–961.
Atwood, H.E. and Bennett, C.O. (1979). Kinetics of the Fischer-Tropsch reaction over iron. Industrial
Engineering Chemistry Process Design Development 18: 163–170.
Bai, B., Liu, M., Lv, X. et al. (2011). Correlations for predicting single phase and two-phase flow pressure
drop in pebble bed flow channels. Nuclear Engineering and Design 241: 4767–4774.
Bansal, V., Perkins, J.D., Pistikopoulos, E.N. et al. (2000). Simultaneous design and control optimisation
under uncertainty. Computer and Chemical Engineering 24: 261–266.
Barker, J.J. (1965). Heat transfer in packed beds. Industrial and Engineering Chemistry 57: 43–51.
Basha, O.M., Sehabiague, L., Abdel-Wahab, A., and Morsi, B.I. (2015). Fischer–Tropsch synthesis in
slurry bubble column reactors: experimental investigations and modeling – a review. International
Journal of Chemical Reactor Engineering 13: 201–288.
Bayat, M., Hamidi, M., Dehghani, Z., and Rahimpour, M. (2014). Sorption-enhanced Fischer-Tropsch
synthesis with continuous adsorbent regeneration in GTL technology: modeling and optimization.
Journal of Industrial and Engineering Chemistry 20: 858–869.
Bazmi, M., Hashemabadi, S.H., and Bayat, M. (2013). Extrudate trilobe catalysts and loading effects on
pressure drop and dynamic liquid holdup in porous media of trickle bed reactors. Transport in Porous
Media 99: 535–553.
Bechara, R., Balloy, D., and Vanhove, D. (2001). Catalytic properties of Co/Al2O3 system for hydrocarbon
synthesis. Applied Catalysis: A General 207: 343–353.
Benyahia, F. and O’Neill, K.E. (2005). Enhanced voidage correlations for packed beds of various particle
shapes and sizes. Particulate Science and Technology 23: 169–177.
Béttega, R., Moreira, M.F.P., Corrêa, R.G., and Freire, J.T. (2011). Mathematical simulation of radial heat
transfer in packed beds by pseudohomogeneous modeling. Particuology 9: 107–113.
Bey, O. and Eigenberger, G. (1997). Fluid flow through catalyst filled tubes. Chemical Engineering Science
52: 1365–1376.
Bey, O. (1998). Strömungsverteilung und wärmetransport in schüttungen. VDI-Verlag.
Biegler, L.T. (2010). Nonlinear programming: concepts, algorithms, and applications to chemical processes,
1ste. Philadelphia, PA: SIAM.
Bird, R.B., Stewart, W.E., and Lightfoot, E.N. (2001). Transport Phenomena, 3rde. New York: Wiley.
Bischoff, K.B. (1965). Effectiveness factors for general reaction rate forms. American Institute of Chemical
Engineers Journal 11: 351–355.
Bosanquet, C. H. (1944). British TA Report BR-507. Technical Report.
Box, G. and Luceño, A. (1997). Discrete proportional-integral adjustment and statistical process control.
Journal of Quality Technology 29: 248–260.
Brady, R.C. III and Pettit, R. (1981). Mechanism of the Fischer-Tropsch reaction: the chain propagation
step. Journal American Chemical Society 103: 1287–1289.
Brötz, W. (1949). Zur Systematik der Fischer-Tropsch-Katalyse. Journal of electrochemistry and applied
physical chemistry 5: 301–306.
Burcat, A. and Ruscic, B. (2005). Third millenium ideal gas and condensed phase thermochemical database
for combustion with updates from active thermochemical tables. Argonne, IL: Argonne National
Laboratory.
Calderone, V.R., Shijn, N.R., Curulla-Ferre, D. et al. (2013). De novo design of nanostructured iron–cobalt
Fischer-Tropsch catalysts. A Journal of the German Chemical Society 52: 4397–4401.
Caldwell, L. and Van Vuuren, D.S. (1986). On the formation and composition of the liquid phase in
Fischer-Tropsch reactor. Chemical Engineering Science 41: 89–96.
Carbonell, R.G. and Whitaker, S. (1983). Dispersion in pulsed systems—II: theoretical developments for
passive dispersion in porous media. Chemical Engineering Science 38: 1795–1802.
Çarpinlioğlu, M.Ö. and Özahi, E. (2008). A simplified correlation for fixed bed pressure drop. Powder
Chabot, G., Guilet, R., Cognet, P., and Gourdon, C. (2015). A mathematical modeling of catalytic milli-
fixed bed reactor for Fischer-Tropsch synthesis: influence of tube diameter on Fischer Tropsch
selectivity and thermal behaviour. Chemical Engineering Science 127: 72–83.
Chambrey, S., Fongarland, P.V., Karaca, H. et al. (2006). Fischer–Tropsch synthesis in milli-fixed bed
reactor: comparison with centimetric fixed bed and slurry stirred tank reactors. Catalysis Today 171:
201–206.
Chaumette, P., Courty, P., Kiennemann, A., and Ernst, B. (1995). Alcohol and paraffin synthesis on cobalt
based catalysts: comparison of mechanistic aspects. Topics in Catalysis 2: 117–126.
Chen, J., Ring, Z., and Dabros, T. (2001). Modeling and simulation of a fixed-bed pilot-plant hydrotreater.
Chilton, T.H. and Colburn, A.P. (1934). Mass transfer (absorption) coefficients prediction from data on
heat transfer and fluid friction. Industrial and Engineering Chemistry 26: 1183–1187.
Constantinides, A. and Mostoufi, N. (1999). Numerical methods for chemical engineers with Matlab
applications, 1ste. New Jersey: Prentice Hall.
References 423
Cybulski, A. and Moulijn, J.A. (1994). Monoliths in heterogeneous catalysis. Catalysis Reviews 36: 179–270.
Dai, X.P., Liu, P.Z., Shi, Y. et al. (2014). Fischer–Tropsch synthesis in a bench-scale two-stage
multitubular fixed-bed reactor: simulation and enhancement in conversion and diesel selectivity.
Danckwerts, P. (1953). Continuous flow systems: distribution of residence times. Chemical Engineering
Science 2: 1–13.
Das, T.K., Conner, K.W., Li, J. et al. (2005). Fischer−Tropsch synthesis: kinetics and effect of water for a
Co/SiO2 catalyst. Energy and Fuel 19: 1430–1439.
De Jong, K.P. and Geus, J.W. (2000). Carbon nanofibers: catalytic synthesis and applications. Catalysis
Reviews 42: 481–510.
Delgado, J.M.P.Q. (2006). A critical review of dispersion in packed beds. Heat Mass Transfer 42: 279–310.
Dimitriadis, V.D. and Pistikopoulos, E.N. (1995). Flexibility analysis of dynamic systems. Industrial and
Dixon, A.G., Nijemeisland, M., and Stitt, E.H. (2006). Packed tubular reactor modeling and catalyst
design using computational fluid dynamics. Advances in Chemical Engineering 31: 307–389.
Dixon, A.G. (2012). Fixed bed catalytic reactor modelling-the radial heat transfer problem. The Canadian
Journal of Chemical Engineering 90: 507–527.
Donnelly, T.J., Yates, I.C., and Satterfield, C.N. (1988). Analysis and prediction of product distributions of
the Fischer-Tropsch synthesis. Energy and Fuels 2: 734–739.
Dry, M.E., Shingles, T., and Boshoff, L.J. (1972). Rate of the Fischer-Tropsch reaction over iron catalysts.
Dry, M.E. (1976). Advances in Fischer-Tropsch chemistry. Industrial and Engineering Chemistry Product
Research and Development 15: 282–286.
Dry, M.E. (1981). The Fischer-Tropsch synthesis. In: Catalysis. Science and Technology (ed. J.R. Anderson
and M. Boudart). New York: Springer-Verlag.
Dry, M.E. (1983). The Sasol Fischer-Tropsch process. In: Applied Industrial Catalysis, vol. 2 (ed. B.E.
Leach). New York: Academic Press.
Dry, M.E. (1996). Practical and theoretical aspects of the catalytic Fischer-Tropsch process. Applied
Dry, M.E. (2001). High quality diesel via the Fischer-Tropsch process – a review. Journal of Chemical
Technology and Biotechnology 77: 43–50.
Edgar, T.F. and Himmelblau, D. (2001). Optimization chemical processes, 2nde. New York: McGraw Hill.
Eidt, C.M., Bauman, R.F., Elsenberg, B., and Hochman, J.M. (1994). Current developments in natural gas
conversion technology. In: Proceedings of the 14th World Petroleum Congress. Chichester, Stavanger,
Norway: Wiley.
Eisfeld, B. and Schnitzlein, K. (2001). The influence of confining walls on the pressure drop in packed
beds. Chemical Engineering Science 56: 4321–4329.
Ellman, M., Midoux, N., Wild, G. et al. (1990). A new, improved liquid hold-up correlation for trickle-bed
reactors. Chemical Engineering Science 45: 1677–1684.
Erkey, C., Rodden, J.B., and Akgerman, A. (1990a). A correlation for predicting diffusion coefficients in
alkanes. The Canadian Journal of Chemical Engineering 68: 661–665.
Erkey, C., Rodden, J.B., and Akgerman, A. (1990b). Diffusivities of synthesis gas and n-alkanes in
Fischer-Tropsch wax. Energy and Fuels 4: 275–276.
Ermolaev, V.S., Gryaznov, K.O., Mitberg, E.B. et al. (2015). Laboratory and pilot plant fixed-bed reactors
for Fischer-Tropsch synthesis: mathematical modeling and experimental investigation. Chemical
Fazlollahi, F., Sarkari, M., Zare, A. et al. (2012). Development of a kinetic model for Fischer–
Tropsch synthesis over Co/Ni/Al2O3 catalyst. Journal of Industrial and Engineering Chemistry 18:
1223–1232.
Feimer, J.L., Silveston, P.L., and Hudgins, R.R. (1981). Steady-state study of the Fischer-Tropsch reaction.
Industrial Engineering Chemistry Product Research and Development 20: 609–615.
Fischer, N., van Steen, E., and Claeys, M. (2013). Structure sensitivity of the Fischer–Tropsch activity and
selectivity on alumina supported cobalt catalysts. Journal of Catalysis 299: 67–80.
Forghani, A., Elekaei, H., and Rahimpour, M. (2009). Enhancement of gasoline production in a novel
hydrogen-permselective membrane reactor in Fischer–Tropsch synthesis of GTL technology.
International Journal of Hydrogen Energy 34: 3965–3976.
Förtsch, D., Pabst, K., and GroB-Hardt, E. (2015). The product distribution in Fischer-Tropsch synthesis:
an extension of the ASF model to describe common deviations. Chemical Engineering Science 138:
333–346.
Frennet, A. and Hubert, C. (2000). Transient kinetics in heterogeneous catalysis by metals. Journal of
Molecular Catalysis A: Chemical 163: 163–188.
Froment, G.F., Bischoff, K.B., and De Wilde, J. (2011). Chemical Reactor Analysis and Design, 3rde.
United States: Wiley.
Fu, T. and Li, Z. (2015). Review of recent development in Co-based catalysts supported on carbon
materials for Fischer–Tropsch synthesis. Chemical Engineering Science 135: 3–20.
Fuller, E.N., Schettler, P.D., and Giddings, J.C. (1966). New method for prediction of binary gas-phase
diffusion coefficients. Industrial and Engineering Chemistry 58: 18–27.
Gardezi, S.A., Landrigan, L., Joseph, B., and Wolan, J.T. (2012). Synthesis of tailored eggs-hell cobalt
catalysts for Fischer–Tropsch synthesis using wet chemistry techniques. Industrial and Engineering
Chemistry Research 51: 1703–1712.
Gasem, K.A.M. and Robinson, R.L. Jr. (1985). Prediction of phase behavior for CO2 + heavy normal
paraffins using generalized parameter Soave and Peng–Robinson equations of state. American Institute
of Chemical Engineering Journal 24–28.
Gideon Botes, F. (2009). Influences of water and syngas partial pressure on the kinetics of a commercial
alumina-supported cobalt Fischer−Tropsch catalyst. Industrial and Engineering Chemistry Research
48: 1859–1865.
Gideon Botes, F., Van Dyk, B., and McGregor, C. (2009). The Development of a macro kinetic model for a
commercial Co/Pt/Al2O3 Fischer−Tropsch catalyst. Industrial Engineering Chemistry Research 48:
10439–10447.
Gill, S., Tsolakis, A., Dearn, K., and Rodríguez-Fernández, J. (2011). Combustion characteristics and
emissions of Fischer-Tropsch diesel fuels in IC engines. Progress in Energy and Combustion Science 37:
503–523.
Girardon, J., Lermontov, A., Gengembre, L. et al. (2005). Effect of cobalt precursor and pretreatment
conditions on the structure and catalytic performance of cobalt silica-supported Fischer–Tropsch
catalysts. Journal of Catalysis 230: 339–352.
Glasser, D., Hildebrandt, D., Liu, X. et al. (2012). Recent advances in understanding the Fischer–Tropsch
synthesis (FTS) reaction. Current Opinion in Chemical Engineering 1: 296–302.
Guardo, A., Coussirat, M., Recasens, F. et al. (2006). CFD study on particle-to-fluid heat transfer in fixed
bed reactors: convective heat transfer at low and high pressure. Chemical Engineering Science 61:
4341–4353.
Haarlemmer, G. and Bensabath, T. (2016). Comprehensive Fischer–Tropsch reactor model with
non-ideal plug flow and detailed reaction kinetics. Computers and Chemical Engineering 84:
281–289.
References 425
Hall, C.C., Gall, D., and Smith, S.L. (1952). A comparison on the fixed-bed, liquid-phase (“slurry”) and
fluidized-bed techniques in the Fischer-Tropsch synthesis. Journal of the Institution of Petroleum
Technologists 38: 845–875.
Hallac, B.B., Keyvanloo, K., Hedengren, J.D. et al. (2015). An optimized simulation model for iron-based
Fischer-Tropsch catalyst design: transfer limitations as functions of operating and design conditions.
Han, N.W., Bhakta, J., and Carbonell, R.G. (1985). Longitudinal and lateral dispersion in packed beds: effect of
column length and particle size distribution. American Institute of Chemical Engineers Journal 31: 277–288.
Haughey, D.P. and Beveridge, G.S.G. (1969). Structural properties of packed beds – a review. The
Canadian Journal of Chemical Engineering 47: 130–140.
Hicks, R.E. (1970). Pressure drop in packed beds of spheres. Industrial and Engineering Chemistry
Fundamentals 9: 500–502.
Hinchiranan, S., Zhang, Y., Nagamori, S. et al. (2008). TiO2 promoted Co/SiO2 catalysts for Fischer-
Tropsch synthesis. Fuel Processing Technology 89: 455–459.
Hooshyar, N., Vervloet, D., Kapteijn, F. et al. (2012). Intensifying the Fischer-Tropsch synthesis by
reactor structuring – a model study. Chemical Engineering Journal 207: 865–870.
Huang, S.H., Lin, H.M., Tsai, F.N., and Chao, K.C. (1988). Solubility of synthesis gases in heavy
n-paraffins and Fischer-Tropsch wax. Industrial and Engineering Chemistry Research 27: 162–169.
Huff, G.A. and Satterfield, C.N. (1985). Liquid accumulation in catalyst pores in a Fischer-Tropsch fixed-
bed reactor. Industrial and Engineering Chemistry Process Design and Development 24: 986–995.
Ibrahim, H.H. and Idem, R.O. (2006). Kinetic studies of the partial oxidation of isooctane for hydrogen
production over a nickel-alumina catalyst. Chemical Engineering Science 61: 5912–5918.
Iglesia, E., Reyes, S.C., and Soled, S.L. (1993). Computed-aided design of catalysts. Reaction-transport
selectivity models and the design of Fischer-Tropsch catalysts, 1ste. R. Becker and C. J. Pereira.
Iliuta, I., Ring, Z., and Larachi, F. (2006). Simulating simultaneous fines deposition under catalytic
hydrodesulfurization in hydrotreating trickle beds-does bed plugging affect HDS performance?
Iordanidis, A. A. (2002). Mathematical Modeling of Catalytic Fixed Bed Reactors, Ph.D. thesis, Twente
University Press.
Irani, M. (2014). Investigating the production of liquid fuels from synthesis gas (CO + H2) in a bench-
scale packed-bed reactor based on Fe-Cu-La/SiO2 catalyst: experimental and CFD modeling.
International Journal of Industrial Chemistry 5: 1–9.
Irankhah, A., Haghtalab, A., Farahani, E.V., and Sadaghianizadeh, K. (2007). Fischer-Tropsch reaction
kinetics of cobalt catalyst in supercritical phase. Journal of Natural Gas Chemistry 16: 115–120.
Jakobsen, H.A. (2014). Chemical Reactor Modeling, 1ste. New York: Springer Science + Business Media.
Jager, B. and Espinoza, R. (1995). Advances in low temperature Fischer-Tropsch synthesis. Catalysis
Today 23: 17–28.
Jess, A., Popp, R., and Hedden, K. (1999). Fischer–Tropsch-synthesis with nitrogen-rich syngas:
fundamentals and reactor design aspects. Applied Catalysis A: General 186: 321–342.
Jess, A. and Kern, C. (2009). Modeling of multi-tubular reactors for Fischer-Tropsch synthesis. Chemical
Engineering and Technology 32: 1164–1175.
Jess, A. and Kern, C. (2012a). Influence of particle size and single-tube diameter on thermal behavior of
Fischer-Tropsch reactors. Part I. Particle size variation for constant tube size and vice versa. Chemical
Engineering Technology 35: 369–378.
Jess, A. and Kern, C. (2012b). Influence of particle size and single-tube diameter on thermal behavior of
Fischer-Tropsch reactors. Part II. Eggshell catalysts and optimal reactor performance. Chemical
Engineering and Technology 35: 379–386.
Jess, A., Kragl, U., and Wasserscheid, P. (2013). Chemical Technology: An Integral Textbook, 1ste.
Weinheim: Wiley-VCH.
Jia-Yang, Y.Q., De Chen, J.Y., and Anders, H. (2015). Recent progresses in understanding of Co-based
Fischer–Tropsch catalysis by means of transient kinetic studies and theoretical analysis. Catalysis
Letters 145: 145–161.
Kaiser, P. and Jess, A. (2014). Modeling of multitubular reactors for iron- and cobalt-Catalyzed
Fischer-Tropsch syntheses for application in a power-to-liquid process. Energy Technology 2: 486–497.
Kaiser, P., Pöhlmann, F., and Jess, A. (2014). Intrinsic and effective kinetics of cobalt-catalyzed Fischer-
Tropsch synthesis in view of a power-to-liquid process based on renewable energy. Chemical and
Engineering Technology 37: 964–972.
Kang, J., Cheng, K., Zhang, L. et al. (2011). Mesoporous zeolite-supported ruthenium nanoparticles as
highly selective Fischer–Tropsch catalysts for the production of C5–C11 isoparaffins. Chemical
International Edition 50: 5200–5203.
Karimi, Z., Rahmani, M., and Moqadam, M. (2012). A study on vapour-liquid equilibria in Fischer-
Tropsch synthesis. Procedia Engineering 42: 25–33.
Kaskes, B., Vervloet, D., Kapteijn, F., and van Ommen, J. (2016). Numerical optimization of a structured
tubular reactor for Fischer-Tropsch synthesis. Chemical Engineering Journal 283: 1465–1483.
Keyser, M.J., Everson, R.C., and Espinoza, R.L. (2000). Fischer–Tropsch kinetic studies with cobalt–
manganese oxide catalysts. Industrial and Engineering Chemistry Research 39: 48–54.
Keyvanloo, K., Lanham, S.J., and Hecker, W.C. (2016). Kinetics of Fischer-Tropsch synthesis on
supported cobalt: effect of temperature on CO and H2 partial pressure dependencies. Catalysis Today
270: 9–18.
Khodakov, A.Y., Chu, W., and Fongarland, P. (2007). Advances in the development of novel cobalt
Fischer-Tropsch catalysts for synthesis of long-chain hydrocarbons and clean fuels. Chemical Reviews
107: 1692–1744.
Kim, Y.H., Jun, K.W., Joo, H. et al. (2009). A simulation study on gas-to-liquid (natural gas to Fischer-
Tropsch synthetic fuel) process optimization. Chemical Engineering Journal 155: 427–432.
Knottenbelt, C. (2002). Mossgas “gas-to-liquid” diesel fuels-an environmentally friendly option. Catalysis
Today 71: 437–445.
Koekemoer, A. and Luckos, A. (2015). Effect of material type and particle size distribution on
pressure drop in packed beds of large particles: extending the Ergun equation. Fuel 158:
(8.232–8.238).
Kölbel, H. and Ralek, M. (1980). The Fischer-Tropsch synthesis in the liquid phase. Catalysis Reviews 21:
225–274.
Kuipers, E.W., Scheper, C., Wilson, J.H. et al. (1996). Non-ASF product distributions due to secondary
reactions during Fischer-Tropsch synthesis. Journal of Catalysis 158: 288–300.
Kürten, H., Raasch, J., and Rumpf, H. (1966). Beschleunigung eines kugelförmigen feststoffteilchens im
strömungsfeld konstanter geschwindigkeit. Chemie Ingenieur Technik 38 (9): 941–948.
Kwack, S.H., Bae, J.W., Park, M.J. et al. (2011a). Reaction modeling on the phosphorous-treated Ru/Co/
Zr/SiO2 Fischer–Tropsch catalyst with the estimation of kinetic parameters and hydrocarbon
distribution. Fuel 90: 1383–1394.
Kwack, S., Park, M., Bae, J.W. et al. (2011b). Development of a kinetic model of the Fischer-Tropsch
synthesis reaction with a cobalt-based catalyst. Reaction Kinetic and Mechanism Catalysis Journal 104:
483–502.
Larachi, F., Laurent, A., Midoux, N., and Wild, G. (1991). Experimental study of a trickle-bed reactor
operating at high pressure: two-phase pressure drop and liquid saturation. Chemical Engineering
Science 46: 1233–1246.
References 427
Larachi, F., Belfares, L., Iliuta, I., and Grandjean, B.P.A. (2003). Heat and mass transfer in cocurrent gas-
liquid packed beds. Analysis, recommendations, and new correlations. Industrial and Engineering
Chemistry Research 42: 222–242.
Lee, B.I. and Kesler, M.G. (1975). A generalized thermodynamic correlation based on three-parameter
corresponding states. American Institute of Chemical Engineers Journal 21: 510–527.
Lee, H.J., Choi, J.H., Garforth, A., and Hwang, S. (2015). Conceptual design of a Fischer-Tropsch reactor
in a gas-to-liquid process. Industrial and Engineering Chemistry Research 54: 6749–6760.
Lee, T.S. and Chung, J.N. (2012). Mathematical modeling and numerical simulation of a Fischer-Tropsch
packed bed reactor and its thermal management for liquid hydrocarbon fuel production using biomass
syngas. Energy and Fuels 26: 1363–1379.
Levenspiel, O. (1999). Chemical Reaction Engineering, 3rde. New York: Wiley.
Liu, Z.T., Li, Y.W., Zhou, J.L., and Zhang, B.J. (1995). Journal of the Chemical Society. Faraday
Transactions 91: 3255–3261.
Lox, E.S. and Froment, G.F. (1993a). Kinetics of the Fischer-Tropsch reaction on a precipitated promoted
iron catalyst. 1. Experimental procedure and results. Industrial and Engineering Chemistry Research 32:
61–70.
Lox, E.S. and Froment, G.F. (1993b). Kinetics of the Fischer-Tropsch reaction on a precipitated promoted
iron catalyst. 2. Kinetic modeling. Industrial and Engineering Chemistry Research 32: 71–82.
Lozano-Blanco, G., Thybaut, J.W., Surla, K. et al. (2009). Simulation of a slurry-bubble column reactor for
Fischer-Tropsch synthesis using single-event microkinetics. American Institute of Chemical Engineers
Journal 55: 2159–2170.
Luque, R., de la Osa, A.R., Campelo, J.M. et al. (2012). Design and development of catalysts for Biomass-
To-Liquid-Fischer–Tropsch (BTL-FT) processes for biofuels production. Energy and Environmental
Science 5: 5186–5202.
Mabry, J. C. (2014). Utilization of cobalt catalyst for high temperature Fischer-Tropsch synthesis in a
fluidized bed reactor, Ph.D. thesis, College of Engineering in the Graduate School Southern Illinois
University Carbondale
Macías, M. and Ancheyta, J. (2004). Simulation of an isothermal hydrodesulfurization small reactor with
different catalyst particle shapes. Catalysis Today 98: 243–252.
Maitlis, P.M. and de Klerk, A. (2013). Greener Fischer-Tropsch processes for fuels and feedstocks, 1ste.
Weinheim: Wiley.
Mansouri, M., Atashi, H., Mirzaei, A., and Jangi, R. (2013). Kinetics of the Fischer-Tropsch synthesis on
silica-supported cobalt-cerium catalyst. International Journal Industrial Chemistry 4: 1.
Marano, J.J. and Holder, G.D. (1997). Characterization of Fischer-Tropsch liquids for vapor-liquid
equilibria calculations. Fluid Phase Equilibria 138: 1–21.
Marvast, M.A., Sohrabi, M., Zarrinpashne, S., and Baghmisheh, G. (2005). Fischer-Tropsch synthesis:
modeling and performance study for Fe-HZSM5 bifunctional catalyst. Chemical Engineering
Masuku, C.M., Hildebrandt, D., and Glasser, D. (2011). The role of vapour–liquid equilibrium in Fischer–
Tropsch product distribution. Chemical Engineering Science 66: 6254–6263.
Mears, D.E. (1971a). Diagnostic criteria for heat transport limitations in fixed bed reactors. Journal of
Catalysis 20: 127–131.
Mears, D.E. (1971b). Tests for transport limitations in experimental catalytic reactors. Industrial and
Engineering Chemistry Process Design and Development 10: 541–547.
Mederos, F., Elizalde, I., and Ancheyta, J. (2009). Steady-state and dynamic reactor models
for hydrotreatment of oil fractions: a review. Catalysis Reviews-Science and Engineering 51:
485–607.
Mehta, S. and Ricardez-Sandoval, L.A. (2016). Integration of design and control of dynamic systems
under uncertainty: a new back-off approach. Industrial and Engineering Chemistry Research 55:
485–498.
Méndez, C.I., Ancheyta, J., and Trejo, F. (2019). Importance of proper hydrodynamics modelling in fixed-
bed reactors: Fischer-Tropsch synthesis study case. The Canadian Journal of Chemical Engineering 97:
2685–2698.
Méndez, C.I., Trejo, F., and Ancheyta, J. (2022). On the use of steady-state optimal initial operating
conditions for control scheme implementation of a fixed-bed Fischer–Tropsch reactor. Arabian
Journal for Science and Engineering 47: 6099–6113.
Mikhailova, I.A., Fadeev, S.I., Khasin, A.A. et al. (2003). Calculating the liquid-gas equilibrium in the
Fischer-Tropsch synthesis. Theoretical Foundations of Chemical Engineering 37: 167–171.
Mirzaei, A.A., Sarani, R., Azizi, H.R. et al. (2015). Kinetics modeling of Fischer-Tropsch synthesis on the
unsupported Fe-Co-Ni (ternary) catalyst prepared using coprecipitation procedure. Fuel 140: 701–710.
Moazami, N., Mahmoudi, H., Panahifar, P. et al. (2015a). Mathematical modeling and performance study
of Fischer-Tropsch synthesis of liquid fuel over cobalt-silica. Energy Procedia 75: 62–71.
Moazami, N., Mahmoudi, H., Rahbar, K. et al. (2015b). Catalytic performance of cobalt–silica catalyst for
Fischer–Tropsch synthesis: effects of reaction rates on efficiency of liquid synthesis. Chemical
Moazami, N., Wyszynski, M.L., Mahmoudi, H. et al. (2015c). Modelling of a fixed bed reactor for Fischer–
Tropsch synthesis of simulated N2-rich syngas over Co/SiO2: hydrocarbon production. Fuel 154:
140–151.
Moazami, N., Wyzynski, M.L., Rahbar, K., and Tsolakis, A. (2017a). Parametric study and multi-objective
optimization of fixed-bed Fischer–Tropsch reactor: the improvement of FT synthesis products
formation and synthetic conversion. Industrial and Engineering Chemistry Research 34: 9446–9466.
Moazami, N., Wyszynski, M.L., Rahbar, K. et al. (2017b). A comprehensive study of kinetics mechanism
of Fischer-Tropsch synthesis over cobalt-based catalyst. Chemical Engineering Science 171: 32–60.
Montillet, A., Akkari, E., and Comiti, J. (2007). About a correlating equation for predicting pressure drops
through packed beds of spheres in a large range of Reynolds numbers. Chemical Engineering and
Processing 46: 329–333.
Mosayebi, A., Mehrpouya, M.A., and Abedini, R. (2016). The development of new comprehensive kinetic
modeling for Fischer-Tropsch synthesis process over Co-Ru/γ-Al2O3 nano-catalyst in a fixed-bed
reactor. Chemical Engineering Science Journal 286: 416–426.
Mousavi, S., Zamaniyan, A., Irani, M., and Rashidzadeh, M. (2015). Generalized kinetic model for iron
and cobalt based Fischer–Tropsch synthesis catalysts: review and model evaluation. Applied Catalysis
A: General 506: 57–66.
Moutsoglou, A. and Sunkara, P.P. (2011). Fischer-Tropsch synthesis in a fixed bed reactor. Energy and
Fuels 25: 2242–2257.
Mueller, G.E. (1992). Radial void fraction distributions in randomly packed fixed beds of uniformly sized
spheres in cylindrical containers. Powder Technology 72: 269–275.
Nakhaei Pour, A. and Housaindokht, M.R. (2014). Studies on product distribution of nanostructured iron
catalyst in Fischer-Tropsch synthesis: effect of catalyst particle size. Journal of Industrial and
Engineering Chemistry 20: 591–596.
Nakhaei Pour, A., Housaindokht, M.R., Irani, M., and Kamali Shahri, S.M. (2014a). Size-dependent
studies of Fischer-Tropsch synthesis on iron based catalyst: new kinetic model. Fuel 116: 787–793.
Nakhaei Pour, A., Khodabandeh, H., Izadyar, M., and Housaindokht, M.R. (2013). Mechanistic double
ASF product distribution study of Fischer-Tropsch synthesis on precipitated iron catalyst. Journal of
Natural Gas Science Engineering 15: 53–58.
References 429
Nakhaei Pour, A., Khodabandeh, H., Izadyar, M., and Housaindokht, M.R. (2014b). Detailed kinetics of
Fischer–Tropsch synthesis on a precipitated iron catalyst. Reaction Kinetics, Mechanisms and Catalysis
111: 29–44.
Nakhaei Pour, A. and Modaresi, S.M. (2016). Methane formation in Fischer-Tropsch synthesis: role of
nanosized catalyst particles. Journal of Nano Research 35: 39–54.
Novak, S., Madon, R.J., and Suhl, H. (1982). Secondary effects in the Fischer-Tropsch synthesis. Journal of
Catalysis 77: 141–151.
Oukaci, R., Singleton, A.H., and Goodwin, J.G. (1999). Applied Catalysis A: General 186: 129–144.
Outi, A., Rautavuoma, I., and van der Baan, H.S. (1981). Kinetics and mechanism of the Fischer Tropsch
hydrocarbon synthesis on a cobalt on alumina catalyst. Applied Catalysis 1: 247–272.
Ozahi, E., Gundogdu, M.Y., and Carpinlioglu, M.Ö. (2008). A modification on Ergun’s correlation for use
in cylindrical packed beds with non-spherical particles. Advanced Powder Technology 19: 369–381.
Park, N., Kim, J.R., Yoo, Y. et al. (2014). Modeling of a pilot-scale fixed-bed reactor for iron-based
Fischer–Tropsch synthesis: two-dimensional approach for optimal tube diameter. Fuel 122: 229–235.
Park, S., Jung, I., Lee, U. et al. (2015). Design and modeling of large-scale cross-current multichannel
Fischer-Tropsch reactor using channel decomposition and cell-coupling method. Chemical
Pearson, J. (1959). A note on the “Danckwerts” boundary conditions for continuous flow reactors.
Petersen, E.E. (1965). A general criterion for diffusion influenced chemical reactions in porous solids.
Philippe, R., Lacroix, M., Dreibine, L. et al. (2009). Effect of structure and thermal properties of a Fischer-
Tropsch catalyst in a fixed bed. Catalysis Today 147: S305–S312.
Pinna, D., Tronconi, E., Lietti, L. et al. (2002). Review of kinetic models for the Fischer-Tropcsh synthesis.
La Rivista dei Combustibili 56: 69–85.
Pistikopoulos, E.N. and Mazzuchi, T.A. (1990). A novel flexibility analysis approach for processes with
stochastic parameters. Computer and Chemical Engineering 14: 991–1000.
Pöhlmann, F. and Jess, A. (2016a). Influence of syngas composition on the kinetics of Fischer-Tropsch
synthesis of using cobalt as catalyst. Energy Technology 4: 55–64.
Pöhlmann, F. and Jess, A. (2016b). Interplay of reaction and pore diffusion during cobalt-catalyzed
Fischer–Tropsch synthesis with CO2-rich syngas. Catalysis Today 275: 172–182.
Pöhlmann, F., Kern, C., Rößler, S., and Jess, A. (2016). Accumulation of liquid hydrocarbons in catalyst
pores during cobalt-catalyzed Fischer-Tropsch synthesis. Catalysis and Science Technology 6: 6593–6604.
Pollard, W.G. and Present, R.D. (1949). On gaseous self-diffusion in long capillary tubes. Physical Review
73: 762–774.
Post, M.F.M., Van’t Hoog, A.C., Minderhoud, J.K., and Sie, S.T. (1989). Diffusion limitations in Fischer-
Tropsch catalysts. American Institute of Chemical Engineers Journal 35: 1107–1114.
Puskas, I., Hurlbut, R.S., and Pauls, R.E. (1993). Telomerization model for cobalt-catalyzed Fischer-
Tropsch products. Journal of Catalysis 139: 591–601.
Qian, W., Zhang, H., Ying, W., and Fang, D. (2013). The comprehensive kinetics of Fischer-Tropsch
synthesis over a Co/AC catalyst on the basis of CO insertion mechanism. Chemical Engineering Journal
228: 526–534.
Rados, N., Dahhan, M.H., and Dudukovic, M.P. (2003). Modeling of the Fischer-Tropsch synthesis in
slurry bubble column reactors. Catalysis Today 79–80: 211–218.
Rafiei-Shishavan, M., Mehta, S., and Ricardez-Sandoval, L.A. (2017). Simultaneous design and control
under uncertainty: a back-off approach using power series expansions. Computer and Chemical
Rafiq, M., Jakobsen, H., Schmid, R., and Hustad, J. (2011). Experimental studies and modeling of a fixed
bed reactor for Fischer-Tropsch synthesis using biosyngas. Fuel Processing Technology 92: 893–907.
Rahimpour, M.R. and Elekaei, H. (2009). A comparative study of combination of Fischer–Tropsch
synthesis reactors with hydrogen-permselective membrane in GTL technology. Fuel Processing
Rahimpour, M.R., Mirvakili, A., and Paymooni, K. (2011a). A novel water perm-selective membrane
dual-type reactor concept for Fischer–Tropsch synthesis of GTL (gas to liquid) technology. Energy 36:
1223–1235.
Rahimpour, M.R., Mirvakili, A., Paymooni, K., and Moghtaderi, B. (2011b). A comparative study between
a fluidized-bed and a fixed-bed water perm-selective membrane reactor with in situ H2O removal for
Fischer–Tropsch synthesis of GTL technology. Journal of Natural Gas Science and Engineering 3 (3):
484–495.
Rangel, N., Santos, A., and Pinho, C. (2001). Pressure drop in packed shallow beds of cylindrical cork
stoppers. Chemical Engineering Research and Design 79: 547–552.
Rao, V.U.S., Stiegel, G.J., Cinquegrane, G.J., and Srivastava, R.D. (1992). Iron-based catalysts for slurry-
phase fischer-tropsch process: technology review. Fuel Processing Technology 30: 83–107.
Rase, H.F. (1990). Fixed-Bed reactor design and diagnostics, 1ste. New York: Butterworth Publishers.
Reichelt, W. (1972). Zur berechnung des druckverlustes einphasig durchströmter kugel- und
zzylinderschüttungen. Chemie Ingenieur Technik 44: 1068–1071.
Reid, R.C., Sherwood, T.K., and Prausnitz, J. (1977). The Properties of gases and liquids, 3rde. United
states: Mc Graw Hill.
Reuel, C.R. and Bartholomew, C.H. (1984). Effects of support and dispersion on the CO hydrogenation
activity/selectivity properties of cobalt. Journal of Catalysis 85: 78–88.
Rohde, M.P., Schaub, G., Khajavi, S. et al. (2008). Fischer–Tropsch synthesis with insitu H2O removal –
directions of membrane development. Microporous and Mesoporous Materials 115: 123–136.
Rytter, E., Tsakoumis, N.E., and Holmen, A. (2016). On the selectivity to higher hydrocarbons in
Co-based Fischer-Tropsch synthesis. Catalysis Today 261: 3–16.
Saeidi, S., Talebi Amiri, M., Saidina Amin, N.A., and Rahimpour, M.R. (2014). Progress in reactors for
high-temperature Fischer–Tropsch process: determination place of intensifier reactor perspective.
International Journal Chemical Reactor Engineering 12: 639–664.
Salmi, T., Wärna, J., Toppinen, S. et al. (2000). Brazilian Journal of Chemical Engineering 17: 1023–1035.
Sandler, S.I. (2006). Chemical, Biochemical, and Engineering Thermodynamics, 5the. United States: Wiley.
Sarup, B. and Wojciechowski, B.W. (1989). Studies of the fischer-tropsch synthesis on a cobalt catalyst II.
Kinetics of carbon monoxide conversion to methane and to higher hydrocarbons. Canadian Journal of
Chemical Engineering 67: 62–74.
Satterfield, C.N., Huff, G.A. Jr., and Longwell, J.P. (1982). Product distribution from iron catalysts in
Fischer-Tropsch slurry reactors. Industrial and Engineering Chemistry Process Design and Development
21: 465–470.
Schanke, D., Hilmen, A.M., Bergene, E. et al. (1995). Study of the deactivation mechanism of Al2O3-
supported cobalt Fischer-Tropsch catalyst. Catalysis Letters 34: 269–284.
Schulz, H. (1999). Short history and present trends of Fischer-Tropsch synthesis. Applied Catalysis A:
General 186: 3–12.
Schulz, H. and Claeys, M. (1999). Reactions of α-olefins of different chain length added during Fischer-
Tropsch synthesis on a cobalt catalyst in a slurry reactor. Applied Catalysis A: General 186: 71–90.
Sharma, A., Philippe, R., Luck, F., and Schweich, D. (2011). A simple and realistic fixed bed model for
investigating Fischer-Tropsch catalyst activity at lab-scale and extrapolating to industrial conditions.
References 431
Sie, S.T. (1998). Process development and scale-up: part 4. case history of the development of a Fischer–
Tropsch synthesis. Chemistry; Reviews in Chemical Engineering 14: 109–157.
Singh, J. and Gu, S. (2010). Biomass conversion to energy in India – a critique. Renewable and Sustainable
Energy Reviews 14: 1367–1378.
Smith, J. (1981). Chemical Engineering Kinetics, 3rde. United States: McGraw Hill.
Soroush, M. and Kravaris, C. (1993a). Optimal design and operation of batch reactors. 1. Theoretical
framework. Industrial and Engineering Chemistry Research 32: 866–881.
Soroush, M. and Kravaris, C. (1993b). Optimal design and operation of batch reactors. 2. A case study.
Stephanopoulos, G. (1985). Chemical Process Control an introduction to theory and practice, 1ste. New
Jersey: Prentice Hall.
Storsaeter, S., Borg, O., Blekkan, E., and Holmen, A. (2005). Study of the effect of water on Fischer–
Tropsch synthesis over supported cobalt catalysts. Journal of Catalysis 231 (2): 405–419.
Sun, B., Yu, G., Lin, J. et al. (2012). A highly selective Raney Fe@HZSM-5 Fischer–Tropsch synthesis
catalyst for gasoline production: one-pot synthesis and unexpected effect of zeolites. Catalysis Science
and Technology 2: 1625–1629.
Surisetty, V.R., Dalai, A.K., and Kozinski, J. (2010). Alkali-promoted trimetallic Co−Rh−Mo sulfide
catalysts for higher alcohols synthesis from synthesis gas: comparison of MWCNT and activated
carbon supports. Industrial Engineering Chemistry Research 49: 6956–6963.
Tavasoli, A., Trépanier, M., Dalai, A.K., and Abatzoglou, N. (2010). Effects of confinement in carbon
nanotubes on the activity, selectivity, and lifetime of Fischer-Tropsch Co/carbon nanotube catalysts.
Journal of Chemical Engineering Data 55: 2757–2763.
Thiele, E.W. (1939). Relation between catalytic activity and size of particle. Industrial and Engineering
Chemistry 31: 916–920.
Tobis, J. and Ziókowski, D. (1988). Modelling of heat transfer at the wall of a packed-bed apparatus.
Todic, B., Ma, W., Jacobs, G. et al. (2014). Effect of process conditions on the product distribution of
Fischer–Tropsch synthesis over a re-promoted cobalt-alumina catalyst using a stirred tank slurry
reactor. Journal of Catalysis 311: 325–338.
Todic, B., Ordomsky, V.V., Nikacevic, N.M. et al. (2015). Opportunities for intensification of Fischer–
Tropsch synthesis through reduced formation of methane over cobalt catalysts in microreactors.
Catalysis Science and Technology 5: 1400–1411.
Tosun, I. (2007). Modeling in transport phenomena, 2nde. Elsevier Science and Technology Books.
Trépanier, M., Tavasoli, A., Dalai, A.K., and Abatzoglou, N. (2009). Co, Ru and K loadings effects on the
activity and selectivity of carbon nanotubes supported cobalt catalyst in Fischer–Tropsch synthesis.
Tsakoumis, N.E., Rønning, M., Borg, O. et al. (2010). Deactivation of cobalt based Fischer-Tropsch
catalysts: a review. Catalysis Today 154: 162–182.
Tyreus, B.D. and Luyben, W.L. (1992). Tuning PI controllers for integrator/dead time processes.
Valero-Romero, M.J., Sartipi, S., Sun, X. et al. (2016). Carbon/H-ZSM-5 composites as supports for
bi-functional Fischer–Tropsch synthesis catalysts. Catalysis Science Technology 6: 2633–2646.
Van Antwerpen, W.d., Toit, C., and Rousseau, P. (2010). A review of correlations to model the packing
structure and effective thermal conductivity in packed beds of mono-sized spherical particles. Nuclear
Engineering and Design 240: 1803–1818.
Van Berge, P. and Everson, R. (1997). Cobalt as an alternative Fischer-Tropsch catalyst to iron for the
production of middle distillates. Studies in Surface Science and Catalysis 107: 207–212.
Van Der Laan, G.P. and Beenackers, A.A.C.M. (1999). Kinetics and selectivity of the Fischer-Tropcsch
synthesis: a literature review. Catalysis Reviews 41: 255–318.
Van Dijk, H. A. (2001). The Fischer-Tropsch synthesis: a mechanistic study using transient isotopic
tracing, Ph. D. Thesis, Technische Universiteit Eindhoven, Eihdhoven, the Netherlands
Van Vliet, O.P., Faaij, A.P., and Turkenburg, W.C. (2009). Fischer–Tropsch diesel production in a well-to-
wheel perspective: a carbon, energy flow and cost analysis. Energy Conversion and Management 50:
855–876.
Vannice, M.A. (1976). The catalytic synthesis of hydrocarbons from carbon monoxide and hydrogen.
Catalysis Reviews 14: 153–191.
Visconti, C.G., Tronconi, E., Lietti, L. et al. (2007). Development of a complete kinetic model for the
Fischer-Tropsch synthesis over Co/Al2O3 catalysts. Chemical Engineering Science 62: 5338–5343.
Visconti, C.G., Tronconi, E., Groppi, G. et al. (2011a). Monolithic catalysts with high thermal conductivity
for the Fischer-Tropsch synthesis in tubular reactors. Chemical Engineering Journal 171: 1294–1307.
Visconti, C.G., Tronconi, E., Lietti, L. et al. (2011b). Detailed kinetics of the Fischer–Tropsch synthesis on
cobalt catalysts based on H-assisted CO activation. Topics in Catalysis 54: 786–800.
Visconti, C.G. and Mascellaro, M. (2013). Calculating the product yields and the vapor-liquid equilibrium
in the low-temperature Fischer–Tropsch synthesis. Catalysis Today 214: 61–73.
Vollmari, K., Oschmann, T., Wirtz, S., and Kruggel-Emden, H. (2015). Pressure drop investigations in
packings of arbitrary shaped particles. Powder Technology 271: 109–124.
Vosloo, A.C. (2001). Fischer–Tropsch: a futuristic view. Fuel processing Technology 71: 149–155.
Wang, Y.N., Li, Y.W., Bai, L. et al. (1999). Correlation for gas–liquid equilibrium prediction in Fischer-
Tropsch synthesis. Fuel 78: 911–917.
Wang, Y.N., Xu, Y.Y., Xiang, H.W. et al. (2001). Modeling of catalyst pellets for Fischer-Tropsch synthesis.
Wang, Y.N., Ma, W.P., Lu, Y.J. et al. (2003a). Kinetics modelling of Fischer-Tropsch synthesis over an
industrial Fe-Cu-K catalyst. Fuel 82: 195–213.
Wang, Y.N., Xu, Y.Y., Li, Y.W. et al. (2003b). Heterogeneous modeling for fixed-bed Fischer–Tropsch
synthesis: reactor model and its applications. Chemical Engineering Science 58 (3-6): 867–875.
Wärnå, J. and Salmi, T. (1996). Dynamic modelling of catalytic three phase reactors. Computers and
Chemical Engineering 20: 39–47.
Welty, J.R., Wicks, C.E., Wilson, R.E., and Rorrer, G.L. (2007). Fundamentals of momentum, heat, and
mass transfer, 5the. USA: Wiley.
Whitaker, S. (1973). The transport equations for multi-phase systems. Chemical Engineering Science 28:
139–147.
White, S.M. and Tien, C.L. (1987). Analysis of flow channeling near the wall in packed beds. Wärme- und
Stoffübertragung 21: 291–296.
Wijngaarden, R.J., Kronberg, A., and Westerterp, K.R. (1998). Industrial catalysis: optimizing catalyst and
processes, 1ste. Germany: Wiley.
Wilhelm, R.H. (1962). Progress towards the a priori design of chemical reactors. Pure and Applied
Chemistry 5: 403–421.
Wilhelm, D.J., Simbeck, D.R., Karp, A.D., and Dickenson, R.L. (2001). Syngas production for gas-to-
liquids applications: technologies, issues and outlook. Fuel Processing Technology 71: 8.139–8.148.
Wilke, C.R. (1949). Estimation of liquid diffusion coefficients. Chemical Engineering Progress 45: 218–224.
Wojciechowski, B.W. (1988). The kinetics of the Fischer-Tropsch synthesis. Catalysis Reviews 30:
629–702.
Yagi, S. and Wakao, N. (1959). Heat and mass transfer from wall to fluid in packed beds. American
Institute of Chemical Engineers Journal 5: 79–85.
References 433
Yan, F., Qian, W., Sun, Q. et al. (2014). Product distributions and olefin-to-paraffin ratio over an iron-
based catalyst for Fischer–Tropsch synthesis. Reaction Kinetics, Mechanisms and Catalysis 113:
471–485.
Yan, Z., Wang, Z., Bukur, D.B., and Goodman, D.W. (2009). Fischer–Tropsch synthesis on a model
Co/SiO2 catalyst. Journal of Catalysis 268: 196–200.
Yang, C.H., Massoth, F.E., and Oblad, A.G. (1979). Kinetics of CO + H2 reaction over Co-Cu-Al2O3
catalyst, Hydrocarbon Synthesis from Carbon Monoxide and Hydrogen. American Chemical Society
178: 35–46.
Yang, J., Liu, Y., Chang, J. et al. (2003). Detailed kinetics of Fischer-Tropsch synthesis on an industrial
Fe-Mn catalyst. Industrial and Engineering Chemistry Research 42: 5066–5090.
Yates, I.C. and Satterfield, C.N. (1991). Intrinsic kinetics of the Fischer-Tropsch synthesis on a cobalt
catalyst. Energy and Fuel 5: 168–173.
Zennaro, R., Tagliabue, M., and Bartholomew, C.H. (2000). Kinetics of Fischer–Tropsch synthesis on
titania-supported cobalt. Catalysis Today 58: 309–319.
Zhang, Q., Kang, J., and Wang, Y. (2010). Development of novel catalysts for Fischer-Tropsch synthesis:
tuning the product selectivity. ChemCatChem: The European Society Journal for Catalysis 2:
1030–1058.
Zhang, Q., Deng, W., and Wang, Y. (2013). Recent advances in understanding the key catalyst factors for
Fischer-Tropsch synthesis. Journal of Energy Chemistry 22: 27–38.
Zhen, Y., Zhou, J.W., Bukur, D.B., and Goodman, D.W. (2009). Fischer-Tropsch synthesis on a model
Co/SiO2 catalyst. Journal of Catalysis 268: 196–200.
Zhou, L., Froment, G.F., Yang, Y., and Li, Y. (2016). Advanced fundamental modeling of the kinetics of
Fischer–Tropsch synthesis. American Institute of Chemical Engineers Journal 62: 1668–1682.
Zhu, X., Lu, X., Liu, X. et al. (2010). Study of radial heat transfer in a tubular Fischer-Tropsch synthesis
reactor. Industrial and Engineering Chemistry Research 49: 10682–10688.
Zimmerman, W.H. and Bukur, D.B. (1990). Reaction kinetics over iron catalyst used for Fischer-Tropsch
synthesis. The Canadian Journal of Chemical Engineering 68: 292–301.
434
Computational Fluid Dynamics Modeling of Mass Transfer Processes

in Structured Beds of Microfibrous Catalysts
Sergey Lopatin1,2, Andrey Elyshev2, and Andrey Zagoruiko1,2
1
2
9.1 Introduction
Different and practically important catalytic processes are significantly influenced by heat and mass
transfer limitations, and so the development of new catalysts with improved mass transfer proper-
ties is an important task. A possible promising solution in this area is the use of structured catalytic
systems using a catalyst based on microfiber carriers made in the form of webs or textiles. Micro-
fibers can be made from glass, minerals, ceramics, carbon, polymers, and other materials. The
active components for such catalysts can be selected from a wide range of different noble metals
or transition metal oxides depending on the requirements of the particular process.
The most widely used and well-studied microfiber catalysts are glass-fiber catalysts (GFCs),
which use glass-fiber cloth as a carrier (Balzhinimaev et al. 2010; Zagoruiko and Lopatin 2019).
Such catalysts have demonstrated promising catalytic properties in reactions, such as deep oxida-
tion of hydrocarbons, organic and organohalide compounds (Balzhinimaev et al. 2010; Paukshtis
et al. 2010; Lopatin et al. 2015), oxidation of SO2 and H2S (Mikenin et al. 2016; Golyashova et al.
2019), low-temperature reduction of NOx (Li et al. 2014; Golyashova et al. 2023), oxychlorination of
light olefins (Shalygin et al. 2013), synthetic gas production (Aldashukurova et al. 2013), selective
hydrogenation (Salmi et al. 2000; Gulyaeva et al. 2015), and so on.
The features of microfiber catalysts include their original geometric shape, high mechanical
strength and flexibility, as well as the ability to be arranged into structured systems of various geo-
metries and sizes, which make it possible to implement fundamentally new options for catalytic
processes.
Various structured GFC packages have been proposed earlier. The most intuitive packaging
design for GFC is the multilayer mat or multilayer GFC textiles set. The reaction fluid in these
packages flows through the GFC layers. Although it was expected that such a design should be char-
acterized by a fairly high mass transfer, in reality, such packages may have insufficient efficiency
due to the suboptimal structure of the internal flow of the reaction mixture (Zagoruiko and Lopatin
2019) and may also have an unreasonably high pressure drop.
GFC-structured cartridges with axial movement of the reaction mixture along the catalytic
textile surface are made of parallel layers of GFC alternating with layers of flat and corrugated metal
wire-mesh to ensure the regular channel structure and mechanical strength of the cartridge. Such
cartridges may have the outer shape of a circular cylinder with spirally wound layers of glass-fiber/
mesh or a prism with square, rectangular, triangular, and other cross-sectional shapes. Spiral
shapes are suitable for round reactors with a relatively small diameter, while prismatic cartridges
can be used to build catalyst beds of almost any shape and size in large reactors. GFC axial flow
cartridges were experimentally studied in our earlier work (Zagoruiko et al. 2017; Zagoruiko
and Lopatin 2019). It was shown that such structures are characterized by a high intensity of mass
transfer and an extremely low pressure drop. Many peculiar features and properties of such systems
have been discovered as well. It has been suggested that the high mass transfer efficiency may be
due to the flexibility of the GFC fabric, making possible the mechanical interaction between the
textile and the reaction flow, resulting in the variability of the effective area of GFC washed by
the flow. The study also revealed a high sensitivity of the cartridge’s mass transfer characteristics
to the details of the cartridge’s internal geometry, which is a possible cause of problems with the
reproducibility of experimental results. On one hand, such non-reproducibility and some other
complications typical of such catalysts can be considered as a disadvantage of GFCs, but on the
other hand, a deep understanding and correct presentation of these details are potential sources
of new solutions and improvements in the field of GFC packaging and reactor designs based
on them.
Experimental work is an important part of a mass transfer research; however, further optimiza-
tion of such cartridges and the development of new designs also require the use of modern
advanced modeling methods, including those based on computational fluid dynamics (CFD) sim-
ulation. This method has already found application in the simulation of catalytic systems of various
structures and scales, ranging from traditional catalyst pellets in fixed-beds and tubular reactors
(Dixon et al. 2008; Karthik and Buwa 2019; Yang et al. 2021; Bouras et al. 2022) to structured cat-
alysts, including monolithic blocks (You et al. 2019; Klenov et al. 2019), catalysts based on foamed
carriers (Bracconi et al. 2018; Dong et al. 2018; Sinn et al. 2019; Wehinger et al. 2019; George et al.
2022), wire-mesh catalytic blocks (Iwaniszyn et al. 2021b; Zazhigalov et al. 2022), and structured
catalysts of other forms (Iwaniszyn et al. 2021a; Jurković et al. 2021; Tedesco and Moraes 2021).
The importance of CFD modeling of structured beds of micro-fibrous catalysts lies in its ability to
enhance understanding and optimization of the catalytic processes. By accurately simulating the
complex flow and reaction phenomena within these structured beds, CFD modeling enables
researchers and engineers to analyze the performance of the catalysts, improve their design, and
ultimately increase their efficiency and effectiveness in various industrial applications. This mod-
eling approach also facilitates the identification of potential issues, such as pressure drop and heat
transfer, and supports the development of innovative solutions to overcome them. Overall, CFD
modeling plays a crucial role in the advancement of the development and application of structured
beds of micro-fibrous catalysts, leading to significant advancements in catalytic technology.
Until recently, the number of CFD studies of GFC-based catalytic structures was close to zero.
CFD has been used to simulate multilayer GFC packaging with a propagating flow of gaseous reac-
tion mixture (Chub et al. 2005; Zagoruiko and Lopatin 2019), but the simulation of GFC cartridges
with axial flow has been done for the first time quite recently (Lopatin et al. 2023). This study made
it possible to reproduce the main experimentally observed regularities, as well as to provide qual-
itative confirmation of experimental hypotheses about the unusual hydrodynamic properties
of GFCs.
This chapter outlines the most recent achievements in the field of CFD modeling of GFC-based
catalytic cartridges.
436 9 Computational Fluid Dynamics Modeling of Mass Transfer Processes in Structured Beds of Microfibrous Catalysts
9.2 Mathematical Model
The object of the simulation was structured catalytic cartridges consisting of glass-fiber catalytic
fabrics and structuring metal meshes. Two basic designs of such cartridges were considered: one
with plain and corrugated meshes (Figure 9.1a) and the other with only simple meshes
(Figure 9.1b).
It is assumed that the gaseous reaction flux moves parallel with the GFC planes in the spaces
formed by the structuring grids.
Two main types of glass fabric were used in the modeling: satin (Figure 9.2a), consisting of single,
tightly intertwined threads, and openwork (Figure 9.2b), made of bundles of threads with signif-
icant holes between them.
(a) (b)
GFC
Corrugated
mesh
Plain
mesh
Figure 9.1 Structured cartridges with glass-fiber catalyst: (a) with corrugated and simple structuring meshes
and (b) with simple meshes only. The arrows show the direction in which the fluid is moving.
(a) (b)
Figure 9.2 The geometric types of the GFC under consideration: (a) satin and (b) openwork.
9.2.1 Model Description

A complete system of equations describing the motion of a viscous fluid in a Cartesian system is
discussed.
The coordinate system can be represented in tensor form, as follows:
• Equation of continuity
∂U j
∂x j
=0 91
• Equations of momentum change (Navier–Stokes equations) averaged by Reynolds:
• ρ
∂U i
∂t
+ ρU j
∂U i
∂x j
= −
∂P
∂x j
+μ
∂2 U i
∂x j ∂x i
−
∂Pui uj
∂x j
92
where t is time; xi are Cartesian coordinates (i = 1, 2, 3), Ui are the components of the average veloc-
ity of fluid flow in the direction xi, P is the average static pressure, ρ is the density, and μ is the
dynamic viscosity coefficient.
The Boussinesq conjecture offers a relationship relating Reynolds stresses to average velocity
gradients:
• − ρui uj = μt
∂U i
∂x j
+
∂U j
∂x i
−
2
μ
∂U i
3 t ∂x i
+ ρk δij 93
ui uj
where k = – kinetic energy of turbulent pulsations; μt – turbulent (or vortex) viscosity, and
2
δij – delta function:
• δij =
1, i = j
0, i j
94
The turbulent viscosity coefficient was determined from the selected turbulence model. To close
the Reynolds-averaged Navier–Stokes equations, SST transition turbulence models were used.
Uniform distributions of gas flow velocity, temperature, and mass fraction of each component were
used as boundary conditions on the input section of the simulated system. In the flow outlet area, the
conditions of the “pressure outlet limit” were established, and the gauge pressure value was zero.
Temperature dependencies of thermodynamic quantities and viscosity are also considered.
9.2.2 Computing Domain

The commercial software, Ansys Fluent (version 2021 R2, license No. 1148156 for Tyumen State
University), with a mesh module was used for CFD simulation of described objects. The parameters
of the computational grid were selected in such a manner that the vortices were resolved over the
volume of the cartridge, and the resolution of complex geometry was satisfactory. The mesh is built
evenly over the volume to allow vortex movements throughout the cartridge volume. A volumetric
mesh is created using the Mosaic method with polyhedral elements. The minimum mesh size is
10−3 mm. The minimum spelling quality is 0.30, and the maximum aspect ratio is 22.88.
9.2.3 Simulation Object Geometry

The GFC cartridge is made up of similar layers. Assuming that the inlet stream is evenly distributed
among the layers, only one layer can be considered for the simulation of the entire cartridge. A part
of this GFC cartridge layer is shown in Figure 9.3a. The corrugated mesh forms a wave with a
(a)
(b)
Glass-fiber
catalyst
Inlet Corrugated
mesh
Outlet
Figure 9.3 Part of the GFC cartridge layer (a) and a reduced simulation segment (b). Arrows show the gas
flow movement.
repetitive geometry; so it is possible to minimize the segment for modeling using the symmetry of
the channels, thereby reducing the simulated segment to the section of the channel between the
upper and lower positions of the corrugated mesh (Figure 9.3b).
To simulate a cartridge with only flat structures, one can use the same geometry after removing
the corrugated wire-mesh. The wire-meshes were modeled with a detailed reproduction of their
structure as a combination of round wires.
The simulated segment corresponds to the GFC cartridge used in the experiments. The length of
the GFC textile was 44 mm, and the total length of the simulated channel with the addition of inlet
and internal voids was 66 mm. The width of the channel was taken to be 8.0 mm, and the total
height was 5.7 mm.
GFC fabrics have complex geometry. Intuitively, threads in textiles can be thought of as cir-
cular in shape in the cross section, but they are actually much closer to flattened ellipsoids. To
make it easier to build a geometric model, they can even be considered rectangular. The satin-
type fabric model can be further reduced to almost a layer of 0.4 mm thickness, formed by inter-
secting rectangular bodies of 0.8 mm wide with small holes (0.1 × 0.1 mm) and small grooves
between them (Figure 9.4a). Openwork textiles, which are easier to model, can be represented
by a more complex geometry with threads in the form of flattened ellipsoids (1.0 mm wide
and 0.45 mm high) with larger holes (1.5 × 1.5 mm) between them (Figure 9.4b). These values
correspond to the geometric parameters of real GFC textiles used in mass transfer
experiments.
(a) (b)
y
y
x
x
z
z
Figure 9.4 Geometric model of satin (a) and openwork and (b) GFC textiles.
Figure 9.5 An example of an internal cartridge of detailed geometry with a computational mesh.
Figure 9.5 shows an example of the geometric structure of a cartridge showing the use of a com-
putational mesh. The total number of cells in the created mesh has reached 83696753, the face has
reached 295431252, and the number of nodes has reached 135098549.
Similar models of cartridges with varying channel heights were built for simulations.
9.2.4 Reaction
The deep oxidation reaction of toluene vapor in air on a Pt-based GFC was taken as a test reaction
for CFD simulation:
• C7 H8 + 9O2 7CO2 + 4H2 O 95
This reaction was previously used for an experimental study of mass transfer in GFC-based car-
tridges (Zagoruiko et al. 2017; Zagoruiko and Lopatin 2019); so this reaction was chosen to simplify
the comparison of simulated and experimental data.
The rate of this reaction can be calculated using the first-order velocity equation with respect to
the concentration of toluene:
• W = k0 e − RT CC7H8
E
96
where W is the reaction rate, k0 is the pre-exponent of rate constant, E is the activation energy, R is
the universal gas constant, T is the temperature and, CC7H8 is the molar concentration of toluene.
The reaction was assigned to the textile surface GFC (Wall Surface reaction type). The kinetic
parameters were taken from our experimental data (Zagoruiko et al., 2017; Zagoruiko and Lopatin
2019) with a corresponding conversion from units related to the mass of GFC to units related to the
external surface area of GFC: k0 = 9.0∗108 kmol/m2/s, E = 125.6 kJ/mol.
9.2.5 Model Parameters

The simulation was performed for a flow with an inlet temperature of 350 C. Experimental studies
(Zagoruiko et al. 2017; Zagoruiko and Lopatin 2019) have shown that this temperature ensures the
achievement of the mass transference regime, when the apparent reaction rate is completely deter-
mined by external mass transfer constraints. The basic inlet fluid velocity, recalculated to standard
conditions, was assumed to be 0.55 m/s with further variation in some simulation studies. The basic
inlet toluene content was set at 100 vppm.
9.3 Simulation Results
9.3.1 Cartridge Channel with Corrugated Structuring Mesh

Figure 9.6 illustrates the distribution of flow parameters in a cartridge channel with satin GFC and a
corrugated structuring mesh. The contours are presented in 2D for the middle section of the chan-
nel, in a plane that is positioned vertically along the channel cartridge, intersecting the corrugated
mesh at half its height.
The static pressure (Figure 9.6a) decreases along the channel from the inlet to the outlet, being
almost uniform over the entire height of the channel. This uniformity may be due to the very low
pressure drop (below 5 Pa) in the channel. The fluid velocity contours (Figure 9.6b) show intense fluid
flows above and below the corrugated mesh with very low flow rates in the mesh and near the GFC
surface. A very slight increase in temperature is observed along the channel due to heat emission in
exothermic oxidation reaction, though the heat effect is very low due to insignificant concentration of
toluene (Figure 9.6c). The concentration of toluene (Figure 9.6d) decreases along the length of the
channel due to its consumption in the reaction. Minimum concentrations are observed near the
GFC surfaces and maximum concentrations near the channel axis. Obviously, such a distribution
is quite expected, and it is formed by the interaction of toluene consumption in the surface reaction
to GFC with the mass transfer of toluene from the flow core in the axial region to these surfaces.
The features of the spatial distribution of flow parameters may be better seen in the 3D represen-
tation (Figure 9.7). It can be seen (Figure 9.7a) that the corrugated mesh divides the flow into two
streams on both sides of the mesh. The maximum flow rate is observed at the cross-sectional corners
of the channel furthest from the mesh, and the minimum velocity is observed near the corrugated
mesh and GFC surfaces; this deceleration is the result of the surface friction of the flow.
The concentration of toluene (Figure 9.7b), as already shown in Figure 9.6d, is maximized in the
fluid core and minimized near the surfaces of the GFC. 3D representation shows that such a
(a)
Static Pressure
[Pa]
1.86e+00
1.66e+00
1.46e+00
1.26e+00
1.06e+00
8.64e–01
6.63e–01
4.63e–01
2.63e–01
6.26e–02
–1.38e–01
(b)
Velocity Magnitude
[m/s]
1.10e+00
9.89e–01
8.79e–01
7.69e–01
6.59e–01
5.49e–01
4.39e–01
3.30e–01
2.20e–01
1.10e–01
0.00e+00
(c)
Static Temperature
[C]
350.00009
350.00003
349.99997
349.99991
349.99985
349.99979
349.99979
349.99973
349.99966
349.99960
349.99954
(d)
Mass fraction of c7h8
1.00e–04
9.00e–05
8.00e–05
7.00e–05
6.00e–05
5.00e–05
4.00e–05
3.00e–05
2.00e–05
1.00e–05
2.37e–14
Figure 9.6 Spatial distribution of fluid parameters in the middle sequence of a channel with a corrugated
structuring mesh and a satin-type catalyst. (a) Static pressure, (b) velocity of the fluid, (c) static temperature, and
(d) mass fraction of the toluene. The dashed lines at the top and bottom of the channel correspond to the GFC
layers, the dotted-dotted line in the middle is a cross section of the corrugated mesh.
(a)
Velocity Magnitude
[m/s]
1.09e+00
9.78e–01
8.69e–01
7.61e–01
6.52e–01
5.43e–01
4.35e–01
3.26e–01
2.17e–01
1.09e–01
0.00e+00
Contour-3
(b)
1.00e–04
9.00e–05
8.00e–05
7.00e–05
6.00e–05
5.00e–05
4.00e–05
3.00e–05
2.00e–05
1.00e–05
6.57e–15
Contour-3
(c)
Kinetic Rate of Reaction-1
[kgmol/(m^2s)]
3.64e–07
3.28e–07
2.92e–07
2.55e–07
2.19e–07
1.82e–07
1.46e–07
1.09e–07
7.29e–08
3.64e–08
6.91e–16
Figure 9.7 Spatial distribution of fluid parameters in a channel with a corrugated structuring mesh and a
satin-type catalyst. (a) velocity of the fluid, (b) mass fraction of toluene, (c) reaction rate.
distribution is more complex: In addition to the obvious vertical gradient, it shows some gradient
along the width of the channel with a decrease in concentration in the region near the structuring
mesh. This pattern is observed in all cross sections along the channel. This is most likely due to the
lower fluid velocity in this region, which provides more favorable conditions for mass transfer in its
competition with axial mass transport with convective flow.
The maximum reaction rate (Figure 9.7c) is observed on the GFC surface in the inlet part of the
channel with a further decrease in the velocity through the channel due to a decrease in the con-
centration of toluene. A gradient in the reaction rate can also be observed throughout the GFC tex-
tile. For example, the lower GFC textile in Figure 9.7c shows the decrease in reaction rate from
channel inlet to its outlet. This distribution correlates quite well with the distribution of fluid velo-
cities near the GFC surface (Figure 9.7a): A higher velocity on the left side of the textile results in a
shorter residence time of the fluid and thus a lower conversion of toluene, resulting in a higher
concentration of toluene (as can be seen in Figure 9.7b) and, in general, a higher reaction rate.
9.3.2 Influence of GFC Textile Shape

Satin and openwork GFCs have different surface shapes, which can cause differences in their mass
transfer properties. CFD simulations have shown that the overall pattern observed in the case of the
satin-type GFC is approximately the same in the case of the openwork weave catalyst (Figure 9.8).
The most significant difference is observed in the surface distribution of the reaction rate
(Figure 9.8c): Although the general trends here (decrease in the reaction rate from the upper left
to the lower right corner of the lower GFC textile) are similar to those observed in the previous case
of the satin-type catalyst (Figure 9.7c), the overall distribution of the rates in the case of openwork
textiles is more uniform. The surface of openwork GFC is much less smooth compared with satin;
this can increase the turbulence of the flow near the textile surface and adjacent areas, and thereby
improve the overall mass transfer efficiency in the openwork geometry. This hypothesis is sup-
ported by the values of the Nusselt number characterizing the intensity of external mass transfer,
calculated in CFD studies: For these conditions, they are equal to 7.1 for satin and 9.2 for the open-
work catalyst, that is, the estimated increase in apparent activity is 30%. A similar trend can be
traced in experimental data: The apparent reaction rate with an openwork catalyst under such con-
ditions is higher than that of the satin type, and this difference ranges from 7–14% to 25–28%.
9.3.3 Cartridge Channel Without Corrugated Structuring Mesh

Figures 9.9 and 9.10 demonstrate the distribution of fluid parameters in a cartridge channel with
openwork GFC and without a corrugated structuring mesh.
Although the pressure distribution over the height of the channel is still uniform in this case
(Figures 9.9a and 9.10a), the fluid has a well-defined velocity profile with a maximum velocity
of the fluid was along the axis of the channel and minimum velocity was on the GFC surfaces. Com-
pared to the corrugated mesh channel, which breaks down the fluid core into two smaller nuclei,
this situation is less favorable from the point of view of external mass transfer. The worst diffusion
of toluene from the axial region on the GFC surface is observed in Figures 9.9d and 9.10b. As a
result, lower apparent reaction rates lead to reaction propagation along GFC textiles, as can be seen
from the temperature distribution (Figure 9.9c) and reaction rate (Figure 9.10c). Unlike the system
with a corrugated mesh, where this mesh creates an asymmetry of fluid velocities across the entire
width of the textile surface, the distribution of the reaction rate in this case is symmetrical in rela-
tion to the channel axis.
(a)
Velocity Magnitude
[m/s]
6.61e–01
5.29e–01
3.97e–01
2.64e–01
1.32e–01
0.00e+00
contour-3
(b)
1.00e–04
8.00e–05
6.00e–05
4.00e–05
2.00e–05
0.00e+00
contour-3
(c)
[kgmol/m^2s)]
1.09e–07
9.81e–08
8.72e–08
7.63e–08
6.54e–08
5.45e–08
4.36e–08
3.27e–08
2.18e–08
1.09e–08
1.88e–28
contour-3
Figure 9.8 Spatial distribution of fluid parameters in a channel with a corrugated structuring mesh and an
openwork catalyst. (a) velocity of the fluid, (b) mass fraction of toluene, (c) reaction rate.
(a)
Static Pressure
[Pa]
0.70451
0.62271
0.54091
0.45911
0.37731
0.29551
0.21370
0.13190
0.05010
–0.03170
–0.11350
(b)
Velocity Magnitude
[m/s]
1.11637
1.00473
0.89309
0.78146
0.66982
0.55818
0.44655
0.33491
0.22327
0.11164
0.00000
(c)
Static Temperature
[C]
350.00009
350.00003
349.99997
349.99991
349.99985
349.99979
349.99973
349.99966
349.99960
349.99954
349.99948
(d)
0.00032
0.00029
0.00026
0.00022
0.00019
0.00016
0.00013
0.00010
0.00006
0.00003
0.00000
Figure 9.9 Spatial distribution of fluid parameters in a channel without a corrugated structuring mesh
and an openwork catalyst. (a) static pressure, (b) velocity magnitude, (c) static temperature, (d) mass fraction
of toluene.
(a)
Velocity Magnitude
[m/s]
1.10e+00
9.90e–01
8.80e–01
7.70e–01
6.60e–01
5.50e–01
4.40e–01
3.30e–01
2.20e–01
1.10e–01
0.00e+00
contour-2
(b)
3.20e–04
2.88e–04
2.56e–04
2.24e–04
1.92e–04
1.60e–04
1.28e–04
9.59e–05
6.39e–05
3.20e–05
2.80e–10
contour-2
(c)
[kgmol(m^2 s)]
3.41e–07
3.07e–07
2.73e–07
2.39e–07
2.05e–07
1.71e–07
1.36e–07
1.02e–07
6.82e–08
3.41e–08
1.10e+12
contour-2
Figure 9.10 Spatial distribution of fluid parameters in a channel without a corrugated structuring mesh and
an openwork catalyst. (a) velocity of the fluid, (b) mass fraction of toluene, (c) reaction rate.
Table 9.1 Mass transfer characteristics of GFC-based structured cartridges.
Mass transfer
coefficient, m/s
Cartridge/GFC type Computed Nusselt number Calculated Experimental
Corrugated mesh + satin-type GFC 7.1 0.025 0.034

Corrugated mesh + openwork GFC type 9.2 0.031 0.036–0.044
Without corrugated mesh + satin-type GFC 6.6 0.023 0.030–0.032
Without corrugated mesh + openwork-type GFC 7.8 0.026 0.035–0.038
According to the quantitative analysis of the calculated mass transfer data, the Nusselt number
decreases to 6.6 for satin and 7.8 for the openwork catalyst, respectively, showing 7–14% less-
intensive mass transfer than a similar system with a corrugated structuring mesh. The general
behavior of the system (better mass transfer in the presence of a corrugated mesh) agrees well with
the experimental data, although in experiments, the increase in the Nusselt number under similar
conditions can be even more significant (up to 20–30%).
The general quantitative comparison (Table 9.1) revealed that the mass transfer coefficients
obtained in CFD calculations are slightly (30%) lower than those obtained in the experiment.
On one hand, such a discrepancy may be attributed to the relatively low accuracy of mass transfer
experiments. However, this distinction appears to be quite systematic, so it cannot be completely
ignored. This issue deserved a further study, including detailed analysis of fluid interactions with
GFC, resulting from its flexibility and specific internal structure.
9.3.4 Two-Sided Washing of GFC Textiles

In our previous mass transfer studies (Zagoruiko et al. 2017; Zagoruiko and Lopatin 2019), we dis-
cussed possible differences between the behavior of GFC-based structured systems and conven-
tional solid catalysts. It has been suggested that such differences may arise due to the
mechanical flexibility of GFC, which is unusual for virtually all other known forms of catalysts.
In particular, such flexibility can lead to a specific mechanical interaction between the catalyst
and the moving reaction stream, resulting in variability in the external specific surface area of
GFC available to the reactants as a function of the fluid velocity. It has been suggested that, in addi-
tion to the normal contact between the convective flow of the fluid and the face of the GFC textile
oriented to the channel space, part of the flow may also pass through the gap between the GFC and
the supporting plain mesh, thus coming into contact with the back side of the textile. This situation
is completely different from conventionally structured catalysts; for example, from monolith blocks
where the solid washcoat with active component is accessible to reactants from external side only.
This hypothesis has been tested experimentally. For this purpose (Zagoruiko et al. 2017; Zagor-
uiko and Lopatin 2019), the fluorescent paint was sprayed in the airstream at the inlet of the GFC
cartridge using corrugated wire-meshes. It was found that the outer surface of the textile, oriented
toward the gas passage and the corrugated nets, contained clearly visible traces of paint applied
along the channels between the corrugated elements. Surprisingly, the inner textile surface,
oriented toward a flat structural mesh, also contained a significant number of paint particles.
The ink particles are relatively large (up to tens of microns in size), and the possibility of its diffusion
through the fabric has been excluded. These results made it possible to state that convective gas flow
exists in GFC cartridges not only in the cartridge channels but also between the textile and the sup-
porting GFC wire-mesh. This behavior essentially turns the surface area of GFC, which is always
assumed to be constant in the traditional approach, into a variable value that depends on the fluid
velocity and catalyst cartridge geometry. This phenomenon is solely due to the mechanical flexi-
bility of GFC, which is impossible for the packaging of solid conventional catalysts.
In the current study, an attempt is made to reproduce the described flow structure based on CFD
simulations. The first calculation showed that in the case of direct attachment of the GFC textile to
the support mesh (or, in terms of the geometry of the model, to the upper and lower boundaries of
the modeling area), no flow was observed along the back of the textile. This is reasonable, because
the CFD model does not consider the flexibility of the GFC and the ability to change its shape, so the
height of the gap between the GFC and the support grid was set to zero and, technically, the gap for
the backside flow was simply absent in the model. To explain the existence of this gap, it was
assumed that the GFC is located at a short distance from the supporting grid. Figure 9.11 shows
the result of a CFD simulation for a system with a GFC plane located 0.3 mm from the supporting
smooth mesh.
In this case, there are convective backside flows between the GFC textile and the support meshes
in both the upper and lower parts of the channel (Figure 9.11a). These flows provide a response not
only on the external side of the GFC but also on its internal one (Figure 9.11b). The calculated fluid
structure agrees well with the experimentally observed pattern: While the main frontal lateral flow
moves along the GFC surface mainly within the contact lines between the corrugated mesh and the
GFC (the “trace” of the corrugated mesh paint is clearly visible in Figure 9.11d), the flow of the
reverse side is more uniform along the textile width and mainly observed in the very inlet part
of the GFC (Figure 9.11c).
In general, it can be stated that the proposed approach with controlled positioning of GFC textiles
is applicable for modeling the complex structure of the fluid arising from the flexibility of the GFC.
Of course, the real picture is more complex, and the distance from the reference mesh can be
(c)
(a) (b) (d)
Figure 9.11 The distribution of a calculated values of the fluid velocity (a) and the reaction rate (b) in the inlet
part of the channel (Lopatin et al. 2023) and the experimentally measured distribution of paint on the back (c),
front, and (d) sides of the GFC textile in the area of the inlet duct (Zagoruiko et al. 2017; Zagoruiko and
Lopatin 2019).
Figure 9.12 Values of S/S0 ratio, 1,8

providing the best possible agreement the
y = 1,3075×+1,027
experimental and calculated data in each 1,7
R2 = 0,9864
experimental point, upon the gas flow
velocity in the GFC cartridge channel. 1,6
1,5
S/S0
1,4
1,3
1,2
1,1
1
0,00 0,10 0,20 0,30 0,40 0,50 0,60
Flow velocity, m/sec
considered varying depending on many factors (flow rate, shape and density of textiles, details of
structuring the mesh geometry, etc.). This issue requires more in-depth research.
Another option to reproduce the backside flow was to directly set the value for the washable sur-
face area of the GFC textile. We have performed the variation of the accessible specific GFC surface
area S in the range from the surface of the one side of GFC textile (S0) and up to doubled S0 value,
corresponding to full two-side washing of the textile by the fluid. For each fluid velocity, we varied
the values of S/S0 ratio using the surface area washcoat factor to provide the best possible agree-
ment between the experimental and calculated data in each experimental point. The optimal S/S0
values are plotted in Figure 9.12.
It is seen that the apparent S/S0 ratio is 1 at low gas speed, equivalent to one-sided flow along
the textile front side, as in case with the solid catalyst surface; it rises practically linear with the rise
of gas flow velocity reaching the value of 1.7. It is equivalent to the two-sided washing of the textile
by the flow with the backside flow as high as 70% of the front side one. This phenomenon may be
attributed solely to the flexibility of the GFC; it is impossible for all known shapes of conventional
solid catalysts.
9.3.5 Convective Flow Inside the GFC Thread

In the beds of conventional catalysts, the transport of reactants in the free layer occurs predomi-
nantly or entirely due to the convective mechanism of the movement of the flow, while the trans-
port of substances within the catalyst material is provided by the diffusion of reactants in the pores
of the catalyst. Such pores are usually very small in size, in the order of a few tens of nanometers, so
convective flow inside them is impossible. The intensity of the diffusion flow is relatively low, and
this can be the cause of significant internal diffusion limitation of catalytic reactions.
In the case of GFC, the catalyst microfibers are nonporous, and reactions take place on the sur-
face of these microfibers. In such systems, the internal mass transfer takes place inside the catalytic
threads. Given that the distances between the microfibers in the GFC threads are relatively large
(up to a few microns), it is possible that the mass transfer by diffusion may be accompanied inside
the threads with the convective flows, thus potentially increasing the overall efficiency of internal
transport of reactants.
This hypothesis was tested using CFD modeling. To do this, a 2D model of the channel segment in
the GFC-based cartridge was built with two catalytic threads located close to each other perpendic-
ular to the observation plane and adjacent to the channel wall. Each thread was represented as a
cross section of an elliptical shape, the internal volume of each thread containing 2512 uniformly
spaced single microfibers with a diameter of 6 microns, and this corresponds to the real GFC thread
parameters. In the simulation, a gas flow with a velocity of 0.5 m/s at the inlet to the channel was
considered.
Figure 9.13a shows the calculated distribution of flow velocities in the channel volume. It can be
seen that in the core of the flow, the velocity can significantly exceed (up to 1.5 m/s) the initial one
due to the narrowing of the channel, while in the lower part of the channel near the filaments and
between them the velocity is close to zero, and there are no convective flows inside the filaments.
However, when the velocity scale is increased (Figure 9.13b), it becomes clear that there is still a
movement of the flow inside the filaments, although the velocities are very small (0.5–1 mm/s).
This velocity is negligible compared with the flow velocity in the channel volume, but its contri-
bution to the overall transport of matter inside the filaments can be noticeable when compared with
the diffusion of internal transport of substances. The mechanisms of these types of mass transfer are
different and difficult to compare with each other directly, but as a rough estimate, we can use the
diffusion Peclet number, which describes the ratio of convective and diffusion mass transfer as
follows:
• Pe =
ul
D
97
where u – gas velocity (m/s), l – typical geometrical size (m), and D – effective diffusion coefficient
(m2/s). We may consider the velocity values obtained in CFD calculations u = 1∗10−3 m/s and use
the thread thickness as a characteristic geometric dimension l = 10−3 m. The value of the diffusion
coefficient of toluene in air was considered at a temperature of 200 C, at which the internal
(a)
Velocity Magnitude [m/s]
contour-5
0.00e+00 1.24e–01 2.48e–01 3.72e–01 4.96e–01 6.20e–01 7.43e–01 8.67e–01 9.91e–01 1.12e+00 1.24e+00
(b)
Velocity Magnitude [m/s]
contour-5
0.00e+00 1.00e–04 2.00e–04 3.00e–04 4.00e–04 5.00e–04 6.00e–04 7.00e–04 8.00e–04 900e–04 1.00e–03
Figure 9.13 Fluid velocity distribution in the void volume of the channel (a) and inside the GFC threads (b).
diffusion limitation in the reaction under consideration becomes noticeable (Zagoruiko and
Lopatin 2019), and it is equal to D = 1.5∗10−5 m2/s. For these conditions, Pe = 0.07.
On one hand, this value indicates the dominance of diffusion transport of matter within the GFC
threads; on the other hand, it indicates that the convective flow within the threads makes a nonzero
contribution to improving the overall efficiency of internal mass transfer. This fundamentally dis-
tinguishes GFC threads from traditional types of catalysts, where such convective transfer within
the catalyst material does not actually exist.
Of course, the aforementioned quantitative estimate is of a particular nature, and the ratio of
diffusion and convective transport inside the GFC threads can vary significantly depending on
the reaction conditions, types of reactants, characteristics of the filaments, etc. For example, it
can be expected that the velocities of internal convective flows will increase with an increase in
the total fluid velocity in the channel. These issues require further study of individual cases, how-
ever, in any case, the information obtained provides a new understanding of the specifics of mass
transfer processes in structured GFC packages, as well as allows for an unexpected approach to
optimizing the structural parameters of such packages and increasing the efficiency of catalytic pro-
cesses based on GFC.
9.3.6 The General Description of Mass Transfer in GFC

Although the proposed CFD-based approach proved to be an efficient tool for simulation of GFC-
based cartridges, it is rather complicated and resource-consuming. Sometimes it is necessary to use
the simplified approach methods for practical applications. To provide such simplification, it is pos-
sible to use the approach based on criterion equations. Within the framework of this approach, the
value of the dimensionless criterion that determines the intensity of mass transfer, the mass Nusselt
number, may be represented as a general equation:
• Num = ARem Sc1 3

98
The Reynolds criterion is a measure of the ratio of the inertial forces acting in the flow to the
forces of viscous friction. Physically, it has the meaning of a dimensionless flow velocity or a dimen-
sionless geometric dimension. The Reynolds test is calculated using the following equation:
• Re =
udeq ρ
μ
99
where deq is the equivalent size of the cartridge channel, m; Re is the dimensionless Reynolds num-
ber; u is the actual fluid in the cartridge, m/s; ρ is the density of the medium, kg/m3; μ is the
dynamic viscosity of the medium, Pa∗s.
The mass Nusselt number has the physical meaning of a dimensionless mass transfer coefficient
and is calculated using the following formula:
• Num =
βdeq
D
9 10
where β is the mass transfer coefficient, m/s; D is the coefficient of diffusion of toluene in air, m2/s.
The Schmidt criterion is a measure of the ratio of viscosity to mass of a substance. The Schmidt
criterion is calculated using the following equation:
•
μ
Sc = 9 11
ρD
100
Nu/SC1/3
10
1
10 100 1000
Re
Figure 9.14 Dependence of mass-related Nusselt number upon Reynolds number. Points – values calculated
by means of CFD, solid line – approximation of calculated values by power function (9.14), dashed
line – approximation of experimental data by power functions (9.12) and (9.13).
Earlier in our work (Zagoruiko and Lopatin 2019), such criterion equations were proposed that
generally describe the intensity of mass transfer in ICS cartridges with corrugated mesh structuring
elements:
• Rе < 80, Num = 2,95 × Rе0,31 × Sc1 3

9 12
• Rе > 80, Num = 0,84 × Rе 0,59 1 3

× Sc 9 13
In this work, a similar dependence was built on the basis of CFD calculations. For this purpose, a
series of calculations was carried out with a variation in flow velocity from 0.07 to 0.54 m/s and a
channel height in the cartridge in the range of 3–9 mm. Results are shown in Figure 9.14. It is seen
that the calculated data obtained with CFD can be approximated with acceptable accuracy over the
entire range of Re values by a power equation with a power of 0.42, which is close to the arithmetic
mean for the powers in Equations (9.12) and (9.13), which are 0.31 and 0.59, respectively:
• NuCFD
m = 2,20 × Rе0,42 × Sc1 3
9 14
Besides, the new equation obtained by the CFD method reproduces with acceptable accuracy and
experimental results are summarized by Equations (9.12) and (9.13).
9.4 Conclusion
The structure of GFC-based catalytic cartridges is really complex as it contains elements (GFC
textiles and structuring meshes) with intricate, fine-shaped parts. At the same time, these details
can affect the performance of the cartridge and thus can be important in cartridge modeling. Of
course, appropriate simplification procedures were applied, that is, the use of symmetry to rea-
sonably reduce the modeling volume (reducing the entire multilayer cartridge to half a channel)
and a simplified representation of the GFC textile structure (especially in the case of satin-type
fabric). On the contrary, reducing the channel geometry from 3D to 2D or using a stationary
References 453
model instead of a transient model proved to be impractical, as such approaches lead to an

obvious loss of modeling quality.
As a result, the constructed model turned out to be quite complex, and the modeling procedure
turned out to be very resource intensive. For example, some calculations required up to 100 hours of
calculations, while modeling a single-dimensional segment of a convection monolithic catalyst
with a square channel and a smooth surface took only 6 seconds on the same computer. Perhaps
the proposed model could be further simplified to reduce computational time, but this study con-
firmed that the current model is nevertheless applicable for research purposes in its current form.
The simulation revealed the specificities of fluid mechanics in GFC cartridges. Mass transfer CFD
studies have successfully reproduced the qualitative effects of rather complex cartridge design
details, such as the GFC fabric structure or the structure of the internal mesh geometry observed
earlier in experimental work.
CFD calculations were able to describe the theoretically proposed and experimentally observed
regularities of mass transfer processes in GFC structures, including those existing only in GFCs. It
was shown that unlike conventional solid catalysts, in case of GFCs, there may exist the flow of
reactants along the back side of GFC textile, additional to the main flow along its external surface.
This may result in variability of the washable surface area of the GFC textile, which may rise with
increase in fluid velocity and reaching up to 170% (and possibly even higher) of the textile external
surface area accessible to reactants in absence of flow movement. This is an exceptional feature of
GFC, with their flexibility which is not present in the case of solid catalysts.
Another unusual phenomenon is the combined internal mass transfer in GFC threads, which
may be provided not only by diffusion, as in usual porous catalysts, but also by convective flow,
which is possible only in GFC due to the relatively high permeability of its threads.
The developed model accounts for many sophisticated but important details, and it can be used to
optimize the design parameters of existing GFC cartridges as well as to develop new designs of
structured systems based on GFC for various practical applications.
Abbreviations
CFD – computational fluid dynamics

GFC – glass-fiber catalyst
Acknowledgement
This work was partially supported by the Ministry of Science and Higher Education of the Russian
Federation within the state assignment for the Boreskov Institute of Catalysis (project FWUR-2024-
0037) and by the Tyumen Oblast Government, as part of the West-Siberian Interregional Science
and Education Center’s project No. 89-DON (3).
References
Aldashukurova G.B., Mironenko A.V., Mansurov Z.A., Shikina N.V., Yashnik S.A., Kuznetsov V.V.,
Ismagilov Z.R. (2013) Synthesis gas production on glass cloth catalysts modified by Ni and Co oxides,
Journal of Energy Chemistry 22: 811–818. http://dx.doi.org/10.1016/S2095-4956(13)60108-4.
Balzhinimaev, B.S., Paukshtis, E.A., Vanag, S.V. et al. (2010). Glass-fiber catalysts: Novel oxidation
catalysts, catalytic technologies for environmental protection. Catalysis Today 151: 195–199.
http://dx.doi.org/10.1016/j.cattod.2010.01.011.
Bouras, H., Haroun, Y., Bodziony, F.F. et al. (2022). Use of CFD for pressure drop, liquid saturation and
wetting predictions in trickle bed reactors for different catalyst particle shapes. Chemical Engineering
Science 249: 117315. https://doi.org/10.1016/j.ces.2021.117315.
Bracconi, M., Ambrosetti, M., Maestri, M. et al. (2018). A fundamental investigation of gas/solid mass
transfer in open-cell foams using a combined experimental and CFD approach. Chemical Engineering
Journal 352: 558–571. https://doi.org/10.1016/j.cej.2018.07.023.
Chub, O.V., Borisova, E.S., Klenov, O.P. et al. (2005). Research of mass-transfer in fibrous sorption-active
materials. Catalysis Today 105: 680–688. http://dx.doi.org/10.1016/j.cattod.2005.06.049.
Dixon, A.G., Taskin, M.E., Nijemeisland, M., and Stitt, E.H. (2008). Wall-to-particle heat transfer in steam
reformer tubes: CFD comparison of catalyst particles. Chemical Engineering Science 63: 2219–2224.
https://doi.org/10.1016/j.ces.2008.01.017.
Dong, Y., Korup, O., Gerdts, J. et al. (2018). Microtomography-based CFD modeling of a fixed-bed reactor
with an open-cell foam monolith and experimental verification by reactor profile measurements.
Chemical Engineering Journal 353: 176–188. https://doi.org/10.1016/j.cej.2018.07.075.
George, G., Bockelmann, M., Schmalhorst, L. et al. (2022). Radial heat transport in a fixed-bed reactor
made of metallic foam pellets: experiment and particle-resolved computational fluid dynamics.
International Journal of Heat and Mass Transfer 197: 123376. https://doi.org/10.1016/j.
ijheatmasstransfer.2022.123376.
Golyashova, K., Mikenin, P., Elyshev, A. et al. (2019). Structured catalytic cartridges for SO2 oxidation in
flue gases of coal-fired powerplants. Chemical Engineering Journal 378: 122194. https://doi.org/
10.1016/j.cej.2019.112194.
Golyashova, K., Glazov, N., Lopatin, S., and Zagoruiko, A. (2023). Glass-fiber catalysts for selective
catalytic reduction of nitrogen oxides by co and hydrocarbons. Catalysis Communications 106765:
1–11. https://doi.org/10.1016/j.catcom.2023.106765.
Gulyaeva, Y.K., Kaichev, V.V., Zaikovskii, V.I. et al. (2015). Silicate fiberglass catalysts: from science to
technology. Catalysis in Industry 7: 267–274. http://dx.doi.org/10.1134/S2070050415040029.
Iwaniszyn, M., Gancarczyk, A., Gąszczak, A., and Kołodziej, A. (2021a). Structured intra-tubular catalyst
carrier for highly exothermic processes: modelling and CFD study. International Journal of Heat and
Mass Transfer 175: 121357. https://doi.org/10.1016/j.ijheatmasstransfer.2021.121357.
Iwaniszyn, M., Sindera, K., Gancarczyk, A. et al. (2021b). Experimental and CFD investigation of heat
transfer and flow resistance in woven wire gauzes. Chemical Engineering and Processing – Process
Intensification 163: 108364. https://doi.org/10.1016/j.cep.2021.108364.
Jurković, D., Jing-Lin, L., Pohar, A., and Likozar, B. (2021). Methane dry reforming over Ni/Al2O3
catalyst in spark plasma reactor: linking computational fluid dynamics (CFD) with reaction kinetic
modelling. Catalysis Today 362: 11–21. https://doi.org/10.1016/j.cattod.2020.05.028.
Karthik, G.M. and Buwa, V.V. (2019). Effect of particle shape on catalyst deactivation using particle-
resolved CFD simulations. Chemical Engineering Journal 377: 120164. https://doi.org/10.1016/
j.cej.2018.10.101.
Klenov, O., Chumakova, N., Pokrovskaya, S., and Noskov, A. (2019). Impact of heat and mass transfer in
porous catalytic monolith: CFD modeling of exothermic reaction. Chemical Engineering Science
205: 1–13. https://doi.org/10.1016/j.ces.2019.04.010.
Li, L., Diao, Y., and Liu, X. (2014). Low temperature selective catalytic reduction of NOx with NH3 over
Mn-based catalyst: a review. Journal of Rare Earths 32: 409–415. http://dx.doi.org/10.3934/
environsci.2016.2.261.
References 455
Lopatin, S.A., Tsyrul’nikov, P.G., Kotolevich, Y.S. et al. (2015). Structured woven glass-fiber IC-12-S111
catalyst for the deep oxidation of organic compounds. Catalysis in Industry 7: 329–334. http://dx.doi.
org/10.1134/S2070050415040121.
Lopatin, S., Elyshev, A., and Zagoruiko, A. (2023). CFD modelling of the structured cartridges with glass-
fiber catalysts. Chemical Engineering Research and Design 190: 255–267. https://doi.org/10.1016/j.
cherd.2022.12.027.
Mikenin, P., Zazhigalov, S., Elyshev, A. et al. (2016). Iron oxide catalyst at the modified glass fiber support
for selective oxidation of H2S. Catalysis Communications 87: 36–40. https://doi.org/10.1016/j.
catcom.2016.08.038.
Paukshtis, E.A., Simonova, L.G., Zagoruiko, A.N., and Balzhinimaev, B.S. (2010). Oxidative destruction
of chlorinated hydrocarbons on Pt-containing fiber-glass catalysts. Chemosphere 79: 199–204.
https://doi.org/10.1016/j.chemosphere.2010.01.050.
Salmi, T., Mäki-Arvela, P., Toukoniitty, E. et al. (2000). Immobile silica fibre catalyst in liquid-phase
hydrogenation. Studies in Surface Science and Catalysis 130: 2033–2038. https://doi.org/10.1016/
S0167-2991(00)80767-3.
Shalygin, A., Paukshtis, E., Kovalyov, E., and Bal’zhinimaev, B. (2013). Light olefins synthesis from
C1-C2 paraffins via oxychlorination processes. Frontiers of Chemical Science and Engineering
7: 279–288. http://dx.doi.org/10.1007/s11705-013-1338-1.
Sinn, C., Pesch, G., Thöming, J., and Kiewidt, L. (2019). Coupled conjugate heat transfer and heat
production in open-cell ceramic foams investigated using CFD. International Journal of Heat and Mass
Transfer 139: 600–612. https://doi.org/10.1016/j.ijheatmasstransfer.2019.05.042.
Tedesco, G. and Moraes, P. (2021). Innovative design of a continuous flow photoelectrochemical reactor:
hydraulic design, CFD simulation and prototyping. Journal of Environmental Chemical Engineering
9: 105917. https://doi.org/10.1016/j.jece.2021.105917.
Wehinger, G., Kolaczkowski, S., Schmalhorst, L. et al. (2019). Modeling fixed-bed reactors from metal-
foam pellets with detailed CFD. Chemical Engineering Journal 373: 709–719. https://doi.org/10.1016/j.
cej.2019.05.067.
Yang, X., Wang, S., Zhang, K., and He, Y. (2021). Evaluation of coke deposition in catalyst particles using
particle-resolved CFD model. Chemical Engineering Science 229: 116122. https://doi.org/10.1016/
j.ces.2020.116122.
You, Y., Wu, Z., Zeng, W. et al. (2019). CFD modeling of unsteady SCR deNOx coupled with regenerative
heat transfer in honeycomb regenerators partly coated by Vanadium catalysts. Chemical Engineering
Research and Design 150: 234–245. https://doi.org/10.1016/j.cherd.2019.07.015.
Zagoruiko, A.N. and Lopatin, S.A. (2019). Structured glass-fiber catalysts, 158. Francis & Taylor Group,
CRC Press https://doi.org/10.1201/9780429317569.
Zagoruiko, A.N., Lopatin, S.A., Mikenin, P.E. et al. (2017). Novel structured catalytic systems - cartridges
on the base of fibrous catalysts. Chemical Engineering and Processing 122: 460–472.
https://doi.org/10.1016/j.cep.2017.05.018.
Zazhigalov, S.V., Shilov, V.A., Rogozhnikov, V.N. et al. (2022). Mathematical modeling of diesel
autothermal reformer geometry modifications. Chemical Engineering Journal 442 (1): 136160. 1-10.
https://doi.org/10.1016/j.cej.2022.136160.
456
Index
a b
AAE see average absolute error (AAE) batch reactor 10, 12, 13, 15, 18, 21, 26, 194
absolute temperature 343 Bayesian information criteria (BIC) 30, 41, 48
acentric factor 23, 344 bed porosity 352
ACP see adsorption–catalytic processes (ACP) bed void fraction 84, 352
ADM see axial dispersion models (ADM) bench-scale pilot plant 96
adsorbed toluene distribution 155, 158 binary interaction coefficient 272
adsorbent-catalyst bed 147 Boussinesq conjecture 437
adsorption–catalytic processes (ACP) 139, 140 bubble column reactor 244
multisectional reactor 156–164 bulk diffusivity 348
reactor with truncated cone entrance
153–156
adsorption constant 62 c
adsorption equilibrium constant 61 carbonaceous compounds 7, 61, 77
agglomeration 46 carbon rejection technologies 2, 3
aguacate heavy oil 17 Caribbean long residue 66
Akaike information criteria (AIC) 30, 41, 48 Cartesian coordinates 437
aliphatic carbons 4 cartridge channel
alkenyl mechanisms 306, 318 with corrugated structuring mesh 440–443
alkylidene hydride-methylidene 306 without corrugated structuring mesh 443–447
alkyl mechanisms 306, 318 catalyst-adsorbent pellets 140
1-alkyl-3-methylimidazolium cations 18 catalyst deactivation 57, 96
α-olefins 318 catalyst decay function 61
ammonium phosphomolybdate catalyst 13 catalyst fraction 84
Anderson–Schulz–Flory (ASF) type 306, 323 catalysts 116
anthropogenic sources 138 catalytic hydrotreating (HDT) reactor, green diesel
apparent reaction rate coefficients 78, 193 production
aromatics 188 adiabatic commercial reactor simulation
carbons 4 hydrogen quenching configuration 232, 233
solvents 4 liquid-phase yields and gas composition
Arrhenius equation 25 232–235
Arrhenius fitting index weight factor 29 bench-scale reactor product 219–222
asphaltenes 4 catalytic deactivation model 204–205
condensation of 58 dynamic reactor model development
flocculation 22–24 equations for, hydrotreating reactor
fraction 22 modeling 210–212
liquefaction 19 gas generation 213–214
molecules 57 hydrogen consumption 213–214
asymmetric arrangement 141 kinetic model, for vegetable oil hydrotreating 213
Athabasca bitumen 10 reactor models solution 215–216
atmospheric gas oil 14 hydrotreating kinetic models and reaction pathways
autothermal synthesis 305 model compounds 190–197
average absolute error (AAE) 28, 204 vegetable oils 197–204
Avogadro’s number 62 non-isothermal reactor simulation
axial dispersion coefficients 260 bench-scale vs. pilot-scale reactor 227–229
axial dispersion models (ADM) 249 isothermal vs. non-isothermal reactor
axial feed 149, 152 model 225–227
Index 457
pilot plant data validation 222–225 corrected liquid molar volume 22

trickle-bed reactor behavior correction factor 48
catalyst effectiveness factor 219 correlation coefficients 61
catalyst wetting efficiency 218–219 corrugated mesh 437
ideal plug-flow 217, 218 critical pressure 23
wall effects 219 critical temperature 23
vegetable oils cumulative feed-to-catalyst ratio 62
conversion into renewable fuels 187, 188 cyclic process 6
hydrotreating reactor model 205–210 cyclization reactions 194
modelling reactors importance 210 cylindrical apparatuses 144
catalytic particle equivalent radius 85
catalytic precursors 245
catalytic reverse-flow processes 140 d
CFD see computational fluid dynamics (CFD) Danclwerts’ boundary condition 335
chain growth surface 306 deactivation
chemical compounds 175, 190, 197 mechanisms
chemical formation 306 coke deposition 59
Chilton–Colburn analogy 342 metal deposition 59–60
closed-loop system 416 models
cobalt-based catalysts 307–308 by coke deposition 60–64, 79
COC see coke-on-catalyst (COC) coke vs. metal deposition 70–78, 81
CO insertion mechanism 318 by metal deposition 65–70, 80
coked cobalt-molybdenum catalyst 117 deactivation constant 61, 70, 274
coke deposition 57, 59 deactivation function 61, 73, 274
coke formation 204 dead zones 331
coke nitrogen 119 dealkylation reactions 45
coke-on-catalyst (COC) 78 decarboxylation/decarbonylation (deCO2/deCO)
coke oxidation 118 reactions 188, 196
collision factor 17 deep oxidation
colloidal model 4 of hydrocarbons 434
combined-based kinetic model 50 reaction 439
commercial-scale fixed-bed reactor model 230 degree of deactivation 66
CoMo catalyst 207 dialkyldithiocarbamate molybdenum 19
compressibility factors 272, 344 diesel
computational fluid dynamics (CFD) 208, 249, 453 properties 188
mathematical model 436 renewable diesel production 189–190
computing domain 437 RTD 190
defined 437 ULSD 188
parameters 440 diffusion limitations 323
reaction 439–440 dimensionless concentration 12
simulation object geometry 437–439 dimensionless constant 347
simulation 435 dimensionless factors 68
cartridge channel with corrugated structuring dimensionless temperature 9, 10
mesh 440–443 dimensionless time 68
cartridge channel without corrugated structuring dimethyl disulfide (DMDS) 99
mesh 443–447 discrete lumping models 8
convective flow, inside GFC thread 449–451 dispersed catalysts 6, 7, 21
GFC textile shape 443 dispersed NiWMo catalyst 10
mass transfer, in GFC 451–452 dispersed oil-soluble catalysts 7, 19
two-sided washing of, GFC textiles 447–449 dispersion coefficients 265–267
computational geometry 141 distillation curve 8, 9
computational regrouping algorithms 10 diverse reactor models 26, 206
COMSOL multiphysics 141 downflow reactor 13
condensation reactions 18, 49 dynamic one-dimensional pseudohomogeneous
ConocoPhillips 189 fixed-bed FTS reactor model 388–390
continuous gas-oil hydrocracking reactor 12 formulation
continuous kinetic model 21 equations and solution 371–372
continuous lumping models 8 parameters and correlations 372–373
continuous stirred tank reactor (CSTR) reactor 15, 62, results
75, 270 CO and H2 conversion 373–378
convective flux gradients 270 experimental data 373
conventional petroleum refining process 360 product selectivity 381–388
cooling jacket dynamic approach 390 temperature profiles 378–381
coordinate transformation 9 dynamic-state temperature profile 381
458 Index
e boundary conditions, of generalized model 334–336

ebullated-bed reactors (EBR) 4, 56, 70, 242 catalyst particles parameters 345–351
ecofining process 189 catalytic bed parameters 352–353
effectiveness factor 68, 69, 82, 85–87, 93 classification 325
Eley–Rideal type 310 concentration profiles 330
elucidation of reaction 190 equations
emission purification 139 energy balance 333–334
empirical deactivation model 63 heat balance equations for fluid phases 329
end-of-run (EOR) 60 mass balance 327–333
energy balance 205, 206, 229 generalized momentum equations 339
enol mechanism 306, 318, 321 model parameters
entropy function 171 heat transfer parameters 341–343
entropy maximization procedure 171, 174, 179 mass transfer parameters 340–341
equilibrium constant 195 one-and two-dimensional pseudohomogeneous
equilibrium ratio 22 model 325–326
Ergun equation 206 one-dimensional (axial dispersion) heterogeneous
for pressure drop 339 models 326
semiempirical 337 phase equilibrium 343–345
Ergun-type constants 338 pressure drop 337–340
Eulerian–Eulerian framework model 327 steady-state optimal initial operating
exothermic reactions 206 conditions 420, 421
experimental reactors control scheme implementation 416–420
dynamic simulations of, CSTR and SPR 275–278 experimental data 414
steady-state simulations, of SPR 278–280 implementation of 413–414
ex situ sulfidation 7 model equations and numerical solution 408–409
external mass transfer resistance 351 nonlinear constrained optimization 409–413
parameters and correlations 409
steady-state nonlinear constrained
f optimization 415–416
fatty acid (FA) esters 187 two-dimensional (axial and radial dispersion)
Fe-based minerals 7 heterogeneous models 326
feed flow rate 62 fixed-bed reactors (FBR) 4, 242
feedstock gas 215 fixed-bed system 327
feed velocity 278 flattened ellipsoids 438
fine-dispersed catalyst 18 flocculation 24, 46, 48, 49
first-order power-law model 193 flow regimes 247, 265, 337
first-order rate equations 213 fluidized-bed reactor 303
first-order reaction 11 fluid velocity
Fischer–Tropsch liquid wax 344 contours 440
Fischer–Tropsch reaction 271 distribution 450
Fischer–Tropsch synthesis (FTS) 188, 303 formate mechanisms 306
catalysts fouling 59 see also coke deposition
cobalt-based catalysts 307–308 four-lump kinetic models 11, 20
iron-based catalysts 308–309 fourth-order Runge–Kutta method 28
support 309 free fatty acids (FFA) 198
catalytic fixed-bed reactors modeling 324 F–T synthesis 261
catalytic mechanisms 315–321
fixed-bed Fischer–Tropsch reactor modeling 324
classification 325 g
development 326–340 Galileo numbers 338
one-and two-dimensional pseudohomogeneous gamma distribution 175, 176
model 325–326 gas components
one-dimensional (axial dispersion) heterogeneous formation and consumption 127
models 326 heat capacities 130
two-dimensional (axial and radial dispersion) gas flow density 131
heterogeneous models 326 gas holdup 260–262
hydrocarbons growth simplification 322 gas hourly space velocity (GHSV) 359
kinetic models gas–liquid equilibrium 212, 264
with cobalt catalyst 310, 315, 319–321 gas–liquid interfaces 194, 206, 264
with iron catalyst 310–314, 316–318 gas–liquid mass transfer coefficients 262–263, 331
water-gas shift, on iron catalyst 315 gas–liquid system 324
product distribution models 321–324 gas Peclet number 266
five-lump kinetic models 11, 20, 21 gas phase 207, 268
fixed-bed Fischer–Tropsch reactor modeling gas-to-liquid technology (GTL) 304
324, 354 gas-water shift (WGS) reaction 307
Index 459
Gaussian function 171 hydrogenation of aromatic rings 45

generalized heat balance equation 256 hydrogen-hydrocarbon systems 344
generalized mass balance equation 255 hydrogenolysis 6
geometric factor 348 hydrogen sulfide 7, 277
glass-fiber catalysts (GFC) 434 hydrogen-to-carbon (H/C) ratio 2, 61
structured cartridges 434, 436 hydrogen-to-oil ratio 12
global kinetic model 16 hydroprocessing catalyst 61, 68
green fuels see renewable fuels hydrotreated asphaltenes 58
GTL see gas-to-liquid technology (GTL) hydrotreating (HDT) 2, 56, 116
catalyst deactivation development model 78–83
h heavy crude oil
Haldor Topsoe 189 activation of, catalyst 98–99
heat balance equations 128, 145, 225, 229 bench-scale reactor 98
heat capacity, of solid components 130 catalyst properties 96–98
heat effects of reactions 131 characterization methods 99–100
heat-transfer coke and metals on catalyst 102–105
coefficient 206 operating conditions 99
parameters 341–343 properties 96, 97
heavy crude oils 1, 3 sulfur and metals concentration in
composition on, SARA fractions 5 products 101–102
hydrocracking catalysts, in slurry phase 6–7 hydrotreatment process 349
reserves and production of 1–2 hydroxyl carbine (CHOH) 306, 318
upgrading process 2–6 hydroxyl surface species 321
heavy gas oil (HGO) 15 hyperbolic function 61, 63
heavy oils 273 hypothetical asphaltene molecule 5
Henry’s coefficients 264, 271
Henry’s constants 272, 330 i
Henry’s law 212, 215, 264 ideal flow pattern deviation
heteroatomic compounds 175 catalyst wetting efficiency 209
heterogeneous bubble flow regime models 253 ideal plug-flow 209
heterogeneous (churn turbulent flow) regime 247 wall effects 209–210
high-molecular-weight liquid hydrocarbons 383 ideal gas constant 272
high-temperature Fischer–Tropsch (HTFT) 303 ideally separated bubble flow 248
homogeneous (bubbly flow) regime 247–253, 265, industrial catalytic processes 57
268, 292 industrial hydrocracking reactor 62
homogeneous suspension 267 industrial slurry-phase reactor
hot gas discharge 140 dynamic simulations of 283–287
hot-spot formation 413 sensitivity analysis 287–291
hydraulic oil refining process 116 initial grain density 65
hydrocarbons selectivity 323 in situ upgrading process 15
hydrocracking (HCR) reactions 2, 4, 6, 16, 56, 242 integral factor technique 25
hydrodeasphaltenization (HDA) 254 internal heat exchange 140
hydrodecarboxylation/hydrodecarbonylation (HDCx) intraphase gradients 26
reactions 194 iron-based catalysts 308–309
hydrodemetalation reaction 67 isothermal bench-scale reactor 210
hydrodemetallization (HDM) catalysts 65, 77 isothermal conditions 83, 219, 226, 268
hydrodenitrogenation (HDN) 254 isothermal hydrodesulfurization (HDS) 71
hydrodeoxygenation (HDO) 188
hydrodesulfurization (HDS) 16, 59, 77, 253 j
hydrodevanadization (HDV) 74
Jacobian matrix 412
hydrodynamics modeling, in fixed-bed Fischer–Tropsch
Jatropha oil hydrotreating 210
reactor 354, 371
mathematical modeling 354–355
kinetics 356–357 k
numerical method 357–358 Karamay atmospheric residue 18
parameters and correlations 357 kinetic equations 10, 17, 121
reactor model 355–356 kinetic models 48–50, 268
one-stage reactor simulations 358–364 accuracy
two-stage reactor simulations 364–370 distillation curves-based models 41–44
hydrogen (H2) SARA-based models 38–41
consumption 213 activation energies
conversion 359, 364, 365 distillation curves-based models 38
conversion profile 359 SARA-based models 38
donor 19, 20 assumptions 30–32
quenching configuration 232, 233 based on
460 Index
kinetic models (cont’d) m

continuous lumping 21–22 maltenes fractions 8
distillation curves 10–18 mass balance equations 11, 25, 26, 84, 229
SARA fractions 18–21 mass fractions
parallel and series reactions 44–45 of carbon 62
reaction rate coefficients of component 11
distillation curves-based models 34–38 mass gradient 332
SARA-based model 32–34 mass Nusselt number 451
thermodynamic model 22–24, 45–48 mass transfer coefficient 451
types mass transfer flux 246, 262, 340
continuous lumping kinetic models mass transfer parameters 260, 340–341
8–10 mathematical models 288
lumping kinetic models 8 ACP 146–148
single-event kinetic models 10 unsteady-state catalytic reverse-flow
kinetic parameters estimation process 145–146
assumptions 26–27 MATLAB software 25, 357
MATLAB software 25 maximum carbon deposition 62
nonlinear optimization 28 Maya crude oil 245
objective function 28–29 mean absolute error (MAE) 224
optimal estimation 28 mechanistic kinetic approach, with FTS reactor model
parameter initialization 27 experimental data 397
sensitivity and statistical analyses 29–30 formulation
mechanistic FTS kinetic model 395–396
model equations and solution 392–394
l model parameters and correlations 394–395
Lagrange multiplier 169, 172–173, 412 PI controller implementation 396–397
Lagrangian function 412 simulations 397–404, 408
Langmuir–Hinshelwood (LH) mechanism 61, Fischer–Tropsch fixed-bed reactor vs. cooling jacket
196, 310 thermal behavior 404–405
large bubbles (LB) 248 light gases and heavy liquid selectivity
large-scale reactor 12 397–404
lateral-axial system 152 surface of, syngas conversion and heavy liquids
lateral feed 149, 150 selectivity 405–407
lateral-orthogonal systems 152 syngas conversion 397–404
Legendre orthogonal polynomials 394 metal deposition 7, 59–60
Levenberg–Marquardt algorithm 213 metal distribution factor 65
Levenberg–Marquardt methodology 25 metallic compounds 56
Levenspiel model 69 metal-on-catalyst (MOC) 74, 78
LHSV see liquid-hourly space-velocity (LHSV) metal removals 71
light gas oil (LGO) 15 metal sulfide deposits 59
lignin-derived bio-oils 187 methylene 306, 318, 359
linear correlations 45 microfiber catalysts 434
liquefied petroleum gas (LPG) 189 microkinetic models 10
liquid-acid catalyst 15 Middle-East long residue 66
liquid axial dispersion coefficient 265 middle-of-run (MOR) 60
liquid density 271 mineral catalysts 7, 50
liquid-hourly space-velocity (LHSV) 70 Mo-based ionic liquid 18
liquid hydrocarbons 369 Mo-based minerals 7
selectivity 384, 386 model compounds 190–197
volatile 327 model parameters 9, 216
liquid molar volume 22 Mo-dispersed catalyst 12
liquid-phase 207 molar balance equation 194, 206
profile 229, 232 molar product selectivity 356
velocities 268 molar volumes 22, 23, 46, 343
liquid radial dispersion coefficient 266 molecular diffusion coefficients 148
liquid–solid equilibrium 22 molecular reconstruction, of complex hydrocarbon
liquid–solid fluidization 267 mixtures 168
liquid–solid interfaces 206 demonstration 169
liquid–solid mass transfer coefficient 264 MTHS 183–184
literature model 91, 361 problem 168–169
lobe-shaped catalyst particles 349 REM 169–174
long-chain hydrocarbons 327 SOL method 181–182
long-chain liquid hydrocarbons selectivity 384 SR 174–178
low-temperature Fischer–Tropsch (LTFT) 303 SR-EM 179–181
lumping kinetic models 8 state space representation method 182–183
Index 461
molecular type-homologous series matrix method oxygen (O2)

(MTHS) 183–184 atom removal 188
molybdate anion 18 concentration 133
molybdenite ore catalyst 15 oxygenated compounds 188, 200, 327
molybdenum compounds 245
Monte Carlo algorithm 27
Monte Carlo method 175, 177, 213 p
Mosaic method 437 packed-bed systems 335, 341
MTHS see molecular type-homologous series matrix palmitic acid 195
method (MTHS) parametric sensitivity 287
multisectional adsorption-catalytic process 144 parity plots 29, 203
multisectional reactor 156–164 partial differential equations (PDE) 210, 249
multi-stage fixed-bed configurations 367 particulate matter (PM) 188
Peneloux correction 22, 23
Peng–Robinson equation 271, 272
n pentadecane 193, 195
naphtha 6, 8, 10, 18, 61, 62, 278, 286 perturbations 29, 34
Navier–Stokes equations 249 Petrobras 189
N gas-phase models 248 petroleum-derived streams 189, 190, 229, 235
Ni-based catalyst 196 Petzold–Gear backward differentiation formulae 25
Ni-porphyrins 59 pilot plants 12, 78, 96, 140
nitrogen-containing compounds 116 poisoning see metal deposition
nitrogen oxides (NOx) 188 polar fraction 4
Ni-W-Mo catalyst 15 polyaromatic compounds 6
non-isothermal condition 360 polyaromatic fractions 45
non-isothermal state 205 polymerization 59, 194, 200, 305, 401
nonlinear optimization 28 polynuclear aromatics 24
algorithms 204 pore diffusion 58, 246, 332, 345, 350
nonmetallic compounds 56 pore-plugging model 72
non-negative Lagrange multiplier method 182 pore radius 60, 65–69
nonnegative matrix factorization 183 Powell methods 87, 88
non-stable oligomerized compounds 198 power-law model 63, 193, 310
nonstationary catalytic process 164 pre-exponential factor 17, 84, 194, 213, 373
n-paraffins 318, 344 pressure profile method 260
process parameters 6, 141, 145, 152, 164
o proportional–integral controller (PIC) 391
octadecanol 193 pseudo-component-based models 8, 273
oil refinery process 168 pseudohomogeneous first-order reaction 346
oil-soluble catalysts 15, 245 pseudohomogeneous model 325
olefins 188, 307, 310 pseudo-steady state 57
oleic acid (OA) 193 pyrite 7, 18, 30, 38
oligomerized compounds 198, 200, 227 pyrite-activated carbon (pyrite-AC) 38
Omsk refinery 117 Python software 394, 412
one-dimensional mathematical model 126
one-dimensional plug-flow 325 q
one-stage reactor simulations 358–364 quasi-adiabatic reactor 121, 123, 124
openwork textiles 438, 443 quasi-dynamic one-dimensional pseudohomogeneous
optimization algorithms 25, 28, 204 reactor model 83, 84
ordinary differential equations (ODE) 28, 193, 274
organic compounds 211, 212, 434
organohalide compounds 434 r
organometallic molecule 60, 67 Racket compressibility factor 23
orthogonal collocation numerical method 394, 409 radial dispersion coefficients 265, 266, 270, 271, 340
oxidative regeneration, of coked catalysts 135 radial dispersive plug flow 355
experimental result 133–134 raw tall diesel (RTD) 190
kinetic parameters 131 reaction kinetics 273–274
mathematical model 126–132 reaction rate coefficients 11, 32, 61
process chemistry reaction rate equation 24–26, 32, 34, 49, 195
catalyst and its reactions 117–119 reactor designs technique
experimental setup 121–124 ACP 142–144
experiments 124–126 unsteady-state catalytic reverse-flow
reaction kinetics 119–121 process 141–142
solution method 132–133 reactor mathematical models 205, 217, 219
oxychlorination, of light olefins 434 reactor model 93–95
advantages 86–87
462 Index
reactor model (cont’d) dispersion coefficients 265–267

coke and vanadium profile, on catalyst 91–93 equations 253–256
concentration profile, of sulfur and vanadium 89–90 gas holdup 260–262
defined 84–86 gas–liquid mass transfer coefficients 262–263
parameter estimation 88–89 gas–solid mass transfer coefficients 264–265
predictions of, literature vs. proposed model 90–91 heat transfer coefficients 267
solution 86 initial and boundary conditions 257–259
usefulness 95–96 liquid–solid mass transfer coefficients 264–265
reactor wall 190, 207, 242, 342 model parameters estimation 260
readsorption 318, 323 simplification 267–268
CO 316 solids concentration 257
olefins 316, 317 numerical simulations
reconstruction by entropy maximization experimental reactors 275–280
(REM) 169–174 industrial-scale reactor 280–291
refined, bleached, and deodorized palm oil simplified models
(RBDPO) 206 CSTR model 270
refined vegetable oils 187 parameters 270–272
relative standard error 91 reaction kinetics 273–274
renewable diesel production 189–190 solution method 274–275
renewable fuels 188, 191 SPR 1D model 268–269
residue fraction 15, 17, 62 SPR 2D model 269–270
residue hydrodesulfurization (RDS) 65 SPR modeling
resin-building micelles 4 classification 246–248
restrictive factor 347 complexity 249
reverse-flow process 139, 141 slurry reactors 249–253
reversible dehydrogenation-decarbonation 195 slurry-phase reactors (SPR) 4, 56, 242, 244
Reynolds numbers 330, 338 SPR 1D model 268–269
Reynolds test 451 SPR 2D model 269–270
Runge–Kutta–Fehlberg method 26 solid catalyst particle 352
Runge–Kutta method 84, 213, 274, 293 solid heterogeneous powders 6
solid–liquid equilibrium 22
solid–liquid mixture 257
s solid-phase temperature profiles 229
SARA fractions 8, 41, 46 solids concentration 257
saturate fraction 23, 24, 46 solubility parameters 22, 23
SBCR see slurry bubble column reactors (SBCR) spatial distribution, of fluid parameters
Schmidt number 262, 264, 266 with corrugated structuring mesh
second kinetic model 20 and openwork catalyst 444
second-order model 63 and satin-type catalyst 441, 442
sedimentation-dispersion model (SDM) 254 without corrugated structuring mesh and openwork
sediment deposition 6 catalyst 445, 446
sediment formation 4, 6, 22–24 squares of the errors (SSE) 200
sediment fraction 24 SR-EM 179–181
selective hydrogenation 434 SRGO see straight-run gas oils (SRGO)
semibatch reactor 16, 18 SSE see squares of the errors (SSE)
sequential quadratic programming (SQP) 87 standard decomposition strategies 183
2400 Series II model Perkin Elmer elemental start-of-run (SOR) 73, 231, 232
analyzer 100 state space representation method 182–183
Shannon entropy 169 steady-state simulations
shape factor 346 of hydrocracking reactions 284
short-chain hydrocarbons 327 of hydrotreating reactions 285
single-event kinetic models 10 stearic acid 193, 195
six-lump kinetic models 13, 21 stirred-tank reactor (STR) 243, 244
SLFA 2100 Horiba X ray fluorescence equipment 99 stochastic reconstruction (SR) 174–178
slug flow regime 247 stoichiometric coefficients 13, 17
slurry bubble column reactors (SBCR) 243, 266 stoichiometric hydrogen 196
slurry phase hydrocracking reactions 6 straight-run gas oils (SRGO) 98, 231
slurry-phase hydrocracking reactor model structural increments 181
advantages 243, 244 structure attributes 175
characteristics structure-oriented lumping (SOL) method 175,
catalyst properties 245–246 181–182
reactor types 242–245 sulfur 2, 4, 7, 66
technologies 243 compounds 116
disadvantages 244 pilot-plant 78
generalized model removals 17, 71
Index 463
sulfur dioxide (SO2) concentration 133 ultradispersed nanocatalyst 17

sulfuric anhydride 119 ultra-low sulfur diesel (ULSD) 188
sulfur oxides (SOx) 188 uncertain/relax constraints 171
surface compound formation 147 unconventional resources 1
syngas 305, 307 unconverted oil (UCO) 15, 62
syngas conversion 308, 356 unidimensional pseudohomogeneous models 360
synthetic gas production 434 uniform distributions 437
universal constants 272
universal gas constant 17, 440
t unsteady-state catalytic reverse-flow process 148–153
tall oil fatty acids (TOFA) 196 mathematical models 145–146
tangential feed 150 novel reactor designs 141–142
tangential-lateral arrangement 151 USC see ultimate storage capacity (USC)
temperature-dependent equilibrium constant 194
temperature-dependent parameters 272
temperature distribution 152, 155, 157 v
temperature runaway 411 vacuum gas oil (VGO) 6, 61, 273
tetra lobular catalyst 96 vacuum residues (VR) 1, 245
tetra-lobular geometry 350 value liquid hydrocarbons 304
thermal conductivity 128, 131 vanadium deposition 93
thermal cracking 6, 198 vanadium removal 66
thermal methods 139 vapor–liquid equilibrium (VLE) 323, 343
thermal wave 124, 143, 153 vegetable oils 187
thermodynamic parameters 264, 272, 340 conversion into renewable fuels 187, 188
Thiele modulus 85, 345, 347 hydrotreating reactor model 205–210
three-phase catalytic reactors 4 modelling reactors importance 210
three-phase fluidized beds 267 viscosity 131
time-dependent nonselective expression 274 volatile organic compounds (VOC) 138–140, 188
time-on-stream (TOS) 57, 60, 307 catalytic oxidation 139
toluene concentration 440 neutralization of 139
distribution in, gas phase 159 V-porphyrins 59
inlet 142
outlet 163
toluene oxidation reaction 145 w
total pressure constant 12 Wagner–Weisz–Wheeler modulus 348
transverse catalyst fracture 127 wall effect coefficients 338
trapezoidal rule 25, 132 water-soluble catalysts 245
trialkylmethylammonium molybdate ionic liquid 18 Watson factor 271
trickle-bed reactor (TBR) model 56, 207, 210 weight fraction 9, 12
trickle-flow regime 353 weight hourly space velocity (WHSV) 275
triglyceride conversion 223 weighting factor 29
triglycerides (Tg) 187 WHSV see weight hourly space velocity (WHSV)
tripalmitin (TP) 195 wire-meshes 438
tristearin (TS) 195
truncated cone inlet system 147
turbulent viscosity coefficient 437 x
two-phase bubble columns 261 X-to-liquid technology (XTL) 304
u z
ultimate storage capacity (USC) 65 zero Aris number 347
ultradispersed catalyst 12, 13

Mathematical Modeling of Complex Reaction Systems in The Oil and Gas Industry

Uploaded by

Copyright:

Available Formats

Mathematical Modeling of Complex Reaction Systems in The Oil and Gas Industry

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mathematical Modeling of Complex Reaction Systems in The Oil and Gas Industry

Uploaded by

Copyright:

Available Formats

Downloaded from https://onlinelibrary.wiley.com/doi/ by ibrahim ragab - Universita Di Firenze Sistema , Wiley Online Library on [29/07/2024].

Instituto Mexicano del Petróleo

Boreskov Institute of Catalysis

Limit of Liability/Disclaimer of Warranty

Library of Congress Cataloging-in-Publication Data is Applied for

Cover Design: Wiley

Set in 9.5/12.5pt STIXTwoText by Straive, Pondicherry, India

List of Contributors xiii

1 Modeling the Kinetics of Hydrocracking of Heavy Oil with Mineral Catalyst 1

2 Modeling Catalyst Deactivation of Hydrotreating of Heavy Oils 56

2.6.7 Parameter Estimation 100

3 Simulation of the Oxidative Regeneration of Coked Catalysts: Kinetics,

4 Modeling of Unsteady-State Catalytic and Adsorption–Catalytic

5 Molecular Reconstruction of Complex Hydrocarbon Mixtures for Modeling

6 Modeling of Catalytic Hydrotreating Reactor for Production of Green Diesel 187

7 Modeling of Slurry-Phase Hydrocracking Reactor 242

8 Modeling of Fischer–Tropsch Synthesis Reactor 303

8.1.3 Fischer–Tropsch Synthesis Kinetic Models 309

8.4.5 Final Remarks 388

9 Computational Fluid Dynamics Modeling of Mass Transfer Processes in

The General Description of Mass Transfer in GFC 451

Osman Abdulla Sergey Lopatin

Guillermo Félix Andrey Zagoruiko

Modeling the Kinetics of Hydrocracking of Heavy Oil with

1.1.1 Reserves and Production of Heavy Crude Oils

Crude oil reserves (billion bbl)

Figure 1.1 Proven reserves of crude oil by country.

1.1.2 Heavy Crude Oil Upgrading Processes

✓ Price of heavy crude oil

✓ Low yield of upgraded oil

✓ Minimizes the coke formation

Figure 1.4 Heavy crude oil composition based on SARA fractions.

Figure 1.3 Hypothetical asphaltene molecule.

1.1.3 Reactions During Slurry Phase Hydrocracking

1.1.4 Catalysts for Hydrocracking of Heavy Crude Oils in Slurry Phase

1.2 Kinetic Models

1.2.1 General Types of Kinetic Models

1.2.1.1 Lumping Kinetic Models

1.2.1.2 Continuous Lumping Kinetic Models

✓ Prediction of the distillation curve with few parameters to be estimated

1.2.1.3 Single-Event Kinetic Models

1.2.2 Kinetic Models Reported in the Literature for Hydrocracking of Heavy

r R = v1R k 1 CLR C LH2 + v2R k 2 CLR CLH2 1 22

r D = v2D k 2 CLR C LH2 + v3D k 3 C LVGO C LH2 + v5D k 5 C LD CLH2 1 24

r N = v4N k 4 C LVGO C LH2 + v5N k 5 C LD C LH2 + v6N k 6 CLN CLH2 1 25

r G = k4 yR + k8 yVGO + k10 yAGO 1 36

r VGO = k 1 y2UCO − k 2 + k4 yVGO 1 39

Complementary reactions rate equations:

r G = k 4 yAs + k 7 yRe + k9 yAr + k 10 ySa 1 47

A fine-dispersed catalyst (nickel nanoparticles in polyethylene glycol, PEG300) was employed by

1.2.2.2 Kinetic Models Based on SARA Fractions

Figure 1.8 SARA- and distillation k3 k2 k1

Figure 1.9 Four-lump kinetic models based on (a)

hydrogen pressure, temperature of 380–400 C, reaction time of 3–9 h, 5:1 asphaltenes-to-hydrogen

Figure 1.10 Five- and six-lump kinetic models (a)

1.2.3 Kinetic Models Based on Continuous Lumping

1.2.4 Thermodynamic Model to Predict the Asphaltenes Flocculation

Substituting this term in Eq. (1.90) gives

The Peneloux correction term (c) can be estimated by