UNIVERSITY OF CALGARY

Thermal Cracking of Gas Oil in a Helical Coil Reactor

by

Dahai Dong

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

I PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE N
DEGREE OF MASTER OF SCIENCE M CHEMICAL ENGINEEFUNG

DEPARTMENT OF CHEMICAL AND PETROLEUM ENGINEERING

CALGARY,ALBERTA
JUNE, 2000

O Dahai Dong 2000

Nationai Library Acquisitions and Bibliographic Services

395 Wellington Street Ottawa ON K I A ON4

Biblioth&que nationale du Canada Acquisitions et services bibliographiques

395, rue Wellington Ottawa ON K1A ON4

Canada

Canada
Your fih Voim r e M m w 8

Our 6Ie U e r e reHrence

The author has granted a nonexclusive licence allowing the National Library of Canada to reproduce, loan, distribute or sell copies of this thesis in microform, paper or electronic formats.

L'auteu. a accorde une licence non exclusive pennettant a la Bibliotheque nationale du Canada de reproduire, przter, distribuer ou vendre des copies de cette these sous la forme de microfiche/film, de reproduction sur papier ou sur format electronique. L'auteur conserve la propriete du droit d'auteur qui protege cette these. Ni la these n des extraits substantiels i de celle-ci ne doivent Ctre imprimes ou autrement reproduits sans son autorisation.

The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts f?om it may be printed or othewise reproduced without the author's permission.

Abstract

Hydrotreated light gas oil derived fiorn Athabasca bitumen has several properties, which

make it a less desirable transportation fuel. Its boiling point is in the range of diesel,
which is too high to be used as gasoiine. However, the gas o l also has a sigmficant i amount of fused ring compounds, which makes its cetane number too low for diesel specifications. Thermal cracking (pyrolysis) is one of the simplest and economically the most important upgrading reactions in industrial application. It was expected that the

thermal cracking method could significantly upgrade the gas oil to naphtha.

A previously designed helical coil reaction system was modified and employed to

thermally crack light gas oil to produce naphtha. Thermal cracking of hydrotreated light gas oil was studied at three different temperature levels, 500C, 530C and 550C.

Extents of cracking or conversions of LGO for experiments were in the range of 15% to
46%.

For the purposes of comparing current results to the literature, a kinetic analysis

assuming a f r t order ovemll decomposition model for the light gas oil thermal cracking is
reaction was examined. The activation energy of this overall reaction was evaluated at 153 kJ/mol. The light gas oil feed used in this study has an estimated average molecular weight of 212. The activation energy found in this study was very close to that of n-

hexadecane (molecular weight of 226) determined by previous researchers (Fairburn et al., 1990, activation energy of 1 64 kl/mol).

A more detailed reaction network involving a four-lump kinetic model was proposed.

Lumps were chosen in order to give a better understanding of the product distribution.
According to the experimental data, the activation energy for the reaction of gas oil to

naphtha was 173 kl/mol, for the reaction of gas oil to gas was 83 kl/mol, for the reaction
of naphtha to gas was 2 15 kl/mol and for the reaction of gas oil to coke was 262 Wmol.

The activation energy for the reaction of naphtha to gas is close to 220.8 kJlmol obtained
by Kumar and K

m (1985) and within the range of 212 to 228 kJ/mol found by van

Damme and Froment (1981). From the proposed model, a set of optimal reaction times
at different temperatures is predicted.

Acknowledgements

The author wishes to express his sincere gratitude to the many individuals who have contributed to this work, and particularly to:

Dr. Caroline L. Hyndman, his supervisor, for her guidance, patience, and friendship throughout the course of this work.

Dr. Gordon R. Moore for his continued interest and on-going encouragement.

Dr. Leo A. Behie for his wonderful course and for teaching me about the important and
practical issues of the reactor design and process modelling.

Dr. IAikhar Huq and Dr. Keng Chung for their helpful suggestions and discussions.

D . Ayodeji A. Jeje, for his helpll discussions on fluid dynamics and heat transfer r
issues.

Dr. B. Milne, for his important help on the EXCEL solver.

Mr. Sean Cuthbert for his design of the helical coil reactor and training.

Mr. Mike Grigg for his help on the data acquisition system, which occurred quite often.

Mr. Derek Verhegge for the construction of the apparatus and timely help.

M .Mark Misunis, Ms.Paige Deitsch and Ms. Claudine Cumow for their on-going help r
and guidance on the gas chromatograph analysis.

Dr. Nancy Okazawa for her help on analysis of some gas products.

Ms. Elizabeth Zalewski for her help on the simulated distillation analysis of the feed and
liquid products.

Appreciations are also extended to Mr. Viet Bui, Mr. Don Mallory and Mr. Jean-Marc

Labontd for their valuable ideas and Wendship.

Financial support fiom Syncrude Canada Ltd. and Natural Sciences and Engineering
Research Council of Canada are gratefully acknowledged.

Finally, the author wishes to express his deep gratitude to his wife, Jing, and his parents,

who gave him unlimited love and support tll during this work.

Table of Contents

Approval Page
Abstract

Acknowledgements Table of Contents

List of Tables vii

xi
xi11

List of Figures Nomenclature

...

CHAPTER 1 Introduction and Objectives I. 1

1.2
Introduction Objectives

CHAPTER 2 Literature Review

Industrial Applications of Thermal Cracking Reactions Pilot Reactors Laboratory Reactors for Thermal Cracking Kinetic Studies
2.3.1 Batch Reactors
2.3.1.1 Conventional Batch Reactors

2.3.1.2 Curie Point Py~olysis Reactor

vii

2.3.2

Flow Reactors

2.4 2.5

The Chemistry of Thermal Cracking

Reaction Kinetics for Complex Mixtures 2.5.1 Black Box Method

2.5.2 Model Compound Method

2.5.3 Monte Carlo Simulation Method 2.5.4 Lumped Kinetics

CHAPTER 3 Experimental Apparatus and Methods

3.1

Process Overview

3.2

Experimental Apparatus
3.2.1 Helical Coil Reactor 3.2.2 Transfer Cylinder and Hydraulic System 3.2.3 Feed Preheaters and Gas Furnace 3.2.4 Product Condenser and Collection
3.2.5 Process Instrumentation and Data Acquisition

3.3 3.4

Light G s Oil Properties a

Gases Used in the Study

Light Gas Oil Thermal Cracking Experiments Procedure Product Analysis
3.6.1 GasAnalysis
3.6.1.1 GC Calibration

3.5
3.6

viii

3.6.2 Simulated Distillation Analysis of Liquid Samples

CHAPTER 4 Results and Discussion

4.1

Material Balance Product Analysis 4.2.1 TCD GC for Product Analysis 4.2.2 Liquid Sample Analysis

4.2

4.3

Residence Time

4.4

Kinetic Model Deveiopment

4.4.1 Lumped Reaction Network 4.4.2 Model Assumptions 4.4.3 Evaluation of Pseudo-Component Fractions 4.4.4 Reaction Network Equations and Rate Constants 4.4.5 Frequency Factors and Activation Energies for the Four-Lump Model

4.5
4.6

Overall Reaction Rate Product Distribution Prediction

CHAPTER 5 Conclusions and Recommendation

5.1

Conclusions Recommendation

5.2

References

Appendix A Appendix B Appendix C Appendix D

Mass Baiance Calculation for Different Runs

Temperature Profile for Different Runs Pump #2 and Pump #3 Calibration

Valve Switching Scheme for Gas Analysis on GC

Appendix E

Estimation of the Reynolds Number and Heat Transfer Coefficient

Appendix F Appendix G Appendix H

Estimation of the Pressure Drop along the Reactor Tube Calculations of the LGO to Nitrogen Ratio
List of Failed Experiments

List of Tables

Table 1.1

Typical Properties of the Light Gas Oil Fraction of Syncrude Sweet Blend

Table 2.1

Thermal Processing for the Purpose of Hydrocarbons Upgrading

Table 3.1 Table 3.2 Table 3.3 Table 3.4 Table 4.1 Table 4.2 Table 4.3 Table 4.4 Table 4.5 Table 4.6 Table 4.7 Table 4.8 Table A. 1 Table A.2 Table. A.3 Table A.4

Properties of Feed Light Gas Oil Product Specification of Gases Used in Experiments

G s Chromatograph Analysis Conditions a

Standard Gas Mixture Composition for GC Calibration

Summary of Successful Experiments

Summary of Experimental Conditions

Summary of Mass Balances

LGO Conversion and Pseudo-Component Distribution

LGO Thermal Cracking Reaction Rate Constants
Activation Energies and Frequency Factors Overall Thermal Cracking Reaction Rate Constant

Optimal Reaction Times at Different Temperatures

Molecular Weight Calculation of Whole Feed Molecular Weight Calculation of Feed (>196OC)

Mass Balance Calculation for Run # 1

Mass Balance Calculation for Run #2

Table A.5 Table A.6 Table A.7 Table A.8 Table A.9 Table A. 10

Mass Balance Calculation for Run #3

M s Balance Calculation for Run #4 as

Mass Balance Calculation for Run #5 Mass Balance Calculation for Run #6

M s Balance Calculation for Run $7 as

Mass Balance Calculation for Run #8
Valve Status and Flow/Sample Status Summary of Failed Experiments

Table D. 1
Table H. 1

xii

List of Figures

Figure 2.1 Figure 2.2 Figure 3.1 Figure 3.2

Activation Energy Evaluation for n-Hexadecane Pyrolysis

Kinetic Model for Fluid Catalytic Cracking Picture of LGO Thermal Cracking Experimental Set-up Simplified Schematic Diagram of LGO Thermal Cracking Apparatus

Figure 3.3 Figure 3.4

Helical Coil Reactor Process Flow Diagram Process Flow Diagram for the Condensing and Product Collection Systems

Figure 3.5 Figure 3.6 Figure 3.7 Figure 3.8 Figure 3.9 Figure 3.10

Photograph of the Reactor Detailed Dimensions of the Reactor Arrangement of Reactor Heaters (Outside) Pumps (#2 and #3), Furnace and Hydraulic Oil Flow Meter

Transfer Cylinder and Pumps (#2and #3)

Picture of Condensing System Picture of Distillation Unit Chromatograph of the Standard Gas Mixture Chromatograph of Reaction Gas Product of Run #7 Chromatograph of Coke Burning Gas Product of Run #7 Simulated Distillation Curves for Feed and Products
Kinetic Model for LGO Thermal Cracking

Figure 3.1 1
Figure 3.12 Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4

Figure 4.5
Figure 4.6
Figure 4.7

Activation Energy Evaluation (LGO

+ Naphtha) Activation Energy Evaluation (Naphtha + Gas)

Activation Energy Evaluation (LGO3 Gas) Activation Energy Evaluation (LGO

Figure 4.8 Figure 4.9

+ Coke)

Overall Activation Energy Evaluation

Component Mass Percent vs. Residence Time at 500C Component Mass Percent vs. Residence Time at 530C Component Mass Percent vs. Residence Time at 550C

Figure 4.10
Figure 4.1 1 Figure 4.12

Figure B. 1 Figure B.2

Figure B.3

Temperature Profile for Run #1

Temperature Profile for Run #2

Temperature Profile for Run #3

Temperature Profile for Run #4 Temperature Prof le for Run #5 Temperature Profile for Run #6 Temperature Profile for Run #7 Temperature Profile for Run #8 Calibration Curve for Pump #2 and Pump #3

Figure B.4
Figure B.5

Figure B.6
Figure B.7

Figure B.8
Figure C. 1 Figure D. 1

GC Column Connection Sequence as Both Valves

in OFF Position

Figure D.2

Valve Timing Logic

xiv

Nomenclature

cross-section area of the reactor tube, m2

frequency factor for reaction j, f1

heat capacity of reaction fluid, J/kgK reactor tube ID, m

activation energy for reaction j, J/mol

friction factor fluid mass flux, kg/m2-s heat transfer coefficient, w / m 2 ~

thermal conductivity, W/mK

first order rate constants for reactions 1,2,3,4, s*'

first order rate constant for reaction j (j= l,2, 3,4), s-'
first order rate constant for the overall reaction, s" reaction constant in Equation (2.1) feed mass flow rate, g/s mass of pentene and heavier hydrocarbons, g mass of coke pseudo-component, g demister weight increase after experiment, which is included in
the mass of naphtha pseudo-component, g

mass of gas pseudo-component, g

mass of gas oil pseudo-component, g

mass of liquid product, g

mass of naphtha pseudo-component, g

average molecular weight of inlet feed and outlet product, g/mol molecular weight of component i in the gas bag, g/mol order of the reaction in Equation (2.1) NusseIt number reaction pressure, Pa Prandtl number gas constant, Urn01 K Reynolds number, duplp residence time, s reaction half-life, s reaction temperature, K room temperature, K fluid velocity, d s volumetric flow rate at reactor conditions, m3/s nitrogen volumetric flow rate at room temperature, rn3/s

reactor volume, m3
volume of gas bag, recorded by wet test meter, m3
initial (feed) weight percent of pseudo components 1 (naphtha)

and 2 (LGO)

xvi

weight percent of pseudo-component 1 (naphtha), 2 (LGO),

3 (gas), 4 (coke)

a property of the feed in Equation (2.1)

naphtha pseudo-component mass fraction fiom SimDist gas oil pseudo-component mass hction from SimDist
C02volume percent in the gas bag

volume percent of component i in the gas bag

Greek

fluid viscosity, Pas fluid density, kg/m3

Abbreviations

ASTM

American Society for Testing and Materials

Coke

Gas
Gas Chromatograph

Hydraulic Oil
Hewlett-Packard

ICFB

Internally Circulating Fluidized Bed

ID

Inside Diameter

L
LGO

Liquid

Light G s Oil a
Liquid Hourly Space Velocity

LHSV MS

Mass Spectrophometer
Molecular Weight Plug Flow Reactor
Simulated Distillation Thermal Conductivity Detector
Wet Test Meter

MW

PFR
SimDist

TCD

CHAPTER 1 Introduction and Objectives

1.1

Introduction

Oil sand is composed of sand, bitumen, mineral rich clays and water. In its raw state, bitumen is a black, asphalt-like oil. It requires upgrading to make it transportable by pipeline and usable by conventional refineries. The upgraded bitumen product consists

of naphtha, light and heavy gas oils that are combined to produce a light, sweet crude
oil.

The oil sands deposits in Alberta are estimated to contain over 1.7 trillion barrels - more

than five times larger than Saudi Arabia's hydrocarbon reserves.

As Canada's

conventional reserves of light and medium oil continue to diminish while consumption rises, the need for such an alternative oil source is apparent. (Syncrude Canada Ltd.,

2000b)

In the Syncrude Canada oil sand plant at Fort McMurray, bitumen is fed into either one

of two coken or a hydroprocessing unit. It is thermally cracked into hydrocarbon gases,

naphtha and gas oil. While the hydrocarbon gases are treated for use as refinery &el, the

naphtha and gas oils are treated and blended into sweet blend crude oil. The oil is then

transported via pipeline to refineries in Edmonton, eastern Canada and the mid-western United States.

Coking processes have been selected by many refiners as their preferred choice for
bottom of the barrel upgrading, because the process has the inherent flexibility to handle even the heaviest of residues. The process provides essentially complete rejection of

m t l while providing partial conversion to naphtha and diesel cuts. eas

The valuable products of the coker are naphtha and gas oil; the by-products are fuel
gases and coke. The coker naphtha and gas oil are unstable because they contain relatively high quantities of sulfur, nitrogen and unsaturates. Naphtha and gas oil

streams are sent to catalytic hydrotreating units to turn nitrogen into ammonia and s u h
to hydrogen sulfide and to stabilize or saturate the unsaturates.

However, to utilize the hydrotreated gas oil derived from Athabasca bitumen is difficult.

Its boiling point is in the range of diesel, however, the synthetic crude diesel h c t i o n has
a cetane number of 34 (Yui, 1999), which is too low for diesel specifications. The
minimum cetane number required for a general purpose, middle distillate fuel for

automotive diesel engines is 40 (ASTM D975, 1998). The low cetane number of the

bitumen derived gas oil fraction is caused by a significant amount of f h e d ring

compounds. Table 1.1 shows properties of the light gas oil fraction of Syncrude Sweet Blend (Yui, 1999).

Table 1.1 Typical Properties of the Light Gas Oil Fraction

of Syncrude Sweet Blend (Yui, 1999)

Property Density at 1 SC, g h l Carbon, w ? t! Hydrogen, WYO Nitrogen, wppm

Light Gas Oil

0.878
87.7

12. t
99

Sulfur, w ! t ?

0.023
47

Aniline Point, O C
Aromatics (Supercritical Fluid

42.2 Chromatography), wt%

Cetane Number
34.0

Kinematic Viscosity at 20C, cSt

5.3

Hydrocarbon cracking is the process by which higher molecular weight hydrocarbons are converted to lower molecular weight hydrocarbons through carbon-carbon bond scission. There are three general types of hydrocarbon cracking: thermal cracking, catalytic cracking, and hydrocracking.

Each process has its own characteristics

concerning reaction conditions and product compositions. Among them, hydrocarbon thermal cracking (pyrolysis) is the simplest and economically the most important industrial reaction. A wide spectrum of feedstock, ranging from ethane to vacuum gas oil, is pyrolyzed to produce olefins vital to the petrochemical industry. Thermal cracking has also been applied to the upgrading of bitumen or heavy oil. At present, a total daily production of about 330,000 barrels of upgraded oil has been achieved at the northern Alberta oil sand plants. It is expected that after on-going expansion projects, a
daily production of 830,000 barrels may be reached in 2008 (Syncrude Canada Ltd,

2000a and Suncor Energy Inc., 2000).

1.2

Objectives

The objective of this work is to study thermal cracking reactions of hydrotreated light
gas oil for production of liquid products over the temperature of 500 tr. 550C. T i hs

temperature range was chosen as it is within the range of reactor temperatures for fluid

coking (Kunii and Levenspiel, 1991). The experiments were also designed to obtain
measurable quantities of all products. For a range of operating conditions, yield of products, especially naphtha, will be measured. Lumped reaction kinetics will be

determined and using the experimentally determined kinetic parameters, the effect of reaction temperature and residence t h e on the optimal production of naphtha will be
investigated.

CHAPTER 2
Literature Review

Thermal cracking reactions of hydrocarbons have been of great interest to chemists even before the petroleum industry began. The industry of producing chemicals from petroleum started when chemists in the 1920's produced alcohols from ethylene and propylene via thermal cracking (Stolfa, 1980). From this point, thermal processes for

the production of basic chemical feedstocks: ethene, propylene, butenes, butadiene and
aromatics, have been increasingly important, Nowadays a vast industry throughout the

world is based on olefins and aromatics produced by thermal reactions of hydrocarbons.

The by-products? such as coke and gas, are available as commercial items and all are
produced by thermal means. Thermal reactions are still playing an important role in the

production of hydrocarbons from shale, oil sands, coal and lignite.

Before a kinetic study of light gas oil (LGO) thermal cracking can be initiated, a good understanding of the industrial and the theoretical aspects of hydrocarbon thermal

cracking is essential.

2.1

Industrial Applications of Thermal Cracking Reactions

The thermal processing technologies in the petroleum refining industry refers to the
upgrading method that utilizes heat to initiate reactions of feedstocks. Thermal cracking

is distinct from other processes such a s catalytic and solvent processes.

Thermal

processes have good flexibility for different feeds because the reactions depend only on heat effects without the need for catalyst or solvent. Thermal processes can deal with
residues as well as distillate oil, and can treat "cleaner" feed or feed with large amounts

of impurities (sulfur, nitrogen, oxygen, heavy metals, asphaltenes, etc.).

One of the purposes of heavy oil thermal processing is to convert the heavy end of the crude, including vacuum distillate and residue, into useable products. For example, to
turn high viscosity, high pour point vacuum residue into fuel oil. However, the more

important aspect is upgrading, to transform low value heavy oil into valuable light oils.

Although facing competition from other processes, thermal processes remain important.

The following processes are used by many refineries (Cheng, 1994):

(1) low severity visbreaking to produce fixel oil f o atmospheric or vacuum rm residue; (2) thermal cracking to produce gasoline, diesel and fuel oil from gas oil;
(3) coking to produce gasoline, diesel, gas oil and coke fiom vacuum residue.

Table 2.1 shows some of the thermal operations used in indmtry (Stolfa, 1980). In contemporary petroleum refining industry, the thermal processes tend to have larger capacity, simpler operation, and lower utility costs. These are the essential points that help the modem thermal processes survive and continue developing.

Table 2.1 Thermal Processing for the Purpose of Hydrocarbons Upgrading

(adapted f?om Stolfa, 1980)

Operation
1. Thermal cracking

Feed
Gas oil, decant oil,

Products

Gas, gasoline, thermal tar

furfuraI extract,
coker gas oil,

or vacuum gas oil

Atmospheric bottoms Gas, gasoline, middle distillate,

fuel oil

2. Visbreaking

Atmospheric bottoms Atmospheric bottoms

Gas, gasoline, he1 oil

Gas, gasoline, middle distillate, fuel oil

Atmospheric bottoms

Gas, gasoline, middle distillate,

vacuum gas oil, vacuum bottoms

Vacuum bottoms

Gas, gasoline, middle distillate, fuel oil

I
3. Visbreaking & gas oil cracking

Atmospheric bottoms

Gas, gasoline, middle distillate,

vacuum bottoms

4. Coking
(partial recycle)

Atmospheric bottoms

Gas, gasoline, middle distillate, coke

5. Coking
(low pressure)

Atmospheric bottoms
or vacuum bottoms

Gas, gasoline, middle distillate,

heavy coker gas oil, coke

2.2

Pilot Reactors

Pilot reactors are scale down u i s of a commercial or potentially commercial process nt

designed to test a particular process. They are designed based on kinetics obtained fkom
laboratory or bench scale reactors similar to those obtained in this study. Two examples

of pilot scale thermal cracking reactors are given below.

A fast pyrolysis reactor called the thermovortactor has been developed at the University

of Western Ontario for heavy hydrocarbon thermal cracking. Vogiatzis et 01. (1 988)
used this reactor to study the pyrolysis of Cold Lake heavy oil. In this reactor, high temperature nitrogen and preheated heavy oil feed are rapidly mixed in a cyclone-like
device prior to entering a reaction tube. The pyrolysis products exiting the reaction tube

are quenched wt liquid nitrogen in a second cyclone-like device called a cryovortactor. ih The system is capable of achieving temperature in the range of 550-580C and residence
time of 100-500 ms.

Milne et al. (1992, 1999) developed an internally circulating fluidized bed (ICFB)
reactor. The ICFB is a modification to the gas spouted bed. The main advantage of this design is that the ICFB successMly and simultaneously overcomes gas bypassing and

high temperature instability of gas spouted bed. With a pilot scale facility, heavy oil
pyrolysis experiments were carried out and it is shown that the ICFB reactor is capable

of achieving the high temperature (800C) and short reaction time (450 ms) required to

minimize coke by-product formation in the thermal conversion of heavy oils to olefins

and light liquids (Milne et al., 1994).

With pilot scale results, the next step is a demonstration unit, which is designed to show that a process can operate at near to commercial scale. Kinetics fi-om studies such as this one are u s e l l for reactor modelling at all scales of operation: pilot, demonstration and commercial. However, the laboratory reactors used for kinetic studies do not necessarily resemble the commercial unit and for good reason. Residence time and temperature must be more tightly controlled than is possible with a scale down unit.

2.3

Laboratory Reactors for Thermal Cracking Kinetic Studies

Chemical kinetic study is a very important approach to analyze the rates of chemical

transformations and predict rates of reactions under new conditions. A proper kinetic
model combined with suitable reactor model can predict performance for industrial scale production.

Proper design of the reactor is crucial to a kinetic study. When intrinsic rate equations

are not available, experimental data is required. It is preferred to obtain these data from
a bench-scale laboratory reactor, rather than fiom a pilot plant or commercial-scale reactor. The reason for this is with a laboratory reactor, the design as well as operating

as conditions can be chosen to reduce or eliminate the effects of m s and heat transfer. It

should also be noted that these bench scale reactors suitable for kinetic studies are not scaledown versions of units in commercial operation but often designed especially to

obtain kinetic data.

The following are different types of reactors that have been used to study kinetics of hydrocarbon thermal cracking. The advantages and restrictions of each type of reactor

are reviewed.

23.1

Batch Reactors

Before 1970's, most laboratory gas-phase pyrolyses of hydrocarbons were carried out in batch reactors.

However, they are seldom employed on a commercial scale as

productivities are low.

2 . . Conventional Batch Reactors 311

In a conventional batch reactor, the reactant is placed in a closed vessel and pyrolyzed
under constant volume conditions. The reactor is usually made of high temperature

glasses such as quartz or metals such as stainless steel. Temperature is maintained

constant by an electric furnace or liquid metal bath. Batch reactors are generally quite large to hold sizeable sample and to minimize surface effects. It is common that a batch reactor has a few hundred milliliters capacity.

It is not surprising that given their large size, batch reactors do not meet the ideal laboratory criteria. Temperature is not uniform throughout the reactor and there can be dead volumes within the reactor.

The most difficult problem that the conventional batch reactor faces is that it can only used to study slow reactions whose half-lives (tin) are between 15 minutes and 100 is hours (Fabuss et al., 1964). For a first order reaction, if the t ~ n 15 minutes, the reaction rate constant will be 7.7
x

lo4 s? For the high temperatures used in the current

study, the reaction is fast and the heat-up time and cool-down time become nonnegligible. However, the batch reactor does have a definite merit over other reactors:

the reaction time is accurately known.

Hayashitani (1978) carried out an extensive study of the thermal cracking of Athabasca
bitumen using boiling point range criteria and solvent solubility to define pseudocomponents.

M n cracking models were proposed that predicted the general ay

experimental trends. A batch reactor with 8 mm OD and 118 mrn in length was used and the reaction temperatures were in the range of 300C to 450C. Activation energies obtained in all the models wele in the range 188 to 293 kJ/mol. Over the temperature

range 360 to 422OC, the activation energy for the reaction of middle distillates to light
oils was found to be 240 W o l (Model A-3 in the thesis).

23.13 Curie Point Pyrolysis Reactor

Curie point reactor has a radically different heating and quenching concepts from
conventional batch reactors. Even at very high temperature (600 to 800C),it provides shorter time for feed heat up and product quenching than other types of reactors after a precisely controlled pyrolysis time (Fairbum introduced by Simon and Giacobbo (1965).
et

al., 1990). This technique was

In Curie point pyrolysis, a thin ferromagnetic wire is coated with a microgram quantity
of sample and then placed inside a low volume glass tube. The glass tube is iriserted into a high fiequency induction coil. The ferromagnetic wire is composed of magnetic nickel-iron-cobalt alloy. When the high fiequency field is energized, the ferromagnetic
wire is almost instantaneous heated, and hence the sample. At a temperature which

depends only on the alloy composition, termed the Curie point temperature, a transition from ferromagnetism to paramagnetism occurs. At this time, the high fiequency energy

intake by the wire drops dramatically. Consequently, the wire temperature stabilizes at
the Curie point and this temperature is then maintained. After a precisely timed interval, an electronically activated hammer mechanism is used to break open the glass

microreactor. The pyrolysis products are rapidly quenched by m x n with a stream of iig cold helium and the product vapors are flushed into a G C M S system for complete component identification and quantification.
Different ferromagnetic alloys have

different Curie point temperatures. Therefore, the pyrolysis reaction temperature can be

controlled by using different alloy compositions. Wires are available with Curie point temperatures fiom 300 to 1000C.

Fairbum et al. (1990) studied the kinetics of n-hexadecane pyrolysis using the Curie
point pyrolyser. n-Hexadecane has been pyrolysed over conditions which overlap the conventional and ultrapyrolysis regimes, specifically at 576 to 842C and 100 to 3200

ms. Peak ethylene production (28 wt%) was exhibited at 842C and 500 ms. Figure 2.1
depicts the results used to determine the activation energy of n-hexadecane pyrolysis,

which w s determined to be 164 klimol. The desired product of the reaction was a
ethylene, but olefins up to C15 were obtained in appreciable concentrations under lower conversion (60 to 75%) conditions.

23.2 Flow Reactors

For reaction times faster than those that can be accurately studied by conventional batch reactors, flow reacton can be used. Thus, especially since the 1970's, tubular flow reactors have predominantly been employed as laboratory reactors for pyrolysis studies.

The added benefit of using this type of reactor is that many commercial pyrolysis u i s nt,
such as those used for ethane cracking, are large-scale m l i t b reactors. ut-ue

As wt conventional batch reactors, tubular reactors are built with glass or metal tubes. ih

Heat is supplied by the media around the reactor and the reactor wall temperature is

usually deemed constant. If the gas flow is turbulent, it is assumed that there is no mixing in the direction of flow and complete mixing in the radial direction. This type of reactor is called a plug flow reactor (PFR).

An equivalent reactor volume concept was introduced by Hougen and Watson (1 947) to account for slight temperature differences in flow reactors. The equivalent reactor volume is that volume which, at a chosen reference temperature and reference pressure, yields the same conversion as the actual reactor with its temperature and pressure profile. The reference pressure is always one bar and the reference temperature is the average of the process gas temperatures in the last 40% of the reactor. The calculation

of the equivalent reactor volume requires an accurate knowledge of the temperature

profile along the tube. Strictly speaking, the concept is only valid for a single reaction with a single activation energy or for a set of parallel reactions having identical values of activation energies (van Darnme, et al., 1975).

For flow reactors, the key point that requires attention is m i t i i g the ideality of the anann

PFR operation. In a laboratory, it is difficult to realize plug flow because the Reynolds
numbt .- is generally less than 100 (CBme, 1983). Therefore, a temperature profile and a
velocity profile across the cross section of the tube exist, which means that no single

residence time but rather a distribution of residence times exist. At the entrance and exit
region, because of the poor heat transfer, reactants may travel a long distance before reaching the reactor wall temperature. To improve the operating conditions, the PFR

tube diameter should be made small to reduce the temperature gradient. The reactor
lengthldiameter ratio should be kept as large as possible so that the effects of thermal entrance and exit regions are negligible.

Gas flow through a tubular reactor experiences a pressure drop along its length. When
the reactor tube is long and gas flow rate is high to maintain turbulent conditions, the pressure drop along the reactor may become non-negligible and an equation of momentum conservation is required.

Zhou et al. (1987) studied thermal cracking kinetics of several straight-chain alkanes and
their mixtures, including C9, C 12, C13, C16, and C22. The reactor was a straight

tubular reactor under atmospheric pressure and temperature range was 350 to 620 O . C
The reactor w s 305 mm i length with an inside diameter of 8 1 mm (a diameter of 81 a n

mm was quoted in the article, however, the diameter is more likely 8.1 mm to give a
reasonable length to diameter ratio). For dodecane (C12) thermal cracking, the

activation energy was determined to be 164 kllmol.

Depeyre et ui. (1989) studied atmospheric gas oil thermal steam cracking kinetics over a range of 625 to 800C w t the a m of producing olefins, using a helical coil tube reactor ih i

at atmospheric pressure. The reactor used was 4 m in length and the inside diameter w s a
4 mm. A mechanistic radical and molecular model consisted of 138 reactions, 18

species and 24 radicals were proposed. The model made use of a simplified composition

of gas oil. Pure compounds were w d to represent each family of hydrocarbons in the
gas oil, i.e., n-hexadecane represented parafZns, iso-hexadecane represented iso-

p a d f h , C13H26 represented naphthenes and C I SH24 represented aromatics.

Kumar and K u m (1985) studied naphtha pyrolysis over the temperahue of 720 to

800C. An annular tubular reactor was used. The length of the reactor was 60 cm. A
model including 21 molecular reactions and an overall reaction of naphtha to gases w s a
reported. The activation energy of the overall reaction for naphtha to gases was found to

be 220.8 kllmoi.

Froment and co-workers conducted a wide range of studies on the thermal cracking of

Light hydrocarbons and naphtha.

They found that the rate of cracking of single

hydrocarbons closely followed a fint order law (Froment et a[., 1976, 1977; van Damme
et al., 1975). This is true for naphtha cmcking also (van Damme and Froment, 1981).

Further, the first-order rate coefficient was shown to be independent of the conversion.

In other words, fiom a kinetic point of view, naphtha cracks as if it were a single
component and its disappearance can be described by a conversion. For typical naphtha

(boiling point range: 40-195OC), the activation energies were determined to be in the
range of 212 to 228 kJ/rnol.

2.4

The Chemistry of Thermal Cracking

In thermal processes, there are two types of reactions taking place. One, scission, is an
endothermic reaction that makes larger molecules break to smaller molecules; the other,

called condensation, is an exothermic reaction that converts the smaller molecules to larger molecules. Without catalyst, isomerization reactions or olefin polymerization reactions rarely occur. The generally accepted model for paraffin cracking was first presented 60 years ago and is based on a &ee-radical mechanism (Kossiakoff and Rice, 1943).

The free-radical mechanism asserts that when a certain molecule undergoes a thermal reaction, it breaks apart hrst at the position with weakest chemical bond and forms two free radicals. C-C and C-Hbond energy data indicate that the reactions most likely take place at C-C bonds of chain paraffins, as well as at side chains of naphthenes and aromatics.

Smaller fiee radicals such as H e, CH3 and C2Hs may exist for a short time and can

consequently collide with other molecules to form new free radicals. However, larger

&ee radicals are very active and are easily broken to form olefins and smaller fiee
radicals. Thus chain reactions occur. When products leave the reaction system, free radicals will combine to form a paraffin and reactions terminate. Therefore, the ultimate

products of scission will be pamfEns and olefins whose molecular weight is lower than the feed.

This mechanism is illustrated using n-hexadecane thermal cracking as an example

(Cheng, 1994). Many fke radical reactions are possible for a given reactant; several
examples are shown below.

(1) Initiation

Free radicals are formed from the large molecule:

C16H34- CH3 '

+CISH~ 1

(2) Propagation

I this stage, a series of reactions occur that converts reactants to products while n
leaving the concentration of radicals unchanged.
reactions are included: The following types of

reactions of radicals with other molecules,

CH3.

+ C 6 3 -* C 1H4

b + Cla33

larger h e radicals release H and fonn olefins,

C15H31

C15H30 +

H'

C63 1H3

C 6 3 +H 1H2

and larger free radicals crack to smaller ones.

C53 1H1

'

C7H14 + C8H17

C8H17 * C a 8 + C4H9
C4Hp-+C3&+C&.

C4H9

C2H4+ C2H5

(3) Termination

Free radicals combine with wch other.

CH3*+Ha-+Cb

C2H5+ CH3

C3H8

2.5

Reaction Kinetics for Complex Mixtures

Petroleum and its upgraded products are very complex hydrocarbon mixtures. Even the
lighter hctions such as LGO and naphtha contain a continuous spectrum of hydrocarbons and make the reaction feed and product characterization very difficult. A

number of approaches have been applied to deal with complex mixtures. The simplest
approach is to treat the reactions as a black box; this approach does not incorporate any
chemical insight.

Others try to understand the reaction chemistries using model

compounds. More recently, Monte Carlo simulations have been used to construct

ensembles of molecules to represent complex mixtures, and then calculate their reaction behaviour. Lumped pseudo components are also applied to represent important classes of compounds with simplified reactions. (Gray, 1997)

2.5.1

Black Box Method

This empirical method treats the reaction process as a black box. equation is used to determine the extent of reaction in a reactor:

The following

dx
d (LHSV)

= k(T)Xn

where X can be a property of the feed, LHSV is the liquid hourly space velocity in the reactor, k(T) is the reaction constant, and n is the order of the reaction. This method has some limitations (Gray, 1997):

(1) the initial composition is not represented in detail, and therefore, the kinetic
constants are unique to a specific feedstock;

(2) the subsequent reactions of the mixture are not represented in detail, and
therefore, it is not suitable to predict performance outside the initial data set;
(3) this model is reactor-dependent so that it can only be scaled to reactors of similar
geometry.

2.52

Model Compound Method

In this method, the composition of a complex mixture is first studied in order to select

structures of interest and then a model compound is established to represent the

feedstock. Then the selected model compound is reacted at the desired conditions. The

kinetic study of properly selected model compound can provide insight into the
behaviour of more complex or larger molecules.

The m i problem that the model compound method faces is that the key interactions an
between components in the feedstock may be omitted. Another problem is the

hypothetical model compounds are sometimes very difficult to obtain.

An important example of using the model compound method is thermal cracking of n-

hexadecane. ~Hexadecaneis a widely chosen model compound (Voge and Good,

1949; Fabuss, et al., 1964; Depeyre and Flicoteaux, 1991;Khorasheh and Gray, 1993).

2.53 Monte Carlo Simulation Method

This method first uses statistical methods to characterize the structural attributes of a complex hydrocarbon mixture. At least six attributes have been used: pa&
length,

number of naphthenic rings, number of side chains, side-chain length, number of

aromatic rings and number of naphthenic rings per aromatic molecule (Petti et aI., 1994;

Trauth et al., 1994). Then a Monte Carlo method is used to sample these distributions randomly and construct molecules. An ensemble of these molecules represents the
heavy hydrocarbon mixhue. Finally, each molecule is reacted according to the known

behaviour of model compounds and the molecuies are collected into appropriate grouphgs to define the product yield. However, this method is extremely

computationally intensive and with the difficulty of modelling different feeds, the method has not yet been used to predict the observable kinetic or product properties (Gray, 1997).

2.5.4

Lumped Kinetics

In a mixture with many components, it is often impossible, and almost invariably useless
to have a complete description of all components. In fact, industry is often only interested in some sort of lumped concentration.

The lumping method, as introduced by Wei and Kuo (1969) and Kuo and Wei (1969),
seeks to represent large monomolecular reaction systems by low-order, linear

differential equations involving lumped pseudo-components. These pseudo-components are created by grouping chemical species having s m l r behaviour. A detailed analysis iia of lumped-system reaction dynamics was given by Wei and Kuo (1969). Once an appropriate lumping scheme is established, the new lumped-system rate matrix is constructed fiom the response data (Coxson and Bischoff, 1987a and 1987b).

The kinetic model for fluid catalytic cracking by Jacob et a[. (1976) is shown in Figure
2.2. Kinetic constants were determined by experiments on sub hctions with controlled
composition. The subhctions were resolved in enough detail to have similar reactivity,

and therefore, be related to known model compound reactions. The reaction network
was then validated by experiments using actual feeds.

The model can predict

performance even with changes in the feed composition and the operating conditions in the reactor.

One of the limitations of lumped kinetics is that information about individual components, md thus details of the product distribution, is lost. The greatest challenge for applying lumped kinetic methods for a catalytic process is that whenever the underlying catalyst s s e changes, either fiorn changes in composition or aging, all of ytm

the rate constants could change (Weekman, 1979). Fortunately, kinetic studies of a

thermal cracking process avoid this disadvantage. Another problem with lumped
kinetics is that the structures present in the lumps evolve as the reactions proceed, thus

changing the reactivity of the lumps. As a result, the rate constants associated with

lumped models are frequently functions of feedstock properties and conversion.

Naphthenes 342'0-

komatic substituents 342"C+

rings

CIto C4 + Coke

Light Components (220-342OC):

PL=paraffins
AL = aromatic substituents

NL= naphthenes Ca = aromatic rings

Figure 2.2 Kinetic Model for Fluid Catalytic Cracking (Jacob et a[., 1976)

CHAPTER 3
Experimental Apparatus and Methods

This section outlines the experimental set-up and the method. Figure 3.1 is a photograph of the experimental apparatus. Figure 3.2 is a simplified schematic diagram of the apparatus, while a detailed diagram is shown in Figure 3.3. Solid black valves are always closed in the current study and the main process flow is shown by a thicker line. Figure 3.4 is a detailed diagram of the condensing and product collection systems.

3.1

Process Oveniew

A helical coil reactor w s used for the light gas oil thermal cracking kinetic study. a

Pumps pushed the hydraulic oil (HO) and therefore moved the piston of a transfer
cylinder to force the hydrotreated light gas oil (LGO)feed into the preheaters. M e r the feed oil was preheated to about 350C, it was combined with 350C nitrogen coming fiom the furnace. Nitrogen was used to dilute the reactant oil and its flow rate was adjusted to achieve different gas oil residence times in the reactor. Once inside the reactor, feed gas oil w s rapidly heated to reactor temperature. Reynolds number was a found to be about 2500 and the temperature rose at a rate of approximately 800 Wm (see Appendix E for detailed calculations).
At the exit of the reactor, products were

quenched in a series of three water-cooled condensers (Figure 3.4). A portion of the reactor outlet fluid was trapped after each condenser in a flask submerged in an ice bath.

Figure 3.1 Picture of LGO Thermal Cracking Experimental Set-up

Transfer Cylinder

I
Gas Bag 4
I

Condensing System

WTM*

Demister

Figure 3 2 Simplified Schematic Diagram of LGO Thermal Cracking Apparatus .

A demister r l l of rayon fibre (fibre made by Kendall Canada Inc.) absorbed the uncondensed aerosol before gas products entered a wet test meter (WRVI).

The volume of gaseous products was measured by a wet test meter, and gases were
collected in a 200-litre gas bag. After the feed was stopped, nitrogen continued to flow through reactor to sweep the remaining hydrocarbons to the reaction gas bag. Coke

make was evaluated using a bum procedure once the reactor had been well purged.
Another gas bag was employed for the gas product from the coke bum procedure.

The liquid product was collected f o the flasks. Flasks, condensers and pipes were rm
cleaned completely using dichloromethane (Anachemia Canada Inc., > 99.5% purity). Product oil was then separated from the dichloromethane by a small laboratory distillation unit inside a fume hood.

32 .

Experimental Apparatus

321 ..

Helical Coil Reactor

The reactor coil is a total of 18 m in length. It is comprised of two pieces of stainless

steel tubes welded together. The inside diameter (ID) of the first 9 m was 3.34 mm
(3/16" tubing), while the second half of the reactor had a larger ID of 6.16 mm (5116"

tubing) to accommodate the increased volume of cracked gases. The two parts were

welded together. As shown in Figure 3.5, the helical coil was wound into a groove around a 1Ocm diameter stainless steel bar stock. Detailed dimensions of the reactor are

shown in Figure 3.6. T i assembly was contained within a larger cylindrical section of hs
bar stock (see the lower section of Figure 3.5).

Reactor length is long enough

(lengtkddiameter > 1000) to eliminate the effects of entrance and exit region. The long reactor tube was coiled to allow it to fit in a smaller bar stock,

The bar stock was heated both from the centre and the exterior. The large thermal mass provided energy for the endothermic reactions. Figure 3.7 shows the arrangement of reactor heaters. Heaters are made by ARI Industries, Inc. Two heaters were coiled around the larger cylindrical section of bar stock holding the reactor assembly. They

had a heated length of 3.20 m,6.4 mm in diameter, 7200 W power and 1.132~ wlm2 10'
power density at 230 volts. A third smaller heater made by A N was used to heat the

bottom section only. This heater accounts for convection effects. Six others were
installed at the centre of the reactor assembly.

3.2.2

Transfer Cylinder and Hydraulic System

As shown in Figure 3.3, the hydraulic system comprised a hydraulic oil reservoir, an

overflow t n ,hydraulic oil filter, flowmeters and three pumps. ak

Figure 3.5 Photograph of the reactor

Figure 3.6 Detailed Dimensions of the Reactor

Heater #I

64.14 cm

The two pumps (pump #2 and #3 in Figure 3.3) used in the current study were made by

Milton Roy Company (product number 9-2015003). They are displacement type and
can raise pressure up to 40 MPa Both pumps operating together can generate a

maximum flow rate of 0.3 gls of LGO. A hydraulic oil filter was installed before the
pumps to remove the possible particles or debris. As shown in Figure 3.3, pressure safety valves were used to protect the rest of the system fiom overpressure. They were
set at 0.5 MPa The pump had a knob with scale ranging from 0 to 100. These t o w

pumps were set i parallel and the calibration was done with both pumps on and at the n

same setting on the scale. The calibration c w e for the LGO flow rate is given in
Appendix C. A photograph of the two pumps is shown in Figure 3.8.

A transfer cylinder bridged the hydraulic system and the LGO feed. A photograph of

the transfer cylinder can be found in Figure 3-9. Pumps pushed the hydraulic oil to move the piston in the transfer cylinder. The volume of the transfer cylinder was 2.3 litres and the piston diameter w s 10 crn. a

As shown i Figure 3.3, Hydraulic oil was stored in a reservoir. By opening valves HI, n

H ,H9,H 0 H11, H12 and closing valve F4T,hydraulic oil was able to circulate within 5 1,
a closed circuit. Once valve H12 was turned off and valve F4T was turned on, pumps could push hydraulic oil to move the piston in the transfer cylinder.

Figure3.8 Pumps (#2 and #3), Furnace and Hydraulic Oil Flow Meter

Figure 3.9 Transfer Cylinder and Pumps (#2 and #3)

3.23

Feed Preheaters and Gas Furnace

Feed LGO was heated from room temperature to about 350C by two preheaters. Before

nitrogen was mixed with the preheated feed LGO, it was heated to 350C by a furnace.

The fimace was also employed to preheat air to 350C before it entered the reactor
during coke burning.

The fiunace was a Series 31 10 tube type fiunace made by Applied Test Systems Inc.
The maximum operating temperature was 10IOC and the power was 1,332 W. It had single-zone construction and the heated length was 0.7i m. The dimensions of the furnace were 9.5 crn I x 79 cm length. A photograph of the fiunace can be found in D Figure 3.8.

3.2.4

Product Condenser and Collection

The reaction effluent was quenched, and then fixher cooled in a series of water-cooled
condensers. The first condenser had a 40 cm long copper water jacket outside a stainless steel tube and the other two condensers were regular glassware condensers 30 cm in
length. As shown in Figure 3.4, after each condenser, there was a flask to collect the

condensed products. Cold water flowed through the shell side of each condenser. Each
flask was placed in an ice bath. A demister 111 of rayon was installed after the

condensers to coalesce any aerosol before gas products entered a wet test meter. A photograph of the condensing system and demister is shown in Figure 3.1 0.

As shown in Figure 3.2, the volume of the gaseous reaction product combined with nitrogen w s measured by a WTM (DM3C-2.5; G. H. Zeal Ltd.; drum capacity 2.5 a litres). The gas w s then collected in a gas bag. The gas bag had a capacity of 200 a

litres. Coke burning product was also measured by the WTM and collected using
another bag with the same capacity. A photograph of the WTM and gas bag can be seen

in Figure 3.1.

After each run, the condensers, flasks, demister and pipes connecting them had to be completely cleaned using dichloromethane (normal boiling point is 4 ) 0C. Dichloromethane and oil product w s separated in a simple batch laboratory distillation a

unit. A photograph of this unit is shown in Figure 3.1 1.

33.5

Process Instrumentation and Dab Acquisition

All the thermocouples used in the current study were calibrated K type made by Alltemp

Sensors. They were installed at the preheaters, furnace and reactor. Temperatures at the
reactor inlet and outlet were measured by thermocouples in the fluid. The inlet

thermocouple was 1.6 mm (1116 inch) i diameter and was placed about 30 cm into the n reactor tube. The inlet thennocouple may actually be touching the reactor tube wall.

Figure 3.1 0 Picture of Condensing System

Figure 3.1 1 Picture of Distillation Unit

However, because of the rapid heat transfer (see Appendix E for detailed calculations),
the wall and fluid temperature equalizes within approximately 20 cm. The outlet

thermocouple was 1.6 mm (1116 inch) in diameter and w s inserted into a T-connection a about 10 cm away fkom the reactor bar stock. The T-connection was covered by a 10

cm thick insulation to minimize heat losses. Because of the T-connection, the outlet
thermocouple was measuring the fluid temperature.

Temperature control was realized by UT350 general-purpose controllers for the reactor,
preheaters and fimace. The controllers were made by Yokogawa Electric Corporation
with an input sampling cycle of 250 ms, an input accuracy of H.1% 111 scale for of

process variables and an output accuracy of H.3% the span. The controllers can be of operated in "Auto" or "Manual" mode. Prior to an experiment during the heat up stage, a certain percentage of power was set and once the temperature was close to the set point, it was then switched to Auto control mode. LabView (Version 4.1) with a pentiurn computer was used as an on-line temperature data acquisition device. The reactor inlet and outlet temperature were measured and stored in a file every two seconds.

33

Light Gas Oil Properties

Hydrotreated light gas oil, used as feed in the current study was provided by Syncrude

Canada Ltd. Some properties of the hydrotreated LGO are given in Table 3.1.

Table 3.1 Properties of Feed Light Gas Oil

hydrotreated light gas oil Density at l!iC, g/ml

0.878'

SimDist '
5% off, "C
180

10% off, "C

210

90% off, "C

95% off,
O C

370
400

Mean MW, g/rnol

212

(estimated fiom SimDist)

' (Yui, 1999)

(Table A. 1)

3.4

Gases Used i the Study n

Nitrogen used in LGO thermal cracking experiments, air used in the coke burning
procedure, and helium used in GC analysis were all produced by Praxair. Table 3.2 gives the product grades and quality specifications of the gases used in this study. Air

used in the coke burning procedure was tested for C02 and CO content on the GC; only
trace quantities were found. Therefore, coke analysis will not be affected.

3.5

Light Gas Oil Thermal Cracking Experiments Procedure

The light gas oil thermal cracking experiments were conducted using a procedure developed in this study. The steps outlined below were strictly followed in the given order for all experiments in this study. A detailed process flow diagram of the apparatus is shown in Figure 3.3. Solid black valves are always closed in the current study.

Prior to an experiment, condensers were disassembled and meticulously cleaned

including the product collecting flasks, demister and process piping. The last one of
three liquid product-collecting flasks (refer to Figure 3.4) and the rayon for the demister were weighed. WTM water level was checked and adjusted as instructed in the WTM

m n a . The transfer cylinder was detached and checked to find out how much feed was aul

left, refilled as necessary and then reinstalled. Rayon was evenly put into the dernister.
It was very important that the demister should be assembled upside down so that the

Table 3.2 Product Specification of Gases Used i Experiments n

Nitrogen

Air

Helium

Product Grade

Oxygen Free
< 0.5 ppmv
99.998%

Hydrocarbon Free
19.5 - 23.5%

5 0 UHP .
< 1 ppmv
< 5 ppmv

0 2

N 2

< 3 ppmv

Helium, %

99.999"
< 2 ppmv < 0.5 ppmv

H20
THC

< 0.1 ppmv

HZ

< 0.1 ppmv < 0.1 ppmv

' 1 ppmv
< 1 ppmv
< 1 ppmv

co2
CO

* Neon fiee basis.

rayon could be packed more tightly at the entrance and exit. Finally, the demister, condensers and flasks were reassembled. SS-30 copper anti-seize compound (Jet-Lube of Canada Ltd.) was used to thoroughly coat the fitting between the first condenser and the reactor. Vacuum grease was put onto the joints of all glass condensers and flasks.

The gas bags were marked and always used for the same purpose. Gas bags for reaction

and coke burning were vacuumed using the building vacuum line, refilled with nitrogen
and then vacuumed again. This should effectively get rid of the poteatially remaining
hydrocarbons or C02 and CO. The reaction gas bag w s then attached to the apparatus. a

The valves and piping were checked and then nitrogen was introai~ced a small flow at rate of about eight ml/s before raising the reactor temperature. LabView software was then loaded and a data file was created to monitor and record the temperatures. The reactor heaters power were set to 20% and h a c e power to 25% using the temperature controllers on the control board. The cooling water flow to the condensers was started at

this time and the exhaust gas w s allowed to flow to the wet test meter. a

The reactor required four to five hours to reach its final operating temperature.
Although the temperature controllers could be used to adjust power in the "Auto" mode to stabilize at the desired set point, there was oscillation within a range of 10C and a period of approximately 40 minutes. Experience showed that at 46% power, after the reactor reached the set point, could maintain the reactor temperature constant for at least
2 hours. Manual power settings were therefore used for experiments. Preheaten were

then turned on with 20% power, it took t e about 2 hours to reach 350C. Once the hm preheaten reached the set points, they were set to "Auto" mode.

Half an hour before the reactor reached its desired temperature, the oil feed pumps #2
and #3 were started in order to circulate hydraulic oil within a closed loop and stabilize pump operation. Special attention should be paid to the low flow rate operation. A large flow rate should be used, and then reduced to desired flow rate. T i ensured there hs was flow going through the pumps. These pumps needed 20 minutes to be stabilized.

Ice was added to containers surrounding the flasks. Current atmospheric pressure was
recorded fiom a barometer. Nitrogen flow rate was increased to the desired value by

measuring flow volume going through the WTM within a known time period. Valve P6
in Figure 3.3 was opened to stop the flow going through the WTM so that it could be zeroed.

Valve F6 w s opened to prepare to bleed the feed oil to a waste oil tank. Immediately a

after valve F4T was opened, valve H w s closed to stop hydraulic oil circulating and 5 a
start to feed preheaters. Feed oil thus went through both preheaters and exited fiom

valve F6 to the waste oil tank (a 4 litre metal can). After about two minutes, the feed flow coming from F6 becomes stable. While waiting for a stable feed flow, nitrogen flow was directed to the gas bag (close P6,P7 and open P8).

The experiment started when valve F6 was closed and F7 was simultaneously opened to
feed the reactor with Light gas oil. An assistant operator started the timer to record the feed time.

After about 15 to 30 minutes of run time, valve F7 was closed and the bleed valve F6

w s simultaneously opened. The timer was stopped and feed time recorded by the a
assistant operator. Valve F6 was closed after pumps were turned off. Nitrogen flow through the reactor was continued until condensers became clear. During nitrogen

purging, the gas chromatograph (GC) w s prepared for analysis of the gas samples. To a
stop the flow of nitrogen to the gas bag and switch it to the vent, valves P6 and P7 were opened and P8 w s closed simultaneously. One should take care not to overpressure the a

WTM during this operation. Reaction gas volume was read from WTM. Data
acquisition with LabView was stopped and data saved.

A Hamilton

SAMPLELOCK syringe was used to obtain a 10 rnl gas sample from the

reaction gas bag, which was analyzed by GC.

Before burning, the reactor temperature w s raised to 600C and coke-burning gas bag a
was attached to the system. LabView w s started for monitoring purposes only. Ice was a
added into bucket with cooling coil for coke burning. Valve P4 was then opened. The

M condenser (shown in Figure 3.4) was detached fiom the system and the rest of the

condensers were isolated. Before the cooler for coke burning was attached to the
reactor, anti-seize compound should be thoroughly coated to its fitting.

AAer the WTM was zeroed again, valve P8 was opened and valves P6 and P7 were closed to direct nitrogen into the gas bag. Air was then introduced by closing valve F10 and opening valves F 11 and F 12. Coke burning continued for about one hour. Then air

flow was directed to vent. Total gas bag volume was recorded by the WTM. The
Hamilton SAMPLELOCK syringe w s also used to obtain this gas sample for GC a analysis.

After the run, the condensers were again disassembled and completely cleaned with
dichloromethane. Liquid product was first mixed and collected in the flask immediately

before the demister, and then the flask was weighed. After weighmg, the liquid product

w s collected in sample bottles for fuhw analysis. a

The mixture of oil and

dichloromethane was then distilled to remove dichloromethane and residue was weighed. Rayon in the demister was also weighed.

36 .

Product Analysis

Product analysis included GC analysis of gaseous products and simulated distillation

analysis of the reaction liquid product. Because the coke make was very small, no

attempt was made to analyze the coke properties.

361 ..

Gas Analysis

In the current study, a Hewlett-Packard (HP) 6890 gas chromatograph (GC) and HP
ChemStation software have been employed. The GC has a large column oven, dual injection ports and a thermal conductivity detector (TCD). The H ChemStation P software contains all parameters for instnunent control, data acquisition and evaluation,

including integration, quantification and reporting.

Two columns were employed: KP - Plot Molecular Sieve (MS) 5A, 15 m

50.0 pm film thickness and

0.53 mm x

H - Plot Q, (Divinylbenzene/styrene Porous Polymer), 30 P

0.53 mrn x 40.0 pxn film thickness. These fbsed silica capillary columns provide a

ih quicker analysis time wt better resolution than the traditional packed columns.
However, these columns should not be exposed to temperatures greater than 250C.

Thus, the maximum GC oven temperature was 250C. The MS column was used to
separate inerts (9 and N2),methane and carbon monoxide. separated carbon dioxide, ethane and heavier hydrocarbons. The Plot Q column

An analytical method was set up to analyze the reaction gas bag and coke burning bag.

T i method was a collection of set points required to run a single gas sample on the hs
GC. Table 3.3 summarizes the operating conditions used in gas sample analysis.
Details on the GC method can be found in Appendix D.

Table 3.3 Gas Chromatograph Analysis Conditions

Reference Flow
Temperature Program

Helium, 22 d m i n
Initial temperature 60C

0-2min, 60C
2-4 min, ramp to 80C at 1 OC/min

4-9.7min, ramp to 250C at 30C/min

Total Analysis Time

9 7 min .

System Temperature:
Inlet
140C

TCD

260C

TCD Detector:
Range

0.025 mV Ml scale

D t Range aa
Minimum Peak Width Cylinder Pressures Helium
Air

20 Hz
0.01 min

620 kPa gauge (90psig)

480 kPa gauge (70psig)

3.6.1.1 GC Calibration

Calibration is the process of determining response factors used to calculate absolute component concentrations by injecting specially prepared calibration samples. The calibration table is also used for peak identification.

Single-level calibration method was used for hydrocarbons and C02 was calibrated

using a multi-level method. The response of the detector is assumed linear over the
working range of concentrations for the samples of interest. The response factor for a
given component peak is given by the inverse of the slope of the calibration curve line
through the point and the origin.

A calibration sample, also referred to as a calibration standard or a standard mixture, is a

sample containing a lcnown amount of the compound to quantify. Figure 3.12 shows a chromatograph of the standard gas mixhue and its composition is shown in Table 3.4.

362 ..

Simulated Distillation Analysis of Liquid Samples

Simulated distillation methods are chromatographic techniques which provide data that emulates glassware distillation through gas chromatography. ASTM method D2887 was used for the current study. The column (Series Number 2892 15) is 10 m
x

0.9 micron

film thickness and 0.53 mm ID. The external standard is ASTM D2887 reference gas

. . . .

r . .

d
I
ld

13.241

I--

1
3
2.073 - Oxygen

3 d

- -

+* e e g c n M

31

- 2.763 - Methane

6.521 ISO-BUTANE 6.740 - N-BUTANE

7.626
7.784

ISO-PENTANE N-PENTANE

Table 3.4 Standard Gas Mixture Composition for GC Calibration

Component

moi %
1.5 10

Component i-C4Hla

mol %
0.5
0.5

0 2

CO

n-caio

cb
CzH6
c2b

5
5

i-CSHI2
n-CsH12 i-c6H 14
nC6H 14

0.5
0.25

0.25
0.25
50.25

co2
C3H8

20
2
2

N 2

cH 36

oil, and the retention time standard is SD-SS3E-05 (carbon number ranges fiom C4 to

C66). The column and the standards are made by Separation Systems Inc.

CHAPTER 4
Results and Discussion

In keeping with the objectives of this study, hydrotreated Light Gas Oil (LGO) has been
pyrolysed at temperatures varying fiom 500 to 550C. Because of the complexity of the reaction system, a number of runs with and without LGO feed were performed in order to identify and debug various operating problems with the reactor, condenser, demister,
pumps, instrumentation and other equipment. The results of these studies prompted

many design additions and modifications, such as using different condensers, a demister
and a gas bag, as well as using the two pumps providing the smallest flow rate. Besides those tests, eight successful runs were conducted where light gas oil w s fed to the a reactor. A summary of all successful experiments appears in Table 4.1. Reactor

temperature profiles are given in Appendix B. The feed time shown in Table 4.1 is the interval between the feed starting and feed stopping (see Figure B. 1). A summary of the operating conditions used in each of the successll experiments is shown in Table 4.2. Unsuccessful experiments are listed in Appendix H.

A simple four-lump kinetic model for the LGO thermal cracking reaction was developed

and kinetic parameters are calculated and discussed. The simple model allows us to
determine the amount and rate at which the LGO pseudo-component is converted to the desired product (naphtha) and to by-products (gas and coke).

Table 4.1 Summary of Successfbl Experiments

Run

Date

Reactor Temperature, O C

Feed Time,
s

Comments

1 2
3

99-08-31 99-11-09 99-11-17

550
530 500

1256
1212
1596

First successfbl run at 550C. Successful run at 530C. Longest pyrolysis experiment performed in study. Over 200g of LGO processed.

4
5

99-12-01
99-12-15

550

1584

Successfbl repeated run. Successfbl repeated run. Successfbl repeated run. Successful run with higher feed f o rate. lw
Successful run with higher N2flow rate.

500

1386 836

00-02-02
00-03-27 00-04-03

550
550
530

7
8

779
912

Table 4.2 Summary of Experimental Conditions

Run

T, O C

Feed Flow

NzFlow
Rate (L/s)
0.058 0.058 0.058 0.075 0.058 0.058 0.058 0.058

LGOMz
mass ratio
2.0 2. O

Residence Time

Rate (g/s)
3
5 500 500 530 0.1314 0.1314 0.1314 0.1314 0.1314 0.1314 0.1314 0.1838

6)
1.73
1.73
1.64

2
8

2 .O
1.5 2.0 2.0

530
550 550 550
550

1.35 1.58 1.58

I
4

6
7

2.0
2.7
1.41

4.1

Material Balance

Products fiom each run included the gas in reaction gas bag, coke in the form of COz in
the bum gas bag, liquid product directly collected from the flasks after each condenser,

liquid product recovered f o remaining oil in the condensers and pipes, and liquid rm

product in the form of mist trapped by cotton in the demister. Table 4.3 gives a
summary of the m s balances. More details on the material balance can be found i as n

Appendix A.

4.2

Product Analysis

As outlined in Chapter 3, the reaction gas products were analysed by an HP6890 TCD

gas chromatography. The liquid products were analysed by an HP6890 simulated

distillation system.

4.2.1

TCD GC for Gas Product Analysis

An HP6890 TCD GC was employed to analyse gas in the gas bags after each run.
Example chromatograms of the reaction gas product and the coke burning gas product
are shown in Figures 4.1 and 4.2. Although a GC equipped with a flame ionization

detector would be more accurate, the potential error in overall mass balance was small

Table 4.3 Summary of Mass Balances

Run

Feed

L
(g)

G
(g)

C
(g)

Recovery

L
Wt%

G
of
W o of t Y

("C) (g)

of
feed

product
93.6
94.1
I

product
6.4 5.9 13.4

3
5
2

500 500 530 530 550 550 550

550

209.7

188.1

12.7 10.3

0.05

95.8 96.0 95.1 93.6 96.8 96.0 95.3

182.0 164.4

0.05
0.20 0.10

159.2 130.9 20.3 119.8

86.6
87.2
74.6

8
1
4

97.7

14.3 40.5 52.0 29.0 34.4

f 2.8
25.4 26.1 27.8 24.9

165.0 118.8 208.2 147.4 109.9 75.4

0.35
0.40 0.30

73.9
72.2 75.1

143.2 103.6

0.30

96.6

Figure 4.1 Chromatograph of Reaction Gas Product o f Run #7

since the mass of the heavier components (C5+) for which the TCD would be less
accurate formed only a small percentage of the total gaseous product.

4.23 Liquid Sample Analysis

Resolution of mixtures by gas chromatography is limited to hydrocarbons up to 12

carbons (molar mass of 170). Beyond this size, the numerous isomers are too similar in
their properties to be separated by a chromatographic column. Mass spectrometry can
provide information on homologous series up to a molar mass of 1500, after extensive preparative separation of the samples. However, the currently available G C N S device could not resolve the light gas oil feed and its thermal cracking products because of the large number of isomers.

PONA (Paraffins, Olefins, Naphthenes and Aromatics)

analysis was considered in the study. However, with currently available devices, a satisfactory PONA analysis was not possible because the column was not long enough (40 m) and could not separate the peaks effectively. NMR analysis is u s e l l for characterizing the average molecular characteristics, but it cannot give boiling distributions of species (Gray et a[., 1991; Ternan et a[., 1994). Consequently, the
mixture of components in the feed and the products defy any attempt at complete

component-by-component analysis. Therefore, simulated distillation (SimDist) is the

best alternative for the liquid analysis.

As explained previously, the liquid products collected were analyzed by simulated distillation gas chromatography (SimDist) according to ASTM method D2887-89.From

the results of SimDist, it is found that the boiling point range and end boiling point have

been reduced sigmficantly after the feed light gas oil was thermal cracked. The higher
the reaction temperature, the lighter the product. Simulated distillation curves for feed

and products at three different reaction temperatures are shown in Figure 4.3. SimDist
results for all eight experiments are shown in Appendix A.

No attempt w s made to mix the light condensate collected in the demister with the a product oil prior to the liquids analysis. However, the oil trapped by the demister was considered part of the naphtha component because after the second condenser the s r a tem temperature had already dropped to room temperature and the demister was located after the third condenser.

43

Residence Time

The residence times were calculated by assuming the molecular weight of the fluid in
the reactor is the average value of the inlet feed and outlet stream. Gases in the reactor

were at low pressure (slightly above atmospheric pressure) and high temperature (500550C), so the compressibility factor is e q d to one and the gases can be assumed to

behave ideally. The molecular weights of feed and liquid products were estimated by the results of the simulated distillation (SimDist) analysis. The molecular weight of gas

---

Feed Product at 500C Reaction-Run 4

-Product at 530C Reaction-Run 2

. - - - - . Product at 550C Reaction-Run 6

20

40

60

Weight Percent Off, %

Figure 4.3 Simulated Distillation Curves for Feed and Products

product w s calculated &om the gas chromatograph results. (See Appendix A for details a
of the calcdations.)

where:
feed mass flow rate, g/s

average molecular weight of inlet feed and outlet product, g/rnol reaction pressure, Pa

gas constant, Jlmol K

residence time, s reaction temperature, K room temperature, 296 K volumetric flow rate at reactor conditions, m3/s
nitrogen volumetric flow rate at room temperature, rn3/s

reactor volume, m3

4.4

Kinetic Model Development

4.4.1

Lumped Reaction Network

Lumped pseudo components were applied to represent classes of compounds with

simplified reactions (for example, Wei and Kuo, 1969; Jacob et al., 1976; Weekman, 1979). A set of four lumps divided by boiling point range is suggested. The four lumps

in the reaction s s e include naphtha @oiling point less than 196OC), light gas oil ytm
(boiling point greater than 196OC), gas (n-butane and lighter components) and coke. According to these categories, C5 and C6 in the reaction gas bag and the mist trapped by the dernister are included in the naphtha boiling range.

A four-lump cracking reaction scheme was chosen to represent the cracking reactions

taking place in the reactor because of its simplicity and its ability to predict the product
distribution. Figure 4.4 shows the four-lump reaction network.

There was a large volumetric expansion in the helical coil reactor due to the cracking reactions. In order to be consistent and to be realistic, in this study all reaction rate calculations were performed on a m s basis. as

n
Naphtha (1)

n
Coke (4)
Figure 4.4 Kinetic Model for LGO Thermal Cracking

(numbers i brackets are pseudo-components) n

4.43

Model Assumptions

According to this scheme, gas oil cracks to form naphtha, gas and condenses to coke

while naphtha may crack further to form gases. It is assumed that the coke formed by
condensation reactions f?om the naphtha firaction is negligible compared to coke formed

from gas oil cracking. It is believed that the activation energy of gas producing coke is
much higher. Therefore, the coke formation reaction rate is very low and it is reasonable
to assume that gas will not condense to form coke at the temperature range of interest.

In thermal cracking studies, nitrogen or steam has been employed as a diluent in tubularflow reactors in order to control hydrocarbon partial pressure and residence time
(Depeyre et al., 1989; Khorasheh and Gray, 1993). These inert gases can reasonably be assumed to have no direct effect on the homogeneous thermal cracking reactions if the reactions are approximately first order and if the reaction rate constants are independent of feed concentration. Steam, however, may have an indirect effect on the reactions by reacting with coke via the following reaction.

C (Coke)+ H 0 (Steam) -3 z -

CO + H2

(4.3)

Steam also contributes the formation of metal oxides on the reactor surface. The steam treatment of temperature-resistant Incoloy 800 at 800C for 24 hours resulted in
significant formation of NiO, Fe304and hydrogen. (Tsai and Albright, 1976)

(metal)surfaet+ H 0 2

.-

(metal oxides),de

+ H2

Because steam was not employed in the study, it is unlikely that surface effects played a

major role.

At least seven surface reactions are thought to occur during most industrial pyrolysis. In

a complex while not completely understood manner, they all contribute directly or
indirectly to coke formation. Coke formation is undesirable in general because it

decreases yields of desired products and increases heat transfer resistance. If there were significant coke formation in our study, it could also increase reactor tube pressure drop.

For the hydrotreated light gas oil feed at the reaction temperature range of this study, it
has been found that the coke formation reaction rate is very small compared to that of other cracking reactions. Therefore, reactor wall surface effect can be assumed to be

negligible.

As indicated by Kays and Crawford (1993), the effect of including the hydraulic and

thermal entrance regions on the tube temperature profile is small for lengthldiameter ratios greater than 50. Thus, the entrance regions are neglected because of their minor
effect in the long reactor tube used in this project (length/diameter >1000). At reactor

temperatures of 500 to 550C, light gas oil is vaporized essentially instantaneously (99% vaporized at 470C;see Figure 4.3). Reynolds numbers for the experiments ranged from

2500 to 3000. These are i an acceptable range of turbulent flow to ensure plug flow n

operation and rapid heat transfer for LGO heating up and quenching (see Appendix E for detailed calculations of heat up rate and Reynolds number). Detailed calculations for pressure drop are included in Appendix F.

4.43 Evaluation of Pseudo-Component Fractions

As explained previously, the four pseudo-components are naphtha, gas oil, gas and coke.

The mass of gas pseudo-component was calculated using methane to n-butane concentrations fiom the GC analysis results of the reaction gas bag and the gas bag

volume &om the WTM. The mass of gas is calculated as follows.

Assuming the C/Hmole ratio to be 1:1, using C02 concentration obtained from GC analysis results of the coke burning gas bag and the gas bag volume fiom WTM, the

m s of coke pseudo-component was calculated as follows: as

With the result of liquid product SimDist analysis, concentration of C5+ from GC

analysis results of the reaction gas bag and the weight increase of the demister, the mass
of naphtha pseudo-component can be calculated as following:

rnleid

is the mass of total liquid product trapped in the demister and the three

condensers.

The gas oil pseudo-component is simply the heavier portion (SimDist > 196OC) of the liquid product (exclusiw demister weight increase).

Therefore, the weight percent of each pseudo-component in the total experimental products can be obtained and is shown in Table 4.4. Conversions of LGO pseudocomponent are also listed in Table 4.4.

4.4.4

Reaction Network Equations and Rate Constants

The reactor w s assumed a plug flow reactor. By considering a small volume of a reactants flowing through the reactor tube, the kinetic equations of the plug flow reactor

Table 4.4 LGO Conversion and Pseudo-Component Distribution

Run

T,

Naphtha
xl,wt0/0

LGO

Gas

Coke

LGO Mass
Conversion,%

"C
Feed
3 5

x ~ , w t % x3,wtYo

~q,wtO/o

500 500 530 530 550

550

6
15.16

94

0
6.12

0
0.02 0.03
0.13 0.09

16.3
14.2
31.8

78.69
80.61 64.13 66.19 52.13

13.69
22.99
22.31 23.54 24.84

5.67
12.75 11.42 24.12
24.52

2
8 1 4

29.6

0.22
0.20

44.5
46.3 48.1

50.44
48.78

550
550

2.0 61
26.97

24.83
22.05

0.29

50.76

0.22

46.0

and those of the batch reactor are essentially the same as long as the volume is constant. For thermal cracking reactions, volume expansion is significant (about 3 times under the conditions in this study). However, the batch reaction equations may still be used on a
mass basis. According to the reaction network shown in Figure 4.4, first order reaction

kinetic equations for the four reactions are outlined below. The reaction network results
in a system of four ordinary differential equations, which m s be integrated. ut

The integrated form of equations (4.9) to (4.12), become:

where:

ki, k 2 . 4 , kq
t
Xl0, x20

first order rate constants for reactions 1,2,3,4, s"

residence time, s

initial (feed) weight percent of pseudo components 1 (naphtha) and 2 ( L O ) weight percent of pseudo-component 1 (naphtha), 2 (LGO),
3 (gas), 4 (coke)

xlo and x2o are known f o the SimDist analysis of the feed, rm

XI,

x2, x3

and

are

determined fiom the experimental material balances, and the residence time is calculated
assuming an average molecular weight and ideal gas flow.

Based on Equations (4.13) to (4.17), a non-linear regression method was used to determine the reaction rate constants at different reaction temperatures. The reaction rate c o ~ t s kl to ,

14, were calculated using the Microsoft EXCEL

2000 solver

function. Although more than one result could be obtained, only one set was reasonable for a particular run. Negative results were rejected due to physical constraints and the order of magnitude of reaction rates was compared to literature on thermal cracking. The initial estimate supplied to the solver function could be changed by 30 percent without affecting the converged results. The results are shown in Table 4.5. A set of detailed tables for the calculation of average molecular weight of feed and products, feed
and product component weight percent and residence time can be found in Appendix A.

4.4.5 Frequency Factor8 and Activation Energies for the Four-Lump Model

Once the reaction rate constants at different reaction temperatures have been determined, activation energies and fiequency factors can be calculated by the Arrhenius relationship:

where:
*j

frequency factor for reaction j, s"

Table 4 5 LGO Thermal Cracking Reaction Rate Constants .

Run

Temperature, O C

activation energy for reaction j, Jlmol

first order rate constant for reaction j, s-'

gas constant, 8.3 14 Jlmol K

reaction temperature, K

The linear relationship of Equation (4.18), for each of the four reactions, is shown in
Figures 4.5 to 4.8. A summary of the Arrhenius parameters is given in Table 4.6. The activation energy for the reaction of naphtha to gas was 215 k.J/mol, which is close to

220.8 kJ/mol obtained by Kumar and Kunzru (1985) and within the range of 2 12 to 228
kl/mol found by van Damme and Froment (1 98 1).

4.5

Overall Reaction Rate

Although the activation energy for the naphtha lump reacting to form gas compares favourably with the literature, there is no data in the literature with which the results of the four-lump reaction scheme can be directly compared. Therefore, calculations have been performed based on a simplified overall reaction for which comparison to the literature is possible. This frt order overall reaction is assumed to be: is

ko
Gas Oil
- - )

Products (Naphtha, Gas and Coke)

The initial weight percent of LGO was 94 and naphtha was six percent. Repeating
similar calculations and regressions on the Microsoft EXCEL 2000 solver function, an

Table 4.6 Activation Energies and Frequency Factors

LGO +Naphtha

El,kJ/mol
A*,S"

173 3 . 2 2 10" ~ 215 2.4 1 x loL3 83 1 . 2 3 1o4 ~

Naphtha +Gas

E2, kJ/mol
AZ,s"

LGO

Gas

E3, H/mol
Ap,S-I

LGO ++

Coke

Eq, Id/mol

262
8.81~10'~

h,s-'

overall reaction rate constant can be obtained and results are given in Table 4.7. Based on these results, the overall reaction frequency factor and activation energy were

0 determined to be 1-9 x 1' s- and 153 Wmol, respectively (see Figure 4.9).

'

Table 4.7 Ovcrall Thermal Cracking Reaction Rate Constant

Runs Temperature, O C
1000xb, s-'

3
500 103

1
550

7
550

500

530
233

530
260

550 393

550 420

89

373

437

>

The molecular weight of the gas oil fraction (>196OC) in the feed gas oil is approximately 221, which is very close to n-hexadecane (MW
= 226).

Fairburn et ul.

(1990) used a Curie point pyrolyser to study ultrapyrolysis of n-hexadecane. She found that the activation energy of n-hexadecane pyrolysis was 164 W o l . The overall activation energy of this study (at 153 Wmol) is quite close to this value.

It should be noted that LGO thermal cracking is a very complex process and it is almost impossible to identify all the chemical species present in the feedstock and study their chemical reactions. The lumping scheme represents an important tool to simplify the model. These simplitications lead to some errors in the model predictions. However, as results of overall reaction kinetic parameters are very close to those of previous

researchers using similar model feed, it can therefore be inferred that the proposed fourlump model is reasonable and can be used to predict the product distribution.

4.6

Product Distribution Prediction

Using calculated activation energies and frequency factors of four hypothetical reactions, shown in Figure 4.4, a reaction rate constant at any temperature within the range tested experimentally can be obtained.

Figures 4.10 to 4.12 are the results of model predictions at several typical operating
temperatures. According to the proposed model, the desired product, naphtha, is an intermediate product &om gas oil to gas. Therefore, there should exist optimal

conditions for which maximum production of naphtha is possible. A summary of the optimal reaction time at different temperatures is shown in Table 4.8.

Table 4.8 Optimal Reaction Times at Different Temperatures

Temperature, "C
Residence time, s

500
10 29

530
3 -6

550

1.8

M x m m Naphtha Yield, wt% of product aiu

28

26

Residence Time, s

Naphtha
Naphtha Data

/ --- Coke

LGO LGO Data CokeData

--***.

Gas Gas Data

,
j

Figure 4.10 Component Mass Percent vs. Residence Time at 500C

Naphtha
--- Coke
i

Naphtha Data

-LGO
LGO
A

Residence Time, s

-...-. Gas
0

Data CokeData

GasData

1
1

Figure 4.1 1 Component Mass Percent vs. Residence Time at 530C

Residence Time, s Naphtha NaphthaData

A
---.I.

LGO Data Coke Data

CI

Gas Gas Data

Figure 4-12 Component Mass Percent vs. Residence Time at 550C

From Figures 4.10 to 4.12, it is also found that coke formation is very low. This is

expected because the feed has been hydrotreated. Gas is the other ultimate by-product
aside fiom coke. Gas production increases continuously with residence time.

CHAPTER 5
Conclusions and Recommendation

5.1

Conclusions

A helical coil reaction system was capable of studying the reaction kinetics of the

hydrotreated light gas oil thermal cracking. The hydrotreated light gas oil thermal

cracking reaction was studied at three different temperature levels, 500C, 530C and

550C. A kinetic analysis has been performed by assuming a fust order overall
decomposition model for the light gas oil thermal cracking reaction. The activation

energy of this reaction was evaluated to be 153 idlmol. The result is very close to the 164 kJ/mol obtained by Fairbum et a[. (1 990) for n-hexadecane pyrolysis in a Curie

point pyrolyser. A four-lump kinetic model w s developed and evaluated to give a a

better understanding of the product distribution. According to the experimental data, the activation energy for the reaction of gas oil to naphtha was 173 kJ/mol, for the reaction of naphtha to gas was 215 kJ/mol, for the reaction of gas oil to gas was 83 kJ/mol and for the reaction of gas oil to coke was 262 kJ/mol. The activation energy for the reaction of naphtha to gas is close to 220.8 kl/mol obtained by Kumar and Kunrm (1985) and

within the range of 212 to 228 kl/mol found by van Damme and Froment (1981).
Product distributions for different reaction temperatures were predicted using the proposed kinetic model. It was found that at 500C, the optimal residence time was 1 0 seconds with a maximum naphtha production of 29 Wto?; at 530C, the optimal

residence time was 3.6 seconds with a maximum naphtha production of 28 wt% arid at
550C, the optimal residence time was 1.8 seconds with a maximum naphtha production

of 26 wt%.

5.2

Recommendation

Although boiling point range is of primary concern to industry, a model based on reactivity of different hydrocarbon structures may result in a more predictive model.

With a more detailed analysis of feed and products, it may be possible to understand the
roles that different types of carbon play in thermal cracking reactions. Among them,

PONA (paraffins, olefins, naphthenes, aromatics) analysis on a GC and NMR

spectroscopy may be the m s usell tools. ot

References

ASTM D975-98% "Standard Specification for Diesel Fuel Oils" (1998).

Cheng, Z., "Heavy Oil Processing Technology", China Petrochemical Press, Beijing,

360-372 (1994).

CBme, G., "Laboratory Reactors for Pyrolysis Reactions" in Pyrolysis: Theory and

Industrial Practice, L.F. Albright, B.L. Crynes and W.H.Corcoran (eds.), Academic
Press, Toronto (1983).

Coulson, J.M. and J.F. Richardson, "Chemical Engineering volume 1, Fluid Flow, Heat Transfer and M s Transfer", 4' ed., Pergamon Press, New Y r (1990). as ok

Coxson, P.G.and K.B. Bischoff, "Lumping Strategy. 1. Introductory Techniques and Applications of Cluster Analysis", Ind. Eng. Chem. Res., 26, 1239- 1248 (1987a).

Coxson, P.G.and K.B. Bischoff, "Lumping Strategy. 2. A System Theoretic Approach",

Ind. Eng. Chem. Res., 26,2151-21 57 (1 987b).

Depeyre, D., C. Flicoteaux, F. Arbabzadeh and A. Zabaniotou, "Modeling of thermal

steam cracking of an atmospheric gas oil", Ind. Eng. Chem. Res., 28,967-976 (1989).

Depeyre, D. and C. Fhcoteawc, "Modeling of Thermal Steam Cracking of n-

Hexadecane", Ind. Eng. Chem. Process Des. Dev., 30, 1116-1 130 (1991).

Fabuss, B.M., J.O. Smith and C.N. Satterfield, "Thermal Cracking of Pure Saturated
Hydrocarbons" in Advances in Petroleum Chemistry and Refining, J.J. McKetta (ed), New Y ork: Interscience (1 964).

Fairburn, J.A., L.A. Behie and W.Y. Svrcek, "Ultrapyrolysis of n-Hexadecane in a Novel Micro-Reactor", Fuel, 69, 1537- 1545 (1990).

Frornent, G.F., B.O. van de Steene, P.S. Van Damme, S. Narayanan and A.G. Goossens,

"Thermal Cracking of Ethane and Ethane-Propane Mixhues", Ind. Eng. Chem. Process

D s Dev., 15,495-504 (1976). e.

Froment, G.F., B.O. van de Steene, P.J. Vanden Berghe and A.G. Goossens, "Thermal

Cracking of Light Hydrocarbons and Their Mixtures" AIChE J., 23,930106(1977).

Gas Processors Suppliers Association, "Engineering D t Book", 10' ed. (1 994). aa

Gray, M.R., P. Jokuty, H Yeniova, L. Nazarewycz, S.E. Wanke, U. Achia, A. .

Knywicki, E.C. Saoford and O.K.Y. Sy, "The Relationship Between Chemical Structure

and Reactivity of Alberta Bitumens and Heavy Oils", Can. J. Chem. Eng., 69, 833-843
(199 1).

Gray, M.R., 'CThrough Glass, Darkly: Kinetics and Reactors for Complex Mixtures", a

Can. J. Chem. Eng., 75,481-493 (1997).

Hayashitani, M., "Therrnd Cracking of Athabasca Bitumen", Ph.D. Thesis, Department

of Chemical and Petroleum Engineering, University of Calgary, Calgary, Alberta,

Canada (1 978).

Hougen, O.A. and K.M.Waston, "Chemical Process Principles. Part Three - Kinetics

and Catalysis", John Wiley, New Y o 4 884-886 (1947).

Jacob, S.M., Gross, S.E. Voltz and V.W.Weekman, "A Lumping and Reaction B. Scheme for Catalytic Cracking", AIChE J., 22,70 1-703 (1976).

Kumar, P. and D. KUIIZN,'LM~deling Naphtha Pyrolysis", Ind. Eng. Chem. Process of

Des. Dev., 24,774-782 (1985).

Kays, W. M. and M. E. Crawford, "Convective Heat and Mass Transfer", 3rd ed.,

McGraw-Hill Book Co., New York, N.Y. (1993).

Khorasheh, F. and M.R Gray, "High-Pressure Thermal Cracking of n-Hexadecane", Ind. Eng. Chem. Res., 1853-1863(1993). 32,

Kossiakoff, A. and F.O.Rice, "Thermal Decomposition of Hydrocarbons, Resonance Stabilization and Isomerization of Free Radicals" J. Am. Chem. Soc., 65, 590-595

(1 943).

Kunii, D. and 0. Levenspiel, "Fluidization Engineering", 2nd ed., Butterworth-

Heinemann, Toronto, p.39 (199 1. )

Kuo, J.C.W.and J. Wei, "A Lumping Analysis in Monomolecular Reaction Systems",

Ind. Eng. Chem. Fundam., 8,124-1 33 (1 969).

Milne, B J , F. Bemti, L.A. Behie and T.J.W. De Bmijn, "Internally circulating ..

fluidized bed (ICFB):a novel solution to gas bypassing in spouted beds", Can. J. Chem.

Eng., 70,910-9 15 (1992).

Milne, B.J., F. Berruti and L.A. Behie, b'Ultrapyrolytic Heavy Oil Upgrading in an
Internally Circdating Fluidized Bed Reactor", U.S.Patent 5,370,789 (1 994).

Milne, B J ,L.A. Behie and F. Berruti, "Recycling of Waste Plastics by Ultrapyrolysis ..

Using an Internally Circulating Fluidized Bed Reactor", Joumal of Analytical and

Applied Pyrolysis 51, 157-166 (1 999).

Perry, R.H. and D.W. Green, "Perry's Chemical Engineer's Handbook", 7'h ed., McGmw-Hill, Toronto (1 997).

Petti, T.F., D.M. Trauth, S.M. Stark, M. Neurock, M. Yasar and M.T. Klein, "CPU Issues in the Representation of the Molecular Structure of Petroleum Resid through Characterization, Reaction, and Monte Carlo Modeling", Energy & Fuels, 8, 570-575
(1 9 4 . 9)

Simon, W. and H. Giacobbo, '4Themal Fragmentation and the Determination of the

Structure of Organic Compounds", Angewandie Chemie, International Edition, 4, 938943 (1965).

Stolfa, F., "New Roles for Thermal Cracking", Hydrocarbon Processing, May, 10 1 - 109
(1980).

Suncor Energy Inc., website:

htt~://www.suncor.codnewsrelease/Fin~

Financeandstock-Quarted8-317.html, last updated on April 19,2000 (2000).

Syncrude Canada Ltd., website: http://www.syncrude.com/, "Library-News

Releas-

January 26,2000" (2000a).

Syncrude Canada Ltd., website: htt~://www.syncrude.com/, "Who We Are-Securing Canada's Energy Future---A Canadian Success Story-Syncrude in the 21st Century", shown in May 2000 (2000b).

Tsai, C. H. and L.F. Albright, "Surface Reactions Occuning During Pyrolysis of Light Paraffh", in Industrial and Laboratory Pyrolysis ACS Symposium Series 32, L.F.

Albright and B.L. Crynes (eds.), Washington, D.C., American Chemical Society, 274295 (1976).

Teman, M., P.M. Rahimi,D.M. Clugston and H.D. Dettman, "The +525 Residue Before and After Hydrocracking with Bimodal Catalysts of Varying Macropore Volumes", Energy & Fuels, 8,518-530 (1994).

Trauth, D.M., S.M. Stark, T.F. Petti, M. Neurock and M.T.EUein, "Representation of the Molecular Structure of Petroieum Resid through Characterization and Monte Carlo
Modeling", Energy & Fuels, 8, 576-580 (1994).

van Damme, P S , S.Narayanan and G.F.Froment, "Thermal Cracking of Propane and ..

Propane-Propy lene Mixtures: Pilot Plant versus Industrial Data", AIChE J., 21, 10651073 (1975).

van Damme, P.S. and G F Froment, "Scaling Up of Naphtha Cracking Coilsy', Ind. Eng. ..

Chem. Process Des. Dev., 20,366-376 (1981).

Voge, H.H. and G.H. Good, "Thermal Cracking of Higher P a r e " , J. Am. Chem. SOC.,71,593-597 (1949).

Vogiatzis, A.L., S. Afara, C.L. Briens and M.A. Bergougnou, "Ultrapyrolysis of Cold Lake Bitumen", Circulating Fluidized Bed Technology II, P. Basu and I.F. Large (eds), Peragarnon Press, Toronto (1988).

Weekman, V. W., "Lumps,Models, and Kinetics in Practice", AIChE Monograph Series no 1l,75, AIChE,New York (1979).

Wei J. and J.C.W.Kuo, "A Lumping Analysis in Monomolecular reaction Systems", Ind. Eng. Chem.Fundam., 8,114-123 (1969).

Yui, S., "The Role of catalysts in Syncrude's Operation", Can. Chem. News, 51,
JulyIAugust, 25-27 (1999).

Zhou, P., O.L. Hollis and B.L. Crynes, "Thermolysis of Higher Molecular Weight
Straight-Chain Alkanes (C9-C22)", Ind. Eng. Chem. Res., 26, 846-852 (1987).

Appendix A

Mass Balance Calculation for Different Runs

Table A.1 is the SimDist analysis of the whole feed light gas oil. An estimation of its average MW is given based on the MW of normal paraffins. The weight percents of

<9') naphtha hction ( 1 6 C and gas oil hction (>196"C) are also shown. Table A.2 is
the average MW of LGO pseudosomponent (>196"C) in the feed and a comparison to n-hexadecane. Tables A.3 to A.10 are mass balance calculations for the Runs #1 to #8, including SimDist results of reaction liquid products and GC analysis results for the reaction gas bags. The sum of Yi for the gas analysis includes only the hydrocarbons; the balance is nitrogen. Compositions of pseudo-components are calculated using

Equations (4.5) to (4.8). The residence time was calculated based on Equations (4.1) and (4.2).

A Erlenmeyer flask containing 0.5 M Zn(Ach solution w s installed immediately before a

the WTM (see Figure 3.4) to absorb any H2Swhich could potentially be produced. A

KOH solution had been used for Runs #1 and #2. A switch was made to t!e Z ~ ( A C ) ~
solution, because C02 could be absorbed by KOH solution. Runs #1 showed similar
CO2 production to replicate runs. There was only 2 cm contacting length (the KOH

solution was shallow), the KOH concentration in the solution was very low (about 0.03

M) and the gas flow rate was high so that not much C02 was absorbed. Zn(Ac)l

solution absorbs H2S but not C02. For hydrocarbon feed such as the one used in this study, this absorption step was probably not necessary.

Table A. 1 Molecular Weight Calculation of Whole Feed

Feed C#

wt%

mol

mol%

ave MW

4
5 6 7
8

0 0 0 0 0.01 0.0001389 0.0002938 0 0 1 1567 .2 0 0 0.0003488 0.000738 0.06347 .3 0.0004 0.0008463 0.0846267 0 6 0.0057895 0 0122486 f ,3963407 .6 . 1 .% 0 01 3125 0.0323962 4.1467087 . 5 3.58 0.0252 1 0.0533387 7.5740904 1 3 5 7 0.036923I 0.0781 17 12,186246 .6 8.32 0.04894 0.1035433 17.602355 12 8.87 0.0482065 0.101989 18.765973 .91 1 10.35 0.0522727 01 105917 21 8 7 6 . 9.57 0.0451 15 0.0955044 20.24694 4 9 71 0.0429646 0.0908988 20.5431 3 . 3 8.81 0.0367083 0.0776626 18.639032 1 . 2 0.0512598 0.1084488 27.545993 30 8 6 0.0306383 0.0648205 18.279369 .4 4 8 0 0 55484 0.0328952 1 . . 2 .1 0 I9751 8 2 6 0.0078698 0 0 66499 5.6276761 .6 .1 1 6 0.0044536 0.0094222 3.4485384 .3 7 0.87 0.0022081 0.004671 1.8406309 0 4 0 0 1 1 137 0.0023563 0.9943638 .7 .0 0 2 0.0005778 0.0012224 0.5500736 .6 0 1 0.0002929 0.0006197 0.2961935 .4 0 0 0.0001779 0.0003763 0.19041 1 .9 0 0 0 9,363E-05 0.0001981 0,1057834 .5 . 0.04 71 17E-05 0.0001506 0.0846267 0 0 0 0 1 0 3 0.472664 0.6 1 212.32841
0.04

M W
XlO, wt% X20,wt% X30, wt% X40,wt%
<196OC >196OC Gas Coke

9 1 0 11 1 2 1 3 1 4 1 5 1 6 1 7
18 3.0

SUM

2 2 24 26 28 30 32 34 36 38 40 42
SUM

Table A.2 Molecular Weight Calculation of Feed (>196"C)

Feed (>196OC)

C#
11 1 2

MW
13
1 4 156 170 184 198 212 226 240 254 282 310 338 366 394 422 450 478 506 534 562 590

wt %

15
1 6 17 1 8 20 22 24 26 28 30 32 34 36 38 40 42

SUM

normalized mol mol% 5.76 0.06122449 0.0003925 0.0867833 8.32 0.088435374 0,0005202 0 1150304 . 8.87 0.094281463 0.0005124 0 1133037 . 10.35 0.110012755 0.0005556 0.1228608 9.57 0.101721939 0.0004798 0.1060997 9 71 0.103210034 0.0004567 0.1009832 . 8.81 0,093643707 0.0003902 0.0862785 13.02 0.138392857 0.0005449 0.1204801 7 8 6 0.091836735 0,0003257 0.07201 1 .4 4.82 0.05 1232993 0.0001653 0.0365446 2.66 0.02827381 8.365E-05 0 0184971 . 1 6 0 0 732568 4.734E-05 0.0104675 . 3 .1 0.87 0.009247449 2.347E-05 0.0051899 0.47 0.004995748 1.184E-05 0.0026177 0.26 0.002763605 6.14 1 E-06 0.001358 0.14 0.001488095 3 1 13E-06 0.0006884 . 0.09 0,000956633 1.891 E-06 0.000418 1 0.05 0.00053 1463 9.952E-07 0.0002201 0.04 0.0004251 7.565E-07 0.0001673 7 0 0 0 0 94.08 1 0.0045224 1

ave M W 13.5381 1 9 19.555165 20.847874 24.326437 22.4931 4 22.8221 93 20.70685 1 30,601952 20.307286 1 1.328833 6.25201 1 7 3.831 1 2 2.0448309 1.1046788 0.61 10989 0.3290532 0 21 15342 . 01 1 7 1 . 5 9 0.0940152 0 221.12378

Table A.3 Mass Balance Calculation for Run #1

outlet Iiq C#

outlet gas

W V
4 5
6

wt%

mol

7 8 9 10 11 12 13 14 I5
16

17 18 20 22 24 26 28 30 32 34 36 38 40 42

58 72 86 100 114 128 142 156 170 184 198 212 226 240 254 282 3 10 338 366 394 422 450 478
506

0 0.02 0.69 3.45 5.03 5.46 7.02

9.68
11.84

10.94 10.51 8.78 7.5 5.68 7.02 3.61 1.71 0.87 0 0 0 0

0

0 0.0002778 0.0080233 0.0345 0.0441228 0.0426563 0.0494366 0.06205 13 0.06%471 0.0594565 0.0530808 0.0414151 0.033 1858 0.0236667 0.0276378 0.0 128014 0.0055 161 0.002574 0 0 0 0
0

534 562 590

SUM
T
550

0 0 0 0 99.81 0.5700493
0 0 0

mol YO 0 0.0004873 0.0140747 0.06052 1 1 0.0774017 0.0748291 0.0867234 0,1088525 0.1221773 0.1043007 0.0931 162 0.0726518 0.0582157 0.04 15169 0.0484832 0.0224567 0.0096766 0.00451 53 0 0 0 0 0 0 0 0 0

ave MW

MW
CI
C2=

Y i
16 28 30 42
56

0 0.0350847 1.2 104216 6.0521082 8.8237984 9.5781 191 12.314725

6.52 4.48
2.62

C2
C3=

C4= C4

C4H6
C5 Cyc-CS C6 SUM
Ave MW (Gas) =

16.980988
20.770134 19.191323 18.437002 15.402177 13.156757 9.9640507 12.314725 6.3327857 2,9997406 1,5261838 0 0
0

58 54 72 70 86

4.1 2.16 0.02 1.22 0.38

0.06

MiYi 104.32 125.44 78.6 172.2 120.96 1.16 65.88 27.36

4.2 4.3 704.42

0.05 21.61
32.5%946

outlet fluid

G% L% MW(G) MW(L) Ave M W 25.423729 74.57627 1 32.596946 175.09012 139

Material Balance fed, g 165 SimDist for Liquid product g 118.8 <196OC,% 0.22 0.78 G,g 40.5 >l%"C, % c,g 0.35 trapped by rayon, g recovery 0.968 >C4 in gas bag, g
I

0 0 0 0 0 0 1 175.09012

12.1 2

<196C
> 196OC

X I0 6

X20 94

G V(Reactor) v, mA31s MW(feed) Rest, s C 0.0003471 0.0002193 212 1.5829728 SUM

37.574 83.226 38.5 0.35 159.65

X1
X2 X3 X4

wt?! 0.2353523
0.5213028

0.2411525 0.0021923 I

0
4

Table A.4 M s Balance Calculation for Run #2 as

outlet Iiq
C#

oullet gas

MW
4 5 6 58 72 86 100 114 f 28 142 156 1 70 14 8 18 9 212 226 240 254 282 31 0 338 366 394 422 450 478

wt%

7
8 9 1 0

II
1 2 1 3 1 4 I5 1 6 1 7 1 8 20 22 24 26 28 30 32 34 38 40 42

mol mol% ave M W 0.03 0.0005172 0.0009387 0.0544427 0I .8 0.0025 0.0045369 0.3266559 1 0 0 0125581 0.0227899 1.9599355 .8 . 2.46 0.0246 0.044643 4.4642975 . 5 % 1 6.1520198 3.39 0.0297368 0 0 3 5 4.48 0.035 0.0635164 8.1301028 6.25 0.0440141 0.0798748 1 1.34221 9 8 7 0.0558333 0.1013238 15.80651 .1 7 1 . 2 0.0654118 0.1187063 20.180077 11 1 . 4 0.0578261 0.1049402 19.308994 06 1 . 6 0.0543434 0.09862 19.526765 07 9 0 0.0425943 0.0772983 16.387239 .3 81 1 0.035885 0,0651224 14.717664 . 6.55 0.0272917 0.0495277 1 1.886646 8.58 0.0337795 0.0613016 15.570599 4 7 0 0169858 0.0308251 8.6926769 .9 . 2.26 0.0072903 0.0132302 4.1013465 1 1 0.0033 3 0.0060134 2.0325257 .2 16 0.57 0.00 15574 0.0028263 1.03441 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 181.67513

SUM

Ave M W (Gas) =

1.3 26 29.33%67

370.56

outlet fluid

G% L% MW(G) MW(L) Ave MW 13.425926 86.574074 29.339667 181.6751 3 11 6

Material Balance feed, g 159.2 SimDist for Liquid product 0.18 1 0 9 <196OC, % 3. L, g G,g 2 . >196"C,% 03 0.82
rccovery
<196OC >196OC

534 562 590

SUM

0 0 0 0 0 0 1 0 11 0.5510385 0. X20 94

0.951 >C4 in gas bag, g

g
tWo

T
530

X10 6

V(Reactor) v, rnA3/s MW(feed) Rest, s C 0.0003471 0.0002118 212 1.6387516 SUM

34.812 97.088 1. 93 0.2 114 5.

X1
X2 X3 X4

0.2299339 0.6412682 0.1274769 0.00132 1

Table A.5 Mass Balmce Calculation for Run #3

outlet liq

C#

MW
4
5 6 7 8 9 10 11 12 13 14

wt%

mol

58 72 86
100 114 128 142 !56 170 184 198 212 226 240 254

IS
16 17 18

0.01 0.07 0.3 0.73 1.4 1 2.78 4.79 7.56 10.7 1 1,12 12.3 10.86 10.14 8.4 1 10.48

0.0001724 0.0009722 0,0034884 0.0073 0.0 123684 0.0217188 0.0337324 0.04846 15 0.0629412 0.0604348 0.0621212 0.0512264 0,0448673 0.03504 17 0.04 12598

mol % 0.0003346 0.0018868 0.00677 0.0141674 0.0240038 0.0421503 0.0654657 0.09405 1 0.1221522 0.1 172879 0.1205608 0.0994169 0.0870755 0.0680066 0.0800744

sve MW 0.0194074 0.1358515 0.5822206 1.4 167369 2.736437 5.3952446 9.2961229 14.67196 20.76587 2 1.580979 23.871046 21.076387 19.679058 16.321585 20.338908

c2 c3=
C4=
C4

WH6
C5 Cyc-C5

C6
SUM
Avc MW (Gas) =

outlet fluid

MW(G) MW(L) Ave MW G% L% 6.371329 93.628671 32.688616 194.71399 184

28 394

30 32 34 36 38 40 42

422

450 478 506 534 562

0 0 0 0 0

590

SUM

0 0 0 100.33 0.5152686
X20
94

0 0 0 0 0 0

0 0 0 0 0 0 0

0 0 0 0
0 0 0 0

Material Balance 209.71~ i r n ~ ifor Liquid product st fd, g 0.1 LI g 188.1 <196OC, % 0.9 G i I ! 12.8 >196"C, % CI 6 0.05 trapped by rayon, g recovery 0.958 >C4 in gas bag, g
g

12.4 0.5

W!
XI

1 194.71399
212

<1%OC >196OC

G
T
51WI

XI0
6

V(Reactor) v, mA3/s MW(fecd)

0.0003471 0.0002006

Res 4 s C 1.7307183 SUM

30.47 158.13 12.3 0.05 200.95

X2
X3 X4

0.1516298 0.7869122 0.0612093 0.0002488

1

Table A.6 M s Balance Calculation for Run #4 as

outlet liq
C#

outlet gas

M W
4
58

wt%

mol

0.03 0.0005172

mol% avc MW 0.000903 0.0523742

M W
Cl 16

Y i
6.61

MiYi
105.76

a=
C2

c3=

C4=
C4

C4H6 C5 Cyc-CS

C6
15 16 17 18 20 22 24 26 28 30 32 34 36 38 40 42 212 226 240 254 282 3 10 338 366 394 422 450 478 506 534 562 590 8.7 7.72 6.04 7.62 3.93 1.75 0.88 0.42 0.22 0.13 0.08 0.06

SUM
T 550

X 10
6

Material Balance feed, 13 208.2 SirnDist for Liquid product L, L: 147.4 C 196OC,% 0.22 >196"C, % 0.78 G,g 52 0.04 W F by rayon, g ~ (A 0.4 recovery 0.960 X 4 in gas bag g 0.03 0.02 W? & <196OC 49.624 XI 0,2483684 0.02 >196OC 100.776 100.38 X2 0.5043844 X3 0,2452452 G 49 X20 V(Reactor) v, mA3/s M W ( f d ) Res 4 s C 0.4 X4 0.002002 199.8 1 94 0.0003471 0.0002194 212 1.5821185 S U M

0.0410377 0.034 1593 0.0251667 0.03 0.0139362 0,0056452 0.0026036 0.0011475 0.0005584 0.0003081 0.0001778 0.0001255 7.905E-05 5.618E-05 3.559E-05 3.39B05 0.5728009

0.071644 0.0596355 0.0439362 0.0523742 0.0243299 0.0098554 0.0045453 0.0020034 0.0009748 0.0005378 0.0003 104 0.0002191 0.000138 9.808E-05 6.2 13MS 5.918E-05 1

15.188524 13.477633 10.544676 13.303052 6.8610229 3.0551629 1.5363105 0.7332391 0.3840776 0.226955 0.1396646 0.1047484 0.0698323 0.0523742 0.0349161 0.0349161 175.24414

SUM Ave MW (Gas) = . .

32.924579

outlet fluid (3% L% MW(G) MW(L) Ave MW 138 26.078235 73.92 1765 32.924579 175.24414

18.2

Table A.7 Mass Balance Calculation for Run #5

outlet lq i C# outlet gas

MW 58 72 86 100 14 1 128 12 4 156 170 1 84 18 9 212

226

wt%

mol

mol%

ave MW

MW

Yi

MiYi

4 5 6 7 8 9 1 0 11 1 2 1 3 1 4 1 5 1 6 1 7 1 8 20 22 24 26 28 30 32 34 36 38 40 42

240 254 282 310 338 366 394 422 450 478
506

534 562 590

SUM

0 0 0 0 0.04 0.0005556 0.0011 1 1 0.0792818 0 0.13 0.001511 0.0029%1 0.257666 6 0.29 0.0029 0.0057479 0.5747934 O % 0.008421 1 0.0166909 1.9027644 . 2.56 0.02 0.0396409 5.0740383 4.78 0.033662 0.06671% 9.4741808 7.65 0.0490385 0.0971965 15.162653 10.77 0.0633529 0.1255685 21.346638 1 1 0.0597826 0 1184919 21.802508 . 1 1 9 0.0605051 01 199238 23.744913 .8 . 1 . 9 0.0490094 0.097139 20.59346 03 9.67 0.0427876 0.084807 19.166387 81 0.03375 0.0668941 16.054574 . 7 1 . 1 0.041771 0.0827933 21.02951 06 5 9 0.0212411 0.0421009 1 1.872457 .9 2.86 0.0092258 0 0 8286 5.6686521 .1 1.37 0.0040533 0.0080337 2 7154033 . 0.74 0.0020219 0.0040074 1.4667142 0 3 0.0009391 0.0018613 0.7333571 .7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100.26 0.5045291 1 198.71995
,DO

C1

C2=
C2 C3=

0%
C4

C4H6
C5
C~C-C5

C6

1 6 28 30 42 56 58 5 4 72 70 86

SUM Ave MW (Gas) =

17 .1 10 .3 16 .1 10 .2 08 .3 0.19 0 0.095 0.012 0 6.497

32.710482

27.36 28.84 48.3 42.84 46.48 1 1.02 0 6.84 0.84 0 212.52

outlet fluid

G% L% MW(G) MW(L) Ave MW 5.8958214 94.1041 9 32.710482 198.71995 7 189

Material Balance

feed, t3
Sg mg e (2 g

recovery

182 SimDist for Liquid product 1 4 4 <I 96OC.% 6. 0.09 09 .1 1 . >196"C,% 03 0.05 trapped by rayon, g 0.960>C4 i gas b g g n a
I3

96 . 04 .

<1 %OC > 196C

T
500

X 10
6

94

G V(Reactor) v, mA3/s MW(feed) Res f s C 0.0003471 0.0002007 212 1.7292161 SUM

23.932 140.868 9.9 00 .5 174.75

X1
X2

X3 X4

wtO/o 0.1369499 0.8061116 0.0566524 0.0002861

1
C 1
C-r

Table A.8 Mass Balance Calculation for Run #6

outlet liq outlet gas
MW
wt%

C#

mol

mol?40

ave MW

MW

Yi

MiYi

C1

a=
Q

C3= C3 C4= C A C4H6 C5

C6

16 28 30 42 44 56 58 54 72 86

SUM
Ave MW (Gas) =

5.87 3.82 1.61 3.58 0.42 0 0.23 0 0.72 0 16.25

29.735385

93.92 106.96 48.3 150.36 18.48 0 13.34

0

51.84 0 483.2

17 18 20 22 24 26 28 30 32 34 36 38 40 42

240 254 282 3 10

338

5.2 1 6.27 3.1 1.37 0.75

0

366 394 422 450 478 506 534 562 590

0.02 17083 0.024685 0.0109929 0.0044194 0.0022189 0 0

0

0.0363345 0.0413167 0.0183995 0.0073969 0.003714

SUM
T 550
XI0 6

0 0 0 0 0 0 0 0 0 100.36 0.5974584 XO 2 94

0 0 0 0 0

8.7202717 10.494454 5.1886454 2.2930465 1.2553174 0 0 0 0 0 0

0 0

outlet fluid
G% L% MW(G) MW(L) Ave M W 27.777778 72.222222 29.735385 167.9782 1 130

Material Balance
feed, g

L&?
G, 6

0 0 0 0 0 0 0 0 0 1 167.9~~1

6s
recovery

109.9 SimDist for Liquid product 75.4 < 196C % 0.25 0.75 29 >196"C, % 0.3 trapped by rayon, g 0.953 >C4 in gas bag, g
g

7.3 3

<196C >196C

27.325 51.075
26 0.3

G
V(Reactor) v, mA3/s M W ( f d ) Rcs t, s C 0.0003471 0.0002219 212 1.5645645 SUM

XI X2 X3 X4

wt?? 0.2609838 0.4878223

0.2483286

0.0028653

104.7

c .
C-,

Table A.9 Mass Balance Calculation for Run #7

outlet liq
C #

outlet gas

M W

wt%

mol

mol YO

avc M W

M W
C1

Y i
16 28 30 42 44 56 58 54 72 86 7.29 4.78 2.1 4.3 0.44
0

MiY i
116.64 133.84

C2= C2 C3= C3 C4= C4

C4H6
C5

63
180.6 19.36 0 18.56 0 69.12 0 601.12

0.32
0 0.96 0 20.19
29.773 155

C6 SUM
Ave M W (Gas) =

outlet fluid
22
3 10

1.39 0.0044839 0,0075935 2.353979

(3%

L%

MW(G)

MW(L)

Ave MW

Material Balance

38
40 42

534 562 590

SUM
T
XI0

0 0 0 0 0 0 100.32 0.5904896
X20

0 0 0 0 0 0 1 169.89293

recovery
<196C > 196C G Res f s C

0.9661>~4 gas bag, g in

6
wt%

3-91

V(Reactor) v, mA3/s MW(feed)

37.3 70.2 30.5 0.3

XI

X2 X3 X4

0.2697035 0,5075922 0,2205351 0.0021692

Appendix B
Temperature Profde for Different Runs

Figures B l to 8.8 show the temperature profile along with time during the Runs #I to .
#8 respectively. The temperature data were obtained by the LabView on-line data

acquisition system once every two seconds from thermocouples #19 (inlet) and #20
(outlet). The average of the reactor inlet and outlet temperature (dashed lines) in the figures was used as reaction temperature.

Figure B.1 shows the different stages during an experiment and how the reaction
temperature was evaluated.

Figure B.l shows the temperature of the inlet and outlet fluid. Nitrogen had been

purging before feed oil started. After feed oil entered the reactor, it affected the reactor

fluid flow rate and caused the dynamic stage. The reaction temperahue was determined
using the steady state interval shown in the figure B. 1. Nitrogen continued to purge after

the feed oil was stopped.

Appendix C

Pump #2 and Pump #3 Calibration

Figure C.l shows the calibration results for the pumps #2 and #3. The pump had a knob
with scale ranging from 0 to 100. These two pumps were set i parallel and the n

calibration w s done with both pumps on and at the same setting on the scale. The a
pumps were stabilized with hydraulic oil circulating in a recycle loop for one half hour and then the flow w s opened to the transfer cylinder for 10 minutes before the a

calibration for each point.

Appendix D

Valve Switching Scheme for Gas Analysis on GC

Two valves (#I and #2) were installed in a heated valve box on top of the HP 6980 GC oven. Valve #I has ten ports and valve #2 has six ports. These pneumatic valves were automatically actuated.

Normally the valves were in the OFF position, with the shaft rotated counter-clockwise.

Figure D. 1 shows column connection sequence with both valves in OFF position. When
the valve is switched clockwise, it is the ON mode. For example, for valve #1 in Figure

D.1, when it is in OFF mode, port 2 and 3, 4 and 5, 6 and 7, 8 and 9, 10 and 1 are
connected. Once valve #1 is turned on, port 1 and 2,3 and 4,s and 6,7 and 8,9 and 10 are connected. Therefore, the columns' relative position in the sequence changes. This
strategy was used to realize an effective separation of CO and CO2. Table D. illustrates I

the operating valve switching sequence, corresponding hydrocarbon position and

emerging sequence in the columns at different times. Figure D 2 shows the valve timing .
logic. The MS column and its contents are shown on the left, while the PLOT Q column and its contents are shown on the right of Figure D 2 ..

Table D l Valve Status and FlowISmpie Status .

Time

Valve #I

Valve #2

Flow/Sample Status

Initid
Injection
0.0 min 1.35 rnin

OFF
OFF

OFF OFF
OFF

Helium f o through MS to Plot Q. lw

Sample trapped in 25 pl sample loop. Light gases to CO through Plot Q to MS. Plot Q isolated, containing C 0 2 and
heavier hydrocarbons. Light gases to CO

ON ON

ON

are analyzed by TCD.

4.0 min

OFF

ON

Flow reversed on MS, COz and heavier

hydrocarbons still trapped on Plot Q.

4.1 min

OFF

OFF

Flow through MS then Plot Q, C02 and

heavier hydrocarbons go to TCD.

Valve 2 -OFF, Valve 2-OFF

MS

PLOT Q

TCD

CO Cl N2O2
Adjustable Restrictor

TCD

1.35 min LON, 2-ON

TCD

--I

TCD

b TCD

Figure D.2 Valve Timing Logic

Appendix E Estimation of the Reynolds Number and Heat Transfer Coefficient

Equation E. 1 and E.2 were used to estimate the Reynolds number.

where:
Re d
u
P Ir

Reynolds number
reactor tube ID, m
fluid velocity, m/s fluid density, kg/rn3

fluid viscosity, Pas

fluid mass flw, kg/m2~s fluid m s flow rate, kg/s as

G
k

cross-section area of the reactor tube, m2

For Run #1, the LCO mass f o rate w s 0.13 14 g/s and the LGO/N2 mass ratio was 2.0 lw a
(see Appendix G for detailed calculations). To give an estimate of the Reynolds number

at the inlet end, the smaller reactor tube ID of 3.34 mm was used to obtain a reactor fluid

mass flux of 22.5 kg/m2-s. Nitrogen viscosity at 500C was found to be 3 . 6 ~

Pa*s

(Perry and Green, 1997, p.2-321). The viscosity of the hydrocarbon gases in the reactor
was estimated to be 1.6x10-~ Paas based on molecular weight (200g/mol) and

temperature (GPSA data book., 1994, p.23-39). An average reactor fluid viscosity of
2 . 6 lo-' Paas w s used to calculate the Reynolds number. It was found to be 2900. By ~ a

considering the enhancement of helical geometry, it can be assumed that the reactor fluid is in the turbulent regime.

The Colbum equation was used to calculate the heat transfer coefficient (Coulson and
Richardson, 1990, p. 343).

where:

Nu
Re

Nusselt number Reynolds number

Prandtl number heat transfer coefficient, w/rn2~ thermal conductivity, W/mK

Pr h

When gases are heated or cooled, the Prandtl group usually has a value of about 0.74, substitution of Pr = 0.74 in equation E.3 gives the equation (Coulson and Richardson,
1990, p. 344):

The thermal conductivities of nitrogen and hydrocarbon gases at 500C are 0.056 WlrnK

and 0.05 WImK, respectively (GPSA data book, 1994, p.23-47). An average value of
0.053 WImK was used for the reactor fluid thermal conductivity. From the above

equations, the heat transfer coefficient can be calculated as 187

w/rn2~ straight tube. for

With a helical coil, somewhat improved heat transfer is obtained for the same physical

conditions. The following equation w s used to account for this effect (Coulson and a

Richardson, 1990, p. 399) and the heat transfer coefficient for the coil geometry was
found to be 21 0 w/rn2~.

h(coi2) = h(straight pipe)

d(tube ID) d(he1ir)

From the energy balance equation:

AT -=

d h ( T d - TJua
mc,

where:
CP

heat capacity of reaction fluid, J/kgK

AT -

fluid temperature change along the reactor tube, Wm

The heat capacities of nitrogen and hydrocarbon gases at 500C are 1050 JIkgK and
3100 J/kgK (GPSA data book, 1994, p.23-3), respectively. An average value of 2075

JIkgK was used for the reactor fluid heat capacity. From Equation E.7, the fluid
temperature rose at a rate of 8 10 W m along the reactor tube. To heat the fluid entering
the reactor f o 350C to 500C would require 0.185 m or approximately 20 cm. rm

It should be noted that the proceeding calculations are based upon wall to gas heat transfer. Boiling heating transfer rates are generally higher or on the same order of magnitude as wall to gas heat transfer rates. Therefore, the lower bound of wall to gas
heat transfer has been used in the analysis.

Appendix F Estimation of the Pressure Drop along the Reactor Tube

The pressure drop along the reactor tube can be calculated using the following equation
(Coulson and Richardson, 1990, p. 54).

where:

reactor tube ID, rn fiction factor

fluid mass flux, kg/m2+s

f
G

L
u

reactor length, rn

fluid velocity, rn/s

For Run #I, the average fluid velocity can be calculated using the residence time obtained in Appendix A and the reactor length of 18 m. Here a residence time of 1.73 s

was used. G was calculated to be 22.5 kg/m2*s(Equation E.2). The Blasius equation
(Coulson and Richardson, 1990 p. 53) was used to calculate the friction factor.

Assuming a Reynolds number of 2900 (Appendix E , the pressure drop along the reactor )

tube was estimated as 27 kPa.

The actual pressure drop must be based upon the actual fluid velocity, which changes

along the length of the reactor. The velocity tends to increase due to cracking but decreases because of increased diameter in the last half of the reactor. These two effects counteract each other. 27 kPa is estimated to be an upper bound on the pressure drop over the reactor. This is low enough to be neglected in the current study.

Appendix G

Calculations of the LGO to Nitrogen Ratio

The ideal gas equation of state was used to calculate the nitrogen mass flow rate because
of the high temperature and the low pressure.

where:

mass flow rate of nitrogen, kg/s

volume flow rate of nitrogen at room temperature, m3/s

gas constant, J/mol K

room temperature, 296 K

For Run #1, the nitrogen volume flow rate is 0.058 Lls at 23OC which is equivalent to a mass flow rate of 0.067 gls can be obtained. If the m s flow rate of the feed oil is as
0.13 14 g/s, the LGO to nitrogen mass ratio will be 2.0.

Appendix H

List of Failed Experiments

Table H. 1 Summary of Failed Experiments

Run

Date

Reactor Feed Time, s Temperature, "C

Comments

1 2
3
4

99-05-26 99-06-05
-

500 500 600 620 580

550

40

Run failed due to high feed f o rate and very short feed time. Poor mass lw
balance, Forgot to refill the transfer cylinder; feed ran out during bleeding. Thermocouples were out of order after heating up to 600C.

99-06-17 99-06-23 99-06-25 99-07-16 99-07-28

62
57

Run failed due to high temperature and insufficient packing of rayon. Some liquid products went into piping after the demister and gas bag.
Oil went into the piping after the demister. Feed flow rate was too high.

5
6 7

60

Run failed because no feed entered the reactor.

Run failed due to no bleeding of feed prior to directing it to the reactor. Feed went through reactor very fast because of elevated preheater pressure. The SimDist results showed no thermal cracking occurred.

500

877

99-08-08 99-08-1 9 99-13-23

540

9 10

540

No products exited the reactor.

Pump problem, no oil enter the reactor.
LabView did not record temperatures properly (large fluctuations).