Fuel Processing Technology 94 (2012) 131–144

Contents lists available at SciVerse ScienceDirect

Fuel Processing Technology

journal homepage: www.elsevier.com/locate/fuproc

Review

Kinetic modelling of thermal cracking of petroleum residues: A critique

Jasvinder Singh a,⁎, Surendra Kumar b, Madhukar O. Garg a
a
Indian Institute of Petroleum, Dehradun 248005, Uttarkhand, India
b
Indian Institute of Technology Roorkee, Roorkee 247667, Uttarkhand, India

a r t i c l e i n f o a b s t r a c t

Article history: Several models have been reported in the literature for thermal cracking of petroleum residues. Most of these
Received 20 May 2011 models are either highly empirical or first principles based models requiring detailed analysis of the feed-
Received in revised form 18 October 2011 stock. The present paper is an attempt to put forward a critical appraisal of various published models. The au-
Accepted 28 October 2011
thors have also generated the experimental data on a batch reactor for different feedstocks, and multi-lump
Available online 1 December 2011
parameter models (reported elsewhere) have been developed with the generated data. Three different
Keywords:
models for prediction of experimental yields in terms of gas and distillate fractions (i.e. kerosene, LGO and
Visbreaking VGO) have been critically evaluated in the light of experimental data.
Pyrolysis © 2011 Elsevier B.V. All rights reserved.
Residues
Kinetics
Modelling

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
1.1. Processes for residue upgradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
1.1.1. Hydrogen based processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
1.1.2. Non-hydrogen based processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
2. Kinetic modeling of thermal cracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
2.1. Lumping schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
2.2. Models with two to five lumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
2.3. Complex models with more than five lumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3. Critical evaluation of three latest models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
3.1. Model evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
4. Estimation of yields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5. Four lump model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6. Further directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

1. Introduction distillates has resulted in the increased popularity of processes for

In the present day scenario the availability of more heavy crude
oils has resulted in the increased production of atmospheric and vac-
uum residues and simultaneous decrease in light and middle distillate
fractions. In recent years an increased demand of light and middle
⁎ Corresponding author. Tel.: + 91 135 2525784; fax: + 91 135 2660202.
E-mail address: jassa28@yahoo.co.in (J. Singh).
upgradation of these heavy fractions to more useful lighter products.
The presence of large amount heavy molecules as well as high metal
contents of residual feedstocks renders these unsuitable for proces-
sing via catalytic routes. The high molecular weight compounds e.g.
resin and asphaltenes contribute in coke formation and metals work
as catalyst poison. In view of these factors, thermal cracking is the
preferred route for upgradation of the crude oil residues (bottom of
the barrel) as well as for other heavy oils e.g. shale oils, tar sands
and heavy crude oils.

0378-3820/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.fuproc.2011.10.023
132 J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144
1.1. Processes for residue upgradation
Hus [1] mentioned in his review paper that the cracking of a hy-
drocarbon compound could be represented by monomolecular reac-
tion mechanism and observed that the velocity constant depends
upon the following:
• Type of hydrocarbon to be cracked — Cracking reactivity decreases in
the following order: Normal Hydrocarbons, Iso-paraffins, Cycloparaf-
fins, Aromatics, Aromatic naphthenes, and Poly-nuclear aromatics.
• Molecular weight or boiling points of the hydrocarbons — higher
the hydrocarbon, larger the molecule.
• Temperature at which the reaction occurs.
• Pressure at which cracking takes place. For the pressures commonly
used in visbreaking process, the influence of the pressure may be
neglected except for its effect on the residence time.
The essential requirement of the residual upgradation is to im-
prove its properties, especially H/C ratio. This can be achieved by add-
ing hydrogen or by rejecting carbon. Thermal upgradation processes
may be classified in following two categories namely hydrogen
based processes and non-hydrogen based processes.
1.1.1. Hydrogen based processes
These processes use hydrogen to upgrade hydrogen deficient
heavy molecules under suitable processing conditions. The exam-
ples of hydrogen based thermal cracking processes are Hydrovis-
breaking, Donor solvent visbreaking and DYNA cracking. The
essential details about these processes are described in the follow-
ing subsections.
1.1.1.1. Hydrovisbreaking. This is a thermal conversion process [2,3],
operated in the presence of hydrogen. It is based on the concept
that the solubility of hydrogen in oil increases with the increase in
pressure and temperature and part of it takes part in reaction to
yield products of higher stability.
1.1.1.2. Donor solvent visbreaking. This process is the outcome of fur-
ther improvement of hydrovisbreaking process. In this process the
addition of a hydrogen donor solvent ensures comparatively more
conversion than the one obtained in the presence of hydrogen
alone. This process seems to be ideal for conversion of heavy oils
from conventional crude oils [4,5].
1.1.1.3. DYNA cracking. It is a thermal hydrocracking process that con-
verts a wide variety of heavy oil feedstocks into distillate products
and high heating value fuel gas in the presence of internally generat-
ed Hydrogen [6]. The reactor in the process consists of three different
sections, i.e. thermal hydrocracking, stripping, and coke gasification.
Operating conditions can be varied to provide product slate flexibility.
1.1.2. Non-hydrogen based processes
As the name implies, these processes facilitate hydrocarbon mo-
lecular transformation by heat and catalyst or heat transfer alone, in
the absence of hydrogen or hydrogen donor solvents.
The thermal conversion processes in this category are carbon re-
jection processes, which have the flexibility to operate with all type
of heavy feedstocks for their upgradation into light distillates/
liquid products. In fact these processes have staged a comeback in
the refinery industry to process heavy residues for the production
of feedstocks for secondary conversion processes. Brief details of
the important processes in this category are given below, while
the kinetic modeling of visbreaking has been reviewed separately
thereafter.
Delayed coking and flexicoking [7] are the two commercially
applied residue upgradation processes. In delayed coking, the coke
formed remains in the coke drum and is removed by hydraulic
water jets, while vapour leaves the top of the coke drum. Depending
upon the feedstock quality (mainly metals and sulfur contents) the
green coke produced may either be used as fuel or after calcination
as metallurgical grade coke in aluminum industry. Flexicoking is gen-
erally based on high sulfur feedstock, and is currently practiced with
gasification process to obtain better yields than delayed coking; it
drastically reduces the coke yield. Principal process disadvantages in-
clude the disposal of the drag coke to specialty markets.
A modification of fluid cracking reactor technology is Asphalt
Resid Treating (ART) [8,9]. It is used to demetallize and decarbonize
heavy oils such as heavy crudes, atmospheric residues, tar sands
and bitumens and like by contacting the feedstocks with ARTCAT
(trade-mark) fluidizable contact material in a dilute phase riser envi-
ronment. During the process, over 95% of the metals are removed, and
products are essentially free of asphaltenes. Here an inert solid is cir-
culated instead of FCC catalyst. Oil, metal and coke precursors deposit
on the solid surface while a substantial proportion of the feed is va-
porized and cracked.
Visbreaking is another popular thermal conversion process widely
used for residue upgradation [10–12]. It is a comparatively mild ther-
mal cracking process mainly used to lower the viscosity and pour
point of the residual oil, and has shown a revival worldwide[1] as a
convenient and cheap tool for residue upgradation. This process
makes an important contribution in reducing the amount of cutter
stocks (i.e. straight run gas oil) used in adjusting the viscosity of res-
idues for their conversion to fuel oils in order to meet standard spec-
ifications. The most recent problem for refiners is the development of
process(es), which may reduce the production of heavier fuel oil and
increase the production of lighter distillates [13].
In spite of the thermal cracking being quite an old process, the de-
tailed kinetic modelling of this process initially did not get due atten-
tion in literature. The models reported in literature were either
developed with pilot scale or industrial data [14,15] or based on de-
tailed structural analysis [28] of the feedstock, normally not carried
out in a refinery in routine. Most of these models did not present de-
tailed analysis of in terms of distillate fraction. Some earlier attempts
were made to develop the kinetic models based on a very simple de-
scription with a single first order reaction [1,12,17]. Later attempts
were made using more comprehensive reaction schemes using up
to sixteen pseudo-components [15]. Dente et al. [18] have critically
examined the pyrolysis models for various hydrocarbon fractions in
gas as well as liquid phase. They feel that further theoretical as well
as experimental analysis will lead to still better understanding of de-
tailed chemistry of pyrolysis and fouling processes. Recently some
studies have been reported on non isothermal kinetics of heavy hy-
drocarbon pyrolysis [19–21].
This paper presents a critical evaluation of the available models for
kinetic modeling of thermal cracking process. The models reported in
literature have been reviewed in the light of various parameters e.g.
no. of lumps, coke formation, nature of feedstock, and data generation
method. Finally three most recent models have been evaluated with
the lab data generated by authors and reported earlier [22].
2. Kinetic modeling of thermal cracking
The available kinetic models can be classified into two major parts
on the basis of their complexity. In the present analysis we have taken
the models with one to five lumps as simple models, and the models
with more than five lumps or models requiring more computational
effort have been classified as complex models. A number of models
available in literature have been reviewed and critically examined.
Table 1 list some significant kinetics and modelling studies reported
in the literature. A brief description of the reviewed models has
been presented in following sub sections.
 J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144
2.1. Lumping schemes
A number of lumped parameter models have been proposed.
Some studies are reported with one feed and one product lump
[14,23,24] whereas number of lumps has been reported up to sixteen
[15]. The criteria for lumping have been either based upon statistical
information [16,25], characteristic information [26], or pseudo-
components [27–30]. The pseudo-component lumping schemes sug-
gested by Del Bianco et al. [27] do not include distillate as separated
lumps of gasoline and gas oil fractions, and therefore, it is not suitable
for the design and optimization of a mild thermal cracking process to
maximize distillates. The coke lump has been considered, which
makes it suitable for describing more severe thermal cracking only.
Molecular modeling of heavy oil pyrolysis has been reported by Yan
et al. [31]. They feel that molecular modeling of petroleum processes
may be more useful for producing refinery products requiring tighter
specifications. Recently some authors have reported models with two
gas lumps based on flue gas and LPG [32,33].
2.2. Models with two to five lumps
The simplest form of reported kinetic model is one reactant and
one product lump [14,23,24]. Although all of these models describe
cracking of residue as a first order reaction, the studies are differ-
ent in terms of feedstocks, experimental setup, and experimental
conditions.
Di Carlo and Janice [14] have reported two lumps model based on
data obtained on a continuous pilot plant using three different Atmo-
spheric residue feedstocks. In addition to the kinetic parameters, the
authors also discussed the stability of the cracked residue. Influence
of the operating parameters has been discussed on the cracking be-
havior of three feedstocks. The reactivity of the studied feedstock
has been discussed at low and high temperatures. The observed be-
havior has been analyzed on the basis of characteristics of the feed-
stocks as well as reaction mechanism of the cracking reactions. The
paper provides a reasonably good estimate of the kinetic parameters.
Krishna et al. [24] studied Aghajhari long residue on a continuous
bench scale coil type reactor unit. They have also analyzed the data
assuming similar two lump model. The temperatures and flow rates
studied were in the range of 427–500 °C and 2.04–2.91 lit/h respec-
tively at a pressure of 17 kg/cm 2(g). The reaction was found to ob-
serve first order kinetics in terms of the conversion of IBP-150 °C
cut, at low conversion range (b 7%). Beyond 7% conversion the yield
pattern showed marked deviation, indicating higher order kinetics.
Such deviation on the order of reaction at severe conditions has also
been later reported by Martinez et al. [34] in their study on decompo-
sition of an asphaltenic residue. The additional study reported in
Krishna et al. [24] model is effect of feedstock conversion on viscosity
reduction, which is not studied by Di Carlo and Janice [14] and Al
Soufi et al. [23].
Further to two-lump model are three lump models. Del Bianco et
al. [27] presented the kinetic analysis of thermal cracking reaction
taking three lumps, with one Intermediate I. They conducted lab
scale studies on Belayam crude in a 30 ml capacity stainless steel au-
toclave at temperatures 410, 430, 450 and 470 °C for reaction times
up to 120 min. The concentration of three pseudo components name-
ly vacuum residue, distillate and coke were observed and data from
the experiments have been used for defining a three lumps, three pa-
rameter model. Distillate conversion has been assumed to be a first
order reaction whereas coke formation seems to be the consequence
of secondary reactions, in particular asphaltenes. A study of structural
changes of asphaltenes as a function of severity was also performed
and the variation in average molecular parameters with conversion
were correlated. The advantage of their study is inclusion of coke
lump. The estimation of liquid product shall be more precise as com-
pared to two lump models with no coke was accounted for separately.
133
Martinez et al. [34] studied the kinetics of thermal decomposition
of an asphaltenic residue, obtained from the synthetic crude by coal
liquefaction. The kinetics of formation of oil + gas and coke (toluene
insoluble) from the conversion of the asphaltenic fraction were deter-
mined. A three lump model is proposed, which considers parallel re-
action for oil + gas and coke formation. Conversion data has been
reported fitting second order kinetics throughout for asphaltene con-
version and oil + gas and coke formation. The activation energies ran-
ged from 63.14 to 86.13 kJ/mol, the highest being for asphaltene
conversion. These values are comparatively less as compared to
those reported by Del Bianco et al. [27]. They have also compared
and found the obtained values are less than those reported in some
earlier studies [35]. The highest asphaltene conversion (55.9%) was
obtained at 475 °C. At short reaction times, the asphaltenic fraction
converted to coke (~30%) was much less than that converted to oil
(~70%) at the four temperatures used, whereas at long reaction
times this trend was reversed, especially at 475 °C. Structural analysis
shows that lower aromaticity and higher H/C and N/C ratios in the oil
from products than from feed. The difference in activation energies
obtained above, may be attributed to these structural characteristics.
Yasar et al. [29] studied the effect of reaction environment on
pathways and selectivity by experiments on resids and isolated
asphaltenes from Arabian light and Arabian heavy feedstocks. The re-
sults were compared with previous experiments with Hondo and
Maya feeds to determine the effect of these parameters. The experi-
ments were conducted at temperatures of 400, 425 and 450 °C for
holding times ranging from 20 to 180 min in a micro-batch reactor.
Reaction products were recovered as gas, maltene, asphaltene and
coke lumps. The goal of these experiments was to determine the ef-
fect of the variation among the four feeds and also to provide a com-
parison of asphaltene behavior when pyrolysed neat and in a resid
environment. The comparison of relative kinetics and apparent acti-
vation energies yielded insight into the thermal reaction pathways,
feedstock effects and asphaltene environment effect. At 400 and
425 °C, asphaltenes reacted predominately to coke. Isolated maltene
pyrolysis indicated that the asphaltene and coke formed in series, i.e.
Maltene → Asphaltene → Coke
It was also concluded that resids with higher asphaltene content
are more reactive than resids with low asphaltene content. The
range of obtained activation energies are slightly higher than those
reported by Martinez et al.[34], but in good agreement with those
reported by Al Soufi et al. [23].
Recently, the authors have generated and published experimental
data on a batch reactor (Singh et al., [36]) and also suggested a five
lump model (Singh et al., [37]) and a four lump reduced model
(Singh et al., [38]) for the prediction of the product yields. An impor-
tant consideration in development of these models was to include the
industrially important lumps, fixed on the basis of boiling ranges of
the distillate fractions. It is believed that the inclusion of LGO, VGO
and Kerosene fractions will improve the applicability of this model
for estimation of operating conditions to maximize the yield of
lower boiling distillate fractions. A five lump model has also been
reported by Kataria et al., [39], taking similar lumps.
A critical review of the reported models of less than five lumps re-
veals that these models have been developed around the experimen-
tal data obtained on feedstocks with wide range of asphaltene
contents and isolated asphaltenes. The main purpose of most of
these studies was to study the effect of nature of feed stock on the re-
action kinetics. These models are simple to use and provide a reason-
able estimate for preliminary evaluation of a feedstock in terms of
obtained product. First order reaction kinetics has been assumed in
most of these studies except few researchers (Krishna et al. [24] and
Martinez et al. [34]) who have explained the kinetics of coke forma-
tion from maltene / oil fraction as more than first order. These
 134
Table 1
Kinetic Models.

S.N. Authors Feed Characteristics Process / Experimental Conditions Proposed Model Conclusions / Remarks

1. Al-Soufi et al. Atmospheric Residue from Iraqi Continuous lab scale unit with 2 l /h Atmospheric Residue → Visb. Product
k
E = 99.78 kJ/mol (Visbreaking)
(1988) crude Composition, wt % CCR feed capacity E = 252.6 – 378.9 kJ/mol (condensation
14.76 Asphaltene 9.50 T = 435–480 °C Res. Time: Coil 67 s and coking)
Soaker 297 s P = 7 bar
2. Castellanos Atmospheric Residue / Vacuum Azcapotzalco Refinery (Mexico) coil • May be used for analysis of the effect
et al. (1991) Residue (TBP available) visbreaker data taken. (Operating of temperatue and residence time.
conditions not mentioned) • Successfully used for the design of
new units and improvement of
existing units.

J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144

All reactions are first order.
3. Chakma and Bitumen feed stock. Composition, wt% Bitumen pre-heated at 400 °C is atomized • Continuity equation, Momentum equations (three dimensional) in • Validated the predicted and
Islam (1990) Gas : 0.0 as the droplets of the size 20–30 micron cylindrical co-ordinate system, species mass balance, and k-ε experimental viscosities.
Coke: 0.0 diameter in the form of a jet. turbulence model constitutes model of a visbreaking vertical jet reactor. • May be modified to include thermal
Maltenes: 80.0 • solved by using SIMPLE algorithm of Patankar (1981). effects and chemical reaction.
Asphaltenes: 15.3
4. Del Bianco et. Vacuum Residue (Belayam crude), Thermal cracking in a micro reactor, k1 D E1 = 207.97 kJ/mol • Good agreements between calculated
al (1993a) Composition, wt% Temp Range: 410 – 470 °C; and experimental results.
Asphaltene 18.6 Residence time: up to 120 min. E2 = 174.29 kJ/mol • Same model can be used to correlate
Saturates 9.9 P = 10 kg/cm2 VR k2 E3 = 269.02 kJ/mol variation of available model
Aromatics 35.6 k3 parameters of asphaltene conversion.
Resins 54,5 I C

Temp., oC Rate constants (10 -2 min-1 )

k1 k2 k3
410 1.19 0.91 0.21
430 3.36 2.17 0.80
450 8.95 4.93 2.84
470 22.58 10.71 9.40
VR : Vacuum residue; D : Distillates; I : Intermediate -
Product for Coke formation; C : Coke,
5(a). Dente et al. Used experimental data from • Simplified property predictions equations formed by grouping similar • Good agreements between calculated
(1993) Di Carlos et al. (1992) ; class of compounds. A statistical function defined such as the fraction of and experimental results.
Al-Soufi et al. (1988) ; and pseudo component formed is given by fi = Aclass exp (− kclass ni), where
Krishna et al. (1988) the index kclass depends upon the class of component (paraffin, aromatic
etc.) ; ni is number of C atoms, and Aclass is determined by re-normalisation
on the basis of pertaining class.
5(b). Dente et al. Composition, wt% Coil Inlet Temp. 326 °C Same as above. • Compared predicted properties with
(1995 and Asphaltenes 5–17 Coil Outlet Temp. 450 – 468 °C their own results.
1997) CCR 17–23 Res. Time: 24–26 min. (with soaker) • Good agreement.
Sulphur 3.0–3.8 9 min (without soaker)
6. Di Carlo and Atmospheric Residue: Experimental Data on a Atm. Residue →
k
Visb. product • The feedstocks rich in asphaltenes, resins
Janis (1992) ROSPO DI MARE Continuous Pilot Plant and polars show higher selectivity towards
nC5 insol. 40.76 T = 460–470 °C; gas and lower selectivity towards gasoil as
Saturates 8.67 compared to light and paraffinic crudes.
Aromatics (N + P) 50.57 • Qualitatively correlated the difference
BELAYAM between the aromatic carbon of asphaltes
nC5 insol. 13.67 and aromatic carbon of maltenes in the
Saturates 25.75 feedstock with the conversion of residue.
Aromatics (N + P) 60.58
ES-SIDER
nC5 insol. 4.15
Saturates 48.03
Aromatics (N + P) 47.82
(All compositions in wt. %)
7. Filho and Vacuum residue feedstock, Pilot Plant Experiments A sixteen lump model proposed for predicting yields. • Global kinetic rate constants and their
Sugaya (2001) characterised by correlations given T = 480–530 °C dependence on temperature obtained.
by other authors. P = 1 atm.
Atmospheric Residue → Visb. Product
k
8. Krishna et al. Aghazari long residue Continuous flow bench scale reactor, E = 224.8 kJ/mol
(1988) CCR 7.9 flow rates 2.04 – 2.90 l /h
n-C5 insol. 5.66 T = 427–500 °C
(All compositions in wt. %) P = 17 bar

J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144

9. Martinez et al. Asphaltenic Residue from Synthetic Tubular SS Reactor Lumped parameter model proposed • Both the parallel reactions from
(1997) Crude. Length = 40 cm Asphaltene to oil + Gas and also to
Composition, wt% ID = 1.3 cm coke follow second order kinetics.
Maltenes 19.6 T = 425–475 °C • Estimated kinetic constants and
Asphaltenes 65.5 P = 1 atm. activation energies.
Coke 15.0

10. Xiao et al. Heavy Oil Lab Scale Microreactor A twelve lump model is proposed using temperature cuts as a lump with first order • Good agreement with experimental
(2002) kinetics. (conversion) data at low temperatures.
T = 400–500 °C • E = 268 kJ/mol (cracking reaction) =
531 kJ/mol (condensation reaction)

11 Yasar et al. Resids and asphaltenes from: Micro batch reactor A four lump model • Values of kinetic constants and activation
(2001) Hondo T = 400 – 450 °C energies are estimated.
Asphaltenes 23% P = 5 atm.
Saturates 13%
Aromatics 43%
Resins 21%
Maya
Asphaltenes 12% Saturates 26%
Aromatics 51% Resins 11% (All
compositions in wt. %)
Arabian Light
Asphaltenes 6% Saturates 30%
Aromatics 45% Resins 19%
Arabian Heavy
Asphaltenes 15% Saturates 30%
Aromatics 44% Resins 16%
(All compositions in wt. %) Reaction from Maltene to asphaltene is condensation reaction.

135
(continued on next page)
 136
Table 1 (continued)
S.N. Authors Feed Characteristics Process / Experimental Conditions Proposed Model Conclusions / Remarks

12. Zhou et al. Daqing Vacuum residue Lab scale reactor. An eleven lumped model has been proposed with following lumps: Values of kinetic constants and
(1999) Saturates 41.47 T = 400–460 °C, S : Saturates; Al: Light aromatics; Ah: Heavy aromatics; Rl : Light resins; activation energies are estimated.
Aromatics 23.30 P = 5 kg/cm2. Rh: Heavy resins, B: Asphaltenes; G: Gas; L: Naphtha, V1 : Middle fraction1;
Resins 35.23 V2: Middle fraction 2; C: Coke.
Asphaltenes 0.01
Guanshu Vacuum residue
Saturates 24.14

J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144

Aromatics 20.63
Resins 52.13
Asphaltenes 2.67
Liaohe Vacuum Residue
Saturates 26.84
Aromatics 20.84
Resins 47.52
Asphaltenes 4.80
(All compositions in wt. %)

13. Singh et al. (All compositions in wt. %) Lab scale batch reactor of A five lump seven kinetic parameter model with following lumps Values of kinetic constants are
(2005a) 400 ml capacity, listed in Table 2(a)
North Gujarat Short Residue T = 400 – 430 °C, F: Feedstock (500 °C +)
Saturates 10.30 P = 12 kg/cm2(g) G: Gas (C5-),
Aromatics GLN: Gasoline (IBP-150 °C),
Naphthenic 58.71 LGO: Light Gas Oil (150–350 °C),
Polar 29.20 VGO: Vacuum Gas Oil (350–500 °C)
Asphaltenes 1.85
Bombay High Short
Residue
Saturates 32.89 Aromatics -
Naphthenic 59.82
Polar 4.23 Asphaltenes 3.03
Mathura Refinery Visbreaker Feed
Saturates 14.76
Aromatics -
Naphthenic 68.07
Polar 11.86 Asphaltenes 7.72
Haldia Refinery Asphalt
Saturates 5.63 Aromatics
Naphthenic 68.07
Polar 11.86
Asphaltenes10.15
(All compositions in wt. %)
14 Singh et al. Same as above Same as above A four lump four kinetic parameter mode l(VGO lump dropped from above) Values of kinetic constants are listed
(2005b) in Table 2(d)

15 Kataria et al. North Gujarat Short Residue Lab scale batch reactor of 400 ml A five lump seven kinetic parameter model with following lumps Values of kinetic constants for four feedstocks
(2004) capacity, are listed in Table 2(c). Values for other
Saturates 10.30 T = 400 – 430 °C, VR: Feedstock (500 °C +); Gas (C5-); feedstocks have also been estimated by the
Aromatics Gasoline (IBP-150 °C); authors.
Naphthenic 58.71 LGO: Light Gas Oil (150–350 °C);
Polar 29.20 P = 12 kg/cm2(g) VGO: Vacuum Gas Oil (350–500 °C)

J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144

Asphaltenes 1.85
Bombay High Short Residue
Saturates 32.89
Aromatics -
Naphthenic 59.82
Polar 4.23 Asphaltenes 3.03
Arabian Mix Short
Residue
Saturates 15.00 Aromatics -
Naphthenic 66.60
Polar 6.74
Asphaltenes 6.50
Mathura Refinery
Visbreaker Feed
Saturates 14.76
Aromatics -
Naphthenic 68.07
Polar 11.86
Asphaltenes 7.72
Haldia Refinery Asphalt
Saturates 5.63 Aromatics
Naphthenic 68.07
Polar 11.86
Asphaltenes10.15
Arabian Mix Asphalt
Saturates 5.12 Aromatics -
Naphthenic 66.05
Polar 17.81 Asphaltenes 9.61
(All compositions in wt. %)

(continued on next page)

137
 138
Table 1 (continued)

J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144

S.N. Authors Feed Characteristics Process / Experimental Conditions Proposed Model Conclusions / Remarks

16. Mohaddecy Mixture of VR and slop oil Commercial soaker A six lump, fourteen kinetic parameter model with following lumps • Values of frequency factors and
and Sadighi from vacuum tower, visbreaker. Activation energies have been
(2011) determined for fifteen reactions
Sp. Gr. 1.006 Temperature: • Reduced model has been
S, wt% 3.19 Inlet 325 °C presented with fourteen
Ni + Va 188 (ppm) Outlet 440 °C parameters
 J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144
authors also have observed first order kinetics for conversion to oil /
maltene fraction.
2.3. Complex models with more than five lumps
The models based on less than five pseudo lumps are quite useful
for understanding the cracking behavior of the heavy feedstocks, but
do not quantify effect of structural parameter or hydrocarbon type on
the kinetic parameters. In an effort to overcome this shortfall, some
authors have developed the kinetic models using more detailed struc-
tural analysis and lumping based on class of compound and boiling
range of the constituent lumps. This sub section discusses few com-
plex models developed using more detailed characteristics and hav-
ing more than five lumps.
A kinetic model has been proposed by Castellanos et al. [26] for
the prediction of visbreaker yields. Their proposed model is based
on the sets of feed and product hydrocarbon components character-
ized by their true boiling points, API gravity and molecular weight.
According to this scheme, if n is the number of discrete pseudo-
components, and i equals the number of carbon atoms in the com-
pound of given pseudo component; it follows that i can be any num-
ber between 1 and n, both inclusive. Any i-carbon atom component
can have two degrees of saturation – Si and Oi. An S-component has
a C/H ratio corresponding to the average degree of saturation of the
feed components, and an O component has an equivalent structure,
but with the unsaturation incorporated into the molecule in shape
of two fewer hydrogen atoms than corresponding S component.
Each Si component undergoes a first order reaction producing,
through i-2 parallel reactions, the set of lighter S components from
Si-2 to S1, with simultaneous production of the corresponding O com-
ponents. For a system represented by n pseudo-components, there
would be (n-2)(n-1)/2 such reactions. The main limitation of the pro-
posed model is that it does not consider reactions of cracking, poly-
merization or condensation of O components to yield lighter di-
olefins or heavier hydrocarbons and coking reactions of certain
asphaltenic fractions.
Dente et al. [16,25] conducted a comprehensive study on the
modelling of kinetics and reactor, following lumping methodology de-
veloped by them (Dente et al. [16]). They followed the approach of
lumping real components into the pseudo-components. Initially they
suggested a model containing 150 equivalent pseudo-components
and 100 reactions. For classification of compounds of same class, a sta-
tistical distribution function was used, which is given below:
f i ¼ Aclass expð−kclass :ni Þ
where f is the fraction of a class of component, kclass depends on the
class of component, viz. paraffins, olefins etc. Aclass is determined by
re-normalization on the basis of the total amount of the pertaining
class. Typical considerations adopted for kinetics into the model are
as follows:
• The chain initiation reactions and the β-Scission, i.e. radical decom-
position of olefins and smaller radicals, is more or less strongly en-
dothermic reaction and because of formation of two molecular
bodies, have the following expression for kinetic constant.
# ! # !
ΔSg;liq ΔHg;liq
kliq ¼ kgas exp − exp
R RT
# #
Where ΔSg, liq and ΔHg, liq are respectively the difference of the var-
iation of entropy and enthalpy in forming the transition state from
the reactants to be decomposed into the products.
139
• The Hydrogen abstraction and the substitution reactions in the liq-
uid phase have practically the same constants of their equivalent
into gas phase.
• The “radicals’ recombination reactions” (chain terminations) are so
exothermic to be limited by an oriented diffusion of the colliding
radicals. The sterical factors for the re-combinations were consid-
ered practically to be the same as for the equivalent gas phase,
and the liquid phase collision frequency is well established, so
that the kinetic constants for the radicals’ recombinations rate
could be deduced.
The model was evaluated using data available in the literature for
lab scale, pilot scale and commercial units [14,23,24,27] and the pre-
dicted results have been found in satisfactory agreement with the ex-
perimental data.
Dente et al. [16] model therefore have two advantages over the
one reported by Castellano et al. [26] First is the model overcomes
the one major limitation of the latter [26] in respect of cracking, poly-
merization or condensation reactions to yield lighter di-olefins or
heavier hydrocarbons and coking reactions. Secondly, its good agree-
ment with the earlier data firms up the validity of its application and
gives further strength to the model. The authors further extended
their work to give a comprehensive model (Dente et al. [25]) to pre-
dict product amounts and their property predictions. The additive
corrective contributions of the activation entropies and energies
were applied to the previous ones.
Chakma and Islam [40] reported modelling of visbreaking of bitu-
men in a jet reactor. In their experiment, bitumen preheated to about
400 °C was atomized into very fine droplets of size of 20–30 micron
diameter through a nozzle at very high velocities. The heat input
was through heating elements placed around the reactor wall. The ef-
fluents at the reactor exit are cooled down and separated. The flow
and mass balance equations were formulated and solved by using
SIMPLE algorithm given by Patankar [41]. Dawson et al. [42] has
also obtained a patent on a viscosity reduction process, using this
type of jet reactor. Being a theoretical model the accuracy of estima-
tions would be solely dependent on the accuracy of numerical solu-
tion of the governing equations.
Zhou et al. [28] presented a kinetic model for thermal conver-
sion of vacuum residues. An eleven lump kinetic model was
presented taking six feedstocks and five product lumps. The feed-
stock lumps were defined as vacuum residue (VR), Saturates(S),
Light Aromatics (Al), Heavy Aromatics (Ah), Light Resins (Rl),
Heavy resins (Rh) and Asphaltenes. The product Lumps were
taken as Gas (G), Naphtha (L), Middle fraction-1 (V1), Middle
ð1Þ
fraction-2 (V2) and Coke (C). Parameters for the kinetic model
were determined from the experimental data on thermal cracking
of three different vacuum residues, namely Daqing, Guanshu and
Liaohe vacuum residues.
Filho and Sugaya [15] presented a model for the development of a
computer-aided tool for the simulation of heavy oil thermal cracking
process. A dual plug flow reactor (DPFR) representation was pro-
posed for the light pyrolysis of petroleum distillate residue in a coil
type reactor. The proposed reactor model (DPFR) consists of two par-
allel plug flows each for vapour and liquid, traveling at different
speeds in a coil. Reaction is assumed to be rate-controlling process
with equilibrium between the phases. Because of the pyrolysis reac-
tions and pressure drop, vaporization takes place continuously
along the coil so there is a decrease in the liquid holdup. The derived
model uses a heuristic lumping approach based on the pilot plant
data. The resulting pseudo kinetic scheme presents a certain feed in-
ð2Þ
dependence within the range of stocks available for the study. An in-
dustrial case study was explored to provide insight into the problem
of reconciling the kinetics of pyrolysis and carbonization for the
upgradation of distillation residues. The kinetics of reaction has
been modeled assuming first order kinetics and 16-lump scheme.
140 J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144

The kinetic parameters have been determined using pilot plant data Table 1. The subscripts of rate constants in the following discussion
available elsewhere [43]. refer to the corresponding rate constants defined in the respective
The findings of the study indicate that the coil reactor modelling at models.
low conversions can be performed with a single set of Arrhenius pa- The estimated values for rate constants for these models have
rameters within a limited feedstock range. A case study on delayed been listed in Tables 2a–2d. Table 2a lists the values of the rate con-
coker, where pyrolysis is performed in two sequential units in order stants of the five lump model of Singh et al. (Model-A) reproduced
to optimize the delicate balance between selectivity and coke drum from their earlier publication [37]. Table 2b lists the estimated values
cycle time, caused by the parallelism between radical decomposition of Model-B using DE technique. The values of this model reported in
and condensation reactions. The results suggest that the changes re- their publication [39] are listed in Table 2c. The reported kinetic pa-
sponsible for the difference takes place mainly in the downstream rameters of Model-C have been listed in Table 2d.
unit, in spite of the fact that substantial conversion can occur in the It seems clear from Tables 2a–2c that there is a significant varia-
mainstream unit. tion in the three sets of values. The values reported by Kataria et al.
Wang and Anthony [44] extended the study by Martinez et al. [39] for Model-B are in reasonable agreement with the estimated
[34]. They found that the Martinez et al. [34] model could not explain values by DE algorithm. The variation in the estimated values can
the behavior of asphaltene cracking at higher temperatures. So they be attributed to the higher dependence of k values on temperature.
re-examined the data and gave the expressions for oil + gas yield as Also it is evident from the estimated values of rate constants that the
a function of asphaltene conversion or residence time; these de- values for k7 in Model-B are very low in case of NGSR and MVBF
scribed the data well. Their proposed approach is not dependent on feedstock. Also, the values estimated for this rate constant do not
the assumed reaction orders of the cracking. exhibit a regular trend with variation of temperature. It is thus indi-
A six lump model has been proposed by Mohaddecy and Sadighi cated that this route of reaction may not be dominant for all the
[33] for simulation of a soaker Visbreaking plant. They have charac- feedstocks.
terized the visbroken product into five lumps as Gas (C1–C2), LPG Singh et al. have established the reaction pathways using Delplot
(C3–C4), gasoline (IBP-180 °C), gas oil (180–320 °C), and fuel analysis [46] coupled with parameter estimation by DE algorithm.
(320 °C +). The proposed kinetic model is based on fifteen reactions This approach is more rational for determination of reaction path-
and thirty kinetic parameters, which is finally reduced to seven reac- ways. However the accuracy in the estimated yields of various frac-
tion pathways, as listed in Table 1. Thus fourteen kinetic parameters tions by Model-A does not indicate any improvement as compared
(frequency factors and activation energies for seven reactions) were to those estimated by Model-B. However, the possibility of absence
found appropriate to simulate the performance of the reactor with of the conversion route VGO to Gas is well supported in the kinetic
reasonable accuracy. The advantage of this model as reported by au- model (Model-A) presented by Singh et al. [38]. It is also reported
thors is considering gaseous product as two separate lumps as gas by Steacie and Bywater [47] that at higher pressures, the C–C chain
and LPG. It would be useful for better economical evaluation of the normally breaks from the middle. If it is taken to the consideration,
process, as LPG production may be a value addition to the process. the possibility of cracking of VGO fraction to Gas again seems to be
Nine sets of industrial data have been used for development of the very less.
model.
The complex models are more accurate and sometimes help in
further explanation of the process, as above for Wang and Anthony's
model. But due to the requirement of a detailed analysis of feedstocks
Table 2a
and products, these have limited applicability. Normally such detailed Estimated Rate Constants of Model-A.
analysis is not always possible to obtain for industrial feedstocks in a
refinery. Another problem with complicated models is comparatively Feedstock Rate Temperature
Constant
large computational effort required for their implementation. Thus, a min-1
400 °C 410 °C 420 °C 430 °C
simple model requiring least characteristic information and simple
NGSR k1 0.001236 0.001798 0.00289 0.005629
analysis would be more useful for an industrial application.
k2 0.000108 0.000252 0.001572 0.003515
k3 0.004628 0.007057 0.012457 0.022350
3. Critical evaluation of three latest models k4 0.009187 0.011023 0.015704 0.033992
k5 0.019956 0.038174 0.044900 0.178453
The two models given by Singh et al. [37,38] and a five lump model k6 0.010195 * * *
k8 * 0.006477 * *
reported by Kamal et al. [39] have been critically evaluated in the light BHSR k1 0.000498 0.001101 0.003226 0.005295
of available data, and the analysis is being presented in following sec- k2 0.000364 0.000719 0.001994 0.003630
tions. In subsequent discussions the five-lump model reported by k3 0.004328 0.004681 0.010592 0.024238
Singh et al. [37] has been referred as Model-A, the model reported by k4 0.011561 0.017333 0.020086 0.020233
k5 0.005431 0.043201 0.045147 *
Kamal et al. [39] has been referred as Model-B and the four lump
k6 0.007075 0.009923 * *
model reported by Singh et al. [38] has been referred as Model-C. k8 * * * 0.009689
MVBF k1 0.001060 0.002043 0.003363 0.004584
3.1. Model evaluation k2 0.000208 0.002054 0.003220 0.003844
k3 0.005431 0.011832 0.014652 0.015224
k4 0.005410 0.011959 0.013360 0.016973
The kinetic parameters for the Model-B reported by Kataria et al. k5 0.051968 0.081568 0.082126 0.10025
[38] were estimated using experimental data reported by Singh et k6 0.019675 * * *
al. [36], by differential evolution (DE) algorithm [45]. The estimated k8 * * * 0.038158
yields were compared with the obtained experimental data as well HRA k1 - 0.002013 0.002778 0.006671
k2 - 0.000910 0.000931 0.001451
as values obtained by using Model-A. Since the lumps chosen for
k3 - 0.009735 0.009885 0.013439
both the models are identical on boiling ranges, direct comparison k4 - 0.013170 0.017241 0.032437
of results could be possible. The lumps taken for both the models k5 - * 0.042068 0.107765
were – Gas, Gasoline (IBP – 150 °C), LGO (150–350 °C), and VGO k6 - 0.004800 * *
(350–500 °C). The reaction pathways suggested by these models k8 - * 0.015830 0.018484
–9
have been schematically shown in Figures at S. Nos. 13–15 in * These values are of the order of 10 and, therefore, are not mentioned here.
 J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144 141

Table 2b Table 2d
Estimated Rate Constants of Model-B. Estimated Rate Constants of Model-C.

Feedstock Rate Temperature Feedstock Rate Temperature

Constant Constant
400 °C 410 °C 420 °C 430 °C 400 °C 410 °C 420 °C 430 °C
min-1 min-1

NGSR k1 0.001236 0.001798 0.00289 0.005513 NGSR k1 0.000428 0.001015 0.002825 0.005015
k2 8.93E-05 2.49E-04 1.57E-03 3.44E-03 k2 0.000390 0.000756 0.001961 0.003630
k3 0.004985 0.007062 0.012457 0.01785 k3 0.003832 0.007551 0.012320 0.023256
k4 0.008855 0.011023 0.015704 0.031267 k4 0.002311 0.004256 0.005150 0.006569
k5 1.84E-02 1.91E-02 4.49E-02 1.67E-01 BHSR k1 0.001205 0.001672 0.002658 0.005162
k6 7.27E-03 7.31E-03 * * k2 0.000097 0.000261 0.001446 0.003223
k7 7.87E-09 2.69E-08 * * k3 0.005494 0.008181 0.013838 0.026695
BHSR k1 0.000499 0.000231 0.000108 0.001675 k4 0.010472 0.008241 * *
k2 0.000359 5.68E-05 1.41E-03 2.55E-03 MVBF k1 0.001021 0.001975 0.003102 0.004347
k3 0.004945 * 8.88E-03 2.40E-02 k2 0.000221 0.002003 0.003040 0.004434
k4 0.020207 0.014904 0.020833 0.016463 k3 0.006968 0.011534 0.015452 0.020661
k5 0.001969 5.27E-02 3.06E-02 * k4 0.017452 * * *
k6 0.001845 7.24E-03 5.68E-03 1.04E-02 HRA k1 - 0.001361 0.002718 0.005987
k7 * 5.75E-03 1.43E-02 2.06E-02 k2 - 0.000208 0.001106 0.001721
MVBF k1 0.001059 0.00012 0.003291 0.001202 k3 - 0.006921 0.010700 0.020689
k2 2.38E-10 2.06E-03 3.15E-03 4.09E-03 k4 - 0.013528 0.004425 0.011251
k3 0.005702 1.18E-02 1.30E-02 2.19E-02 –9
* These values are of the order of 10 and, therefore, are not mentioned here.
k4 0.013227 0.007584 0.010576 0.01327
k5 3.43E-02 8.65E-08 3.15E-02 2.98E-07
k6 0.016318 1.28E-08 * 4.07E-03
lump models do not vary much. Also, the amount of error of estima-
k7 1.27E-09 2.27E-02 * 2.24E-02
HRA k1 - 0.001177 0.002699 0.003034 tion is also of the same order of magnitude for the product lumps
k2 - 1.88E-04 4.31E-04 1.49E-03 namely GAS, GLN and VGO. But in case of LGO fraction, the errors of
k3 - 0.006661 0.009104 0.017807 estimation are higher with Kataria et al. model [39] than those esti-
k4 - 0.009563 0.017238 0.030309
mated from Singh et al. [37] model. This may be due to the reason
k5 - 1.63E-03 3.84E-02 5.49E-02
k6 - 1.10E-02 1.85E-02 1.57E-02
that in Model-B, the conversion of LGO to other lower boiling frac-
k7 - 5.79E-03 3.23E-03 3.81E-02 tions has been omitted. In Model-A the conversion of LGO to gasoline
–9 seems to compensate for this error. It may be observed here that the
* These values are of the order of 10 and, therefore, are not mentioned here.
additional route of conversion of VGO to GAS in Model-B does not sig-
nificantly contribute in increase or decrease in error of estimation for
4. Estimation of yields VGO fraction. It is therefore again established that the conversion of
VGO to Gas lump will not be significant, as indicated by Delplot anal-
The yields of converted products were estimated using the three ysis of Singh et al. [37].
models. These results have been tabulated in Tables 3–6. It may be It seems clear from these tables that in most of the points the
seen from these tables that the estimated yields from the two five values estimated by rate constants for Model-B reported by Kataria
et al. [39] in their paper show slightly higher error as compared to
Table 2c those estimated using rate constants obtained by DE minimization.
Rate Constants of Model-B reported by the authors in their publication [56]. It may be due to wide variation in the rate constants with variation
in temperature. Also, the parameters estimated by a gradient method
Feedstock Rate Temperature
Constant may sometimes converge to local minima, whereas the parameters
400 °C 410 °C 420 °C 430 °C estimated using a random search method such as DE is most likely
min-1

NGSR k1 0.00117 0.00168 0.00277 nd

to yield global minima.
k2 0.00035 0.00069 0.00151 nd
k3 0.00425 0.00810 0.01356 nd 5. Four lump model
k4 0.00565 0.00792 0.01248 nd
k5 0.00083 0.00099 0.00133 nd
A four lump kinetic model with four possible reaction pathways
k6 0.00084 0.00085 0.00121 nd
k7 1.74E-05 0.00012 0.00151 nd (Model-C) have been proposed by Singh et al. [38]. The reaction
BHSR k1 0.00044 0.00122 0.00272 0.00567 scheme has been depicted in Table 1 at S. No. 14. There are six possi-
k2 0.00052 0.00112 0.00261 0.00402 ble reaction pathways, out of which only four were found to be logi-
k3 0.00343 0.00756 0.01524 0.02382 cally existing on the basis of mathematical (parameter estimations),
k4 0.00533 0.00588 0.01362 0.01218
k5 3.76E-08 0.00108 0.00233 0.00451
Chemical (Flash calculations), and Delplot analysis [46].
k6 3.76E-08 0.00010 0.00059 0.00145 The results of the estimated values of the rate constants k1–k6 re-
k7 3.76E-08 0.00009 0.00007 0.00259 veal that the rate constants k5 and k6 do not assume reasonable
MVBF k1 0.00104 0.00205 0.00299 0.00487 values. Their estimated values are of the order of magnitude of 10 –9
k2 0.00098 0.00212 0.00308 0.00424
to 10 − 11. The Delplot analysis [46] carried out on the experimental
k3 0.00690 0.01182 0.01476 0.02070
k4 0.00356 0.00587 0.00756 0.01050 data also indicates the absence of reactions indicated by these path-
k5 0.00017 0.00044 0.00062 0.00172 ways. The flash calculations performed for the final reaction mixtures
k6 6.78E-05 0.00010 0.00034 0.00170 in ASPEN plus process flowsheet simulator using Chao-Sieder ther-
k7 0.00012 0.00021 0.00158 0.00214 modynamic property set reveal that the gasoline fraction exists most-
HRA k1 - 0.00143 0.00277 0.00660
k2 - 0.00130 0.00145 0.00654
ly in vapor phase (69–95%) at reaction temperatures, which rules out
k3 - 0.00618 0.01194 0.01866 the reaction via path represented by k5. Pyrolysis of gasoline is very
k4 - 0.00816 0.01218 0.01932 difficult in vapor phase and may not take place without a catalyst.
k5 - 0.00069 0.00083 0.00217 The flash calculations also indicate the liquid fraction of LGO between
k6 - 0.00107 0.00129 0.00351
13 and 60%, at studied temperatures. In thermal cracking, the paraffin
k7 - 0.00092 0.00134 0.00450
chain is more likely to break near the middle, as compared to near the
142 J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144

Table 3
Experimental and Predicted Yields by Five Lump Models (Feed: BHSR).

GAS* GLN* LGO* VGO*

T, °C t, min Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3

400 3 0.1880 0.1438 0.1441 0.1320 0.1200 0.1200 0.1200 0.1538 2.1770 1.3846 1.4446 1.0130 9.4590 5.7524 5.7980 1.5753
400 6 0.3100 0.2769 0.2773 0.2602 0.0420 0.2612 0.2629 0.3032 2.1770 2.7297 2.8147 1.9968 11.4790 10.9862 11.0948 3.1054
400 9 0.3660 0.4002 0.4006 0.3847 0.2530 0.4233 0.4266 0.4483 3.2320 4.0353 4.1157 2.9524 12.2640 15.7411 15.9287 4.5914
400 12 0.4080 0.5143 0.5146 0.5056 0.3800 0.6060 0.6093 0.5892 4.4110 5.3016 5.3523 3.8805 14.0040 20.0542 20.3354 6.0348
400 15 0.6200 0.6200 0.6200 0.6231 0.8090 0.8090 0.8090 0.7261 6.5290 6.5290 6.5290 4.7819 15.5590 23.9595 24.3475 7.4367
410 3 0.3190 0.3073 0.2677 0.4185 0.2460 0.2174 0.2460 0.3601 2.3020 1.5945 1.6846 3.0912 7.0760 4.7706 4.6852 9.9125
410 6 0.4540 0.5935 0.5439 0.7592 0.3770 0.4679 0.5179 0.6759 2.8570 3.7246 3.7122 5.0354 8.5190 8.5190 8.5190 11.2997
410 9 0.7310 0.8599 0.8250 1.0896 0.6080 0.7638 0.8091 0.9826 5.6190 6.2381 6.0075 6.8957 12.3600 11.4164 11.6198 12.6058
410 12 1.1080 1.1080 1.1080 1.4102 1.1140 1.1140 1.1140 1.2806 8.5060 9.0102 8.5060 8.6758 11.7030 13.6076 14.0910 13.8346
410 15 1.9400 1.3390 1.3903 1.7213 1.6850 1.5244 1.4278 1.5704 11.9400 11.9400 11.1531 10.3794 14.4300 15.2149 16.0228 14.9899
420 3 0.5150 0.9173 0.5150 0.8031 0.5970 0.5970 0.5970 0.7306 3.5800 3.3610 3.5799 5.2187 9.1740 5.3742 5.5576 11.6210
420 6 0.7840 1.7407 1.1881 1.4385 0.8690 1.1996 1.2218 1.3613 5.9290 6.9915 7.2808 8.8580 7.4120 9.5168 9.7578 14.6727
420 9 1.4340 2.4796 1.9731 2.0129 1.4330 1.8174 1.8591 1.9379 9.9410 10.7673 11.0210 12.1717 9.5570 12.6403 12.8515 17.3894
420 12 3.5660 3.1428 2.8330 2.5322 3.4670 2.4576 2.4966 2.4655 18.4510 14.5898 14.7372 15.1912 8.6440 14.9244 15.0478 19.8042
420 15 3.7380 3.7380 3.7380 3.0018 3.1250 3.1250 3.1250 2.9490 18.3810 18.3810 18.3810 17.9451 16.5210 16.5210 16.5210 21.9469
430 3 1.2090 1.4865 1.2090 1.5957 1.0910 1.0910 1.0910 1.1352 7.5700 6.7804 6.7279 7.9201 3.1980 3.1980 3.5093 11.7621
430 6 2.1800 2.7864 2.5599 2.9299 2.0950 2.1837 2.1889 2.0661 12.1930 12.6623 12.6123 13.2764 7.5540 5.9028 6.1415 14.0789
430 9 3.9230 3.9230 3.9982 4.1227 3.2750 3.2750 3.2750 2.8941 17.7590 17.7590 17.7590 17.9976 6.8380 8.1791 8.0610 16.0020
430 12 6.7260 4.9169 5.4811 5.1914 5.1610 4.3621 4.3350 3.6321 25.2640 22.1693 22.2604 22.1639 7.2090 10.0831 9.4049 17.5843
430 15 6.9750 5.7861 6.9750 6.1511 4.3690 5.4429 5.3586 4.2912 25.8110 25.9797 26.1975 25.8449 11.6640 11.6640 10.2869 18.8720

*Legends:
Exp. Experimental values.
Pred-1 Values determined using Model-A.
Pred-2 Values determined using Model-B, kinetic parameters estimated using DE technique.
Pred-3 Values determined using Model-B model, using kinetic parameters reported by Kataria et al. [56].

ends [47] at higher pressures. One or both of these may be the reasons
for cracking of LGO fraction only to Gasoline fraction and absence of
reaction pathway along rate constant k6.
Based on these analyses, a four lump model has been suggested by
four rate constants, namely k1 to k4. The predictions by the developed
model are well within acceptable range. The advantage of this model
is that it requires boiling range data of only three product lumps
against four in the previous model. The limitation of this model is
that it is applicable for prediction of the results of cracking of long res-
idue, due to feedstock lump being the fraction boiling above 350 °C.
The data used for parameter estimation for this model are same as
in five lump models.
The Model-C predicts nearly 70% data points within less than 10%
error and 75% data points are predicted with less than 15% error [38].
The predictions with more than 25% error are reported to be only 15%.
6. Further directions
An important shortfall of these three models is their limited
dependence on the structural characteristics of the feedstock. The

Table 4
Experimental and Predicted Yields by Five Lump Models ( Feed : MVBF).

GAS* GLN* LGO* VGO*

T, °C t, min Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3

400 3 0.4320 0.3085 0.3085 0.3076 0.0030 0.1124 0.0905 0.2881 2.0000 1.8203 1.8505 2.0320 4.6750 3.5990 3.5706 1.0489
400 6 0.5990 0.5990 0.5990 0.6043 0.3380 0.3380 0.3380 0.5658 3.9340 3.9340 3.9340 3.9898 6.5070 6.4683 6.4305 2.0580
400 9 0.8530 0.8725 0.8726 0.8904 0.6480 0.6909 0.7104 0.8335 5.5880 6.2284 6.1889 5.8761 6.2130 8.7255 8.6920 3.0290
400 12 1.0860 1.1301 1.1303 1.1664 0.7830 1.1791 1.1808 1.0916 6.7590 8.6132 8.5640 7.6937 7.1530 10.4704 10.4506 3.9631
400 15 1.4410 1.3726 1.3730 1.4327 1.8060 1.8060 1.7267 1.3404 11.0170 11.0170 11.0170 9.4449 11.7880 11.7880 11.7880 4.8617
410 3 0.3830 0.5937 0.3830 0.6307 0.7340 0.5969 0.5974 0.6156 4.6620 3.4385 3.4411 3.9172 1.5720 1.5720 1.8554 5.4947
410 6 0.7540 1.1506 0.8829 1.2013 0.9520 1.1568 1.1573 1.1853 5.7190 6.6637 6.6663 7.0656 1.7990 3.0465 3.4535 6.8796
410 9 1.1150 1.6730 1.4816 1.7463 1.6820 1.6820 1.6820 1.7239 9.6890 9.6890 9.6890 10.0369 4.8210 4.4296 4.8210 8.1380
410 12 2.1630 2.1630 2.1630 2.2670 2.0080 2.1746 2.1738 2.2334 11.5930 12.5267 12.5220 12.8417 7.9970 5.7269 5.9824 9.2789
410 15 3.3080 2.6226 2.9128 2.7647 3.0790 2.6367 2.6347 2.7154 15.9560 15.1883 15.1771 15.4900 6.5750 6.9438 6.9597 10.3106
420 3 1.0700 0.9604 0.9604 0.8924 1.0830 0.9193 0.9193 0.9028 6.4720 4.5758 4.5758 4.8989 3.0230 3.0230 3.0230 6.7385
420 6 1.9050 1.8298 1.8298 1.6723 1.7110 1.7515 1.7515 1.7496 9.3740 9.3740 9.3740 8.5875 5.9310 5.1033 5.1033 8.4913
420 9 2.4810 2.6168 2.6168 2.3995 2.5530 2.5048 2.5048 2.5508 12.6290 14.2312 14.2312 11.9792 5.9620 6.4726 6.4726 10.0623
420 12 2.8570 3.3291 3.3291 3.0778 2.9250 3.1867 3.1867 3.3099 13.7610 19.0302 19.0302 15.0982 7.1750 7.3100 7.3100 11.4670
420 15 3.9740 3.9740 3.9740 3.7113 3.8040 3.8040 3.8040 4.0298 17.4940 23.6893 23.6892 17.9665 7.7530 7.7530 7.7530 12.7196
430 3 1.0560 1.2921 1.0308 1.5011 1.2770 1.2770 1.2770 1.2200 7.0570 4.9836 5.2456 6.9990 4.4530 4.2431 4.3434 10.2712
430 6 2.0990 2.4318 2.2545 2.6758 2.2980 2.4471 2.4628 2.2359 11.9490 10.4814 10.7039 11.7375 9.4360 6.8399 6.9570 12.4550
430 9 3.4370 3.4370 3.5600 3.7380 3.5660 3.5303 3.5514 3.1531 16.1160 16.1160 16.1160 15.9534 6.6020 8.2939 8.3853 14.3346
430 12 5.4620 4.3237 4.8741 4.7003 4.5420 4.5420 4.5420 3.9825 20.8270 21.6402 21.3189 19.7060 8.3300 8.9659 9.0136 15.9455
430 15 6.1500 5.1057 6.1500 5.5736 4.6730 5.4943 5.4375 4.7340 21.9330 26.8985 26.2160 23.0477 9.1130 9.1130 9.1130 17.3192

*Legends:
Exp. Experimental values.
Pred-1 Values determined using Model-A.
Pred-2 Values determined using Model-B, kinetic parameters estimated using DE technique.
Pred-3 Values determined using Model-B model, using kinetic parameters reported by Kataria et al. [56].
 J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144 143

Table 5
Experimental and Predicted Yields by Five Lump Models ( Feed : NGSR).

GAS* GLN* LGO* VGO*

T, °C t, min Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3

400 3 0.3770 0.3626 0.3626 0.3469 0.0540 0.0540 0.0540 0.1046 1.6930 1.4848 1.5326 1.2567 3.5170 2.5448 2.4988 1.6609
400 6 0.7090 0.7090 0.7090 0.6821 0.1310 0.1546 0.1580 0.2098 3.1300 3.1300 3.1300 2.4752 4.7010 4.7010 4.7010 3.2574
400 9 0.8860 1.0400 1.0400 1.0062 0.3060 0.3060 0.3060 0.3156 4.4940 4.8923 4.7794 3.6567 5.5400 6.5158 6.6336 4.7917
400 12 0.9850 1.3564 1.3563 1.3194 0.2020 0.5113 0.4923 0.4220 4.4630 6.7345 6.4695 4.8026 5.8520 8.0309 8.3214 6.2661
400 15 1.7610 1.6586 1.6586 1.6222 0.7970 0.7726 0.7120 0.5289 8.8230 8.6251 8.1898 5.9139 9.7870 9.2834 9.7870 7.6825
410 3 0.5540 0.5234 0.5235 0.4908 0.1070 0.1070 0.1070 0.2043 2.9140 2.1429 2.1464 2.3677 4.3030 3.0839 3.0839 2.3048
410 6 0.8970 1.0162 1.0163 0.9560 0.2020 0.2663 0.2728 0.4036 4.3370 4.3370 4.3370 4.6151 5.7520 5.7520 5.7520 4.4723
410 9 1.1640 1.4801 1.4802 1.3971 0.5100 0.4790 0.4889 0.5979 6.5150 6.5570 6.5563 6.7487 6.6660 8.0465 8.0465 6.5101
410 12 1.9510 1.9168 1.9169 1.8154 0.7460 0.7460 0.7479 0.7875 8.9410 8.7810 8.7907 8.7746 10.0060 10.0060 10.0060 8.4251
410 15 2.3280 2.3280 2.3280 2.2120 1.0430 1.0672 1.0430 0.9727 10.6940 10.9902 11.0281 10.6985 7.7570 11.6653 11.6653 10.2240
420 3 1.1960 0.8260 0.8260 0.8031 0.5130 0.4493 0.4493 0.4401 5.7220 3.8539 3.8539 3.8957 5.8910 4.1943 4.1943 3.5568
420 6 1.3520 1.5750 1.5750 1.5523 0.7100 0.8566 0.8566 0.8548 7.8720 7.8770 7.8770 7.4669 7.4690 7.4690 7.4690 6.7613
420 9 2.3280 2.2542 2.2542 2.2521 1.2810 1.2260 1.2260 1.2460 11.9870 11.9870 11.9870 10.7415 11.1120 9.9764 9.9764 9.6447
420 12 2.8700 2.8700 2.8700 2.9067 1.5610 1.5610 1.5610 1.6157 13.3320 16.1175 16.1175 13.7452 11.5770 11.8463 11.8463 12.2355
420 15 3.3110 3.4284 3.4284 3.5197 1.8440 1.8647 1.8647 1.9656 15.3320 20.2158 20.2158 16.5014 13.1890 13.1890 13.1890 14.5596
430 3 2.0280 1.5178 1.5178 na 1.1980 0.9478 0.9478 na 9.2610 6.7958 6.7966 na 8.1220 6.7265 6.7264 na
430 6 2.7930 2.7930 2.7930 na 1.7440 1.7440 1.7440 na 14.4440 15.1506 15.1519 na 9.7320 9.7320 9.7320 na
430 9 3.7460 3.8643 3.8643 na 2.3870 2.4129 2.4129 na 16.7300 23.7747 23.7763 na 10.6520 10.6520 10.6520 na

*Legends:
Exp. Experimental values.
Pred-1 Values determined using Model-A.
Pred-2 Values determined using Model-B, kinetic parameters estimated using DE technique.
Pred-3 Values determined using Model-B model, using kinetic parameters reported by Kataria et al. [56].

experimental data used for the development of model only character-
izes the feedstocks on the basis of their hydrocarbon type in terms of
polar and non polar aromatics, paraffins and resins or asphaltenes. A
detailed structural analysis of the feedstocks as well as cracked prod-
ucts, in terms of breakage of bonds and change in nature of hydrocar-
bon species would be an added advantage for the description of
cracking behaviour in more detail. This will also be helpful in the de-
velopment of a model based on structural changes during visbreaking.
7. Conclusions
A critical analysis of kinetic models of thermal cracking of petro-
leum residues has been presented in the present paper in terms of
their suitability for application, as well as error in their predictions.
It is clearly indicated in the above discussion that both the five
lump models, Model-A and Model-B are simple and relatively easy
to be used for the prediction of the product yields with reasonable
accuracy. The model-B suggested by Kataria et al. [39] is evolved on
the basis of certain assumptions which are concluded from the past
experiences of the cracking behavior. The model developed by Singh
et al. [37] has been developed on a strong theoretical basis (Delplot
analysis), which has been established mathematically by Bhore et al.
[46]. However the yields predicted by the earlier model are also in
the acceptable range, and do not show much variation with respect
to the values estimated by latter. But the model suggested by Singh
et al. is more realistic and can be justified by stronger theoretical
basis. The four lump model suggested by Singh et al. (Model-C) is ba-
sically a reduced parameter model derived from the five lump model
and is supported by same theoretical basis. The overall advantage of
all of these models is their simplicity, minimum characteristics infor-
mation requirement, and less computational expense. However, these
models can be made more useful if the variation in kinetic parameters
can be estimates as a function of some simple structural characteris-
tics e.g. hydrocarbon type, resin and asphaltene contents.

Table 6
Experimental and Predicted Yields by Five Lump Models ( Feed : HRA).

GAS* GLN* LGO* VGO*

T, °C t, min Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3 Exp Pred-1 Pred-2 Pred 3

410 6 0.7630 0.5812 0.7630 0.8309 0.2830 0.2830 0.2830 0.7542 3.8190 2.7900 3.8190 3.5342 5.5230 3.8017 5.1507 4.6157
410 9 1.1190 1.1190 1.1811 1.2251 0.4940 0.5848 0.5386 1.1145 5.3320 5.3320 5.6012 5.1773 7.3200 7.3200 7.3200 6.7241
410 12 1.5530 1.6167 1.6189 1.6062 0.9450 0.9034 0.8591 1.4643 7.5230 7.6452 7.3033 6.7433 9.2470 10.5760 9.2470 8.7085
410 15 2.0730 2.0774 2.0730 1.9747 1.2370 1.2370 1.2370 1.8040 8.8040 9.7472 8.9290 8.2358 9.4650 13.5892 10.9512 10.5752
420 3 0.7980 0.7980 0.7980 0.8026 0.2550 0.2550 0.2550 0.4244 3.4180 2.7934 2.8864 3.4384 5.2570 4.5229 4.5226 3.4846
420 6 1.1310 1.5292 1.5686 1.5537 0.7180 0.7180 0.7180 0.8277 5.8750 5.8750 6.0054 6.6051 7.9160 7.9160 7.9160 6.6493
420 9 3.6640 2.1992 2.3075 2.2574 1.8540 1.3335 1.3335 1.2114 13.2000 9.1340 9.2585 9.5223 12.8840 10.3986 10.3989 9.5205
420 12 2.7930 2.8132 3.0119 2.9173 1.4370 2.0570 2.0570 1.5770 11.4140 12.4831 12.5687 12.2101 12.1420 12.1510 12.1514 12.1225
420 15 3.6800 3.3757 3.6800 3.5370 1.9760 2.8529 2.8528 1.9257 13.9370 15.8547 15.8764 14.6873 13.3210 13.3210 13.3210 14.4776
430 3 1.2850 1.8475 1.2850 1.8725 0.5960 0.5960 0.5960 1.8477 5.8490 5.0428 5.5800 5.2076 7.6860 7.4689 7.1477 5.2901
430 6 2.8820 3.4188 3.0791 3.5490 1.7240 1.3992 1.3913 3.4877 11.3490 11.3490 11.3490 9.7088 11.7240 11.5070 11.2639 9.6688
430 9 4.7550 4.7550 5.1141 5.0555 2.1750 2.2869 2.2772 4.9478 15.5050 18.1046 17.0001 13.6029 13.3440 13.3440 13.3440 13.2705
430 12 7.2150 5.8914 7.2150 6.4139 3.1830 3.1830 3.1830 6.2520 18.9250 24.8110 22.3491 16.9753 12.2240 13.8040 14.0844 16.2103

*Legends:
Exp. Experimental values.
Pred-1 Values determined using Model-A.
Pred-2 Values determined using Model-B, kinetic parameters estimated using DE technique.
Pred-3 Values determined using Model-B model, using kinetic parameters reported by Kataria et al. [56].
144 J. Singh et al. / Fuel Processing Technology 94 (2012) 131–144
References
[1] Martin Hus, Visbreaking process has strong revival, Oil and Gas Journal (April13
1981) 109–120.
[2] J.F. Le Page, F. Morel, A.M. Trassard, J. Bousquet, Thermal cracking under hydrogen
pressure: Preliminary step to conversion of heavy oils and residues, Proceedings
of the American Chemical Society April 1987 (1987) 470–476.
[3] J.F. Le Page, S.G. Chatila, M. Davidson, Resid and heavy oil processing, Editions
Technip, 1992 Paris.
[4] Michael K. Porter, Glenn A. Clausen,, Jeffrey B. Harrison, Ajit K. Bhattacharya,
Hydroconversion of heavy petroleum fractions in the presence of a solid sup-
ported heterogeneous catalyst and an oil-miscible metal catalyst, Canadian Pat-
ent, 1993 CA 2073417 AA.
[5] P. Rahimi, T. Gentzis, C. Fairbridge, C. Khulbe, Chemistry of petroleum residues in
the presence of H- donors, Preprints - American Chemical Society, Division of Pe-
troleum Chemistry 43 (4) (1998) 634–636.
[6] F.N. Dawson Jr., Dynacracking to go commercial, Hydrocarbon Processing May
1981 (1981) 86–92.
[7] W.L. Nelson, J.D. Mc Kinney, D.E. Blasser, Coking, in: J.J. McKetta, W.A. Cunning-
ham (Eds.), Petroleum, Encyclopedia of Chemical Processing and Design, Vol.
10, Marcel Dekkar Inc., New York, 1979, pp. 1–41.
[8] A. Billon, E. Fehr, E. Lorenz, SDA key to upgrading heavy crudes, Oil and Gas Jour-
nal 75 (Jan. 24 1977) 43–48.
[9] D.B. Bartholic, R.P. Haseltine, New process cleans up FCC feed, Oil and Gas Journal
79 (June 01 1981) 89–92.
[10] A. Rhoe, C. de Blignieres, Visbreaking: a flexible process, Hydrocarbon Processing
(January 1979) 131–136.
[11] J. Bousquet, T. Barbou des Courieres, J.P. Mermoz, Thermal treatment of hydrocar-
bon charges in the presence of additives which decrease the formation of coke,
Belgium Patent. No. BE 901092 A1, 1985.
[12] V.D. Singh, Visbreaking Technology, Erdol und Kohle – Erdgas – Petrocheimie ver-
einigt mit Brennstoff-Chemie, 1986, pp. 19–23, Bd 39, Heft 1.
[13] Jorge A. Bonilla, V.K. Kapoor, M.M. Kumar, Updated visbreaking technology for
the new millennium, Ninth annual refining seminar, October 7 1997 New
Orleans, Louisiana.
[14] S. Di Carlo, B. Janis, Composition and visbreakability of petroleum residues,
Chemical Engineering Science 47 (9–11) (1992) 2695–2700.
[15] R. Maciel Filho, M.F. Sugaya, A computer aided tool for heavy oil thermal cracking
process simulation, Computers & Chemical Engineering 25 (2001) 683–692.
[16] Mario Dente, Giulia Bozzano, M. Rossi, Reactor and kinetic modeling of the vis-
breaking process, Proceedings of ICheaP1, 1993, pp. 163–172.
[17] P.B. Crawford, W.A. Cunningham, Carbon formation in cat cracking, Petroleum
Refiner (January 1956) 169–173.
[18] M. Dente, G. Bozzano, T. Faravelli, A. Marongiu, i S. Pierucc, E. Ranzi, Kinetic
modeling of pyrolysis processes in gas and condensed phase, Advances in Chem-
ical Engineering 32 (2007) 51–166.
[19] Y. Liu, X. Chen, H. Zhao, C. Yang, Establishment of kinetic model for catalytic py-
rolysis of Daqing atmospheric residue, Chinese Journal of Chemical Engineering
17 (1) (2009) 78–82.
[20] Maria Luisa A. Goncalves, Deusa Angelica P. Mota, Wildson V. Cerqueira, D. Andre,
L.M. Saraiva, Maria I.F. Coelho, Ana Maria R.F. Teixeira, Marco Antonio G. Teixeira,
Knowledge of petroleum heavy residue potential as feedstock in refining process
using thermogravimetry, Fuel Processing Technology 91 (9) (2010) 983–987.
[21] E. Alvarez, G. Marroquin, F. Trejo, G. Centeno, J. Ancheyta, Jose A.I. Diaz, Pyrolysis
kinetics of atmospheric residue and its SARA fractions, Fuel 90 (12) (2011)
3602–3607.
[22] Singh, Jasvinder, (2004a), Thermal cracking of petroleum feedstocks: kinetics and
process modelling studies, Ph.D. Thesis, Indian Institute of Technology Roorkee.
[23] H.H. Al Soufi, Z.F. Savaya, H.K. Moahummed, I.A. Al-Azami, Thermal conversion
(visbreaking) of heavy Iraqi residue, Fuel 67 (12) (1988) 1714–1715.
[24] R. Krishna, Y.K. Kuchhal, G.S. Sarna, I.D. Singh, Visbreaking studies on Aghajari
long residue, Fuel 67 (3) (1988) 379–383.
[25] Mario Dente, G. Bozzano, G. Bussani, A comprehensive program for visbreaking
simulation: product amount and their properties prediction, Computers & Chem-
ical Engineering 10 (1997) 1125–1134.
[26] Julian Castellanos, Jose Luis Cano, Rodolfo Del Rosal, Victor M. Briones, Ruben L.
Mancilla, Kinetic model predicts visbreaker yields, Oil and Gas Journal 89 (11)
(1991) 76–82.
[27] A. Del Bianco, G. Garuti, C. Pirovano, R. Russo, Thermal cracking of petroleum res-
idues: 1. Kinetic analysis of the reaction, Fuel 74 (5) (1993) 756–760.
[28] Shao Long Zhou, Zhen Shao Zhou, Ma Bai Wen, Chen Li Lieh, Chang Ke Qia, Re-
search study on dynamic model of complex reaction of vacuum residual oil,
(Translated from Chinese), ACTA Petrolei Sinica (Petroleum Processing Section)
15 (2) (1999) 73–78.
[29] M. Yasar, D.M. Trauth, M.T. Klein, Asphaltene and resid pyrolysis. 2. The effect of
reaction environment on pathways and selectivity, Energy & Fuels 15 (2001)
504–509.
[30] Jiazhi Xiao, Lanjuan Wang, Qinglin Chen, Daoming Wang, Modeling for product
distribution in thermal conversion of heavy oil, Petroleum Science and Technolo-
gy 20 (5&6) (2002) 605–612.
[31] L. Yan, X. Zhang, S. Zhang, The study of molecular modeling for heavy oil thermal
cracking, Chemical Engineering and Technology 30 (9) (2007) 1166–1175.
[32] T. Wang, A. Geng, X. Li, Pyrolysis of one crude oil and its asphaltenes: Evolution of
gaseous hydrocarbons and carbon isotopes, Journal of Petroleum Science and En-
gineering 71 (2010) 8–12.
[33] S.S. Reza Mohaddecy, S. Sadighi, Simulation and kinetic modeling of vacuum res-
idue soaker–visbreaking, Petroleum & Coal 53 (1) (2011) 26–34.
[34] Maria T. Martinez, Ana M. Benito, Maria A. Callejas, Thermal cracking of coal res-
idues: Kinetics of asphaltene decomposition, Fuel 76 (9) (1997) 871–877.
[35] R. Zou, L. Lujia, in: T.K. Yen, G.V. Chilingarian (Eds.), Asphaletenes and asphalts, I,
Elsevier, Amsterdam, 1994.
[36] Jasvinder Singh, M.M. Kumar, A.K. Saxena, Surendra Kumar, Studies on thermal
cracking behaviour of residual feedstocks in a batch reactor, Chemical Engineer-
ing Science 59 (2004) 4505–4515.
[37] Jasvinder Singh, M.M. Kumar, A.K. Saxena, Surendra Kumar, Reaction pathways
and product yields in mild thermal cracking of vacuum residues: A multi-lump
kinetic model, Chemical Engineering Journal 108 (2005) 239–248.
[38] Jasvinder Singh, M.M. Kumar, Y. Ram, A. Mateen, Surendra Kumar, A four-lump ki-
netic model for visbreaking, Hydrocarbon Asia, May/June 2005, 15, 2005, pp. 32–35.
[39] K.L. Kataria, R.P. Kulkarni, A.B. Pandit, J.B. Joshi, M.M. Kumar, Kinetic studies of
low severity visbreaking, Industrial and Engineering Chemistry Research 43
(2004) 1373–1387.
[40] A. Chakma, M.R. Islam, Modelling of visbreaking of bitumen in a jet reactor,
Chemical Engineering Science 45 (8) (1990) 2769–2775.
[41] S.V. Patankar, Numerical heat transfer and fluid flow, McGraw-Hill, New York,
1981.
[42] William H. Dawson, Chornet Esteban, Ralph P. Overend, A. Chakma, Jean - Pierre
Lemonnier, Process for reducing the viscosity of heavy hydrocarbon oils, U.S. Pat-
ent No. 5,096,566, 1992.
[43] M. Bria, R. Filgueiras, Relatório Diter, Report of Petrobras R&D Center, 1982 Rio de
Janerio, Brazil.
[44] J. Wang, Edward J. Anthony, A study of thermal-cracking behavior of asphaltenes,
Chemical Engineering Science 58 (1) (2003) 157–162.
[45] R. Storn, K. Price, Differential evolution – A simple and efficient heuristic strategy
for global optimization over continuous spaces, Journal of Global Optimization 11
(1997) 341–359 Dordrecht.
[46] Nazir A. Bhore, Michael T. Klien, Kenneth B. Bischoff, The Delplot technique: A
new method for reaction pathway analysis, Industrial and Engineering Chemistry
Research 29 (1990) 313–316.
[47] E.W.R. Steacie, S. Bywater, Mechanism for thermal decomposition of hydrocar-
bons, in: B.T. Brooks, Cecil E. Boord, Stewart S. Kurtz Jr., Schmerling Louis (Eds.),
The Chemistry of Petroleum Hydrocarbons Vol-II, Reinhold Publishing Corpn.,
1955, Chapter 22.