Kinetic modeling of high-severity uidized catalytic cracking

q
Mohammad A. Abul-Hamayel
*
Department of Chemical Engineering, Center for Rening and Petrochemicals, The Research Institute, King Fahd University
of Petroleum and Minerals, Box 1483, Dhahran 31261, Saudi Arabia
Received 11 March 2002; accepted 9 January 2003; available online 13 February 2003
Abstract
This paper presents the results of kinetic modeling of high-severity uidized catalytic cracking process using a simplied methodology of
estimating kinetic parameters. A 4-lump model was used to demonstrate the new approach for modeling the kinetics data that was collected
using micro-activity test (MAT) method. MAT experiments were done at 550, 600 and 650 8C. Kinetic parameters for the 4-lump model and
the activation energies are presented.
q 2003 Published by Elsevier Science Ltd.
Keywords: Catalytic cracking; Activation energy; Kinetic parameter
1. Introduction
The Fluid Catalytic Cracking Process (FCC) has been the
workhorse of the petroleum rening industry used to
convert heavy petroleum fractions into gasoline, distillates
and light olens. It plays an important role as a gasoline
producing unit. Growing environmental concerns require
lowering the aromatics content of the gasoline to less than
25 vol% in order to minimize toxic emissions from
automotive exhausts. On the other hand, the value and
demand for light olens (C
2
C
4
) is increasing. These
olens are used in downstream alkylation and etherication
processes and are also used as petrochemical feedstock [1].
The aforementioned factors entailed in extensive research
efforts and lead to the high severity uid catalytic process
(HS-FCC), which maximizes the light olen yield while
maintaining the quality of gasoline. HS-FCC process needs
a reactor optimized for high cracking temperatures, short
residence time and a catalyst system that minimizes
hydrogen transfer reactions [2]. To achieve residence time
shorter than in the conventional FCC riser reactors, a
downer reactor was proposed in a patent by Gross and
Ramage in 1983. Later several authors studied this downer
and their results indicate that the downer reactor has more
advantages compared to riser in terms of back mixing and
residence time. Simulation of a HS-FCC process, which
utilizes a downer, requires kinetic parameters. This work
presents the results of kinetic modeling of uidized catalytic
cracking reactions using a catalyst designed for maximizing
light olens and at micro-activity test (MAT) conditions
severe than conventional.
Many complex reactions occur during the FCC process
and the product consists of a mixture of many compounds.
The description of complex mixtures by lumping large
number of chemical compounds into smaller groups of
pseudo-components has been widely used by researchers to
provide tractable number of kinetic equations [3]. Grouping
of compounds has to be made, taking into consideration the
dynamic behavior of each lump [4]. A 3-lump kinetic model
was rst developed by Weekman [5], comprising of
unconverted feed, gasoline and gas plus coke as three
lumps. Yen et al. [6] and Lee et al. [7] presented a 4-lump
model. In this model, the gas plus coke lump was split into
gas and coke separately. Gas oil cracking was modeled by
considering the gas composition in detail by John and
Wojciechowsky [8]. Kinetic constant independence from
feed composition was achieved using a 10-lump model by
Jacob et al. [9]. Later it was shown that insignicant
improvement, compared to a 4-lump model, is made with a
10-lump model. To predict the catalytic cracking of residual
oil a 6-lump model (residual oil, VGO/HCO, LCO,
gasoline, gas and coke) was presented by Takatsuka et al.
[10]. Coxson and Bischoff [11] presented a 10-lump model
0016-2361/03/$ - see front matter q 2003 Published by Elsevier Science Ltd.
doi:10.1016/S0016-2361(03)00017-6
Fuel 82 (2003) 11131118
www.fuelrst.com
q
Published rst on the web via Fuelrst.comhttp://www.fuelrst.com
*
Tel.: 966-3-860-2266; fax: 966-3-860-2266.
E-mail address: hamayel@kfupm.edu.sa (M.A. Abul-Hamayel).
in which the feed and the products were grouped into six
pseudo-compounds. A similar model was used by Larocca
et al. (1990) [12], where the feed gas oil was split into
parafns, naphthenes and aromatics. Oliveira and Biscasia
[13] presented a 4-lump model for cracking of gasoline.
They incorporated this model into a gas oil cracking model
comprising of six lumps. Ancheyta-Juarez et al. [14]
presented a sequential strategy for determining the kinetic
constants of a 5-lump model. They divided the gas lump of
the 4-lump scheme into dry and wet gas lumps. Recently,
the same authors evaluated the rate constants in combi-
nation with the deactivation function [15].
In all of the above-mentioned studies, kinetic parameters
were determined by simultaneous regression analysis of the
rate equations formulated depending on the lumping scheme
as is conventionally done. In this paper, a simplied
approach of data modeling is presented. Instead of
simultaneous regression of the rate equations involved,
consecutive regression of each rate equation is performed
and the corresponding kinetic constants are determined.
A 4 or 5-lump kinetic scheme seems satisfactory and
reasonable to represent the kinetics of VGO cracking. The
groups can be increased depending upon specic objectives
and availability of accurate analytical tools. In this paper, a
4-lump reaction scheme is used to represent the kinetics of
VGO cracking, using a novel catalyst system and operating
at temperatures higher than conventional. Fig. 1 shows the
reaction scheme.
The governing equation for VGO cracking is given
below in Eq. (1) (since VGO consist of a complex mixture
of hydrocarbons, its cracking is considered as a second order
reaction). The rate of gasoline, gas and coke formation are
given by Eqs. (2) (4) below.
2dY
1
=dt KY
2
1
w 1
where K k
12
k
13
k
14
:
dY
2
=dt k
12
Y
2
1
2 k
2
Y
2
w 2
where k
2
k
23
k
24
:
dY
3
=dt k
13
Y
2
1
k
23
Y
2
w 3
dY
4
=dt k
14
Y
2
1
k
24
Y
2
w 4
w e
2at
deactivation function 5
a deactivation constant
Note: it is conventional to use the same deactivation
function for all reaction steps.
2. Experimental
2.1. Micro-activity tests
MAT were performed in a xed-bed (isothermal and plug
ow) reactor described by ASTM D-3907-92. The feedstock
was hydrodesulfurized VGO of Arabian Light origin. A
novel catalyst, under development stage was used. Weight
hourly space velocity (WHSV) was varied between 24
120 h
21
. Changing the catalyst/oil ratio at a xed feed
injection time of 30 s varied WHSV. Reaction was carried
out at temperatures of 550650 8C. Since these tempera-
tures are higher than the conventional ones, few exper-
iments were done at a catalyst/oil ratio of zero to take into
account the effect of thermal cracking. Table 1 lists the
experimental conditions. A Y-zeolite based proprietary
Nomenclature
a deactivation constant
w deactivation function
C/O catalyst to oil ratio
k
12
rate constant for VGO to gasoline
k
13
rate constant for VGO to gas
k
14
rate constant for VGO to coke
k
23
rate constant for gasoline to gas
k
24
rate constant for gasoline to coke
K k
12
k
13
k
14
n order of reaction
R
2
coefcient of determination
t injection time
Y
1
weight fraction of vacuum gas oil
Y
2
weight fraction yield of gasoline
Table 1
Experimental conditions
Temperature (8C) 550, 600, 650
WHSV (h
21
) 20120
Injection time (s) 30
Feed (g) ,1
Catalyst/oil ratio 05
Fig. 1. Reaction scheme of a 4-lump model.
M.A. Abul-Hamayel / Fuel 82 (2003) 11131118 1114
catalyst having a unit cell size of 24.233 A

and a specic
surface area of 142.8 m
2
/g was used.
2.2. Product analysis
Gas chromatographic techniques were used for the
analysis of liquid and gas products. Hydrocarbon gases
(methane, ethane, propane, iso and normal butane, ethylene,
propylene, isobutene, butenes) were analyzed using Shi-
madzu GC 17A equipped with ame ionization detector.
The components were separated by a 50 m long capillary
column, CP Sil-5. Hydrogen was analyzed in a separate GC
equipped with a thermal conductivity detector. A packed
column of activated carbon was used to separate hydrogen
from methane and nitrogen. Liquid products from MAT
were analyzed in a simulated distillation gas chromatog-
raphy (Shimadzu GC 14B) according to the standard
method ASTM D-2887. The carbon deposited on the
catalyst was determined using a Strohlein carbon analyzer.
The results of two gas chromatographs and carbon analyzer
were combined to obtain weight fraction of unconverted
VGO and yields of gasoline, gas and coke.
3. Results and discussion
3.1. Yield distribution
Weight fraction yields of unconverted VGO Y
1
;
gasoline Y
2
; gas Y
3
and coke Y
4
versus contact time
t were plotted for the three temperatures. Fractional yields
are used instead of other SI units for convenience of
regression analysis. These experimental data were
Fig. 2. Smoothed yields of VGO, gasoline, gas and coke at 550 8C.
Fig. 3. Yields of (a) VGO; (b) gasoline; (c) gas; and (d) coke versus contact time as a function of temperature.
M.A. Abul-Hamayel / Fuel 82 (2003) 11131118 1115
smoothed by suitable equations. Unconverted VGO plot
was t to an equation of the form a=1 bt: A third order
and second order polynomial t gasoline and gas data,
respectively. An exponential form of equation a
x
b
was used to smooth the coke versus time data. Fig. 2 shows
the smoothed curves for VGO, gasoline, gas and coke at
550 8C. The statistical errors in smoothing the experimental
data were within 2%. The equations chosen after extensive
trials are similar to those proposed by Wallenstien and
Alkemade [16] after detailed statistical analysis. Since
the reaction was carried out at higher temperatures than
conventional, a few experimental runs were done without
the catalyst (contact time equal to zero) to account for
thermal cracking. Data needed for modeling purposes (Y
1
;
Y
2
; Y
3
; Y
4
and the rates dY
1
; dY
2
; dY
3
and dY
4
; in Eqs. (1)
(4)) were obtained from the aforementioned equations that
smooth the experimental data.
Fig. 3(a) compares the weight fraction of VGO in
product at different temperatures. At lower contact times,
weight fraction of VGO decreases (conversion increases) as
Table 2
Kinetic constants and activation energies
Temperature (8C) k
12
(wt. frac. h)
21
k
13
(wt. frac. h)
21
k
14
(wt. frac. h)
21
k
23
(h
21
) k
24
(h
21
) a (h
21
)
550 121 35 21 12 3 92
600 446 108 25 52 8 212
650 743 158 60 104 20 281
E (kcal/mol) 28 19 15 34 30 18
Fig. 4. Arrhenius plots of rate constants k
12
; k
13
and k
14
: Fig. 5. Arrhenius plots of rate constants k
23
; k
24
and a:
M.A. Abul-Hamayel / Fuel 82 (2003) 11131118 1116
the temperature is increased from 550 to 650 8C. At contact
times higher than 0.03 h
21
, a conversion of 80% is obtained
at 600 8C and it is not increased as the temperature is
increased to 650 8C. This is due to the fact that at higher
temperature and higher contact time, thermal cracking,
which is thermodynamically limited, is the dominant
reaction. It is seen from Fig. 3(b), the gasoline yield at
600 8C is always higher than that at 550 8C within the
contact time of 0.035 h. The reaction at 650 8C allowed
much lower gasoline yield than three at 600 and 550 8C after
0.005 and 0.015 h, respectively. At contact time of higher
than 0.035 h
21
, gasoline yield is lower at higher tempera-
ture. This decrease in gasoline yield is compensated by
higher gas and coke yields, as are seen in Fig. 3(c) and (d).
3.2. Kinetic parameter estimation
Non-linear regression analysis using least squares
techniques was done for Eq. (1), to determine the cracking
rate constant K; the order of reaction n and the deactivation
rate constant a: The VGO cracking was found to follow
second order kinetics as is conventionally assumed.
The value of a; deactivation constant, obtained from Eq.
(1) was used in Eq. (2) (4) and regression analysis was
performed to obtain the kinetic parameters. k
12
and k
2
were
obtained from the regression analysis of Eq. (2). k
13
and k
23
were obtained from the regression analysis of Eq. (3). k
14
and k
24
were calculated by difference. The kinetic constants
at three temperatures were determined form regression
analysis of rate Eqs. (1) (3) using data obtained from
experiments performed at 550, 600 and 650 8C. Table 2
presents the kinetic constants for the reactions VGO to
gasoline k
12
; VGO to gas k
13
; VGO to coke k
14
;
gasoline to gas k
23
and gasoline to coke k
24
: Activation
energies of the reactions involved were calculated from the
Arrhenius plots of respective rate constants. These plots are
shown in Figs. 4 and 5. Table 3 compares the activation
energies determined in this work with those found in the
literature [14]. Activation energies are reasonably close to
the values found in the literature, except for the activation
energy of the reaction, gasoline to gas. Activation energy
depends on the type of feed, and catalyst used and may vary.
Reconciliation diagram, Fig. 6, shows that yields of product
are predicted reasonably accurately. R
2
value of this plot
was 0.9998. The yields were predicted using kinetic
parameters estimated by consecutive regression analysis
of rate Eqs. (2) (4) instead of conventional simultaneous
regression. Mathematical complexity is avoided and using
consecutive regression instead of simultaneous regression
eliminates errors involved due to initial guess sensitivity.
4. Conclusion
A non-conventional methodology has been successfully
demonstrated to estimate the kinetic parameters of HS-FCC
process. Instead of the traditional approach of simultaneous
regression analysis of the rate equations, consecutive
regression of each rate equation was performed. Prediction
of product yields is reasonably accurate. Kinetic parameters
and activation energies compare well with the literature.
Acknowledgements
The author acknowledges the contribution of M.A. Bari
Siddiqui and wishes to acknowledge the support of the
Research Institute at King Fahd University of Petroleum and
Minerals (KFUPM), Dhahran, Saudi Arabia in this research
work.
References
[1] Avidan A, Shinnar R. Development of catalytic cracking technology.
Ind Engng Chem Res 1990;29:93142.
[2] Chapin L, Letzsch W. Deep catalytic cracking to maximize olen
production. Presented at the National Petroleum Reners Association
(NPRA) Annual meeting, San Antonio, TX; 1994.
Table 3
Comparison of activation energies with literature values
Reaction step E (kcal/mol) (present work) E (kcal/mol) (literature)
VGO to gasoline 28 1036
VGO to gas 24 1221
VGO to coke 16 715
Gasoline to gas 34 1315
Gasoline to coke 30 1627
Fig. 6. Predicted versus experimental yields of gasoline, gas and coke.
M.A. Abul-Hamayel / Fuel 82 (2003) 11131118 1117
[3] Corella J, Frances E. Analysis of the riser reactor of a uid cracking
unit. ACS Symposium Series, Fluid Catalytic Cracking II, vol. 452.;
1991. p. 16582.
[4] Blasetti A. Multitubular reactor for catalytic cracking of hydro-
carbons. Design and kinetic modeling. PhD Dissertation. University
of Western Ontario, London, Ont., Canada; 1994.
[5] Weekman VM. A model of catalytic cracking conversion in xed,
moving and uid-bed reactors. Ind Engng Chem Prod Res Dev 1968;
7:905.
[6] Yen LC, et al. Reaction kinetic correlations for predicting coke yield
in uid cracking. Katalistics Eigth Annual Fluid Vat Cracking
Symposium, Budapest, Hungary; 1987. p. 7:17.
[7] Lee LS, et al. Four lump kinetic model for FCC process. Can J Chem
Engng 1989;67:6159.
[8] John TM, Wojciechowsky BW. On identifying the primary and
secondary products of the catalytic cracking of neutral distillates.
J Catal 1975;37:24050.
[9] Jacob SM, et al. Lumping and reaction scheme for catalytic cracking.
AICHE J 1976;22(4):70113.
[10] Takatsuka T, et al. A reaction model for uidized-bed catalytic
cracking of residual oil. Int Chem Engng 1987;27(1):10716.
[11] Coxson PG, Bischoff KB. Lumping strategy. 1. Introduction
techniques and applications of cluster analysis. Ind Engng Chem
Res 1987;26:123948.
[12] Larocca M, Ng S, de Lasa H. Fast catalytic cracking of heavy gas
oils: modeling coke deactivation. Ind Engng Chem Res 1990;29:
17180.
[13] Olivera LL, Biscaia EC. Catalytic cracking kinetic models, parameter
estimation and model evaluation. Ind Engng Chem Res 1989;28:
26471.
[14] Ancheyta-Juarez J, et al. A strategy for kinetic parameter estimation in
the uid catalytic cracking process. Ind Engng Chem Res 1997;
36(12):51704.
[15] Ancheyta-Juarez J, et al. A simple method for estimating gasoline, gas
and coke yields in FCC process. Energy Fuels 2000;14:3739.
[16] Wallenstein D, Akemade U. Modeling of selectivity data obtained
from micro-activity testing of FCC catalysts. Appl Catal 1996;137:
3754.
M.A. Abul-Hamayel / Fuel 82 (2003) 11131118 1118