You are on page 1of 12

02638762/97/$10.00+0.

00
q Institution of Chemical Engineers

MODELLING AND CONTROL OF A RISER TYPE FLUID


CATALYTIC CRACKING (FCC) UNIT
H. ALI, S. ROHANI and J. P. CORRIOU*
Department of Chemical Engineering, University Of Saskatchewan, Saskatoon, Canada
*Laboratoire des Sciences du Genie Chimique, Nancy Cedex, France

model for a uid catalytic cracking (FCC) unit which describes the dynamic behaviour
of the riser, particle separator vessel, and the regenerator is developed. The model
consists of coupled ordinary differential equations. This facilitates the solution of the
equations and makes the model particularly suitable for control studies. A sensitivity study is
carried out to determine the interactions between the three controlled and manipulated
variables and the elements of the Bristol1 relative gain array matrix. The relative gain array
analysis suggested that the temperature at the top of the riser, the pressure drop between the
particle separator vessel and the regenerator, and the catalyst holdup in the particle separator
vessel should be controlled by manipulation of the ow rates of the regenerated catalyst, ue
gas from the regenerator, and the spent catalyst leaving the particle separator vessel,
respectively. Three PI-controllers were used to achieve reasonable control of the process.
Keywords: dynamic modelling; FCC unit; riser-regenerator modelling; FCC control;
numerical simulation

and El-Hennawi7 and Elshishini and Elnashaie8 described the


kinetics of cracking and recognized the two phase nature of
the reactor and regenerator uidized beds. The cracking
reactor investigated in these models, is a uidized bed instead
of a transported bed (riser) as exists in most of the modern
FCC units. Arandes and de Lasa9 incorporated the kinetics of
coke combustion over cracking catalysts. These models are
considered valuable tools for the simulation and investigation
of the multiplicity behaviour of the FCC units over their
operating range.
The steady-state analysis is not suf cient for the
development and testing of control schemes. Nevertheless,
very few dynamic modelling and simulation investigations
have been carried out on FCC units. McFarlane et al.1 0
presented a dynamic simulator which includes a number of
empirical correlations to be used to capture the major
dynamics of the model IV reactor regenerator. Elnashaie
and Elshishini1 1 extended their steady-state model to an
unsteady state model and investigated the dynamic
responses of the unit in open-loop and closed-loop feedback
controlled modes. Lopez-Isunza and Ruiz-Martinez1 2
investigated the dynamic response of a riser type FCC
unit. In this model, the open-loop transient behaviour was
used to nd the optimum operating conditions that give
maximum gasoline yield in the open-loop mode. Zheng1 3
presented another dynamic model for the riser type FCC
unit. This model simulates the start-up, shut-down and
routine operation of the unit.
Grosdidier et al.1 4 discuss the implementation of a threelevel hierarchy control system on a commercial FCC unit
including the main fractionator and the air and wet gas
compressors. PI controllers are used for the lowest level

INTRODUCTION
Fluid catalytic cracking process is one of the most
extensively used operations for the conversion of gas oil
and certain atmospheric residues to higher octane gasoline.
The unit consists of two reactors, the riser reactor, where
almost all the endothermic cracking reactions and coke
deposition on the catalyst occur, and the regenerator reactor,
where air is used to burn off coke. The regeneration process,
in addition to reactivating the catalyst pellets, provides the
heat required by the endothermic cracking reactions. The
development of new, highly active cracking catalysts, and
the introduction of the additives which greatly enhance the
productivity and the selectivity of the catalyst, allow the
cracking reactions to be completed in the riser. The particle
separator vessel acts as a disengaging chamber to separate
the catalyst from the gaseous products by stripping steam.
The complexity of the FCC unit, from both the process
modelling and control points of view, is attributed to the
strong interactions between the riser and the regenerator
reactors, and the uncertainty in the cracking reactions, coke
deposition and coke burning kinetics. This complexity
together with the important economic incentives, provide
reasons for the extensive research work being performed on
the design and operation of the FCC units.
The literature is relatively rich in the modelling and
simulation investigations of the uid catalytic cracking
units2 6 . Most of these investigations are based on empirical
or semi-empirical models, and present a number of con icting views. The empirical nature of these models limits their
predictive capabilities, and their use in studying and
analysing the behaviour of industrial FCC units. Elnashaie
401

402

ALI et al.

control and the commercially available IDCOM-SETPOITS


model predictive control software is used for the second
level control. The highest level control consists of a steadystate plant wide optimization algorithm which incorporates
economical and operational constraints. Six manipulated
variables are used to control seven controlled variables.
Process identi cation is used to obtain linear rst and
second order dynamics with and without time delays among
the input and output variables. No attempt is made to derive
rigorous models based on rst principles to explain the
process behaviour. Consequently, the results are speci c to
the unit studied and have little general signi cance.
Khandalekar and Riggs1 5 implemented a cascade controller on a type IV FCC model. The reactor temperature,
regenerator temperature and the ue gas oxygen concentration were controlled by the feed temperature entering the
riser reactor, the catalyst circulation rate, and the regenerator air ow rate, respectively. A nonlinear Generic Model
Control (GMC) algorithm was used for the high level
control and PI controllers for the regulatory level. The
model uses the three-lump kinetics for the hydrocarbon
cracking. In the riser model, an empirical correlation is used
to estimate gas oil conversion. A nonlinear least squares
method is used to obtain a model constant and the activation
energy of the reaction by minimizing the the difference
between the model predictions and the steady-state plant
conversions. Another parameter is treated as adjustable and
is continuously updated during the control runs to minimize
the difference between the model and process reactor
temperatures. Two more adjustable parameters are also
updated continuously in the regenerator model by minimizing the difference between model and process regenerator
temperature and the ue gas oxygen concentration. In brief,
the use of a three-lump kinetics scheme and three adjustable
parameters limit the applicability of the model.
Arbel et al.1 6 ,1 7 , in a two part article, developed a model
for the current generation FCC units and studied the stability
and existence of multiple steady-states of such units. Their
model predicts both the steady-state and dynamic behaviour
of the FCC units. It incorporates a complete description of
CO and CO2 combustion kinetics (partial and full combustion) including the effects of combustion promoters. The
reactor kinetics are described by the ten-lump kinetic model
of Gross et al.1 8 . The model assumes plug ow, quasi
steady-state adiabatic operation with no slip velocity
between the catalyst particle and the gas phase and a
constant super cial velocity in the riser. The stripper is
modelled as a well-mixed tank with a linear stripping
function and a constant temperature drop. The regenerator is
modelled as a dense bed and a dilute phase. The gas is
assumed to be in plug ow and the catalyst in the dense bed
is well-mixed. The entrained catalyst in the dilute phase is
assumed to be in plug ow. Quasi steady-state is assumed
for the gas composition and the bed properties. The effects
of uncertainties and new catalyst are also included in the
model. The stability of new generation FCC units is
discussed and at least three steady-states are identi ed.
The upper and the lower steady-states are stable and the
intermediate one is always unstable. Operation with complete
combustion results in a stable condition, while the partial
combustion mode can lead to the unstable steady-state.
The present study aims at developing a reliable and
simple model that limits the use of empirical expressions to

Figure 1. Simulation schematic diagram of the uid catalytic cracking unit.

a minimum and retains the dominant dynamic characteristics of the unit. The selection of this type of the uid
catalytic cracking process is due to its popularity in industry.
The kinetic parameters for coke burning over cracking
catalysts obtained by Morely and de Lasa1 9 - 2 0 were used.
There are ve kinetic schemes available in the literature for
the cracking reactions. Two schemes were investigated in a
previous contribution2 1 , the three-lump scheme2 2 and the
four-lump scheme2 3 . Our conclusion was that the use of the
four-lump scheme gives more reliable and better prediction
of the plant data.
A sensitivity study utilizing the open-loop responses of
major process variables is carried out to verify the
interactions between the three controlled and manipulated
variables and the elements of the Bristol s relative gain
array matrix1 . The calculated values of the relative gain
array showed that the existing control con guration at the
unit under investigation has the appropriate pairing of the
manipulated variables and the controlled variables. Three
PI-controllers with parameters estimated by Lee s method2 4
are used to evaluate the control scheme. In what follows a
brief description of the model and the open-loop and closedloop simulation results are presented.
THE MODEL
The FCC unit model is based on the schematic ow
diagram presented in Figure 1. It is assumed that gas oil is
converted to gasoline range hydrocarbons, light gases and
coke in the riser reactor, which is considered to be a
transported bed. The uidized bed (the particle separator
vessel) immediately above the riser acts as a disengaging
chamber where vapour products and heavy components are
separated from the catalyst using stripping steam. It is
assumed that the stripping process completely removes the
hydrocarbon gases adsorbed inside the pellets before the
spent catalyst is sent to the regenerator. The ow rate of the
stripping steam is small compared to those of the feed oil
Trans IChemE, Vol 75, Part A, May 1997

MODELLING AND CONTROL OF A RISER TYPE FLUID CATALYTIC CRACKING (FCC) UNIT

and the catalyst (the ratio is usually less than 0.25%).


Therefore, the effect of the stripping steam on the energy
balance of the particle separator can be neglected. The
regenerator is divided into a dense region and a dilute
region. The dense region is further divided into two phases,
an emulsion phase and a bubble phase. The effect of the
dilute region (freeboard region) is not included in the
model.

403

four-lump reaction scheme is as follows:

The Riser Reactor


The riser reactor can be divided into three zones. Fresh
gas oil is brought into contact with the hot regenerated
catalyst at the entrance of the riser which leads to the
vaporization of gas oil. The inlet zone is considered to be the
most complex part of the riser. This is attributed to the
presence of high turbulence, large temperature and
concentration gradients, and ow inhomogeneity. In
modern FCC units, studies have shown that a continuous
liquid stream owing through a number of nozzles coats the
catalyst particles and provides an intimate contact between
the feed and the catalyst pellets which leads to rapid
vaporization of the feed. According to plant data (Consumers Co-operative Re neries Ltd., Regina, SK)2 5 , it
takes about 0.1 sec to fully vaporize the feed. This time
represents about 3% of the mixture residence time in the
riser. Therefore, it is justi able to assume instantaneous
vaporization of the feed. Vaporized feed then pneumatically
conveys the solid particles from the bottom to the top of the
riser. An energy balance equation is developed to describe
the heat exchange between the hot regenerated catalyst and
the gas oil feed. The energy balance at steady-state can be
used to determine the dimensionless inlet temperature to the
riser, TR (0), by:
FsR cp s [Td (t) - TR (0)] = FgR [cp l (TB - Tf )
D Hvap
+ t + cpgR (TR (0) - TB )]
ref

(1)

A detailed three-dimensional two-phase modelling study


of the ow patterns and heat transfer in FCC riser reactors
was performed by Theologes and Markatos2 6 . They
concluded that the overall performance of the riser can be
predicted using simple one-dimensional mass, energy, and
chemical species balances. Moreover, the presence of the
high ef ciency feed injection systems in modern units
justi es the assumption of plug ow in the riser. The feed
phase change and the molar expansion as the reactions
proceed in the intermediate and the nal zone of the riser,
results in typically a 34 fold increase in the gas super cial
velocity along the riser. The change in gas super cial
velocity due to the phase change was considered in this
model, whereas changes due to molar expansion were not
accounted for. This contributes to the discrepancy between
the model predictions and the plant data. A xed value for
the gas super cial velocity, calculated immediately after the
inlet zone, is used throughout the riser intermediate and top
sections.
Most of the previous FCC modelling attempts have
employed the three-lump scheme proposed by Weekman
and Nace 2 2 . In this paper, the reaction scheme developed by
Lee et al.2 3 is used to simulate the cracking reactions. The
Trans IChemE, Vol 75, Part A, May 1997

where A, B, C, and D represent gas oil, gasoline, coke, and


light hydrocarbon gases respectively. The riser bed acts as a
transported bed, with a high combined stream velocity and a
short residence time in the order of a few seconds. Thus it
can be assumed that the dynamics of the riser in comparison
with the coke burning and temperature changes in the dense
phase of the regenerator are negligible. Therefore, the mass
and energy balance equations in the riser are considered at
quasi steady-state. In addition, the concentrations of various
hydrocarbon gases in the riser are normalized with respect
to the gas oil feed concentration, and the riser temperature is
normalized with respect to the steady-state regenerator
dense phase temperature (tr e f ). Based on these assumptions,
applying the conservation principles and assuming plug
ow in the riser, the mass and energy balances in
dimensionless form are:
Gas oil:
dyA w R AR e gR LR q gR
[kAB + kAC + kAD ] y2A = 0
(2)
+
dZ
FgR
Gasoline:
dyB w R AR e gR LR q
dZ +
FgR

gR

2
[(kBC + kBD )yB - kAB yA ] = 0

Light Hydrocarbon Gases:


dyD w R AR e gR LR q gR
[kBD yB + kAD y2A ] = 0
dZ FgR

(3)

(4)

Coke:
dyCR w R AR e gR LR q
dZ FgR

gR

2
[kBC yB + kAC yA ] = 0

(5)

Energy Balance:
w R e gR q gR LR AR
dTR
+ (F cp F cp )t
dZ
SR
s +
gR
gR ref
2
(yA [kAB (D HRAB ) + kAC (D HRAC )

(6)

+ kAD (D HRAD )] + yB [kBC (D HRBC )


+ kBD (D HRBD )])+ HIR = 0
with the boundary conditions, at Z=0:
yA (0) = 1
yi (0) = 0

yCR (0) = yCG (s )

= B and D

TR (0) = TRf

The regenerated catalyst ow rate, Fs R , varies based on


the regenerator pressure. This ow rate is calculated in

404

ALI et al.

terms of the pressure drop between the regenerator and the


riser, PG - Pf , and the valve opening, X1 . The equation for
the regenerated catalyst ow rate is as follows:
FsR

----------------

= K X P - P
1

(7)

-1

where K1 (=310 kg sec atm

- 0 .5

) is the valve constant.

The Particle Separator Vessel


In a stacked UOP FCC con guration, the particle
separator vessel or the steam stripper is the upper uidized
bed on the top of the riser. In this vessel, the vapour products
and heavy components are separated from the catalyst using
stripping steam. The ow rate of the stripping steam is small
and its effect on the energy balance of the reactor is assumed
to be negligible. Accordingly, it is assumed that the sole
effect of the particle separator is to introduce a time lag
between the riser outlet and the regenerator. This is
modelled as an ideal continuous stirred tank. A material
balance incorporating the catalyst level inside the particle
separator vessel and normalized with respect to the steadystate catalyst level in the regenerator bed describes the
particle separator dynamics:
dhrv
1
F )
(F
(8)
ds = Arv (1 - e rv )q S Ua sR - sG
with the initial condition at s = 0:
hrv (0) = hrv,ss
The ow rate of the spent catalyst to the regenerator, FsG ,
is calculated from the following equation:
FsG

-----------------

= K X P - P
2

(9)

-1

- 0 .5

where K2 (=310 kg sec atm ) is the spent catalyst valve


constant, and X2 is the spent catalyst valve position.
The Regenerator Reactor
The regenerator reactor consists of two regions, the dense
region and the freeboard dilute region. The freeboard region
is de ned as the section of the vessel between the top
surface of the dense region and the exit of the regenerator
vessel. In this section, the gas stream carries some catalyst
particles. The amount of solids entrainment in this region is
usually very small compared to the total amount of catalyst
retained in the regenerator vessel. Most of the coke on the
catalyst pellets in this region have already been combusted
in the dense region. That is, full combustion of coke to
cabon dioxide in the dense regions is assumed. Accordingly,
the effect of the freeboard region on the overall performance
of the regenerator is ignored.
The dense region is further divided into two phases (Kunii
and Levenspiel2 7 ): a bubble phase and an emulsion phase.
The latter is assumed to be like a bed at incipient uidization
(air ows at minimum uidization velocity Um f ). The
bubbles move in plug ow and exchange mass and heat with
the emulsion phase. In the emulsion phase the air
distributors, the spent catalyst, and the cyclones recycle
pipes produce enough turbulence to justify a continuous
stirred tank reactor behaviour (see Figure 1).
The minimum uidization velocity, Um f , of the emulsion
phase can be calculated from the following equation

proposed by Kunii and Levenspiel2 7 :


2
1.75 dp Umf q gG
150(1 - e mf ) dp Umf q
+
3
l
l
(e mf ) h s
(e mf )2 h 2s

d 3p q

gG

(q

-q

gG

gG

)g

(10)
l
where the volumetric ow rate in the emulsion phase, GI G ,
and in the bubble phase, GC G , are:
GIG = Umf AG
(11)
2

GCG = (Ua AG ) - GIG


(12)
and the volume fraction of bubbles is:
G t
e bG = CG a
(13)
A G LG
The regenerator catalyst bed level is calculated by
performing a mass balance on the catalyst in the
regenerator. The resulting dimensionless equation is as
following:
dLG1
FsG - FsR
(14)
ds = AG (1 - e dG )q s Ua
with the initial condition at s = 0:
LG1 (0) = 1
The regenerator pressure is calculated assuming ideal gas
behaviour. This is justi able since the operating pressure is
about 2.75 - 3 atm (the compressibility factor was found to
be close to unity). Furthermore, in a uidized bed the
pressure drop across the bed is within 5% of the total
pressure and hence it is ignored. The volume of gases inside
the regenerator is calculated as follows:
VG

=A

GG

ZG - AGG LG (1 - e

dG

(15)

The total number of moles of gases inside the regenerator is


calculated as follows:
V G q gG
NGT =
(16)
MWgG
The change in the regenerator pressure with respect to
time is calculated based on the ideal gas law from the
following equation:
dPG
ds

=V

NGT

dTb
dN
Tb GT
+
ds
ds

PG dV G

- V ds
G

(17)

The ow rate of the ue gases to the stack, Ff g , is calculated


based on the regenerator pressure:
Ffg

-------------------

= K X P - P
3

(18)

atm

-1

- 0 .5

where K3 (=1.5 kg sec atm ) is the ue gas valve


constant, and X3 is the corresponding valve opening.
The kinetics of coke burning are simulated using the work
done by Morley and de Lasa 1 9 - 2 0 . It is assumed that coke
only consists of carbon and the overall rate is controlled by
the intrinsic kinetics of combustion. The suggested expressions for the coke burning, catalytic and homogeneous CO
combustion rates are:
kW C
rc = c CG O2d
(19)
MWc
rcod
rcob

=k
=k

cod

0 .5
CCOd CO2d

(20)

cob

0 .5
0.5
CCOb CH2Ob
C02b

(21)

Trans IChemE, Vol 75, Part A, May 1997

MODELLING AND CONTROL OF A RISER TYPE FLUID CATALYTIC CRACKING (FCC) UNIT
The assumptions imposed on the model equations have
been previously used by Elnashaie and Elshishini1 1 and are
also recommended by Lopez-Isunza and Ruiz-Martinez1 2 .
The energy and coke combustion dynamic terms in the
emulsion phase are taken into consideration, while the
dynamic terms associated with the bubble phase are
assumed negligible due to low solids density in this phase.
Therefore, the bubble phase is considered thermally at quasi
steady-state. Moreover, the homogenous CO combustion
reaction taking place in the bubble phase is assumed to be
negligible in comparison with the catalytic CO combustion
in the emulsion phase1 1 ,1 2 , 2 0 . Based on these assumptions,
the bubble phase equations were solved analytically.
Furthermore, the integral terms appearing in the equations
describing the emulsion phase were evaluated analytically.
The concentrations of various gaseous species in the
regenerator were normalized with respect to the oxygen
feed concentration, and both the bubble phase and the
emulsion phase temperatures were normalized with respect
to the emulsion phase temperature. The nal form of the
regenerator model equations is:
Coke balance(Emulsion phase):
dyCG
LGss
(FsG yCR - FsR yCG
C ky y
ds = AG LG1 (1 - e dG )q S Ua - Ua 02f c CG O2d
yCG dLG1
G1 ds

-L

(22)

= y + eCOd

a1 z

- yCOd )

(yCOf

(23)

Emulsion phase:
a
[GIG + GCG (1 - e - 1 )](yCOf

- 2AG LG1 LGssq b k 9COd yCOd y


AG LG1 LGss q s (1 - e dG )FgR k 9C
yCG yO2d = 0
+
MW F
C

(24)

sR

CO2 balance:
Bubble phase:

=y

CO2d

a z
+ e - 1 (yCO2f - yCO2d )

(25)

Emulsion phase:
a
[GIG + GCG (1 - e - 1 )](yCO2f

- yCO2d )

.
- 2AG LG1 LGssLGss q b k 9COd yCOd yO2d
AG LG1 LGss q s (1 - e dG )FgR k C9 9
yCG YO2d = 0
MW F
05

(26)

sR

O2 balance:
Bubble phase:
yO2b

= y + eO2d

Length (m)

Diameter (m)

33
11

0.8
5.8

Riser Reactor
Regenerator Reactor

Energy balance:
Bubble phase:
Tb

= T + ed

a hz

(Tairf

- Td )

(29)

Emulsion phase:
dTd [GIG + GCG (e a h z - 1)]q gG cp gG
(Tairf
ds =
AG LG1 (1 - e dG )q s Ua cpS

- Td )

FsG

+ A L (1 e )q U (TR - Td )
G G1
- dG s a
F L C

gR Gss O2f C
- HlG + F U cp t yCG yO2d (D HRC )
sR a
S ref

(30)

1.5
2LGss q b CO2f

+ (1 e )q U cp t
- dG S a S ref
.5
* kCOd yCOd y0O2d
(D HRCOd ) -

Td dLG1
LG1 ds

with initial conditions, at s = 0:

a 1z

(yO2f

Td (0) = TdSS

THE OPEN-LOOP RESPONSE OF THE FCC UNIT

- yCOd )
0 .5
C0d

yCO2b

Table 1. FCCU dimensions.

yCG (0) = yCGSS

CO balance:
Bubble phase:
yCOb

405

- yO2d )

(27)

Plant data from the Universal Oil Products (UOP) riser


type uid catalytic cracking unit located at Consumer s CoOperative Re neries Ltd., Regina, Saskatchewan2 5 , were
used to verify the model. The unit dimensions and operating
conditions are listed in Tables 1 and 2 respectively. The
physical properties of the reacting species and the catalyst
and the heats of reactions are listed in Table 3. The air and
hydrocarbons physical properties are calculated from the
API Data Book2 8 and the Chemical Engineering Handbook2 9 . The kinetic parameters of the cracking reactions for
the four-lump kinetic scheme were obtained from Lee
et al.2 3 and the coke combustion kinetics from Morley and
de Lasa1 9 - 2 0 . These are presented in Tables 4 and 5,
respectively.
The non-linear algebraic equations were solved by the
Newton-Raphson method and the set of ordinary differential
equations was solved numerically by Gear s method3 0 with
adaptive step size.
The regenerator level equation is solved rst. The
estimated value of the regenerator bed catalyst level is
then used to determine the volume of gases, regenerator

Emulsion phase:
[GIG + GCG (1 - e - a 1 )](yO2f

Table 2. FCCU operating conditions.

yO2d )

+ 2AG LG1 LGssq b kC9 Od yCOd y

0 .5
O2d

AG LG1 LGss q s (1 - e dG )FgR kC9 9 9


yCG yO2d
+
MWC FsR
Trans IChemE, Vol 75, Part A, May 1997

(28)

=0

Gas oil
Air
Catalyst

Flow Rate (kg sec - 1 )

Inlet Temperature (K)

20
16
144

494
378

406

ALI et al.

Table 3. Physical properties and heats of reactions of the reacting species.


Hydrocarbons:
Density:
Speci c heat (gas):
Speci c heat (liquid):
Heat of vaporization:
Vaporization temperature:
Heat of reactions:
gas oil to gasoline
gas oil to light hydrocarbons
gas oil to coke
gasoline to light hydrocabons
gasoline to coke

8.4 kg m 3
3.299 kJ kg- 1 K- 1
2.671 kJ kg- 1 K- 1
156 kJ kg- 1
698 K
710 Cal g- 1
2328 Cal g- 1
5693 Cal g- 1
1618 Cal g- 1
5403 Cal g- 1

Air:
Density:
Speci c heat:
Bubble-emulsion mass transfer coeff.:
Bubble-emulsion heat transfer coeff.:

1.03 kg m 3
1.206 kJ kg- 1 K- 1
0.5 sec 1
0.84 kJ sec- 1 m 2 K- 1

Catalyst:
Bulk density:
dp (particle size):
Heat capacity:
Inventory in the regenerator:

970 kg m 3
75 10 6 m
1.15 kJ kg- 1 K- 1
5000070000 kg

pressure, regenerator temperature, coke on the regenerated


catalyst, and the ue gas stream composition, respectively.
The regenerated catalyst ow rate to the riser is then
determined and the riser reactor mass and energy balances
are solved. The catalyst hold-up and the level of the spent
catalyst bed in the particle separator vessel are evaluated,
from which the spent catalyst ow rate entering the
regenerator vessel is determined.
The Steady-State Results
The deviations between the model predictions and the
plant data are shown in Table 6. The model predictions of
the gasoline yield and the regenerator temperature are in
closer agreement with the plant data. The percent deviations
of the oxygen concentration at the exit of the regenerator
from the plant data obtained is relatively high. The steadystate temperature and the reactant and products pro les are
shown in Figures 2 and 3, respectively.
The Dynamic Response
In this section the responses of the regenerated catalyst,
spent catalyst, and ue gas ow rates due to a step increase
in their valve positions are presented. The subsequent
changes in the riser temperature, the catalyst bed level in the
particle separator vessel, and the pressure difference
between the regenerator and the particle separator vessels
are also discussed. The effect of introducing a step increase

Table 5. Kinetic parameters (coke burning) (Morley and de Lasa19).


Reactions
Coke combustion
CO catalytic
Combustion

Pre-exponential
constant

Activation Energy
(kJ / mol1)

1.4 108 (m3 kmol- 1 sec- 1 )


247.75
((m3)1.5 kmol0.5 sec - 1 kg- 1 )

224.99
70.74

followed by a step decrease to the three valve openings is


also investigated. Dynamic simulations of the open-loop
system to changes in the regenerated catalyst ow rate, the
riser temperature, the spent catalyst ow rate, the catalyst
bed level in the particle separator vessel, the ue gas ow
rate and the pressure difference between the regenerator and
the particle separator vessel are illustrated.
Response to Changes in the Regenerated Catalyst Valve
Stem Position
Figure 4 shows the response of the regenerated catalyst
rate, and the riser temperature to a step increase in the
regenerated catalyst valve position with a magnitude of 2%
and a duration of 11 hrs followed by a step decrease with a
magnitude of 5%.
The increase in the regenerated catalyst valve opening
results in a higher regenerated catalyst ow rate, according
to equation (7). On the other hand, the increase in the
regenerated catalyst ow rate is associated with a lower
catalyst bed level (equation (14)). As the catalyst level in the
regenerator drops, the volume occupied by the gases inside
the regenerator vessel increases. In accordance with
equation (17), the regenerator pressure decreases which in
turn reduces the regenerated catalyst ow rate. The nal
steady state value of the regenerated catalyst ow rate is
higher than the original value before introducing the step
increase.
The higher regenerated catalyst ow rate due to the
increase in the valve opening is associated with more heat
input and active sites available for the endothermic cracking
reactions. Thus, the cracking reactions are accelerated.
Initially, the heat associated with the catalyst dominates,
and the temperature at the exit of the riser increases
(Figure 4a). The higher riser temperature together with the
presence of more catalyst favour the cracking reactions. The
increased reaction rates produce more coke, and the
combustion of higher quantities of coke increases the
regenerator temperature. Subsequently, this results in a
slight elevation in the regenerator pressure which compensates the effect of the level drop and maintains the
regenerator pressure at a steady value.
Table 6. Models predictions and percent deviation from reported plant data.

Table 4. Kinetic parameters (four-lumped scheme) (Lee et al.23).


Reactions
Gas oil to gasoline
Gas oil to coke
Gas oil to light gases
Gasoline to coke
Gasoline to light gases

Pre-exponential
constant

Activation Energy
(kJ/mol- 1 )

554 (m3 kg- 1 sec- 1 )


10.45 (m3 kg- 1 sec - 1 )
1895.4 (m3 kg- 1 sec - 1 )
2210.28 (sec- 1 )
0.904 (sec- 1 )

68.2495
64.5750
89.2164
115.458
52.7184

Model
Riser reactor
Gasoline yield, wt%
Coke wt%
Riser temp, K
Regenerator reactor
Regen. temp, K
CO2, Mol%
O2, Mol%

Plant

%dev.

42.7
6.1
749

43.9
5.8
795

2.7
4.4
5.8

939
17.4
3.7

960
17.7
3.0

2.2
1.6
18.5

Trans IChemE, Vol 75, Part A, May 1997

MODELLING AND CONTROL OF A RISER TYPE FLUID CATALYTIC CRACKING (FCC) UNIT

407

Figure 2. Steady-state temperature pro le in the riser.

On the other hand, the endothermic cracking reactions


cause a reduction in the riser temperature. Furthermore, the
increase in the fresh regenerated catalyst ow rate results in
a higher catalyst level in the particle separator vessel. As the
catalyst level in the particle separator increases, it exerts
hydrostatic pressure and the spent catalyst ow rate
increases at the same valve opening (equation (9)). As a
result of lower riser temperature and higher rates of spent
catalyst owing back to the regenerator at lower temperature, the regenerator temperature and pressure decline. This
continues until they eventually settle at steady-state values.
For the step decrease in the valve opening, the
regenerated catalyst ow rate oscillates before reaching
steady-state. Initially, the closure of the valve results in
>

Figure 3. Steady-state components pro les in the riser.

Trans IChemE, Vol 75, Part A, May 1997

Figure 4. Transient response of the regenerated catalyst ow rate and the


riser temperature to step increase and decrease in the regenerated catalyst
valve position.

lower catalyst ow rate which has two effects. First the


amount of catalyst and the heat available for the cracking
reactions decreases. This in turn lowers the cracking
reaction rates. On the other hand, the amount of catalyst
withdrawn from the regenerator is reduced and the catalyst
level increases. The higher catalyst level in the regenerator
reduces the volume of the gases inside the vessel which,
subsequently, elevates the regenerator pressure. This results
in an increase in the regenerated catalyst ow rate (equation
(4)). As a result of the higher regenerated catalyst
withdrawal rate from the regenerator, the bed level inside
the vessel drops (equation (14)), the gas volume expands
and the regenerator pressure is lowered. Again the decrease
in the regenerator pressure decreases the regenerated catalyst
ow. This cycle is repeated, and as it proceeds the magnitude
of the oscillation decreases until eventually the regenerated
catalyst ow rate settles at a new steady state value.
An observation obtained from Figures 4a and 4b is that
the dynamic responses of either the regenerated catalyst
ow rate or the riser exit temperature to a step decrease in
the regenerated catalyst valve opening are not mirror images
of those obtained from a step increase which demonstrate
the dynamic nonlinearity of the unit.
Response to Changes in the Spent Catalyst Valve Stem
Position
Figure 5 shows a step increase in the spent catalyst valve
position with a magnitude of 4% and a duration of 11 hrs
followed by a step decrease with a magnitude of 5%. The
resulting responses in the spent catalyst ow rate and the
catalyst level in the particle separator catalyst bed level to
these changes are shown in Figures 5a and 5b.

408

ALI et al.

Figure 5. Transient response of the spent catalyst ow rate and the catalyst
bed level in the particle separator vessel to step increase and decrease in the
spent catalyst valve position.

Figure 6. Transient response of the ue gases ow rate and the regenerator


pressure to step increase decrease in the ue gases valve position.

The increase in the spent catalyst valve opening results in


a higher spent catalyst ow rate (Figure 5a). Accordingly,
the catalyst level in the particle separator catalyst bed level
sharply decreases (Figure 5b). The loss of the hydrostatic
head above the valve reduces the spent catalyst ow rate. On
the other hand, as there is more catalyst coming to the
regenerator, the regenerator catalyst bed level increases.
This has more than one effect on the unit. First, the
regenerator pressure rises. Subsequently, the regenerated
catalyst ow rate increases for the same regenerated catalyst
valve opening. The higher regenerated catalyst rate to the

particle separator vessel compensates the higher withdrawal


rate from the latter and eventually the catalyst bed level
reaches a new steady-state value. Furthermore, the increase
in the regenerator pressure negates the increase in the spent
catalyst ow rate which in turn allows the catalyst bed level
in the particle separator to remain at its steady-state.
The introduction of a step increase or a step decrease to
the valve opening produces similar trends in the spent
catalyst ow rate and the catalyst hold-up in the particle
separator. This suggests that the system dynamics with
respect to changes in the spent catalyst ow rate are

Figure 7. The control con guration used for the FCC unit.

Trans IChemE, Vol 75, Part A, May 1997

MODELLING AND CONTROL OF A RISER TYPE FLUID CATALYTIC CRACKING (FCC) UNIT

409

Table 7. The PI controller parameters.


Lee s Method
Loop
Riser temperature
Particle separator
level
Pressure drop

Kc

I, min

Fine Tuned
Kc

I, min

D t, min

0.001 762.11
0.0007 101.36

0.0009
0.0003

504.32
120.3589

1.24
0.41

0.064

0.0015

8.884

0.083

3.91

approximately linear. It is also noticed from the simulation


results that the nal steady-state value of the spent catalyst
ow rate is almost the same.
Response to Changes in the Flue Gas Valve Stem Position
Figure 6 shows the response of the system to a step
increase in the ue gas valve stem position with a small
magnitude of 0.75% and a duration of 11 hrs followed by a
step decrease with a magnitude of 0.85%. Even these small
changes exert relatively large effects on the controlled
variables.
In Figure 6a, the ue gas ow rate initially increases as
the step increase is introduced. Consequently, the pressure
inside the regenerator vessel decreases (Figure 6b). However, according to equation (18), the ue gas ow rate is a
function of the regenerator pressure. Therefore, it is
expected even for a wider valve opening that the ue gas
rate drops as the regenerator pressure decreases. On the
other hand, the lower ue gas rate helps the regenerator
pressure to boost again and eventually settle at a new
steady-state value.
The ue gas ow rate exhibits a large peak before it
settles at the new steady-state value in response to an
increase in the ue gas valve opening. However, when
introducing a step decrease to the valve opening, the ue gas
ow rate does not show this peak. The regenerator pressure
pro le also does not show similar pro les in response to step
increase and decrease in the ue gas valve position. A nal
remark is that although the valve position is not the same,
the nal value of the ue gas ow rate is similar to the
original steady-state value. This is attributed to the higher
regenerator pressure which compensates for the smaller
valve opening.
SELECTION OF THE APPROPRIATE CONTROL
CONFIG URATION
The existing control scheme at the Consumer s CoOperative Re neries Ltd., Regina, Saskatchewan2 5
(Figure 7) consists of three control loops. The riser exit
temperature is controlled by the amount of hot regenerated
catalyst entering the riser. The level of catalyst bed in the
particle separator vessel is maintained by the ow rate of the
spent catalyst leaving the particle separator vessel. The nal
control loop maintains the pressure difference between the
particle separator and the regenerator vessels by manipulating the ue gas ow rate.
A sensitivity study based on Bristol s relative gain array
matrix was carried out to verify the pairing of the three
controlled and manipulated variables. Based on the results
obtained from the open-loop step increases in the three
control valves at their half-open positions the steady-state
Trans IChemE, Vol 75, Part A, May 1997

Figure 8. Closed loop response to a step increase and decrease in the riser
temperature set-point.

gain matrix, K, of the process was determined:

- 4.02 119347 23800


Hrv =
0.1
- 3565 - 655
P
0
0
D
- 147
TR

FSR

FSG

(31)

Ffg

which results in the following Bristol relative-gain array


matrix:

5.87

= - 4.87

- 4.87 0
5.87

(32)

It is noticed that the regenerated catalyst ow rate should


be used to control the riser temperature. Similarly, the spent
catalyst ow rate is the appropriate variable to control the
catalyst level in the particle separator. The ow rate of the
ue gas should be used to maintain the pressure difference
between the regenerator and the particle separator vessel.
One important observation from the relative gain array is the
strong interaction between the rst and the second loops and
no interaction with the third loop.
THE CLOSED-LOOP RESPONSE OF THE FCC UNIT
Three PI-controllers were designed using the method
proposed by Lee2 4 . The controller parameters obtained from

410

ALI et al.

Figure 9. Closed loop response to a step increase and decrease in the level
of catalyst in the particle separator vessel set-point.

Figure 10. Closed loop response to a step increase and decrease in the
regenerator pressure set-point.

IAE set-point tracking are listed in Table 7. The velocity


form of the discrete PI-controller was used:
K D t
Xi,k+ 1 = Xi,k + Kc ,i (ei,k+ 1 - ei,k ) + c,i ei,k+ 1
(33)
s I, i

constraints to overcome this problem. It is also noticed


that changes in the regenerator pressure are independent of
the other two loops. This observation con rms the pairing
suggested by the relative gain array analysis.

The controller parameters listed in Table 7 resulted in


highly oscillatory response which required ne tuning of the
controller parameters. The controller parameters after netuning are also listed in Table 7.
Figures 8 to 10 show the response of the individual
controllers to step changes in the set-point of the riser
temperature, the catalyst level in the particle separator, and
the pressure drop between the regenerator and the particle
separator vessel, respectively. The initial decrease in the
riser temperature as a result of a step increase in its set-point
is an artifact of the computational order used in this study.
An increase in the riser set-point temperature increases the
error which in turn increases the corresponding valve
opening, X1 . This will increase Fs R according to equation
(7). This increase in Fs R decreases the regenarator pressure,
PG , as is evident in Figure 8c, which initially causes a drop
in FsR , i.e. smaller ow of hot regenerated cataltst to the
riser and hence a drop in the riser temperature. Moreover,
Figures 8b and 9b indicate that the particle separator level
controller and the riser temperature controller are not able to
bring these variables to their desired set-points despite the
integral action. In fact, both controllers were saturated at
valves open position. This suggests the necessity of an
advanced multivariable controller which can handle

A simple overall dynamic model for the riser, particle


separator vessel, and the regenerator of an FCC unit was
developed. The steady-state results predicted by the model
were compared with the industrial data and showed good
agreement. Some of the distinctive features of the model
are: it incorporates the four-lump kinetic scheme to simulate
the cracking reactions, it introduces a catalytic kinetic rate
for the CO along with the coke combustion over cracking
catalysts, it does not involve any partial differential
equations, and it minimizes the use of empirical
correlations.
The model simulates the entire FCC unit and internally
calculates all the manipulated and controlled variables.
Bristol relative-gain array method supported the current
pairing of the manipulated and the controlled variables at
the Consumers Co-operative re nery Ltd., (Regina, SK)2 5 .
Strong interactions between the riser temperature and the
level of the catalyst in the particle separator vessel were
observed. The regenerator pressure drop was found to be
independent of the other two loops.

CONCLUSIONS

Trans IChemE, Vol 75, Part A, May 1997

MODELLING AND CONTROL OF A RISER TYPE FLUID CATALYTIC CRACKING (FCC) UNIT
NOMENCLATUR E
A
AG
AGG
AR
Arv
av
B
C
Cjb
Cjd
CO2f
cpl
cpgG
cpgR
cps
D
dp
ei, k
Ffg
FgR
FsG
FsR
GIG
GCG
g
HlR
HlG
h
Hrv
hrv
K
Ki
Kc,i
kij
kc
kcob
kcod
k9 c
k9 c 9
k9 c9 9
k9 cod
kg
LG
LGss
LG1
LR
MSG
MWC
MWgG
NGT
Patm
Pf
PG
PR
R
rc
rcod
rcob
Tairf
Tj
TB
TR
tref
t
ta
Ua
Umf

gas oil feed


area of the emulsion phase in the regenerator, m2
cross sectional area of the regenerator, m2
area of the riser, m2
cross sectional area of the particle separator vessel, m2
speci c hear transfer area, m2 m 3
gasoline range lump
coke lump
concentration of jth component in the bubble phase, where j=CO,
CO2, H2O, and O2 , kmol sec- 1
concentration of jth component in the emulsion phase, where
j=CO, CO2, and O2 , kmol sec - 1
concentration of the oxygen at the feed stream, kmol m 3
heat capacity of the liquid gas oil, kJ kg- 1 . K- 1
heat capacity of gases in the regenerator, kJ kg- 1 K- 1
heat capacity of gases in the riser, kJ kg- 1 K- 1
heat capacity of catalyst, kJ kg- 1 K- 1
light hydrocarbon gases lump
catalyst particles
error signal of the ith loop at the kth sampling interval
ue gases ow rate, kmol sec- 1
hydrocarbon gases mass ow rate in the riser, kg sec- 1
spent catalyst mass ow rate, kg sec- 1
regenerated catalyst mass ow rate, kg sec- 1
gas ow rate in the regenerator emulsion phase, m3 sec - 1
gas ow rate in the regenerator bubble phase, m3 sec- 1
gravitational constant, m sec 2
dimensionless heat losses from the riser
dimensionless heat losses from the regenerator
bubble to emulsion phases heat transfer coef cient, kJ m 2 sec- 1
K- 1
catalyst level inside the particle separator vessel, m
dimension less catalyst level inside the particle separator vessel
steady-state gain matrix
the valve constant , where i=1 for regenerated catalyst, 2 for spent
catalyst, 3 for ue gas
the controller proportional gain of the ith loop
reaction rate constant between species i and j
reaction rate constants for coke burning, m3 kmol- 1 sec- 1
reaction rate constant for the homogeneous CO combustion,
(m3)1.5 kmol0.5 kg of solids- 1 sec- 1
reaction rate constant for the catalytic CO
combustion,(m3)1.5 kmol 0.5. kg of solids- 1 sec- 1
(1/b +1)kc, see equation (24)
(b /b +1)k c, see equation (26)
(b +2/2b +2)kc, see equation (28)
CO2f kcod, see equation (26)
bubble and emulsion phases overall mass transfer coef cient ,
sec 1
regenerator catalyst bed level, m
steady-state regenerator catalyst bed level, m
dimensionless regenerator bed level
riser length, m
catalyst retention in the regenerator, kg
carbon molecular weight, kg kmol- 1
average molecular weight of gases in the regenerator, kg kmol- 1
total number of moles in the regenerator, kmol
atmospheric pressure, atm
riser feed pressure, atm
regenerator pressure, atm
particle separator vessel pressure, atm
universal gas constant
rate of coke combustion in the regenerator bed, kmol sec- 1
rate of catalytic CO combustion in the regenerator bed, kmol sec- 1
rate of homogeneous CO combustion in the regenerator bed,
kmol sec- 1
dimensionless air to the regenerator feed temperature
dimensionless temperature
dimensionless boiling point of gas oil
dimensionless temperature of the riser
reference temperature, K
time, sec
air space time, sec
air velocity, m sec - 1
minimum uidization velocity, m sec- 1

Trans IChemE, Vol 75, Part A, May 1997

VG
WCG
WCR
X
ya
yCG
yCR
yij
yif
Z
ZR
ZG
ZG1
z

411

volume of gases in the regenerator, m3


coke mass fraction from the regenerator, kg coke kg catalyst - 1
coke mass fraction from the riser, kg coke kg catalyst - 1
valve stem position
dimensionless weight percent of hydrocarbons in the riser, where
(a=A, B, C, and D)
coke on regenerated catalyst / feed mass rate, kg coke kg feed - 1
coke on spent catalyst / feed mass rate, kg coke kg feed- 1
dimensionless Concentration of component i in the j phase, where
(i = CO, CO2, H2O, and O2) and ( j= bubble and emulsion)
dimensionless concentration of component i at the feed, where
(i = CO, CO2, and O2)
dimensionless axial distance in the riser, ZR/LR
axial distance in the riser, m
regenerator length, m
axial distance in the regenerator bed, m
dimensionless axial distance in the regenerator, ZG1/LG

Greek letters
a 1
(AG e dG(1-ebG) LGss kg /GCG), see equation (25)
a h
(AG (1-e bG) av h LGss /GCG rgG cpgG), see equation (29)
b
intrinsic molar ratio of CO2 to CO, kmol CO2 / kmol CO
D HRAB heat of reaction for cracking of lump A to lump B, kJ kg- 1
D HRC heat of reaction for coke burning, (kJ kmol- 1 ), see equation (30)
D HRCOj heat of reaction for CO combustion in the j phase, kJ kmol- 1
sampling interval, min
D t
e gR
hydrocarbon gases void fraction in the riser
e bG
bubbles void fraction in the regenerator bed
e dG
emulsion phase void fraction in the regenerator
e mf
bed voidage at minimum uidization velocity
e rv
bed voidage in the particle separator vessel
w
catalyst activity
l
air viscosity, kg (m- 1 sec- 1 )
relative gain matrix
K
q b
catalyst bulk density, kg m 3
q gG
density of gas phase in the regenerator, kg m 3
q gR
density of gas phase in the riser, kg m 3
q s
density of catalyst particles, kg m 3
hs
catalyst particle sphericity
s
dimensionless time, t Ua /LGss
s I,i
integral time of the ith loop, min
Subscripts
b
bubble phase in the regenerator
CO
carbon monoxide
CO2
carbon dioxide
d
emulsion phase in the regenerator
f
feed
G
regenerator reactor
g
gas phase
H2O
water vapour
O2
oxygen
R
riser reactor
rv
particle separator vessel
S
catalyst solid particles
ss
steady-state

REFERENCES
1. Bristol, E. H., 1966, On a new measure of interaction for multivariable
process control, IEEE Trans Autom Control, AC-11: 133134.
2. Kurihara, H., 1967, Optimal control of uid catalytic cracking
processes, PhD Thesis, (MIT, Cambridge, USA).
3. Iscol, L., 1970, The dynamic and stability of a uid catalytic cracker,
Proc Joint Automatic Control Conf, Atlanta, Georgia, 602607.
4. Lee, W. and Kugelman, A. M., 1973, Number of steady-state operating
points and local stability of open-loop uid catalytic cracker, Ind Eng
Chem Proc Des Dev, 12 (2): 197204.
5. Lee, E. and Groves, F. R., 1985, Mathematical model of the uidized
bed catalytic cracking plant, Trans Soc Comp Sim, 2: 219236.
6. Edwards, W. M. and Kim, H. N., 1988, Multiple steady state in FCC
unit operations, Chem Eng Sci, 43: 18251830.
7. Elnashaie, S. S. E. H and El-Hennawi, I. M., 1979, Multiplicity of the
steady state in uidizied bed reactors-IV. Fluid catalytic cracking
(FCC), Chem Eng Sci, 34: 11131121.
8. Elshishini, S. S. and Elnashaie, S. S. E. H., 1990, Digital simulation of

412

9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.

ALI et al.

industrial uid catalytic cracking units I. Bifurcation and its


implications, Chem Eng Sci, 45: 553559.
Arandes, J. M. and de Lasa, H. I., 1992, Simulation and multiplicity of
steady states in uidized FCCUs, Chem Eng Sci, 47: 25352540.
McFarlane, R. C. , Reineman, R. C. , Bartee, J. F. and Georgakis, C.,
1993, Dynamic simulator for a model IV uid catalytic cracking unit,
Comp Chem Eng, 17: 275300.
Elnashaie, S. S. E. H and Elshishini, S. S., 1993, Digital simulation of
industrial uid catalytic cracking units IV. dynamic behaviour, Chem
Eng Sci, 48: 567583.
Lopez-Isunza, F. and Ruiz-Martinez, R., 1991, Dynamic modeling of a
uid catalytic cracking unit, AIChE Ann Conf, Los Angeles, USA.
Zheng, Y.-Y., 1994, Dynamic modeling and simulation of a catalytic
cracking unit, Comp Chem Eng, 18: 3944.
Grosdidier, P., Mason, A., Aitolathi, A., Heinonen, P., and Vanhamaki,
V., 1993, FCC unit reactor-regenerator control, Comp Chem Eng,
17(2): 165179.
Khandalekar, P. D., and Riggs, J. B., 1995, Nonlinear process model
based control and optimization of a model IV FCC unit, Comp Chem.
Eng, 19(11): 11531168.
Arbel, A., Huang, Z., Rinard, H., Shinnar, R. and Sapre, A. V., 1995,
Dynamic and control of uidized catalytic crackers. 1. Modeling of the
current generation of FCCs, Ind Eng Chem Res, 34: 12281243.
Arbel, A., Rinard, H., Shinnar, R. and Sapre, A. V., 1995, Dynamics
and control of uidized catalytic crackers. 2. Multiple steady state and
instabilities, Ind Eng Chem Res, 34: 12281243.
Gross, B., Jacob, S. M., Nace, D. M. and Voltz, Z. E., 1976, Simulation
of catalytic cracking process, US Patent 3960707.
Morley, K. and de Lasa, H. I., 1987, On the determination of kinetic
parameters for the regeneration of cracking catalyst, Can J Chem Eng,
65: 773777.
Morley, K. and de Lasa, H. I., 1988, Regeneration of cracking catalyst
in uence of the homogenous CO post combustion reaction, Can J
Chem Eng, 66: 428432.

21. Ali, H. K. A., Rohani, S., 1995, Effect of cracking reactions kinetic
scheme selection on the model prediction of an industrial uid catalytic
cracking unit, Chem Eng Commun, (in print).
22. Weekman, V. W. and Nace, D. M., 1970, Kinetics of catalytic cracking
selectivity in xed, moving, and uid bed reactors, AIChE J, 16: 397
404.
23. Lee, L-S. , Chen, Y-W. , Haung, T-N. and Pan, W-Y. , 1989, Four lump
kinetic model for uid catalytic cracking process, Can J Chem Eng, 67:
615619.
24. Lee, J., 1989, On-line PID controller tuning from a single, closed-loop
test, AIChE J, 35: 329331.
25. Consumers Co-operative re neries Ltd., Regina, Saskatchewan,
Canada, 1994, (private communications).
26. Theologos, K. N. and Markatos, N. C., 1993, Advanced modeling of
uid catalytic cracking riser-type reactors, AIChE J, 39: 10071017.
27. Kunii, D. and Levenspiel, O., 1991, Fluidization Engineering,
(ButterworthHeinemann Publishers, MA, USA).
28. API Technical Data Book Petroleum Re ning, 1967, (American
Petroleum Institute (API)).
29. Perry, R. H. and Chilton, C. H. (Eds), 1973, Chemical Engineers
Handbook, (McGraw Hill, New York, USA).
30. Gear, C. W., 1971, Numerical Initial-Value Problems in Ordinary
Differential Equations, (Prentice Hall, Englewood Cliffs, NJ, USA).

ADDRESS
Correspondence concerning this paper should be addressed to Professor
S. Rohani, Laboratoire des Sciences du Genie Chimique, BP451-1, rue
Grandville, Nancy Cedex, France.
The manuscript was communicated via our International Editor for
Canada, Professor M. H. I. Baird. It was received 28 February 1996 and
accepted for publication after revision 8 November 1996.

Trans IChemE, Vol 75, Part A, May 1997

You might also like