Cincia/Science

NUMERICAL APPROACHES FOR A MATHEMATICAL

MODEL OF AN FCCU REGENERATOR

J. C. Penteado, ABSTRACT
a
C. O. R. Negro This work discusses a mathematical model of an FCCU (Fluid Catalytic
and L. F. S. Rossib Cracking Unit) regenerator. The model assumes that the regenerator is
divided into two regions: the freeboard and the dense bed. The latter is
composed of a bubble phase and an emulsion phase. Both phases are
a,b
Universidade Tecnolgica Federal do Paran modeled as a CSTR (Continuously Stirred Tank Reactor) in which ordinary
Departamento Acadmico de Mecnica differential equations are employed to represent the conservation of mass,
energy and species. In the freeboard, the flow is considered to be one-
Bairro Rebouas
dimensional, and the conservation principles are represented by partial
CEP 80230-901, Curitiba, Paran, Brasil differential equations to describe space and time changes. The main aim of
a
[email protected] this work is to compare two numerical approaches for solving the set of
partial and ordinary differential equations, namely, the fourth-order Runge-
[email protected]
Kutta and implicit finite-difference methods. Although both methods give
a
Corresponding author very similar results, the implicit finite-difference method can be much
faster. Steady-state results were corroborated by experimental data, and the
dynamic results were compared with those in the literature (Han and Chung,
2001b). Finally, an analysis of the models sensitivity to the boundary
conditions was conducted.

Keywords: Regenerator, FCC, numerical approaches, finite-difference

method, fourth-order Runge-Kutta method.

NOMENCLATURE R ideal gas constant (=8314.26 J kmol-1 K-1)

r''' - coke consumption rate per unit of volume
a coefficient of the discretized equation (24) (-) (kmol m-3 s-1)
A heat transfer area (m-2) Rg rate of gas produced by combustion (kg s-1)
C molar concentration (kmol kg-1) s coefficient in equation (24) (-)
cp specific heat (J kg-1 K-1) S coefficient in equation (26) (-)
d coefficient in the discretized equation (24) (-) sb coefficient in equation (24) (-)
D mass transfer coefficient (s-1) Sb coefficient in equation (26) (-)
H heat transfer coefficient per unit of volume (W t time (s)
m-3 K-1) T temperature (K)
k valve flow rating factor (kg1/2 m1/2) V volume (m3)
kc coke combustion rate coefficient (Pa-1 s-1) vc catalyst velocity (m s-1)
k1j , k2j, k3j, k4j Coefficients for the fourth-order vg gas velocity (m s-1)
Runge-Kutta method (-) U overall heat transfer coefficient (W m-2 K-1)
k1 combustion rate coefficient in reaction i (Pa-1 s-1) w coke on catalyst (-)
k2 combustion rate coefficient in reaction ii ( Pa-1 s-1) x hydrogen to carbon atomic ratio in coke (-)
k3C rate constant in reaction iii (kmol kg-1 s-1 Pa-2) X valve stem position (-)
k4h homogeneous rate constant in reaction iii (kmol y mole fraction (-)
m-3 s-1 Pa-2) z vertical coordinate (m)
L height (m) Z compressibility factor (-)
m mass (kg)
m& - mass flow rate (kg s-1) Greek symbols
M molecular weight (kg kmol-1)
N number of adjacent regions in equation (24) (-) valve head differential at maximum flow/valve
P pressure (Pa) head differential at zero flow [0 1] (-)
Q&los - rate of energy loss to the surroundings (kW) c CO/CO2 on the catalyst surface (-)
Q& reac - rate of energy released by chemical reactions volume fraction [0 1] (-)
mf - void fraction in the dense bed at minimum
(kW)
fluidizing condition (-)
r1, r2, r3, r4 reaction rates (kmol m-3 s-1)

Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86 71

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

- volume fraction of the bubble phase Gas) by the action of a catalyst (generally composed
Hr heat of reaction (difference in the enthalpy of of silica and alumina) at high temperature. In
the products and reactants) (kJ m-3 s-1) addition, carbonic residues (coke) are deposited on
P - pressure difference across the valve (Pa) the surface of the catalyst, reducing its activity. This
t time step (s) endothermic process takes place in a vertical flow
- coefficient in equation (26) (-) reactor (riser). Following this, the products are
- coefficient in equation (26) (-) separated from the catalyst to prevent further
cracking and avoid loss of performance. The spent
a general variable (mc,E, mg,R, wE, Ci,E, TE, Ci,B, TB)
catalyst then flows to a regenerator where the coke is
in the discretized equation (24) (-)
burnt in a fluidized bed. The combustion reactivates
- density (kg m-3)
the catalyst and provides energy for the cracking
reactions.
Superscripts
From the modeling viewpoint, regeneration is
very complex: it is a two-phase (solid - gas) flow
a feed air
combustion problem. Several studies have been
B bubble phase
carried out to model the regeneration process. For
c catalyst
instance, McFarlane et al. (1993) developed a
C cyclone
simulation model for the ESSO model IV FCC Unit.
ck coke
Lansarin (1997) proposed a mathematical model to
CO carbon monoxide
simulate the steady-state behavior of a UOP Stacked
CO2 carbon dioxide
converter. Ali and Rohani (1997) published a
D dense bed
dynamic model of an FCC converter, which was later
E emulsion phase
modified by Malay et al. (1999). Santos (2000)
F freeboard
described different models for the regenerator. Han
g gas
and Chung (2001a and 2001b) developed a dynamic
i different components: O2, CO, CO2, H2O, N2
FCC model in which the regenerator was divided into
j one of the ordinary differential equations in the
two regions.
fourth-order Runge-Kutta method
Other authors published more specific studies of
k adjacent region defined in equation (24)
regenerator models. Errazu et al. (1979) developed a
N2 - nitrogen
grid model for an FCCU regenerator. They concluded
O2 oxygen
that the fluidized bed can usefully be represented by a
r regenerator
CSTR without feed-gas bypass. De Lasa and Grace
R regenerated
(1979) described a model for a fluidized-bed reactor
S spent
with the dense bed treated as a CSTR. They assumed
U Upstream discrete point
piston flow on the freeboard and concluded that
V neighborhood
modeling of the freeboard is very important. Lee et
al. (1989) developed three different models for a
Superscripts
typical regenerator and compared the results of these
models with measurements from an industrial plant.
t previous time
Faltsi-Saravalou and Vasalos (1991) described a
t+t current time dynamic model of a two-region fluidized bed (dense
and freeboard) and gave a rigorous description of the
INTRODUCTION hydrodynamics of Geldarts A and B-type particles.
They divided the dense bed into two phases: the
Nowadays, hundreds of fluidized catalytic bubble phase and emulsion phase. The freeboard
cracking (FCC) units are in operation worldwide, above the dense region was modeled as a one-
producing almost 500 million gallons of gasoline per dimensional flow reactor. Guigon and Large (1984)
day. This process is considered one of the most applied a two-phase flow model to a multistage
important developments in chemical engineering in regenerator. Filho et al. (1996) explored the influence
the last century (NACS, 2003). Because of the large of the air jet near the regenerator air distribution grid.
amount of gasoline produced, small efficiency gains Based on these earlier works, a dynamic model
can lead to significant increases in earnings, and as a for simulating the regeneration process in an FCC
result, major investments have been made to enhance unit is presented. The model is based on the mass and
the processs efficiency. The operational flexibility of energy conservation equations. The differential
the FCC process means that it can easily be adapted equations are solved by two numerical methods: the
to the demands of local markets. Furthermore, the fourth-order Runge-Kutta method and the finite-
process is highly profitable, as heavy distillates with difference method. The main aim of the present work
limited commercial value are converted into more is to investigate and compare the performance of
expensive products. Residues from atmospheric and these numerical methods.
vacuum distillation (gasoil and other residues) are
cracked into gasoline and LPG (Liquefied Petroleum

72 Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

MATHEMATICAL MODEL dmc,E

= m& c,S + m& c,C m& c,R m& c,F (1)
dt
In the proposed model, the regenerator is
divided into two distinct regions: the dense bed and
where m is the mass and m& the mass flow rate. The
the freeboard. The dense bed comprises two phases
indices c, S, C, R, E and F represent catalyst, spent,
(the bubble and emulsion phases), as shown in Figure
cyclone, regenerated, emulsion and freeboard,
1. The model is based on the following assumptions:
respectively.
i) heat and mass are exchanged between the bubble
and emulsion phases; ii) most of the combustion
The coke mass balance within the emulsion
takes place in the emulsion phase; iii) the regenerator
phase, based on coke on catalyst, is given by the
exhaust gas consists of CO, CO2, O2, H2O and N2; iv)
following equation:
the combustion of hydrogen is considered
instantaneous; v) the hydrogen-carbon ratio is
d ( wE )
constant during the coke-burning process; vi) the mc,E = ( wS wE ) m& c ,S + ( wC wE ) m& c ,C
cyclones are 100% efficient; vii) the specific heat of dt (2)
the catalyst is constant; viii) the specific heat of the rck, E M ckVE
gases is a function of the temperature; ix) the
temperature is uniform throughout the dense bed and where w is the coke on catalyst, r ''' the coke
varies with height in the freeboard; and x) solid and consumption rate per unit of volume, M the
gas are in thermal equilibrium within the emulsion molecular weight and V the volume. The index ck
phase. represents coke.

Dense Bed Mass conservation of gases: For a gas

compound, the mass balance can be written as:
In this study, the dense bed comprises: i) the
emulsion phase, composed of gases and catalyst at dCi , E vg , E DB , E ri, E
minimum fluidization condition; and ii) the bubble dt
=
LD
(C
i,a Ci , E ) +
mf (1 )
(Ci , B Ci , E ) +
mf
(3)
phase, consisting of gas in excess of that needed for
minimum fluidization. Both phases are considered as
where C is the molar concentration, D the mass
continuously stirred tank reactors (CSTRs)
transfer coefficient, L the height, mf the dense bed
Emulsion Phase void fraction at the minimum fluidization condition,
the volume fraction of the bubble phase and vg the
This phase is composed of solid particles gas velocity at minimum fluidization condition. The
(catalyst and coke), air (O2 and N2) and combustion gas velocity and the void fraction at the minimum
products (CO, CO2, and H2O). fluidization condition are calculated using Kunii and
Levenspiels (1996) model. The index i represents the
gas components O2, CO, CO2, H2O and N2, and the
indices a, B and D, the feed air, bubble phase and
dense bed, respectively.

Energy conservation: In the energy balance for

the emulsion phase, the model assumes that the gas
and catalyst are in thermal equilibrium:

(m c
c,E p ,c + mg,E c p ,g,E )
dTE
dt
=

c
m& c ,S c p ,c (TS TE ) + m& c ,C c p ,c (TC TE ) + m& ac pg ,E p ,a Ta TE
c
p , g ,E
( )
+VB H B,E + DB,E CO2 ,B CO2 ,E c p ,O2 ( TB TE )

+UV AV (TV TE ) H r ,EVE

Figure 1 The bubble and emulsion phase. (4)

Mass conservation of solids: The rate of where T is the temperature, cp the specific heat, U the
accumulation of catalyst within the emulsion phase is overall heat transfer coefficient, A the heat transfer
equal to the spent and recycled catalyst flow rates area and Hr the heat of the combustion reaction. H
minus the regenerated catalyst flow rate and the is the heat transfer coefficient per unit of volume. The
catalyst flow rate to the freeboard: indices g and V are, respectively, gas and
neighborhood.

Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86 73

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

Bubble Phase The conservation of each gas species assumes

the form:
Mass conservation: The bubble phase comprises
the mass of gas in excess of that needed for minimum (Ci,F ) 1 (vg,F Ci,F ) ri''',F (8)
fluidization. The oxidation of CO into CO2 is + =
t g , F z g ,F
considered to take place within this phase. The rate of
accumulation of component i within this phase can be
written as: where vg,F is the average gas velocity along the
freeboard.
dCi,B vg,B
= (Ci,a Ci,B ) + DB , E (Ci,E Ci,B ) + ri''', B (5) Energy conservation: The gas components are
dt LE considered to be in thermal equilibrium with the
catalyst. The conservation equation includes the
where vg,B is the upward bubble velocity, which is a advective transport, heat of reaction and heat loss
function of the superficial gas velocity and the through the regenerator walls:
average diameter of a single bubble in the bed
(Levenspiel, 1998). TF
(c c p ,c c,F + g,F c p,g,F g,F ) =
t (9)
Energy conservation: In the energy balance for T
the bubble phase, the model assumes that all gas (vc,F cc p,c + vg,F g,F c p ,g,F ) F + H i ri''',F
components within the bubble phase are in thermal z
equilibrium. The rate of energy accumulation
accounts for the exchange of heat and mass with the where Hi ri''',F is the rate of energy released by the
emulsion phase, the energy transported by the inlet combustion reactions. The catalyst fraction in the
and outlet gas streams, and the heat generated by the freeboard, c,F, drops exponentially with height
combustion reactions: according to a correlation described by Kunii and
Levenspiel (1996).
dTB
mg , B c p , g , B = H B , EVB + DB , EVB (Ci , E Ci , B )c p ,i (TE TB )
dt Overall Mass Balance
c p , g ,a
+ m& i ,a , B c pg , B Ta TB H r , BVB
c The overall mass of catalyst within the
p,g,B
regenerator is the sum of the mass in the dense bed
(6)
and freeboard:
The outlet temperature of the dense bed is
considered to be an average value of the emulsion mc,r = mc,E + mc,F (10)
and bubble phases weighted by the product of the
respective specific heats and mass flow rates of these The rate of change of the overall mass of gas
phases. stored in the regenerator is equal to the inlet air flow
rate added to the coke combustion rate minus the
Freeboard exhausted gas flow rate:

The freeboard is modeled as a one-dimensional dmg ,r

flow reactor, and the temperature and concentration = m& a + Rg , D + Rg , F m& g,C (11)
dt
of each component therefore change with regenerator
height as a result of combustion. The model predicts
where Rg is the rate of gas produced by combustion
the coke combustion on the catalyst surface and also
within the dense bed, D, and freeboard, F.
the oxidation of CO into CO2.
Pressure in the Regenerator
Mass conservation: The one-dimensional coke
conservation equation can be written as:
The pressure within the regenerator is calculated
'''
according to the ideal gas law:
(wF ) vc , F (wF ) r M ck (7)
+ = ck , F

t c , F z c, F c g , R RTE
P = Zg (12)
M g ,R
where is the volume fraction and vc the catalyst
velocity, which is based on the catalyst mass flow
where P is the pressure, Z the compressibility factor,
rate and is independent of the height. c is the catalyst
density. R the ideal gas constant and g,R the gas density
computed as the ratio of the mass of gas to the
volume of gas within the regenerator.

74 Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

Control Valves c kc (20)

k1 =
c +1
The catalyst flow rate and gas flow rate
throughout the transport lines are a function of the kc
pressure differences across the valves: k2 = (21)
c +1

kX (13)
m& = P where k c and c are, respectively, the rate
+ (1 ) X 2
coefficient of the coke combustion reaction and the
molar ratio of CO2 and CO on the catalyst surface
where k is the valve flow rate factor, X the valve stem (Arbel et. al., 1995). The oxidation of CO is
position, the valve head differential at maximum dependent on the type of catalyst (composites of
flow/valve head differential at zero flow and P the vanadium, nickel and copper contribute to the
pressure difference across the valve. combustion) and can be homogeneous and/or
heterogeneous. The reaction rates for the oxidation of
Combustion Kinetics Model CO (Eq. (16) and (17)) are given by:

Coke is usually composed of different chemical r3 = (1 ) c k3c yO2 yCO P 2 (22)

compounds (C, H, N, S) but consists mainly of
carbon and hydrogen. In the present model, the
following assumptions are made: coke consists of r4 = k4h yO2 yCO P 2 (23)
only CHx hydrocarbon, with a constant atomic ratio
of hydrogen to carbon (x = 0.9); the hydrogen reacts All reaction constants (c, k c , k 3c and k 4 h ) depend
completely and instantaneously; and the combustion
on the temperature and are calculated according to
of coke is the slowest reaction (the bottleneck). The
Arhenius model (Fogler, 1999). The values of
combustion of coke is divided into two reactions: one
frequency factors and Activation Energy are reported
in which carbon monoxide is formed (Eq. (14)), and
in Table 1.
another, carbon dioxide (Eq. (15)). In addition to the
burning of coke, oxidation of CO into CO2 (Eq. (16)
Table 1 Energy of activation and frequency factors
and (17)) also takes place in the regenerator:
of the reaction constants c, kc, k3c, k4h.
k1
i) CH x + (0.5 + x 4)O2 CO + x 2 H 2 O (14) Constant Energy of Frequency Factor
Activation
k2 c 56501.8 2512 (-)
ii) CH x + (1 + x 4 )O2 CO2 + x 2 H 2 O (15) kJ/kmol
kc 158051.0 1055 (Pa.s)-1
k3c
iii) CO + 1 2 O2 CO2 (16) kJ/kmol
k3c 115479.4 615.65 (kmol/kg/s/Pa2)
k4h
kJ/kmol
iv) CO + 1 2 O2 CO2 (17) k4h 299629.9 2.9619x1016
kJ/kmol (kmol/m3/s/Pa2)
where x is the hydrogen to carbon atomic ratio in
coke, k1 , k 2 and k 3c the respective rate coefficients SOLUTION METHODS
of the heterogeneous reactions, and k 4 h the rate As the differential equations described above do
coefficient of the homogeneous reaction. The not have a solution (because of their non-linearity),
reaction rates for coke combustion are of second they are solved by numerical methods. Two
order and depend on the oxygen and coke approaches are employed: i) the finite-difference
concentrations: method (Patankar, 1980); and ii) the fourth-order
Runge-Kutta method (Patankar, 1980). The fourth-
w (18) order Runge-Kutta method is applied to initial-value
r1 = (1 )ck1 yO P
Mck 2 problems, represented by ordinary differential
equations, and the finite-difference method is
w employed for both ordinary and partial differential
r2 = (1 ) c k 2 yO P (19) equations.
M ck 2
Finite-Difference Method
where y is the mole fraction. The constants k1 and k2
are defined as: Ordinary Differential Equations

Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86 75

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

It can be seen that all the first order ordinary where represents one of the properties
differential equations (1), (2), (3), (4), (5), (6) and
(11) have the same form and can generally be written mc , E , mg , rg , wcq, E , Ci , E , TE , Ci , B , TB . k is the value
as, of a property at an adjacent region (spent, cyclone,
regenerated, emulsion, freeboard, feed air), N
d N indicates the number of interactions with adjacent
a = d k (k ) +s sb (24)
dt k =1 regions, and the coefficients a , d k , s and sb are
defined according to each conservation equation.
Tables 2 and 3 show the coefficients for the mass and
energy conservation equations, respectively.

Table 2 - Properties and k and the coefficients a, d k , s, sb for the mass conservation equations.

Equation (1) Equation (2) Equation (3) Equation (5) Equation (11)
mc,E wE Ci , E Ci , B mg ,r
1 - wS Ci ,a Ci ,a -
2 - wC Ci , B Ci , E -
a 1 mc,E 1 1 1
vg ,E vg,B
d1 - m& c ,S -
LD LE
DB ,E
d2 - m& c ,C DB ,E -
mf ( 1 )
ri,E m& ar + Rg ,D + Rg ,F
s m& c ,S + m& c ,C -
ri,B
mf
sb m& c ,R + m& c ,F rck, E M ck VE - - m& g,C

Table 3 - Properties and k and the coefficients ak , dk , s, sb for the energy conservation equations.

Equation (4) Equation (6)

TE TB
1 TS -
2 TC -
c p ,a c p ,a
3 Ta Ta
c p ,g ,E c p ,g ,B
4 TB TE
5 TV -
a mc,E c pc + mg,E c p ,g,E mg ,B c p ,g ,B
d1 m& c ,S c p ,c -
d2 m& c ,C c p ,c -
d3 m& a c p ,g ,E m& i ,a c p ,g ,B
d4 (
H B ,EVB + DB ,EVB CO2 ,B CO2 ,E c p ,O2 ) H B,EVB + DB ,EVB ( Ci ,E Ci ,B )c p ,i
d5 UV AV -
s H r , EVE H r , BVB
sb - -

76 Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

The first order derivative of equation (24) can point to the dense-bed interface (S is assumed to be
be approximated by the backward finite-difference the interface value) and evolves along the height.
approximation to yield the following expression:

at +t N t +t t +t N t +t t +t
+ d k = d k k +
t k =1 k =1 (25)
t +t
t + ( s t +t sbt +t )
a
t
Figure 2 One dimensional discretization.
where t is the time-step. The index t and t+t
represent past and present values. Discretization provides two set of algebraic
equations, represented by equation (25) for the dense
Partial Differential Equations bed and equation (27) for the freeboard.

The freeboard partial differential equations can Convergence criterion

be written in the general form:
The stability of a numerical method cannot
j j usually be proven for non-linear sets of equations.
j + j = S j Sb , j (26) The ability for the user to set up the grid is
t z particularly important to obtain an accurate solution.
Therefore, the convergence and numerical stability
where the coefficients S j , j , S j and Sb, j are were checked by evaluating the solutions
defined according to one of the conservation dependence on t and z .
equations and can be found in Table 4. The implicit finite-difference method applied to
linear equations is unconditionally stable and
Table 4 Property j and the coefficients j , j , Sj provides a solution for any value of t . The only
and Sb,j . restriction to the value of t is the accuracy of the
solution.
Equation (7) Equation (8) Equation (9) The convergence criterion adopted is the
residues of the energy conservation equation:
wF Ci,F TF
j 1 1 ccp,c c,F + g,F cp, g,F g,F Es = max Ri < tol. (28)
i =1... j
vc,F vg,F
j vc,F ccpc + vg,F g,F cpg,F
c, F g ,F E s is the maximum value between the dense and
Sj - - H r '''
i i,F
freeboard energy conservation equations. The
iteration process stops when Es is smaller than a
rck, F Mck ri,F
Sb, j - tolerance (tol). This criterion was adopted because
c,F c g ,F the last equations to reach convergence were the
energy equations.

Both time and spatial derivatives of equation The Solution Algorithm

(26) are discretized by backward approximations.
This results in the following implicit finite-difference The algebraic equations were solved
equation: sequentially, using an approach similar to the Gauss-
Siedel method. Figure 3 shows the flowchart of the
t +t y algorithm developed. Firstly, initial values are
+ tj +t pt + tj +tst +t (27)
j
established for all the variables ( 10 , 20 ,... ). Then the
t
coefficients of the dense-bed equations and the
(
+ s tj +t sbt +
,j
t
y ) respective values are calculated iteratively. Once
the dense-bed equations have converged, the
where the index P is the discretization point, S the coefficients of the freeboard equations are computed.
nearest point to P in the upwind direction and y Finally, the f values for the freeboard properties
the grid dimension, as shown in Figure 2. As can be ( 1k, z+1 , 2k,+z 1 ,... ) are calculated. Convergence of the
seen, the property at point P can be evaluated as a whole set of equations is checked. If convergence is
function of the property at point S. In other words, not reached, the solution process restarts at the dense-
evaluation of the discrete values starts at the nearest bed equation. Finally, current values are attributed to

Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86 77

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

previous ones and printed. A new time-step is then

established and the simulation evolves until the last k j 2 = t f j (t + 12 t,1n + 12 k j1 , 2n
time-step is reached.
+ 12 k j1 ,K ,nn + 12 k j1 )

10 ,20 ,..., j0 k j 3 = t f j (t + 12 t,1n + 12 k j 2 , 2n

a j , di , j + 12 k j 2 ,K , nn + 12 k j 2 )
k = k +1

k +1
, k +1
,..., k +1 k j 4 = t f j (t + t , 1n + k j 3 , 2n + k j 3 ,K , nn + k j 3 )
1 2 j

Note that the variables depend on the value at

no
Ri < tol . jk = jk +1 the previous time-step, indicating that the method is
yes
an explicit one. Iteration is therefore not required, and
the equations can be solved sequentially.
y = y + y
jk,y = jk,y+1
Partial Differential Equations
j , j

As the freeboard partial differential equations

1,ky+1,2,k y+1,..., jk,y+1 y < Lrg (7), (8) and (9) cannot be solved directly by the
no fourth-order Runge-Kutta method, they are first
no
yes discretized in the z direction, resulting in a set of
tt=+t +t t = k
yes ordinary differential equations (Eqs. (29)). These
t < tmax t +t Es < tol .
no
time-dependent equations are then solved by the
fourth-order Runge-Kutta method.
To solve the freeboard equations, boundary
conditions at the dense-bed interface are needed. The
Figure 3 - Finite-Difference Algorithm. outputs of the dense-bed solution are the boundary
conditions for the freeboard equations. Since the
Fourth-Order Runge-Kutta Method catalyst particles collected by the cyclones are
recycled to the dense bed, there is interaction
Ordinary Differential Equations between the two sections, and their equations are
therefore interdependent. To ensure the
According to the fourth-order Runge-Kutta interdependence of the equations, the finite-
method [20], the system of j first order differential difference solution is carried out iteratively between
equations (Eqs. (1), (2), (3), (4), (5), (6) and (11)) can the dense-bed and the freeboard equations, as
be written as, explained in Section 3.1.4.
As the fourth-order Runge-Kutta method is an
explicit one, iterations are not necessary and the time
d1
= f1(t,1 , 2 ,..., j ) solution is performed sequentially.
dt
d 2
= f 2 (t,1 , 2 ,..., j ) (29) RESULTS
dt
M In this section the results of the proposed model
d j are compared with experimental data and with the
= f j (t,1 , 2 ,..., j ) results of numerical models found in the literature
dt
((Han, 2001a) and (Han, 2001b)).
and its general solution is
Verification of the Model

jn+1 = jn +
1
6
( k j1 + 2 k j 2 + 2 k j 3 + k j 4 ) (30) This case study is based on the start-up of a
PETROBRAS/SIX pilot plant located in So Mateus
do Sul-PR, Brazil. At start-up, the regenerator
where temperature is assumed to be 900K, the pressure, 1.9
bar, and the coke on catalyst, zero. The molar
k j1 = t f j (t,1n , 2n ,K , jn ) , concentrations of gas compounds (CO, CO2, H2O, O2
and N2) are assumed to be identical to their
concentrations in the feed air (see Table 5). For the

78 Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

purposes of the comparisons detailed in this section,

925,0
the regenerator boundary conditions are assumed to 925.0

be constant values, as shown in Table 5. The

(K) (K)
924.0
924,0
simulation was carried out over a 200-minute period.

Temperature
923,0
923.0
A comparison of the fourth-order Runge-Kutta

Temperatura
80
80pontos
points
922.0
922,0 40pontos
40 points
and finite-difference methods is given below. 20
20pontos
points
10pontos
10 points
921.0
921,0

Table 5 - Regenerator boundary conditions. 920.0

920,0
0 1 2 3 4
Regenerator height (m)
Altura (m)

Variables Value Unit

Spent-catalyst mass flow rate 0.13 Kg/s Figure 4 Temperature profile for different meshes
in the freeboard. Fourth-order Runge-Kutta solution.
Spent-catalyst temperature 790 K
As the catalyst mass flow leaving the freeboard
Coke on catalyst 0.14 kgcoke kgcatal
returns by the cyclones to the dense bed, the number
Feed-air mass flow rate 0.05 Kg/s of grid points in the freeboard will affect this mass
Air Temperature 350 K flow and, therefore, the dense-bed temperature. Table
Molar concentration of O2 in 0.20 - 6 shows the sensitivity of the temperature to the
the feed air number of grid points. T is the maximum
Molar concentration of N2 in 0.79 - temperature difference between the result with a
the feed air particular grid (5, 10, 20, 40 or 80-point) and the
Molar concentration of CO2 0.01 - result with a 160-point grid.
in the feed air t in Table 6 corresponds to the maximum time-
Molar concentration of CO 0.00 - step that ensures the method is stable for the given
in the feed air number of grid points. The value that ensures the
Molar concentration of H2O 0.00 - method is stable also provides a time-step-
in the feed air independent solution. In other words, if the time-step
is reduced, the same solution is obtained.
Additionally, as the number of grid points doubles,
Fourth-order Runge-Kutta method
the computational time more than triples. In view of
the small difference between the 40 and 160-point
An analysis of the models sensitivity to time-
grid results (0.1K), the 40-point grid is considered
step size and the number of discrete nodes in the
adequate.
freeboard was carried out. As the fourth-order Runge-
Kutta method is an explicit one, its stability depends
Finite-Difference Method
on the time-step size. In the freeboard, the maximum
time-step that provides a stable solution will also
First, an analysis of the convergence criterion
depend on the mesh size. For instance, only one
(the sum of the residues of the energy equation see
volume is allowed in the dilute region, for a value of
Section 3.1.3) was carried out. Solutions for sums of
t=0.54s. As the number of points increases, the residues of 1.0kW, 0.1kW and 0.001kW were
time-step must decrease to keep the ratio z 5.5. compared. The results were found to be insensitive to
t
This value is the ratio of the coefficients j and j the sum of residues smaller than 0.001kW, and this
value was therefore used henceforth as the
in Equation (26), which is basically the ratio of the convergence criterion.
phase velocity (solid or gas) to the volume fraction
(solid or void) of the phase in the freeboard. By Table 6 - Temperature change for different grids.
discretizing Equation (26) in space and substituting it
in the fourth-order Runge-Kutta Equation (30), it can
Grid t (s) T (K) Computational
be seen that the coefficient of Pn can be negative if
points time (min)
z / t < j / j ; negative coefficients can cause 5 0.1 2.84 4
instability in explicit methods (see Patankar, 1980). 10 0.049 0.4459 7.2
Figure 4 shows the steady-state temperature 20 0.023 0.2533 25.5
along the regenerator height for different numbers of As already0.0115
40 mentioned,0.1186
there is no restriction
78.2 in
grid points. As the number of points increases from 80 0.0060 0.0598 249.7
10 to 20, the temperature at each grid point increases 160 0.0025 - 1333
by approximately 0.25K.
As already mentioned, there is no restriction in
terms of time-step size in the implicit finite-
difference method. Figure 5 shows the transient
response of the spent-catalyst temperature for

Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86 79

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

different time-steps. As expected, the steady-state is The finite-difference and fourth-order Runge-
independent of the time-step. However, the smaller Kutta methods give virtually the same results. For
the time-step, the steeper the dynamic temperature example, the maximum temperature difference is
gradient. Table 7 shows the maximum temperature only 0.04 K for a 0.54s time-step. Any value larger
difference between a certain time-step solution and a than this makes the solution unstable. Table 8 shows
7.5s time-step result. As shown, 7.5s and 15s time- a comparison of the performance of the numerical
step results are quite close; therefore, a 15s time-step methods. For a mesh-independent solution, the finite
solution is considered time-step independent. difference method is much faster. As the time-step is
reduced, the finite-difference method becomes slower
Table 7 - Temperature difference for different time- than the fourth-order Runge-Kutta approach; for
steps. 0.0115s, the computational time for the finite-
Time-step (s) Tmax(K) difference method is 81minutes, and for the Runge-
120 4.1 Kutta method, 78 minutes.
60 2.3
Table 8 - Comparison of the finite-difference and
30 0.7
fourth-order Runge-Kutta methods. Computer
15 0.05
employed: Pentium III, 1.1GHz, 512MB.

930
Fourth-order Finite
Runge-Kutta Difference
920
(K)
Temperatura(K)

Discretization Explicit Implicit

Temperature

910 t = 120 s method

t = 60 s
t = 30 s
Time step Conditionally Unconditionally
900 t = 15 s stable stable
t = 7.5
7,5ss
Convergence - Energy
890
0 50 100 Criterion equation
Tempo(min)
Time
Time (min)
(min) residue 0.001
Figure 5 Transient response of the spent-catalyst kW
temperature for different time-steps and a 40-point Grid dependence t e z are t e z are
mesh in the freeboard. Finite-difference solution. interdependent independent
Grid for mesh 40 points 40 points
Figure 6 shows the catalyst temperature in the independent
emulsion phase for different grids and a 15s time- solution
step. The steady-state is sensitive to the number of Computational 78.2 min 81 min
grid points in the freeboard, showing the dependence time for a
of the dense bed on the freeboard solution. As with t = 0.01115 s
the fourth-order Runge-Kutta method, the 40-point Computational 78.2 min 1 min
grid provides a mesh-independent solution. time for a
Nevertheless, as the method is unconditionally stable, time-step
the number of grid points is not dependent on the independent
time-step size. Therefore, a higher time-step can be solutions
used and a mesh-independent solution still be
obtained, thereby considerably reducing the Experimental Comparison
computational time.
Comparison of the results of the proposed
930
model with the steady-state experimental values from
the PETROBRS/SIX# pilot FCCU plant revealed
(K)

920
Temperatura (K)

that the computed regenerated-catalyst temperature

Temperature

910 points
80 pontos and coke on catalyst are smaller and larger,
points
40 pontos respectively, than their experimental counterparts.
points
20 pontos
900 points
10 pontos The small amount of gas exchanged between the
points
5 7.5s
pontos emulsion and bubble phases is a possible cause of
890 these discrepancies. Falsti-Saravalou and Vasalos
0 50 100 150 200
(1991) explained that this mass exchange is strongly
Time
Time (min)
Tempo(min)
(min)
dependent on the bubble sizes within the dense bed. It
Figure 6 Transient response of dense-bed is probable, therefore, that the correlation used to
temperature for a 15s time-step and different grids in
the freeboard. Finite-difference solution.
#
A business unit of Petrobrs S/A, located in So
Mateus do Sul-PR, Brazil.

80 Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

estimate the bubble diameters gave inappropriate in Table 11. The coke on regenerated catalyst and the
values. Table 9 shows a comparison of plant data outlet molar concentrations of CO and CO2 obtained
with computed values for different bubble diameters. with the complete model are closer to the
experimental data. On the other hand, the temperature
Table 9 Comparison of the results obtained with the of the regenerated catalyst provided by the one-bed
proposed model and experimental data for different model is nearer to that of its experimental
bubble diameters. counterpart. This comparison clearly demonstrates
the influence of the freeboard on the combustion of
Variable Plant Bubble diameter (m) Unit CO. When the model including the freeboard is used,
0.15 0.1 0.04 the values for the outlet molar concentrations and
TE 995.1 944.6 978.3 957.2 K coke on catalyst approximate to the experimental
TF 947.2 960.2 938.4 934.2 K values.
wE 0.048 0.24 0.031 0.026 %
CCO 0.00 0.00 0.06 0.40 % Table 11 Comparison of the two-bed (dense and
CCO2 8.31 7.36 8.65 8.42 freeboard) and one-bed models.
%
CO2 11.17 12.34 10.75 10.82 %
CN2 79.57 80.30 80.54 80.36 % Two-bed Plant One-bed Unit
TE 978.3 995.1 1006.0 K
In the conditions prevailing in the case study,
TF 938.4 947.2 - K most
A bubble diameter of 0.1m yields the closest the plant operates with excess oxygen, so that
results to the experimental data. The bubble size that w E 0.031 0.048 0.023
reactions are complete and take place in the dense %
provides the best results is approximately half that CCO
bed. it is for 0.06
this reason0.000 0.93 for the%
that the results one-
obtained by the empirical correlation. A bubble CCO2 8.65 8.31 7.75 %
diameter less than 0.035m makes the solution CO2 10.75 11.17 11.15 %
unstable and is the lowest limit at which the CN2 80.54 79.57 80.17 %
correlation is still valid. For a bubble diameter of
0.1m, the bubble velocity and void fraction, which In the conditions prevailing in the case study,
also depend on the bubble diameter, agree with the plant operates with excess oxygen, so that most
values given by Santos [6]. reactions are complete and take place in the dense
Two models of combustion kinetics were tested: bed. It is for this reason that results for the one-bed
the model proposed by Arbel et al. [17] and that model and those for the complete model are similar.
described by Han and Chung (2001a) and (2001b). When there is no excess O2, a significant amount of
As shown in Table 10, the results obtained using the CO is oxidized in the freeboard, in which case the
model proposed by Arbel et al. [17] are closer to the complete model is more appropriate.
experimental data than the results obtained with the
model described by Han and Chung for most Numerical Comparison
variables.
Following is a comparison of the proposed
Table 10 Comparison of two kinetic models found model with that proposed by Han and Chung
in the literature. ((2001a) and (2001b)).
Han and Chung (2001b) modeled the whole
Variable Plant Arbel Han and Unit FCC, including the riser, regenerator, reactor and
et al. Chung valves. In order to compare the two models, the
(1995) (2001a e b) values of some variables (the inlet-air and spent-
TE 995.1 978.3 958.5 catalyst flow rates (see Figure 7), coke on spent
K
catalyst and spent-catalyst temperature (see Figure
TF 947.2 938.4 906.1 K
8)) obtained from Han and Chungs (2001b) results
wE 0.048 0.031 0.065 %
were established as the boundary conditions of the
CCO 0.00 0.06 0.54 % proposed model. After 10 minutes of simulation, the
CCO2 8.31 8.65 7.91 % airflow rate was increased from 34kg/s to 35.7kg/s.
CO2 11.17 10.75 11.27 % This value was kept unchanged until the simulation
CN2 79.57 80.54 80.28 % had continued for 150 minutes, when the flow rate
was reduced to its initial value. The air temperature
Therefore, the kinetic model described by Arbel was maintained constant at 432K throughout the
et al. is used for the other comparisons, including that simulation.
in Table 9. In order to check the influence of the
freeboard on the regenerator model, a second
approach was considered, in which the freeboard was
excluded (i.e., a one-bed model). A comparison of the
complete model with the one-bed approach is shown

Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86 81

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

1.0
40
40 Airflow
390 Table 13 Steady-state comparison of the proposed
Vazo de Ar
0.8 370 model and Han and Chungs (2001b) model.

rate (kg/s)
catalisador(kg/s)
30
30 Spent catalyst
(kg/s)

Vazo de catalisador gasto

de Ar(kg/s)

0.6 350
Variable Han and Current Unit
Vazo rate

20
20 330

Vazo de flow
0.4
Chung model
Airflow

310

Catalyst
10
10
0.2 Regenerated
Vazo catalystregenerado
de catalisador
290
(2001b)
0.0
00 270
Coke on the regen. 0.1 0.13 %
0 50 100 150 200 250 300 catalyst
Time
Time (min)
Tempo (min)
Regen.-Catalyst 991.0 1003.4 K
temperature
Figure 7 Air, regenerated-catalyst and spent- Molar concentr. of CO2 14.80 14.86 %
catalyst flow rates (Han and Chung (2001b)) Molar concentr. of CO 0.4 2.31 %
transient boundary conditions. Molar concentr. of O2 0.2 0.06 %
1,0
1.0 890 Molar concentr. of H2O 9.20 7.72 %
(%) %

870

(K)
Cat. Gasto (K)
0.8
0,8
Cat. Gasto

The molar concentrations of CO2 and the

Catalyst temperature
850
on catalyst

0.6
0,6 830 temperatures obtained with both models are in good
de Coque

0.4 810 agreement. The values for the coke on regenerated

Temperatura
0,4
Coke

790 catalyst obtained with both models are small: 0.10%

Frao

0.2
0,2
770 and 0.13% for the present model and the Han and
0.0
0,0
0 50 100 150 200 250 300
750 Chung model, respectively. However, the molar
Time
Tempo(min)
(min) concentrations of CO, O2 and H2O for the proposed
model differ slightly from those obtained by Han and
Figure 8 Coke on spent catalyst and spent-catalyst Chung. These differences can be attributed to the
temperature (Han and Chung (2001b)) transient modeling of the dense bed, as Han and Chung
boundary conditions. (2001b) employed a distributed model and the
approach described here assumes it is uniform.
Beyond the boundary conditions, the proposed
approach also uses the Han and Chung (2001b) Dynamic Comparison
models for combustion kinetics, mass transfer
coefficient and control valves. It should be noted that As shown in Figure 9, the transient response of
Han and Chungs model assumes there are one- the coke on regenerated catalyst and regenerated-
dimensional variations of gas properties within the catalyst temperature obtained with the proposed
whole regenerator, whereas the proposed model only model and those obtained with Han and Chungs
assumes this within the freeboard. These are the model ((2001a) and (2001b)) are similar. Because of
differences between the models. the low molar concentration, the discrepancies
between the two models appear quite large (see
Steady-state comparison Figure 10). However, the maximum molar
concentration is in the order of 2.5% and the
Initially, a steady-state comparison of the differences in concentration are under 2%. Despite
proposed model and Han and Chungs (2001b) model these differences, the models show similar trends.
was conducted. The fixed boundary conditions are
shown in Table 12, and the comparison of the main Sensitivity Analysis
variables is given in Table 13.
Lastly, an analysis of the sensitivity of the
Table 12 Boundary conditions for the steady-state proposed model to all the boundary conditions was
comparison between the proposed model and Han carried out for the operating conditions and
and Chungs (2001b) model. geometry described in Section 4.1 (The
PETROBRAS/SIX pilot FCCU plant).
Variable Value Unit
Spent catalyst flow rate 314 kg/s
Spent catalyst temperature 787 K
Coke on spent catalyst 0.9 %
Air flow rate 34 kg/s
Inlet air temperature 432 K

82 Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86

 Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

(K) (K)
1.0

Frao de coque cat. regenerado %

1,0 1060 1025

(K)
Te 0,8
0.8

(K)
mp
Temperature

Coke on regen. catalyst (%)

Regenerado
Current model

temperature
Modelo
Frao de Coque Cat. Reg. %

1040 era

temperature
1000 tur
Coke on catalyst (%)

0.8
0,8 Han e Chung (2001)

Cat. Regenerado
-------- Han and Chung (2001b) a 0,6
0.6
1020 975 Coke Coque
0.6
0,6

Cat.
1000 0,4
0.4

Regen. catal.
950

Temperatura
Regen. catal.
0.4
0,4 980
0,2
0.2
925

Temperatura
960
0.2
0,2 900 0.0
0,0
940
50 100 150 200 250 300 350 400 450 500 550 600
0.0
0,0 920 Catalyst flow rate (kg/s)
Vazo de catalisador
0 50 100 150 200 250 300
Time (min)
Tempo (min) Figure 11 Regenerated-catalyst temperature and
coke on regenerated catalyst in steady state as a
Figure 9 Numerical comparison of coke on
function of the spent-catalyst flow rate.
regenerated catalyst and regenerated-catalyst
temperature.
Effect of the coke on spent catalyst

2.52,5 CO Up to a limit of 1.0%, the higher the coke on spent

Mole concentration (%)

2.02,0 O2
catalyst the higher the regenerator temperature.
Current model
Frao molar %

Modelo Above that limit, the temperature stabilizes at a

1.51,5 Han
Hane Chung (2001)
and Chung
(2001b) constant value (see Figure 12) because of the lack of
1.01,0 oxygen to burn the coke that is fed in. As a result, the
CO coke on regenerated catalyst starts to increase.
0.50,5
O2
0.00,0
0 50 100 150 200 250 300 Effect of the spent-catalyst temperature
Time
Tempo(min)
(min)
Figure 13 shows the regenerated-catalyst
Figure 10 Numerical comparison of CO and CO2 temperature and coke on regenerated catalyst as a
molar concentrations at the regenerator outlet. function of the spent-catalyst temperature. It can be
seen that both curves are discontinuous, with two
As the boundary conditions were changed one steady states for the spent-catalyst temperature
by one, the transient behavior of the regenerator between 632 and 667K.
temperature, coke on regenerated catalyst and gas
1050 2,0
2.0
composition was similar to a first-order system
(K)

Temperatura
(K)

Temperature
temperature

1000
response. The time for all variables to stabilize was

Coke on catalyst (%)

Cat. Regenerado

1,5
1.5

Frao de coque %
less than 25 minutes. This is of the same order as the 950
ue
Coq
ratio of the mass of catalyst within the regenerator to 900 1,0
1.0
Regen. catalyst

Coke
the mass flow rate of the spent catalyst. As the 850
Temperatura

0,5
0.5
transient responses of all the variables are quite 800
similar, they are not shown. The sensitivity analysis 750 0,0
0.0
0.0
0,0 0.2
0,2 0.4
0,4 0.6
0,6 0.8
0,8 1.0
1,0 1.2
1,2 1.4
1,4 1.6
1,6 1.8
1,8 2.0
2,0
was thus performed only for steady-state conditions. Coke on spent catalyst (%)
Frao de coque no catalisador gasto %

Effect of the spent-catalyst flow rate Figure 12 Regenerated-catalyst temperature and

coke on regenerated catalyst in steady state as a
Figure 11 shows the steady regenerated-catalyst function of coke on spent catalyst.
temperature and coke on regenerated catalyst as a
function of the spent-catalyst flow rate. The The lower temperature curve is obtained by a
maximum temperature can be seen to occur at a flow progressive increase in the spent-catalyst
rate of 250kg/s. For flow rates higher than 250kg/s temperature. When the spent-catalyst temperature
there is insufficient oxygen to burn the coke, and for reaches 667K, a small increase in this temperature
lower flow rates there is excess oxygen. produces a step change in the operating conditions
the regenerated-catalyst temperature jumps from
This is characteristic of the transition from 678.3 to 873.3K. According to Fogler [18], the low
partial to total combustion, which can be achieved by spent-catalyst temperature, where the discontinuity
either increasing the coke on spent catalyst or its flow takes place, is called the ignition temperature.
rate. Both procedures raise the amount of coke in the
regenerator. The higher the catalyst flow rate the
lower the oxygen/coke ratio and, consequently, the
lower the combustion rate.

Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86 83

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

2,5
2.5 8 5
(K)
1000
Regenerado (K)
7 CO

CO mole concentration (%)

molar O 2 % (%)
2.0
2,0
Cat. temperature

coque %(%)
800 4
6

mole concentration

Frao molar CO %
Temperatura

Coke ondecatalyst
Temperature
1.5
1,5 5
600 3
4
Regen. catalyst

1,0
1.0

Frao
400 2
Temperatura

O2Frao
Coque
Coke 0,5
0.5 2
200 1
1
0 0,0 O2
0.0
0 0
500 550 600 650 700 750 800 850 900
20 30 40 50 60
Temperatura
Spent cat. gasto (K)
catalyst temperature (K) Inletdeairflow
Vazo rate (kg/s)
ar alimentado (kg/s)

Figure 13 Regenerated-catalyst temperature and Figure 15 Molar concentrations of carbon

coke on regenerated catalyst in steady state as a monoxide and oxygen at the regenerator outlet in
function of spent-catalyst temperature. steady state as a function of inlet air flow rate.

0.4
Figure 14 shows the molar concentration of 0,4 1050

(K)
temperature(K)
Coque
oxygen at the regenerator outlet as a function of the Temperature 1030

coque %(%)

cat. regenerado
0.3
0,3
spent-catalyst temperature. When the oxygen molar

decatalyst
1010
concentration is low (spent-catalyst temperature 0.2
0,2

Regen. catalyst
Coke on
higher than 800K), an increase in the spent-catalyst 990

Frao

Temperatura
0.1
temperature does not improve the combustion rate. 0,1
970
Coke
Therefore, the regenerated-catalyst temperature only 0.0
Temperatura
0,0 950
rises because of the higher energy of the spent- 20 30 40 50 60
Inlet
Vazo deairflow rate (kg/s)
ar alimentado (kg/s)
catalyst itself. A discontinuity can also be observed
when the spent-catalyst temperature is reduced. At
632K, the steady state of the system jumps to a lower Figure 16 Regenerated-catalyst temperature and
value. This value is called the extinction point (Han coke on regenerated catalyst in steady state as a
and Chung 2001b). As can be seen, there is hysteresis function of the inlet air flow rate.
between the ignition and extinction points. The spent-
catalyst temperature between these two values can A maximum temperature is observed at an air
lead to one or other steady state. Operation near this flow rate of 44kg/s. Above this maximum
region is not desirable, as any disturbance can lead to temperature, the coke on catalyst becomes very low.
instability. High temperatures and excess oxygen are
characteristics of complete combustion. The
25
regenerator therefore works at complete combustion
O 2 mole concentration (%)

20 for air flow rates greater than 44kg/s, when the

Frao molar O2 %

15
temperature is less sensitive to the inlet air flow rate.
The excess oxygen helps to reduce the regenerator
10 temperature once the air is cold, while the excess air
5 significantly improves the combustion process.
0
500 550 600 650 700 750 800 850 900 Effect of inlet air temperature
Spent Temperatura Cat. Gasto
catalyst temperature (K)
A 10K increase in inlet air temperature produces
Figure 14 Molar concentration of oxygen in steady a 1K increase in regenerated-catalyst temperature.
state as a function of spent-catalyst temperature. This relationship is almost linear irrespective of the
inlet temperature, as the energy carried by the air is
Effect of the air flow rate small compared with that produced by combustion.

The outlet composition of the gases as a CONCLUSIONS

function of the inlet air flow rate is shown in Figure
15. The higher the inlet air flow rate, the lower the This paper describes a dynamic model for an
concentration of CO. As the amount of available O2 FCCU regenerator based on both the equations for
increases, less CO is formed, and combustion the conservation of energy and mass and empirical
changes from partial to complete. This change in correlations. The regenerator is composed of two
combustion regime is confirmed by Figure 16, which beds: the dense bed and freeboard. The dense bed
shows the reduction in coke on catalyst and increase comprises bubble and emulsion phases. Both phases
in regenerated catalyst temperature as the air flow are considered to be CSTRs (Continuously Stirred
rate increases. Tank Reactors). The freeboard model is considered
one-dimensional, as its properties vary with
regenerator height. The conservation equations for

84 Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

the dense bed and freeboard are ordinary and partial REFERENCES
differential equations, respectively. Two approaches
are used to solve these equations: the fourth-order Ali, H. and Rohani, S., 1997, Dynamic
Runge-Kutta method and the implicit finite- Modeling and Simulation of a Riser-Type Fluid
difference method. To solve the partial differential Catalytic Cracking Unit. Chemical Engineering
equations by the fourth-order Runge-Kutta method, Technology, Vol. 20, pp. 118-130.
the space derivative is discretized using the finite- Arbel, A., Huang, Z., Rinard, I. H., Shinnai, R.
difference approach and a set of ordinary differential and Sapre, A. V., 1995, Dynamic and Control of
equations that can be solved by the fourth-order Fluidized Catalytic Crackers. 1. Modeling of the
Runge-Kutta method is obtained. Current Generation of FCCs. Industrial and
Comparison of the solution methods revealed Engineering Chemistry Research, Vol. 34, No. 4, pp.
that the fourth-order Runge-Kutta method is slightly 1228-1243.
faster than the finite-difference method when the Carnahan, B., 1990, Applied Numerical
same time-step is used. To avoid instabilities, Methods, Melbourne: Krieger Publishing Company.
however, a very small time-step is needed for the De Lasa, H. I. and Grace, J. R., 1979, The
fourth-order Runge-Kutta method. In contrast, there Influence of the Freeboard Region in a Fluidized Bed
is no restriction on the size of the time-step in the Catalytic Cracking Regenerator, AIChE Journal, Vol.
implicit finite-difference method. Larger time-steps 25, No. 6, pp. 984-991.
can therefore be used, so that the finite-difference Errazu, A. F., De Lasa, H. I. and Sarti, F., 1979,
method becomes much faster than the fourth-order A Fluidized Bed Catalytic Regenerator Model Grid
Runge-Kutta approach. Effects. The Canadian Journal of Chemical
A simulation was carried out with and without Engineering, Vol. 57, pp. 191-197.
the freeboard to investigate the effect of this region. Faltsi-Saravelou, O. and Vasalos, I. A., 1991,
The combustion of coke and oxidation of CO that FBSim: a Model for Fluidized Bed Simulation I.
takes place in the freeboard not only changes the Dynamic Modeling of an Adiabatic Reacting System
composition and temperature of the outlet gases but of Small Gas Fluidized Particles. Computers and
also the coke on catalyst that returns from the Chemical Engineering, Vol. 15, No. 9, pp. 639-646.
cyclones to the dense region. Filho, R. M., Batista, M. F. L. and Fusco, M.,
The steady-state results for the model were 1996, A Fast Fluidized Bed Reactor for Industrial
corroborated by experimental data and found to be FCC Regenerator. Chemical Engineering Science,
very sensitive to the diameter of the bubbles in the Vol. 51, No. 10, pp. 1807-1816.
dense bed. Arbel et als [17] kinetic model showed Fogler, H. S., 1999, Elements of Chemical
better agreement with the experimental data than Han Reaction Engineering. 3rd edition. New Jersey:
and Chungs model ((2001a) and (2001b)). The Prentice-Hall Inc.
model proposed in this paper was also compared with Guigon, P. and Large, J. F., 1984, Application
Han and Chungs dynamic FCCU model ((2001a) of the Kunni-Levenspiel Model to a Multistage
and (2001b)) and was found to be in reasonable Baffled Catalytic Cracking Regenerator. The
agreement with it. Chemical Engineering Journal, Vol. 28, No. 3, pp.
Finally, an analysis of the models sensitivity to 131-138.
the regenerator boundary conditions was performed. Han, I. S. and Chung, C. B., 2001a, Dynamic
In general, regenerator performance depends on the Modeling and Simulation of a Fluidized Catalytic
reactants that are available for combustion. The Cracking Process - Part I: Process Modeling.
influence of a certain variable is therefore dependent Chemical Engineering Science, Vol. 56, No. 5, pp.
on whether the combustion regime is partial or 1951-1971.
complete. Han, I. S. and Chung, C. B., 2001b, Dynamic
Modeling and Simulation of a Fluidized Catalytic
ACKNOWLEDGMENTS Cracking Process - Part II: Property Estimation and
Simulation. Chemical Engineering Science, Vol. 56,
The authors wish to express their gratitude for No. 5, pp. 1973-1990.
the financial support provided by Kunii, D. and Levenspiel, O., 1996, Circulating
i) the Brazilian National Oil Agency (ANP) and Fluidized-Bed Reactors. Chemical Engineering
FINEP through the ANP Human Resources Program Science, Vol. 52, No. 15, pp. 2471-2482.
for the Oil and Gas Sector (PRH-10/UTFPR-PR); Lasarin, M. A., 1997, Modeling and Simulation
ii) FINEP/PETROBRAS (the OCUCC project); of a FCC Converter Unit (Modelagem e Simulao
and iii) the CAPES Foundation. da Seo de Converso de uma Unidade FCC). PhD
The authors would also like to thank Thesis, UNICAMP Department of Chemical
PETROBRAS/SIX for the experimental data Engineering. Campinas, SP, Brazil: (in Portuguese).
provided for the steady-state comparison. Lee, L. S., Yu, S. and Cheng, C., 1989,
Fluidized-Bed Catalyst Cracking Regenerator

Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86 85

Cincia/Science Penteado et al. Numerical Approaches for a Mathematical

Modelling and Analysis. The Chemical Engineering

Journal, Vol. 40, pp. 71-89.
Levenspiel, O., 1998, Chemical Reaction
Engineering. John Wiley and Sons Inc.
Malay, P., Milne, B. J. and Rohani, S., 1999,
The Modified Dynamic of a Riser Type Fluid
Catalytic Unit. The Canadian Journal of Chemical
Engineering; Vol. 77, pp. 169-179.
McFarlane, R. C., Reineman, R. C, Bartee, J. F.
and Georgakis, C., 1993, Dynamic Simulator for a
Model IV Fluid Catalytic Cracking Unit. Computers
& Chemical Engineering, Vol. 17, No. 3, pp. 275-
300.
NACS. Eger Murphree and the four Horsemen:
FCC, Fluid Catalytic Cracking, The North American
Catalysis Society, 2003. See also:
http://www.nacatsoc.org.
Patankar, S. V., 1980, Numerical Heat Transfer
and Fluid Flow. New York: Hemisphere Publishing
Corporation.
Santos, M. G., 2000, Dynamic Model for
Control of a UOP FCC Unit (Modelo Dinmico para
o Controle do Conversor de uma Unidade de FCC
UOP STACKED). MSc Thesis, UFRGS Department
of Chemical Engineering, Porto Alegre, RS, Brazil
(in Portuguese).

86 Engenharia Trmica (Thermal Engineering), Vol. 7 No 01 June 2008 p. 71-86