You are on page 1of 10

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

Contents lists available at ScienceDirect

Chemical Engineering Research and Design


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

Simulation of an industrial turbulent fluidized bed


reactor for n-butane partial oxidation to maleic
anhydride
A. Romano a , A. Di Giuliano a, , K. Gallucci a , P.U. Foscolo a , C. Cortelli b ,
S. Gori b , M. Novelli b
a
b

Industrial Engineering Department, University of LAquila, 18 via G. Gronchi, 67100 LAquila, Italy
Polynt S.p.A, 51 via E. Fermi, 24020 Scanzorosciate, Bergamo, Italy

a r t i c l e

i n f o

a b s t r a c t

Article history:

A fluid-dynamic and chemical model of an industrial-scale turbulent fluidized-bed reactor,

Received 11 April 2016

which performs n-butane partial oxidation to maleic anhydride (MA) catalyzed by (VO)2 P2 O7

Received in revised form 4 July 2016

solid particles, is proposed. Main assumptions are: gas plug-flow in the dense bed because

Accepted 1 August 2016

gas back-mixing is minimized by internals; perfectly mixed particulate phase in the dense

Available online 9 August 2016

bed due to high recirculation rate of entrained particles at the bottom of the reactor, after
separation by cyclones; plug-flow of both gas and solids in the freeboard. Literature kinetic

Keywords:

models are considered to simulate reactions, slightly modified to allow careful prediction of

Maleic anhydride production

industrial performance. The model is organized in ordinary differential and algebraic equa-

Turbulent fluidized bed

tions describing mass and energy balances and constitutive expressions for reaction rates,

Fluid dynamic-kinetic model

heat and mass transport phenomena, and implemented in MATLAB . Numerical results
simulate accurately temperature along the reactor, MA yield and by-products selectivity
in the outlet stream, and allow to estimate the solid circulation rate. Inlet gas flow rate
and n-butane/air feeding ratio are varied within a 10% range around industrial operating conditions, to test the parametric sensitivity of the model: corresponding dense bed
height, n-butane conversion, products selectivity in the outlet stream vary within 5% of
the experimental performances, depicting stable conditions.
2016 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1.

Introduction

The partial oxidation of n-butane to maleic anhydride (MA) catalyzed by


vanadyl pyrophosphate (VO)2 P2 O7 (Reaction 1) is a successful example

scale (Ballarini et al., 2006).


C4 H10 + 3.5O2 C4 H2 O3 + 4H2 O

H0 298K = 1268 kJ/mol


(Reaction 1)

of light alkanes conversion to valuable chemicals by high-temperature


contact with a redox catalyst (Ballarini et al., 2006; Hutchings, 1993).
Hydrocarbon selective oxidation processes are quite popular in indus-

C6 H6 + 4.5O2 C4 H2 O3 + 2CO2 + 2H2 O

try, specifically in the context of alkane functionalization due to their

H0 298K = 1880 kJ/mol


(Reaction 1a)

lower cost compared to olefins (Hutchenson et al., 2010). Since the 80s,
processes based on this reaction have been replacing the more traditional production of MA from benzene (Reaction 1a) on commercial

It is worth mentioning here that Reaction 1 has a higher chemical


efficiency: all carbon atoms in n-butane molecule are converted to the

Corresponding author.
E-mail address: andrea.digiuliano@graduate.univaq.it (A. Di Giuliano).
http://dx.doi.org/10.1016/j.cherd.2016.08.001
0263-8762/ 2016 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

80

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

desired product, with evident economic and environmental advantages


in comparison to Reaction 1a, where two carbon atoms are wasted in
the formation of CO2 for each reacting benzene molecule.
ALMA (Alusuisse Italia-Lummus Crest) process, jointly developed
and licensed by Lummus Technology Inc. and Polynt SpA, operates in
turbulent fluidized bed conditions; Polynt reactor is located in Ravenna
(Italy), with a nominal MA capacity of 65,000 tons per year (Tecnon
OrbiChem, 2014, 2015).
n-Butane partial oxidation to MA occurs in fixed-, fluidized- or
transport bed reactors (Lohbeck et al., 2000). The ALMA process is
heterogeneously catalyzed by vanadyl pyrophosphate (VO)2 P2 O7 fine
particles and uses fluidized bed technology for producing high purity
MA (>99.9%w , downstream the plant purification section) (Stefani et al.,
1990). This reactor choice is dictated by the noticeable heat of reaction,
to limit temperature and concentration gradients within the catalyst
and among different bed regions; the fluid dynamic behavior of these
systems is characterized as Geldart Group A (Gibilaro, 2001). Polynt SpA
have been operating ALMA plant since 1994. In the ALMA reactor unit
in Ravenna the superficial fluidization velocity allows to obtain a turbulent fluidization regime. The dense bed temperature is below 723 K;

On the other hand, gas back-mixing reaches a maximum in correspondence of the transition velocity between bubbling and turbulent
regimes and decreases with increasing velocity (Li and Wu, 1991; Bai
et al., 1992). A number of the fluid-dynamic features of the turbulent
regime, such as good gassolid contact, restrained gas back-mixing,
also due to the presence of several internals in the reactor (Zhang et al.,
2008), and uniform temperature in the dense bed are of paramount
importance to switch selectivity to MA production (heterogeneous catalytic reaction) from gas-phase combustion of n-butane, increasing the
reactor yield and reducing the risk of local temperature runaways. In
particular, the temperature control of the bed reduces MA and n-butane
combustion risks that are favored by an increase of temperature.
As far as chemical reaction models are concerned, some literature
studies address the kinetics for MA production process (Centi et al.,
1985; Schneider et al., 1987; Bej and Rao, 1991; Mills et al., 1999; Sharma
et al., 1991). Most of them assume a simplified triangular reaction
network and obtain the pre-exponential factor and activation energy
values for the relevant reaction kinetic expressions under fixed bed
operating conditions, involving low n-butane partial pressure in air
(1.82.7% n-butane volume percentage (Sharma et al., 1991)). Additional

the reactor pressure is about 24 bar (Dente et al., 2003). An alternative,


Circulating Fluidized Bed (CFB) technology, including catalyst reduc-

studies are performed under high n-butane concentration in aerobic


(ALMA process) and anaerobic (DuPont process (Lohbeck et al., 2000))

ing and oxidizing reactors linked together by a solid circulation loop,


was also developed by E.I. du Pont de Nemours, for partial oxidation
of n-butane to MA (Patience and Bockrath, 2010), however commercial
exploitation of this process has been stopped.

conditions and more detailed kinetic schemes are considered (Gascn


et al., 2006; Lorences et al., 2004). In this paper, to keep reasonable the
burden of calculations, it has been chosen to use the simplified triangular reaction network proposed in different kinetic studies quoted

Air/n-butane mixtures can be flammable in relation to their composition ratios, so in principle it is risky to operate simultaneously inside
the flammability limits and above the ignition temperature (Sharma

below (see Table 2), specifically that described by Sharma et al. (1991);
the values of the overall kinetic parameters have been properly tuned,

et al., 1991). At room temperature, this mixture is characterized by low


and high flammability limits equal to 1.8%v and 8.5%v of n-butane,
respectively (Centi et al., 2012). The spontaneous ignition temperature at the stoichiometric concentration ratio is 704 K (Sharma et al.,
1991). In a turbulent fluidized bed reactor, the presence of a well dispersed particulate solid phase inhibits the mechanisms leading to an
explosive reaction; this has been clearly demonstrated in a study on

taking into account the industrial reactor operating performance and


compared with literature estimates of the corresponding activation
energies.
As regards simulation of turbulent fluidized beds, a quite large number of fluid dynamic models were developed to describe the reactor
behavior. Turbulent-fluidized-bed reactor models often assume single
phase, either one dimensional plug flow (Van Swaaij, 1978; Fane and

methane homogeneous combustion (Van der Vaart, 1992). The ability


of the granular phase to suppress non-selective homogeneous com-

Wen, 1982), or a continuous stirred tank reactor (CSTR) (Wen, 1984) or


axially dispersed plug flow (ADPF) (Wen, 1984; Edwards and Avidan,
1986). Two-phase models, adapted from bubbling fluidization regime,

bustion and at the same time to promote the selective heterogeneous


reactions is documented in the literature also for n-butane catalytic
oxidation (Hutchenson et al., 2010). In such conditions, it is possible
to operate safely with n-butane/air ratios higher (about 45% n-butane

were also developed, with interchange of gas between dilute and dense
phase regions (Krambeck et al., 1987; Foka et al., 1996); nevertheless,
the representation of the transient and fuzzy nature of voids in turbulent regime is not satisfactory (Thompson et al., 1999). In order to

volume percentage) than in a fixed bed where more severe explosion


problems exist, together with more stringent limitations to dissipate
the reaction heat.
A number of commercial gassolid fluidized-bed reactors aimed at

overcome this a Generic Fluidized Bed Reactor model (GFBR) (Abba


et al., 2002, 2003) was developed: different fluidization regimes (bubbling, turbulent, fast) are simulated for a two-regions (i.e. dense bed
and freeboard) fluidized bed, by transitions from a two-phase bubble

different processes, such as FischerTropsch synthesis, acrylonitrile


production, FCC regeneration and silicon chlorination, are routinely

model typical at low superficial velocities (bubbling fluidization), passing through an axially dispersed plug-flow reactor model (turbulent
fluidization), arriving to a core-annulus model at higher velocities (fast

operated in the turbulent fluidization regime (Du et al., 2002). This flow
regime is commonly considered to take place between bubbling and
fast fluidization regimes (Bi et al., 2000). When fine particle beds are
fluidized at a sufficiently high gas flow rate, instead of bubbles, one
observes a turbulent motion of solid clusters and voids of gas of various
sizes and shapes (Kunii and Levenspiel, 1991). This regime offers distinct operational advantages over different fluidization regimes, such
as vigorous gassolids contact, over the bubbling regime, high solid

fluidization).
The model proposed in this paper adopts a two-regions approach
similar to GFBR (Abba et al., 2003), focusing on correspondence between
simulation results and actual industrial measurements and on representation of turbulent fluidization, with a proper deployment of
calculation resources. In fact, the aim of our model is to faithfully
simulate performances of an actual industrial turbulent fluidized bed

holdup, over the fast fluidization regime, and altogether intense solid
mixing determining a prevailing emulsion phase with relative spatial
uniformity in flow properties (Du et al., 2002).

reactor, strictly designed to operate in this fluid dynamic condition.

In turbulent fluidization conditions, the gas volume fraction is


greater compared to the bubbling regime (Bi et al., 2000), assuring a
better expansion of the bed. One of the most attractive features, partic-

2.

ularly with highly exothermic or endothermic reactions, is an uniform


temperature field, due to intense solids back-mixing, greater than in
bubbling and fast fluidization regimes (Du et al., 2002). In a mixing
study on a commercial scale unit, it was suggested that the solids axial
dispersion coefficient is proportional to the square root of the bed diameter (Wei et al., 1993). From the above findings follows that the particle
bed closely approaches perfect mixing conditions in a large industrial
reactor like that considered here (ID > 5 m).

Model description

In the case examined here, the industrial reactor configuration


suggests to assume plug flow behavior for the gas phase, due
to the presence of distributed reactor internals (heat transfer
tube bundles and baffles), and perfect mixing for the particulate phase, according to uniform dense bed temperature
measured in all production campaigns. On the other hand, in
the reactor freeboard above the bed, both gas and entrained
solid phases are described by the plug-flow assumption. Actu-

81

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

Table 1 Model equations.


Mole balances

dFi
= b ri S; i = 1, . . ., 7
dZ

(1)

Fw cp ,w (Tw,in Tw,out ) =

(2)

Ldb

Energy balance for


the pressurized
cooling water in
the dense bed
Pressure drop along
the reactor
Global energy
balance in the
dense bed (Fogler,
2005)
Energy balance in
the freeboard
Energy balance for
the pressurized
cooling water in
the freeboard

hAS (Tw Tdb ) dZ


0

dP
= b g
dZ

(3)
Ldb


Fg,in cp,g Tg,in Tdb + Fw cp,w (Tw,in Tw,out ) + Fs cp,s (Ts,in Tdb ) 

|rij (Z) |b (Z) SdZHij = 0; j = 1, . . . , 3

(4)

j
0

hAS (Tw Tfb ) Sb 3j=1 |rij (Z) |Hij


dTfb
=
Fg cpg + Fs cps
dZ

(5)

Tfb Tw
dTw
= hAS
Fw cp,w
dZ

(6)

Table 2 Activation energy values proposed in this work compared to literature studies.
Kinetic model

EA1 [kJ/mol]

Centi et al. (1985)


Buchanan and Sundaresan (1986)
Schneider et al. (1987)
Sharma et al. (1991)
Mills et al. (1999)
Dente et al. (2003)
This work

45.1
125
72
93
71
79
60

EA2 [kJ/mol]
110
145
7472a
93
1367a
n. p.
45

EA3 [kJ/mol]
57.4
180
n.a.
155
n.a.
243
190

n.a.: In these studies, a parallel reaction model is considered, which does not include MA oxidation, so that this activation energy is not available.
n.p.: Not published.
a

Two different reactions and corresponding activation energy values are considered for n-butane oxidation to CO2 and CO, respectively.

ally, recent studies are focusing on CFD simulation of the


hydrodynamic behavior in the turbulent fluidized bed using
the standard EulerianEulerian two-fluid model and in this
frame the kinetic theory of granular flow (Gidaspow, 1994;
Wang, 2009, 2010; Xu et al., 2012; Yao et al., 2014; Chen et al.,
2015). This approach would be too expensive and time consuming when applied to the simulation of the MA industrial
reactor, due to its dimensions and the presence of distributed
internals (a complex geometry). Only a very fine grid, in fact,
could identify such complicated structures, while a loose grid
would distort the computational results (Wang, 2009).
In the model, the industrial reactor layout is schematized
as in Fig. 1. The air grid is located at the bottom of the reactor,
while a sparger is used to feed n-butane. The gaseous phase
products and the elutriated solid particles leave the reactor
at the top and enter a cyclone, where solids are separated
and then recycled to the bottom of the bed. There are two
distinct zones inside the reactor, dense bed and freeboard, in
which reactions occur and the heat of reactions is transferred
to pressurized water flowing in the tube bundles.

2.1.

Conservation equations

The following modeling assumptions are employed:

reactor layout, bed inventory and inlet conditions are fixed


according to plant data;

plug-flow of the gas phase and complete mixing of the solid


phase in the dense bed;
uniform temperature in the dense bed;
negligible temperature differences between solid and gas
phases;
pressure gradients are entirely due to the weight of the catalyst holdup per unit reactor height, while inertial and wall
effects are negligible;
complete separation of catalyst particles from the product
gas in the cyclone, according to plant data showing negligible catalyst recovery in the subsequent filter unit;
progressive temperature decrease with reactor length in the
freeboard, due to heat transfer to pressurized water flowing
in tube bundles located along the reactor.

The model equations are the following ones:

dense bed region: a system of nine differential conservation equations (seven molar balances for chemical species,
including nitrogenthe inert component, the energy balance for the cooling water flowing in the tube bundles,
the momentum balance along the reactor to determine the
pressure profile), plus a global energy balance referred to the
whole dense bed volume to calculate its temperature level;
reactor freeboard region: ten differential equations to
describe concentrations, temperature, heat transfer and
pressure along the reactor length.

82

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

Cyclone

is considered constant above the TDH (Total Disengagement


Height); TDH is evaluated by means of empirical relations proposed by Zenz and Wei (1958), Amitin et al. (1968) and Horio
et al. (1980). In this case, TDH resulted always lower than the
freeboard height.
Evaluations for the elutriated catalyst flux, Gs , are available
from the plant data and have been utilized in the model equations; they are in fairly good agreement with values calculated
from literature correlations (Choi et al., 1999):
The knowledge of the overall catalyst inventory in the reactor volume, Ms , allows to determine the dense bed height by
a trial and error procedure addressed to match the particulate
solid mass balance:

Product
gas

Freeboard

Water In

Gas and
entrained
catalyst
(PFR)

Water Out

Dip leg for


catalyst recycle

Water Out

Ldb


Lfd
b (Z) SdZ +

Dense bed

Water In

hw =

n-butane

(12)

Air

(13)

In the freeboard, the outer heat transfer coefficient from


the fluidized bed to the tube surface is evaluated as suggested
in the literature (Kunii and Levenspiel, 1991), using the same
decay constant as in Eq. (9):

The corresponding mathematical expressions are summarized in Table 1.


In the equations in Table 1, the index i denotes a gas component, reactant or product, while j denotes the chemical
reaction. The gas flow rate Fg includes also the inert component, nitrogen. In the energy balances i is the reference
component for the heat of reaction, Hij .
To estimate the dense bed height, a modified
RichardsonZaki equation is applied to correlate the average voidage with the superficial gas velocity (Avidan and
Yerushalmi, 1982; Bi et al., 2000):
(7)

where expressions to estimate Ut and n are provided by


Venderbosch (1998) and the superficial gas velocity is evaluated as a function of bed height, assuming ideal gas law:
(8)

The catalyst volume fraction in the freeboard is assumed to


decay exponentially with reactor length, from the value at the
top surface of the dense bed, db (Z = Ldb ) = 1 db (Z = Ldb )
(Kunii and Levenspiel, 1991):
 =  + (db  ) exp (aZfb )



0.44
kg
0.66Prg0.3 g (1 ) /g ReD
Dt

ReD = Dt Ug /g

Fig. 1 Simplified model representation of the industrial


n-butane partial oxidation reactor (nominal MA
capacity = 65,000 tons per year).

U = Ftot RT/ (SP)

(11)

The heat transfer coefficient for immersed horizontal tubes


in the dense bed is evaluated according to Vreedenberg (1958):

Gas Solids
(PFR) (CSTR)

U/Ut = ndb

b (Z) SdZ = Ms
Ldb

(9)

hfb = (hwdb hw ) exp (aZfb /2) + hw

(14)

where hwdb is evaluated by Eq. (12) at the top of the dense bed
while hw is that evaluated at the reactor head.
The tube side convective heat transfer contribution is evaluated by means of the correlation:
Nu = 0.023Re0.8 Pr0.33

2.2.

(15)

Reaction kinetics

Catalyst fine particles and a rather vigorous gas velocity are


adopted so that intra-phase and inter-phase mass transfer
limitations are negligible and consequently there is not a mass
transfer control on the reaction process. As mentioned above,
a simplified triangular reaction network is adopted: together
with the desired heterogeneous reaction of n-butane to maleic
anhydride (Reaction 1), an undesired homogeneous oxidation
reaction is assumed to occur in parallel (Reaction 2). Finally,
the catalytic partial combustion of maleic anhydride closes
the network (Reaction 3). The stoichiometry of secondary reactions and their rate Eqs. (16) and (17) are:

C4 H10 + 5.5O2 2CO + 2CO2 + 5H2 O

H0 298K = 2092 kJ/mol

(Reaction 2)

where:
 = Gs / ((U Ut ) p )

(10)

and a is given by (Abba et al., 2003). Gs = Fs /S is the elutriated catalyst flux at the reactor head and Ut is the particle
terminal velocity. Gs corresponds to the elutriation rate, which

r2 = k2 p0.54
n-but

C4 H2 O3 + O2 4CO + H2 O

(16)

H0 298K = 258 kJ/mol


(Reaction 3)

83

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

Table 3 Pre-exponential factor for the three relevant


reaction kinetics, this work and Sharma et al. (1991).
Parameter

This work

Sharma et al.

k1 673 [mol kg1 s1 atm0.54 ]


k2 673 [mol kg1 s1 atm0.54 ]
k3 673 [mol kg1 s1 atm1 ]
KMA [atm1 ]

2.20 103
0.30 103
0.22 102
185

0.96 103
0.15 103
0.29 102
310

r3 = k3 pMA /(1 + KMA pMA )2

(17)

The rate equation for Reaction 1 is:


r1 = k1 p0.54
/ + KMA pMA )
nbut (1

(18)

The above kinetic expressions are those proposed by


Sharma et al. (1991). The oxygen partial pressure does not
appear explicitly: as the matter of fact, the plant data show
that the reactor always operates with an oxygen excess.
The rate constants are given by an Arrhenius type temperature dependency:
ki = ki673 exp

 E
Ai
673R

(1 673/T)


(19)

Values for activation energies and pre-exponential factors


found in the literature mainly depend on the range of experimental conditions, as already mentioned.
Finally, the reaction rate expressions for each reactant and
product species, to be used in the equations of Table 1, follow immediately from stoichiometric relations of Reactions
(1)(3), together with the corresponding constitutive kinetic
expressions given by Eqs. (16)(18) and the kinetic parameters
reported in Tables 2 and 3, in Section 3.1.1.

2.3.

Numerical procedure to integrate model equations

The numerical integration was implemented in MATLAB


using the integration routine ode45, based on an explicit
RungeKutta numerical scheme of 4th order. The input data
include: reactor geometry; overall catalyst inventory; particle density, equivalent diameter and specific heat capacity;
inlet gas flow rate, air/n-butane ratio, temperature and pressure; cooling water flow rate and heat transfer surface per unit
reactor volume. Before the integration, dense bed temperature
and height, together with solids recirculation rate, need to be
assumed and iteratively adjusted until reactor steady conditions are reached: this preliminary step completes the input
data set to finally solve the system of model equations.

3.

Results and discussion

In order to validate the predictive capabilities of the proposed model, simulations were carried out at typical operating
conditions of the MA reactor. The numerical calculations are
compared with experimental results in terms of pressure drop
and temperature profile, n-butane conversion (20), MA selectivity (21), CO selectivity (22) and the ratio CO/CO2 .
Xn-but = (Fg in xin n-but Fg out xout n-but )/(Fg in xin n-but )

(20)

SMA = (Fg out xout MA )/(Fg in xin n-but Fg out xout n-but )

(21)

Fig. 2 Catalyst bulk density profile along reactor height


(normalized by overall reactor height, L).
SCO = (Fg out xout CO )/(4(Fg in xin n-but Fg out xoutn-but ))

(22)

Then a sensitivity parametric study is carried out to check


the ability of simulations to comply with control strategies
usually adopted at the plant site and to help investigating
process optimization measures.

3.1.
Comparison between numerical calculations and
experimental results
3.1.1.

Kinetic constants

The pre-exponential factor and activation energy for Reactions (1)(3) were obtained starting from the corresponding
values proposed by Sharma et al. (1991) and optimizing them
to allow a good fitting of the industrial performance. The
activation energy values, finally identified as a result of this
procedure, are compared with literature results in Table 2.
As shown in Table 2, the activation energy values fixed in
this study are well inside the range of values proposed in the
literature. This result provides a general indication that the
reactor model proposed here is able to combine correctly the
fluid-dynamic and chemical aspects of the reacting system
under consideration.
As far as the pre-exponential factor is concerned, a direct
comparison is possible only with Sharma et al. (1991), as
reported in Table 3. It should be considered that in Sharma
et al. (1991) the experimental concentration ratio n-butane/air
was different from that adopted in the industrial reactor.

3.1.2. Catalyst bulk density and pressure profile along the


reactor height
In this section, simulation results are presented, in dimensionless form, for the catalyst bulk density and pressure drop
profile along the reactor height and catalyst recirculation rate.
Fig. 2 shows the catalyst bulk density, b , as a function of reactor height. The corresponding pressure drop along the reactor
height is calculated as shown in Fig. 3. Numerical results in
Fig. 2 are in very good agreement with the pressure drop trend
recorded at the plant site (measured values in Fig. 3) and allow
to determine the dense bed height inside the reactor.
As mentioned above, the catalyst recirculation rate predicted by model equations is close to that one evaluated at
the plant site. The renewal time of the solid reactor hold up is

84

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

Fig. 3 Dimensionless pressure profile as a function of


dimensionless reactor height: comparison of numerical
results () with experimental data (). Measured pressures
oscillate between 5% of the respective average values.

Fig. 4 Dimensionless temperature profile as a function of


the dimensionless reactor height: comparison of numerical
results () with experimental data (). Measured
temperatures oscillate between 1% of the respective
average value.
estimated to be of the order of ten minutes. The consequent
considerable flux of solid from cyclones enters the reactor at
its bottom section at significantly lower temperature than the
dense bed temperature: this contributes to avoid hot-spots in
the reactor region where the reactant concentrations are maximum (i.e. the bottom) and to keep the dense bed temperature
reasonably constant, as shown in the next section.

3.1.3.

Temperature profile

Very good agreement is found between calculated and measured temperature in the dense bed and for most of the
freeboard region. The highest heat exchange tube bundle is
located just below 80% of the overall reactor height. The
reaction rate is also quite small there, so that the model predicts substantially constant temperature above 80% of reactor
height. On the other hand, the temperature measured in the
cyclone inlet pipe is somehow lower (Fig. 4, Z/L = 1); this difference is mainly ascribable to local gas cooling at the exit of
the reactor due to an increase of heat losses to the ambient,
close to the reactor ceiling and in the exit pipe.

Fig. 5 (a) Normalized n-butane conversion predicted by the


model as a function of the dimensionless reactor height: (b)
MA and CO selectivity, and CO/CO2 ratio predicted by the
model as functions of the normalized n-butane conversion,
compared with corresponding experimental data recorded
at the exit of the reactor. Measured selectivities and CO/CO2
ratio oscillate between 2% of the respective average value.
The catalyst inlet temperature, Ts,in , at the bottom of the
reactor is chosen to be 20 K below the exit temperature from
the top of reactor. Ts,out , according to plant estimates:
T s,in = T s,out 20

3.1.4.

(23)

MA yield and selectivities

In Fig. 5a, the n-butane conversion is shown as a function of


the reactor height, while in Fig. 5b MA and CO selectivity and
CO/CO2 ratio are reported as functions of the n-butane conversion; activation energy and pre-exponential factor values are
those reported in Tables 2 and 3, respectively. Model predictions are compared to the experimental data recorded at the
exit of the reactor.

3.2.
model

Parametric sensitivity analysis of the simulation

The simulation model is now applied to examine the influence


of small perturbations in inlet parameters on the reactor performance; perturbed parameters are inlet gas flow rate and

85

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

Variaon calculated by the model


[%]

n-BUTANE CONVERSION

MA SELECTIVITY

DENSE BED HEIGHT

-3

-1

-2

-6

-2
-10

-4
-10

10

-10

10

10

Imposed perturbaon of the input variable [%]

Variaon calculated by the


model [%]

CATALYST RECIRCULATION RATE

CO/CO2 RATIO

OUTLET O2 MOLAR FRACTION

40

30

20

15

-20

-1

-15

-40

-2
-10

10

-30
-10

-10

10

10

Imposed perturbaon of the input variable [%]


Fig. 6 Parametric sensitivity analysis of the simulation model: predicted percentage variation of n-butane conversion, MA
selectivity, dense bed height, catalyst recirculation rate, CO/CO2 ratio and outlet O2 molar fraction as functions of 10%
change of input variables: actual gas inlet flow rate () and n-butane/air ratio (), respectively.
10
Variaon of steam producon [%]

n-butane/air feeding ratio. In real circumstances, its important to keep the temperature of the dense bed at its steady
state level because of catalyst sensitivity to this quantity:
the catalyst may change its constitutive phases and its activity (Gabriele et al., 1988), so that kinetic parameters utilized
in model equations would become less reliable. As a consequence, in each simulation the heat transfer surface in the
reactor is appropriately increased or reduced, in such a way
as to keep the temperature profile almost constant: this corresponds closely to the real operative procedure at the plant site
to control the reactor temperature: opening or closing valves
regulating the flow of pressurized water through internal tube
bundles, so controlling the number of tubes actually operating
for heat removal.
Fig. 6 shows effects on n-butane conversion, MA selectivity,
dense bed height, catalyst recirculation rate, CO/CO2 ratio and
oxygen molar fraction at reactor exit, due to a variation within
a 10% range of the above mentioned inlet parameters. Fig. 7
reports the corresponding effects on the steam production
(proportional to the heat subtracted from the reactor by the
pressurized water flow) required to keep constant the dense
bed temperature level at each operating condition.
The reactor results to operate at quite stable conditions:
variations in quantities (except for oxygen molar fraction and
catalyst recirculation rate) at reactor exit (about 5% maximum
relative deviation) are smaller than the amplitude of fluctuation imposed to inlet parameters. On the other hand, the
corresponding variation of steam production required to keep
the dense bed temperature (Fig. 7) shows that the reactor
behavior is quite sensitive to this quantity: the model confirms
the appropriateness and reliability of the temperature control
strategy implemented in the actual reactor and mentioned

-5

-10
-10

-5
0
5
Imposed variaons of the input variable [%]

10

Fig. 7 Parametric sensitivity analysis of the simulation


model: predicted percentage variation of steam production
required to keep constant the dense bed temperature level
at each simulated condition, as a function of 10% change
of input variables: actual gas inlet flow rate () and
n-butane/air ratio (), respectively.
above. It is worth stressing here that simulations take into
consideration variations in the reactor hydrodynamic behavior brought about by each perturbed state: parameter values
are properly adjusted to different operating conditions (Eqs.
(7)(15)) and the constraint of turbulent fluidization regime is
always satisfied, so that model predictions are expected to be
reliable, also because of previous validation with the reactor
real outputs.
In the reactor energy balance, an additional operating
parameter is represented by the heat flow associated to the
particulate catalyst recirculated to the bottom section (Fig. 1):
the exothermic reactions extent is particularly relevant in the

Variaon of dense bed temperature [%]

86

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

1.5

Reference operating conditions


-1.5

-3
-1

0
1
Variaon of catalyst inlet temperature [%]

Fig. 8 Model prediction of dense bed temperature as a


function of the temperature variation of catalyst
recirculated to the bottom of the reactor.
reactants inlet region, so that solid recirculation rate is very
important to assure reactor thermal stability (making available its heat capacity in that region), in addition to a proper
recovery of the elutriated catalyst. Simulations have been performed keeping constant the catalyst recirculation rate, found
in Section 3.1, and assuming small deviations of its inlet temperature. With relation to the reference operating conditions,
small perturbations of the catalyst recirculation temperature
determine an appreciable variation in dense bed temperature,
as reported in Fig. 8, showing the important role of the catalyst
heat capacity to control the reactor temperature in the feeding
zone, in addition to the heat transfer system, the influence of
which is already examined in Fig. 7.

4.

Conclusions

The focus of this paper is on fluid-dynamic and chemical simulation of an industrial scale, turbulent fluidized-bed
reactor for n-butane catalytic partial oxidation with air to
maleic anhydride: model equations and constitutive expressions were fully described, assumptions and input variables
were carefully identified so that numerical solutions can be
obtained for diversified operating conditions of interest.
The model assumes reasonable gas and solid dispersion
conditions in the dense bed and in the freeboard, literature
correlations to estimate turbulent fluidization quality and a
simplified reaction kinetic model, in which activation energies and pre-exponential factors are adjusted to match the
real plant operating data.
The model can simulate faithfully the industrial reactor
behavior and highlight the role of major variables, such as the
heat exchange system or the heat capacity and temperature
of catalyst recirculated at the bottom of the reactor.
A parametric sensitivity analysis was also performed, considering the influence of inlet flow rate and n-butane/air ratio
on n-butane conversion, MA selectivity, dense bed height,
catalyst recirculation rate, CO/CO2 ratio, outlet reactor temperature and steam production by the heat transfer system.
This analysis depicts a reactor operating in stable conditions,
with relatively small changes of these quantities in presence
of inlet parameter perturbations. This result agrees with the
evidences emerged from actual MA production campaigns in
the real plant.
In force of these experimental proofs and despite its simple, however not simplistic, approach, the model developed
in this work is proposed as a useful tool to design and apply
appropriate reactor control strategies.

Notations
A
Heat-transfer surface area per unit reactor volume
[m1 ]
Decay constant [m1 ]
a
Archimedes number [dimensionless]
Ar
Particle drag coefficient based on the superficial gas
CD
velocity [dimensionless]
Heat capacity [kJ mol1 K1 ] [kJ kg1 K1 ]
cp
Cooling tube diameter [m]
Dt
Particle equivalent diameter [m]
dp
fd
Drag force on the particle per projection area [Pa]
Gravity force minus buoyancy force per projection
fg
area of particle [Pa]
Fg
Gas flow rate [mol s1 ]
Solid flow rate [kg s1 ]
Fs
Water flow rate [kg s1 ]
Fw
g
Acceleration of gravity [m s2 ]
Gs
Solids flux [kg m2 s1 ]
Overall bed-to-surface heat-transfer coefficient
h
[kJ s1 m2 K1 ]
hw
Outer heat transfer coefficient [kJ s1 m2 K1 ]
KMA
Maleic anhydride adsorption rate constant [atm1 ]
Gas thermal conductivity [kJ s1 m1 K1 ]
kg
Water thermal conductivity [kJ s1 m1 K1 ]
kw
Overall reactor height [m]
L
Ldb
Dense bed height [m]
Freeboard height [m]
Lfb
Ms
Overall catalyst inventory [kg]
Nusselt number Nu = hwin Dt /kw , dimensionless
Nu
Pressure [atm]
P
Pr
Prandtl number Pr = cpw w /kw , dimensionless
Reaction rate [mol kg1 s1 ]
r
Re
Reynolds number, Re = Dt U/ [dimensionless]
ReD
Reynolds number based on tube diameter [dimensionless]
Particle
Rep
Reynolds
number
Rep =
p dp U/, [dimensionless]
S
Bed cross-sectional area [m2 ]
CO selectivity [dimensionless]
SCO
SMA
MA selectivity [dimensionless]
Temperature [K]
T
Superficial velocity [m s1 ]
U
Ut
Particle terminal velocity [m s1 ]

Effective terminal cluster velocity [m s1 ]


Ut
Xn-bu
n-Butane conversion [dimensionless]
Component molar fraction [dimensionless]
x
Axial coordinate [m]
Z
Greek letters
H
Heat of reaction [kJ mol1 ]
Bed voidage [dimensionless]

Viscosity [Pa s]

Bulk density, b = p (1 ), [kg m3 ]
b
Gas density [kg m3 ]
g
p
Particle density [kg m3 ]
Catalyst volume fraction [dimensionless]

Subscripts
Dense bed
db
Freeboard
fb
g
Gas
Particle size index
h
i
Component index
Inlet
in

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

j
out
s
w

Reaction index
Outlet
Solid
Water

References
Abba, I.A., Grace, J.R., Bi, H.T., 2002. Variable-gas-density fluidized
bed reactor model for catalytic processes. Chem. Eng. Sci. 57
(2223), 47974807,
http://dx.doi.org/10.1016/S0009-2509(02)00289-0.
Abba, Ibrahim A., Grace, John R., Hsiaotao, Bi, Thompson, M.L.,
2003. Spanning the flow regimes: generic fluidized-bed reactor
model. AIChE J. 49 (7), 18381848,
http://dx.doi.org/10.1002/aic.690490720.
Amitin, A.V., Martyushin, I.G., Gurevich, D.A., 1968. Dusting in the
space above the bed in converters with fluidized catalyst bed.
Chem. Technol. Fuels Oils 4 (3), 181184.
Avidan, A.A., Yerushalmi, J., 1982. Bed expansion in high velocity
fluidization. Powder Technol. 32, 223232.
Bai, D., Yi, J., Jin, Y., Yu, Z., 1992. Residence time distribution of
gas and solids in a circulating fluidized bed. In: Potter, O.E.,
Nicklin, D.J. (Eds.), Fluidization VII. Engineering Foundation,
pp. 195202.
Ballarini, N., Cavani, F., Cortelli, C., Ligi, S., Pierelli, F., Trifir, F.,
Fumagalli, C., Mazzoni, G., Monti, T., 2006. VPO catalyst for
n-butane oxidation to maleic anhydride: a goal achieved, or a
still open challenge? Top. Catal. 38 (13), 147156,
http://dx.doi.org/10.1007/s11244-006-0080-z.
Bej, S.K., Rao, Musti S., 1991. Selective oxidation of n-butane to
maleic anhydride. 2. Identification of rate expression for the
reaction. Ind. Eng. Chem. Res. 30 (8), 18241828.
Bi, H., Ellis, T.N., Abba, I.A., Grace, J.R., 2000. A state-of-the-art
review of gassolid turbulent fluidization. Chem. Eng. Sci. 55
(21), 47894825,
http://dx.doi.org/10.1016/S0009-2509(00)00107-X.
Buchanan, J.S., Sundaresan, S., 1986. Kinetics and redox
properties of vanadium phosphate catalysts for butane
oxidation. Appl. Catal. 26, 211226.
Centi, G., Cavani, F., Trifir, F., 2012. Selective Oxidation by
Heterogeneous Catalysis. Springer Science & Business Media.
Centi, G., Fornasari, G., Trifir, F., 1985. n-Butane oxidation to
maleic anhydride on vanadium-phosphorus oxides: kinetic
analysis with a tubular flow stacked-pellet reactor. Ind. Eng.
Chem. Prod. Res. Dev. 24, 3237.
Gabriele, Centi, Trifiro, F., Ebner, Jerry R., Franchetti, Victoria M.,
1988. Mechanistic aspects of maleic anhydride synthesis from
C4 hydrocarbons over phosphorus vanadium oxide. Chem.
Rev. 88 (IV), 5580, http://dx.doi.org/10.1021/cr00083a003.
Chen, Jie, Li, Hongzhong, Lv, Xiaolin, Zhu, Qingshan, 2015. A
structure-based drag model for the simulation of Geldart A
and B particles in turbulent fluidized beds. Powder Technol.
274, 112122, http://dx.doi.org/10.1016/j.powtec.2015.01.010,
Elsevier B.V.
Choi, Jeong-Hoo, Chang, In-Yong, Shun, Do-Won, Yi, Chang-Keun,
Son, Jae-Ek, Kim, Sang-Done, 1999. Correlation on the particle
entrainment rate in gas fluidized beds. Ind. Eng. Chem. Res.
38, 24912496, http://dx.doi.org/10.1021/ie980707i.
Dente, M., Pierucci, S., Tronconi, E., Cecchini, M., Ghelfi, F., 2003.
Selective oxidation of n-butane to maleic anhydride in fluid
bed reactors: detailed kinetic investigation and reactor
modelling. Chem. Eng. Sci. 58 (36), 643648,
http://dx.doi.org/10.1016/s0009-2509(02)00590-0.
Du, Bing, Fan, Liang Shih, Wei, Fei, Warsito, W., 2002. Gas and
solids mixing in a turbulent fluidized bed. AIChE J. 48 (9),
18961909, http://dx.doi.org/10.1002/aic.690480907.
Edwards, M., Avidan, A., 1986. Conversion model aids scale-up of
mobils fluid-bed MTG process. Chem. Eng. Sci. 41 (4), 829835.
Fane, A.G., Wen, C.Y., 1982. Fluidized bed reactors. In: Hetsroni, G.
(Ed.), Handbook of Multiphase Systems. Hemisphere
Publishing.

87

Fogler, H. Scott, 2005. Elements of Chemical Reaction


Engineering. Prentice Hall.
Foka, M., Chaouki, J., Guy, C., Klvana, D., 1996. Gas phase
hydrodynamics of a gassolid turbulent fluidized bed reactor.
Chem. Eng. Sci. 51 (5), 713723,
http://dx.doi.org/10.1016/0009-2509(95)00326-6.
Gascn, J., Valenciano, R., Tllez, C., Herguido, J., Menndez, M.,
2006. A generalized kinetic model for the partial oxidation of
n-butane to maleic anhydride under aerobic and anaerobic
conditions. Chem. Eng. Sci. 61 (19), 63856394,
http://dx.doi.org/10.1016/j.ces.2006.05.023.
Gibilaro, L.G., 2001. Fluidization Dynamics.
Butterworth-Heinemann.
Gidaspow, Dimitri, 1994. Multiphase Flow and Fluidization:
Continuum and Kinetic Theory Descriptions. Academic Press,
Inc.
Horio, M., Taki, A., Hsieh, Y.S., Muchi, I., 1980. Elutriation and
particle transport through the freeboard of a gassolid
fluidized bed. In: Fluidization. Plenum Press, New York.
Hutchenson, Keith W., La Marca, Concetta, Laviolette,
Jean-Philippe, Bockrath, Richard E., 2010. Parametric study of
n-butane oxidation in a circulating fluidized bed reactor. Appl.
Catal. A: Gen. 376 (1), 1091,
http://dx.doi.org/10.1016/j.apcata.2010.01.006.
Hutchings, J.G., 1993. Vanadium phosphorus oxide catalysts for
the selective oxidation of n-butane to maleic anhydride. Catal.
Today 16 (1), 139146,
http://dx.doi.org/10.1016/0920-5861(93)85014-Q.
Krambeck, F.J., Avidan, A., Lee, C.K., Lo, M.N., 1987. Predicting
fluid-bed reactor efficiency using adsorbing gas tracers. AIChE
J. 33 (10), 17271734.
Kunii, Daizo, Levenspiel, Octave, 1991. FLuidization Engineering
II. ButterworthHeinemann.
Li, Y., Wu, P., 1991. A study on axial gas mixing in a fast fluidized
bed. In: Basu, P., Horio, M., Hasatani, M. (Eds.), Circulating
Fluidized Bed Technology III. Pergamon Press, pp. 581586.
Lohbeck, K., Haferkorn, Herbert, Fuhrmann, Werner, Fedtke,
Norbert, 2000. Maleic and fumaric acids. In: Ullmanns
Encyclopedia of Industrial Chemistry. Wiley-VCH.
Lorences, Mara J., Patience, Gregory S., Dez, Fernando V., Coca, J.,
2004. Transient n-butane partial oxidation kinetics over VPO.
Appl. Catal. A: Gen. 263 (2), 193202,
http://dx.doi.org/10.1016/j.apcata.2003.12.023.
Mills, P.L., Randall, H.T., McCracken, J.S., 1999. Redox kinetics of
VOPO4 with butane and oxygen using the TAP reactor system.
Chem. Eng. Sci. 54, 37093721,
http://dx.doi.org/10.1016/S0009-2509(99)00008-1.
Patience, Gregory S., Bockrath, Richard E., 2010. Butane oxidation
process development in a circulating fluidized bed. Appl.
Catal. A: Gen. 376 (1), 412,
http://dx.doi.org/10.1016/j.apcata.2009.10.023.
Schneider, P., Emig, G., Hofmann, H., 1987. Kinetic investigation
and reactor simulation for the catalytic gas-phase oxidation
of n-butane to maleic anhydride. Ind. Eng. Chem. Res. 26,
22362241.
Sharma, R.K., Cresswell, D.L., Newson, E.J., 1991. Kinetics and
fixed-bed reactor modeling of butane oxidation to maleic
anhydride. AIChE J. 37 (1), 3947.
Stefani, G., Budi, F., Fumagalli, C., Suciu, G.D., 1990. Fluidized bed
oxidation of n-butane: a new commercial process for maleic
anhydride. Stud. Surf. Sci. Catal. 55, 537552,
http://dx.doi.org/10.1016/S0167-2991(08)60186-X.
Tecnon OrbiChem, 2014. M.A. Chemical Business Focus, no. 283.
Tecnon OrbiChem, 2015. M.A. Plant and Project Review, no. 53.
Thompson, Michael L., Bi, Hsiaotao, Grace, John R., 1999. A
generalized bubbling/turbulent fluidized-bed reactor model.
Chem. Eng. Sci. 54 (1314), 21752185,
http://dx.doi.org/10.1016/S0009-2509(98)00354-6.
Van der Vaart, Donald R., 1992. Mathematical modeling of
methane combustion in a fluidized bed. Ind. Eng. Chem. Res.
31 (4), 9991007, http://dx.doi.org/10.1021/ie00004a003.
Van Swaaij, W.P.M., 1978. The design of gassolids fluid bed and
related reactors. ACS Symp. Ser. 72, 193222.

88

chemical engineering research and design 1 1 4 ( 2 0 1 6 ) 7988

Venderbosch, Robertus Hendrikus, 1998. The Role of Clusters in


GasSolids Reactors. An Experimental Study. Twente
University.
Vreedenberg, H.A., 1958. Heat transfer between a fluidized
bed and a horizontal tube. Chem. Eng. Sci. 9 (1),
5260.
Wang, Junwu, 2009. A review of Eulerian simulation of geldart a
particles in gas-fluidized beds. Ind. Eng. Chem. Res. 48 (12),
55675577, http://dx.doi.org/10.1021/ie900247t.
Wang, Junwu, 2010. Flow structures inside a large-scale turbulent
fluidized bed of FCC particles: Eulerian simulation with an
EMMS-based sub-grid scale model. Particuology 8 (2), 176185,
http://dx.doi.org/10.1016/j.partic.2009.03.013, Chinese Society
of Particuology.
Wei, Fei, Lin, Shixiong, Yang, Guanghua, 1993. Gas and solids
mixing in a commercial FCC regenerator. Chem. Eng. Technol.
16 (2), 109113.
Wen, C.Y., 1984. Flow regimes and flow models for fluidized bed
reactors. In: Doraiswamy, L.K. (Ed.), Recent Advances in the

Engineering Analysis of Chemically Reacting Systems. Wiley,


pp. 256290.
Xu, M., Chen, F., Liu, X., Ge, W., Li, J., 2012. Discrete particle
simulation of gassolid two-phase flows with multi-scale
CPUGPU hybrid computation. Chem. Eng. J. 207208, 746757,
http://dx.doi.org/10.1016/j.cej.2012.07.049, Elsevier B.V.
Yao, Ya, He, Yi-Jun, Luo, Zheng-Hong, Shi, Lan, 2014. 3D CFD-PBM
modeling of the gassolid flow field in a polydisperse
polymerization FBR: the effect of drag model. Adv. Powder
Technol. 25 (5), 14741482,
http://dx.doi.org/10.1016/j.apt.2014.04.001, The Society of
Powder Technology Japan.
Zenz, F.A., Wei, N.A., 1958. Atheoreticalempirical approach to
the mechanism of particle entrainment from fluidized beds.
AIChE J., 472479.
Zhang, Yongmin, Lu, Chunxi, Grace, John R., Bi, Xiaotao, Shi,
Mingxian, 2008. Gas back-mixing in a two-dimensional
baffled turbulent fluidized bed. Ind. Eng. Chem. Res. 47,
84848491, http://dx.doi.org/10.1021/ie800906n.

You might also like