Notes on Fluidized Bed Reactors

1 Kunii-Levenspiel Model Description

The Kunii-Levenspiel (KL) model assumes four phases and in all of them the catalyst and the gas are
present. These phases are: cloud, bubble, wake and emulsion. The cloud, wake and emulsion have,
practically, the same amount of gas and catalyst, while the bubble is mainly gas and a little amount of
catalyst. There is no frontier gas-liquid type, such as the superficial tension. There is just two regions
of different convective behaviors. The wake is the volume (of emulsion) which is dragged in the lower
part of the bubble and promotes the mixing of the emulsion. The KL model is a three-phase model since
the molar balances consider the bubble, cloud and emulsion phases. The cloud completely envelopes the
bubble and it is the part of the emulsion that strongly recycles along the bubble while this rises.
vb

Emulsion

Bubble

Wake
Cloud

Figure 1: KL Model conceptual description.

The KL model assumes:

1. All the bubbles have the same size;

2. The solids within the emulsion slowly descend (as in plug-flow);
3. For the bubbling regime, the void fraction of the emulsion, e , is constant and equal to the minimum
void fraction of the fluidization, mf , and
4. The void fraction of the wake, w , is equal to that of emulsions and the velocity of the wake is equal
to that of the bubbles.

2 K-L Model Parameters

2.1 Transport Coefficients

In order to estimate the transport coefficients it is necessary to determine some variables such as diameter
of the bubble, db , minimum fluidization velocity and mf . The minimum void fraction of fluidization is
given by
0.029  0.021
2

0.72 g
mf = 0.586 (1)
g g(p p )d3p p
where dp is the diameter of the catalyst particle, p and g are catalyst and gas phase densities, and

Area of a sphere of same particle s volume

= (2)
Particle s external area
The minimum fluidization velocity is estimated by means of the Ergun equation for Re < 10:

(dp )2 3mf
mf = [g(p g )] (3)
150 1 mf

1
where Re is calculated based on the diameter of the bubble. The diameter of the bubble will depend on
the total internal diameter of the reactor above of the gas distributor, Dt , design and type of distributor
or height of the bubble with respect to the distributor, h, superficial velocity of the gas in the whole bed
and properties of the fluid, 0 , of the properties and geometric characteristics of the particles, and on the
internal configuration (tubes and baffles). The diameter of the bubble must be considerable lower than
the reactor. If these two are similar, then the operation regime is no longer bubbling but spouted. It will
be assumed that db is known. The rising velocity is calculated by:
p
b = 0 mf + 0.711 gdb (4)

where the superficial velocity in the whole bed, 0 is given by:

Volumetric gas flowrate (fed at reactor conditions)

0 = (5)
Total transverse area (above distributor)

The area in Eq. (5) is calculated directly from the inside diameter of the tube, Dt . The superficial
velocity 0 is understood as the velocity that would exist inside the reactor if the void fraction was equal
to 1.
The mass transfer is given between the bubble and the cloud, and between the cloud and the emulsion.
Kunii and Levenspiel took the circulating model of Davison for the first case and the theory of Higbie for
the second. The mass transfer coefficient for the interphase bubble-cloud is obtained with:
   0.5 0.25 
Volume exchanged/s mf D g
Kbn , s1 = = 4.5 + 5.85 (6)
Volume of bubble db d1.25
b

where D is the diffusivity of reactant in the gas fluid. The mass transport coefficient for the interphase
cloud-emulsion is obtained with:

mf Db 0.5
 
1 Volume exchanged between interphases/s
Kne , s = = 6.77 (7)
Volume of bubble db

Levenspiel states that the industrial bubbling fluidized reactors operate at a superficial velocity within a
5 and 30 times above the minimum fluidization velocity.
The transport coefficients can be visualized as volumetric flowrates that are exchanged between the
two phases per volume of bubble. The total net flowrate is zero, but it will exist mass transfer of reactant
because each phase has different concentration. For example, the mole flowrate of A (per volume of bubble)
that leaves the bubble will be Kbn CAb , while the one that enters will be Kbn CAn , consequently, the net
mole transfer of A (per volume of A) between both phases is Kbn (CAb CAn ).

2.2 Relationships between the volumes of the phases

In order to design a reactor it has to be estimated the amount of catalyst present along the phases and the
volume of each phase. The basis to relate the mass balances will be the volume of the bubble. Besides,
for the final design it has to be known the fraction of the bed occupied by the bubbles, :

Volume of (all) the bubbles Vb

= =
(8)
Volume of the bed Dt2 h /4

The remaining necessary relations are:

Volume of the wake Volume of catalyst in the bubble

= ; b = ;
Volume of the bubble Volume of the bubble
Volume of catalyst in the cloud Volume of catalyst in the emulsion
n = ; =
Volmen of the bubble Volumen of the bubble

2
If there are experimental measurements for the total porosity of the fluidized bed, , or the heights of the
fluidized bed, h, or minimum fluidiztion, hmf , can be calculated based on geometric considerations:

mf h hmf
= = (9)
1 mf h

The fraction can also be estimated from a gas balance, but it has to be known first the velocity of the
emulsion, e , which requires the velocity of the solids rising in the emulsion, s . The volume of solids per
time unit rising with the wake, that rises with the velocity of the bubble, must be equal to the volume
of solids per time unit that descend with the emulsion; therefore, if we consider that w = e mf and
neglecting the volume fraction occupied by the cloud, we have:

Dt2 Dt2
   
b mf = s mf (1 ) (10)
4 4

simplifying, we get:
b
s = (11)
1
To estimate e we have to visualize the following considerations: i) mf is the referring lineal velocity
or the one that would exist in the empty tube, therefore, mf /mf is the gas velocity already within the
packed-bed that still remains at minimum fluidization conditions; ii) within the emulsion are given the
minimum fluidization conditions; and iii) since the particles descend with a velocity s , mf must be
achieved for the relative velocities between the gas and the particles, therefore:
mf
e = s (12)
mf

We write now the balance of volumetric flowrates of gas neglecting the contribution of the cloud:

(G )0 = (G )b + (G )w + (G )e (13)
Dt2 Dt2 Dt2 Dt2
       
0 = b + b mf + e mf (1 (14)
4 4 4 4

(All the volumetric flowrates are considered as constant). By taking a look on the previous velocities, we
conclude that b and e are the real velocities of the bubble and the emulsion, this is, they are not
superficials. Substituting the expression for e and s on Eq. (14) and simplifying, we get:
0 mf 0 mf
= (15)
b mf (1 + ) b

3 Design Equation for One Reaction

The Fig. 2 shows the existing interaction among the different phases present in the system. The K-L
model is unidimensional and assumes that the feed enters and leaves with the bubbles that rise as plug-
flow; however, along the bubbles trajectory exist mass transfer to the cloud (that includes the wake by
purposes of this development) and, finally, the cloud transports reactants to the emulsion. It has to be
noted that it has been assumed that there is no direct contact between the bubble and emulsion. The dense
phases for modelling purposes are stocked, this is, there is no axial mixing nor convection. Because of
the transport between the phases, the model implies that when there is an axial profile for CAb there will
be also profiles for CAn and CAe . For modelling purposes it will be considered that the particles are small
enough such as 1 or that it is known the apparent velocity coefficient (based on the catalyst weight)
such that: (rpA ) = kapp CA n , and that the internal effects do not affect the apparent reaction. If k
app
were based on particles volume, in the following equations has to be eliminated p . It also be assumed
that there is no temperature profile (there is a contradiction here, the mixing originated by the bubbles is

3
 vb

Bubble Cloud Emulsion

(and Wake)

vb
Kbn Kne

Figure 2: Mass transfer between phases.

such that allows to assume that the temperature is constant along the reactor; however, there is no axial
mixing for the dense phases).
Balance of A in the Bubble. Assuming a bubble as a batch reactor we have:
dCAb n
= Kbn (CAb CAn ) + b p kapp CAb (16)
dt
The terms on the equation are in a bubbles volume basis and correspond to accumulation, the net transfer
that leaves the bubble and what reacts inside of the bubble. Lets keep in mind that it has been neglected
any volumetric expansion. We will relate now the residence time t to the catalyst weight. The residence
time of the bubble is given by: t = h/b and the catalyst weight by
h  i
W = Dt2 h(1 ) (1 mf )p (17)
4
the term in brackets represents the total volume of the dense phases (emulsion, cloud and wake) that when
multiplied by (1 mf ) the total volume occupied by the catalyst (only neglecting the one inside of the
bubble). Solving for h in Eq. (17) and substituting into the residence time equation we have:
W
t=   (18)
b Dt2 (1 )(1 mf )p
4
therefore, the balance of A in the bubble is:
n
Kbn (CAb CAn ) + b p kapp CAb
dCAb
=   (19)
dw b D 2 (1 )(1 mf )p
4 t
where w is the weight of catalyst up to a given height of the fluidized bed.
Balance of A in the Cloud. It is assumed that there is no accumulation in this phase, thus, what it is
entered (net) from the bubble is the same to what leaves (net) out to the emulsion plus what it is reacted:
n
Kbn (CAb CAn ) = Kne (CAn CAe ) + n p kapp CAn (20)

Balance of A in the Emulsion. It is assumed again that there is no accumulation in this phase and that
there is a profile for CAe , this difference is important with respect to the two-phase model that assumes
perfect mixing in the emulsion. The (net) material entering from the cloud is equal to what it is reacted:
n
Kbe (CAn CAe ) = e p kapp CAe (21)

First-Order Reaction
If the catalytic reaction rate is first-order, an analytical solution is obtained provided that concentration

4
profiles exist for the dense phases. (If this is not assumed, we would consider that the concentration in the
emulsion, CAe , is independent of h, and the solution gets complicated because there would be two differential
equations for the profiles of CAb and CAn (actually, it would be one local algebraic equation for CAn , but for
convenience it is transformed into a differential easily), and one algebraic equation for the global balance
for CAe . Such algebraic equation is used as a convergence criteria and regards an integral that considers
the profile of CAn ). Solving for CAe from the last equation for n = 1 and substituting this result on the
right-hand side, we get:
1
Kne (CAn CAe ) = C (22)
1 1 An
+
e p kapp Kne
Including Eq. (22) into Eq. (20):
" 1 #
1 1
Kbn (CAb CAn ) = + + n p kapp CAn (23)
e p kapp Kne

and solving for CAn , substituting it into its right-hand side and rearranging:

1 #1 1
"
k 
p app 1 p kapp
Kbn (CAb CAn ) = + n + + p kapp CAb (24)
Kbn e Kne

Now, this is included,

dCAb KR kapp CAb
=   (25)
dw b Dt2 (1 )(1 mf )
4
1 #1 1
"
k 
p app 1 k
p app
where KR = + n + + + b (26)
Kbn e Kne

Equation (25) is integrated within the limits: CAb = CAb,0 at w = 0 and CAb = CAb,0 (1 xA ) at w = W .
For a first-order reaction with respect to A, the integral has an analytic solution that can be solved to
explicitly calculate the weight of catalyst, thus,
 
b Dt2 (1 )(1 mf )  1 
W = 4 ln (27)
KR kapp 1 xA

Different Reaction Order

If n 6= 1 and kapp is known, then Eq. (19) is numerically integrated and requires the evaluation of CAn
point-by-point. For each value of CAb that requires the Runge-Kutta method (the CAb values are obtained
from the resulting profile of the solution of Eq. (19) is necessary the simultaneous solution of two non-
linear equations, Eqs. (20) and (21) with two unknowns, CAn and CAe . Moreover, if the internal effects
are important, the effectiveness factor will be a function of the concentrations and can be risky to assume
a kapp constant.
Several reactants and/or several reactions
In case of several reactants, the value of Kbn and Kne parameters can be considered the same regardless
each species. This is justified in two ways: 1) the simplistic would be noting that in order to estimate
them the diffusivities of i in the mixture do not appear in the equations but only properties of the
mixture; and 2) the reasoned consideration assumes that these parameters represent the mixture, this is,
a volumetric flowrate that comes and goes among the phases but their net value is zero. This convective
circulation between two phases is integral and, therefore, is reasonable to assume that all the species
will be interchanged in the same proportion.