Está en la página 1de 8

The Effect of Column Diameter and Bed Height

on Minimum Fluidization Velocity
Akhil Rao and Jennifer S. Curtis
Dept. of Chemical Engineering, University of Florida, Gainesville, FL 32611
Bruno C. Hancock
Pfizer Global Research and Development, Groton, CT 06340
Carl Wassgren
School of Mechanical Engineering, Purdue University, West Lafayette, IN 47907
DOI 10.1002/aic.12161
Published online January 20, 2010 in Wiley Online Library (wileyonlinelibrary.com).
Experiments show that the minimum fluidization velocity of particles increases as
the diameter of the fluidization column is reduced, or if the height of the bed is
increased. These trends are shown to be due to the influence of the wall. A new, semi-
correlated model is proposed, which incorporates Janssen’s wall effects in the calcula-
tion of the minimum fluidization velocity. The wall friction opposes not only the bed
weight but also the drag force acting on the particles during fluidization. The
enhanced wall friction leads to an increase in the minimum fluidization velocity. The
model predictions compare favorably to existing correlations and experimental data.
V VC
2010 American Institute of Chemical Engineers AIChE J, 56: 2304–2311, 2010
Keywords: minimum fluidization velocity, fluidized bed, column diameter, bed height,
fluidization
Introduction
As fluidized bed operations have become more popular in
industry, there has been a drive to study these beds in
greater detail at laboratory scales. Reducing the size of fluid-
ized beds is gaining interest because of the better gas distri-
bution and operability at smaller scales. Micro fluidized beds
(MFBs), which were first proposed by Potic et al.,
1
have
good mixing and heat transfer capabilities, making them a
useful tool for studying various processes like drying, mix-
ing, or segregation,
2–6
as well as reaction kinetics.
7
For
example, MFBs are used to investigate fluidization segrega-
tion in the pharmaceutical industry where active ingredients
are expensive and difficult to obtain during the early stages
of development.
One parameter of particular interest when working with
fluidized beds is the minimum fluidization velocity. Many
correlations exist for calculating the minimum fluidization
velocity.
8–21
The Wen and Yu
22
correlation is the most com-
monly used correlation for calculating the minimum fluidiza-
tion velocity and is given as,
24:5
Re
#
_ _
2
þ1650
Re
#
_ _
À
Ar
#
3
_ _
¼ 0 (1)
where
Re ¼
q
g
ð#dÞU
mf
l
and (2)
Ar ¼
ð#dÞ
3
ðq
s
Àq
g
Þq
g
g
l
2
(3)
where Re is Reynolds number, Ar is the Archimedes number, W
is the sphericity, q
g
is gas density, d is particle diameter, U
mf
is
Correspondence concerning this article should be addressed to A. Rao at
akhil-rao@hotmail.com.
V VC
2010 American Institute of Chemical Engineers
2304 AIChE Journal September 2010 Vol. 56, No. 9
the minimum fluidization velocity, l is gas dynamic viscosity,
q
s
is solid density, and g is the gravitational acceleration. None
of the correlations include the effect of the bed height or
diameter of the column, factors which are likely to be of
interest when MFBs are considered.
Two recent investigations, however, have examined the
influence of column diameter and bed height. Di Felice and
Gibilaro
23
described a method for predicting the pressure
drop across a particle bed, taking into account the effect of
column diameter. They considered the column to be com-
prised of two sections: an inner core, where the voidage
remains nearly constant, and an outer annular section, where
the voidage varies because of the presence of the wall. As
there is a difference in the voidage over the cross section of
the column, the velocity also varies across the column’s
cross section. This fact was used to develop a modified
Ergun’s equation, but this effect is only observed at very
small column diameter (D) to particle diameter (d) ratios,
i.e. D/d \15.
Delebarre
24
assumed that the fluidizing gas density is a
function of the instantanious bed height, which in turn influ-
ences the minimum fluidizing velocity. This effect only
occurs if the column is very tall. For MFBs, where the col-
umns are short, this effect does not play much of a role.
As will be shown in the following section, experiments in
which the bed height and column diameter are varied show
that both parameters influence the bed’s minimum fluidiza-
tion velocity. Neither Di Felice and Gibilaro’s
23
or Dele-
barre’s
24
models are able to capture the experimental results.
The remainder of this article describes the theory and experi-
ments in which a new mechanism is proposed for accounting
for bed height and column wall effects.
Experimental Setup
Two fluidization segregation units (Jenike and Johanson,
Fluidization Material Sparing—Segregation Tester and Fluid-
ization Segregation Tester), were used in the experiments.
The first tester has a column diameter of 1.6 cm and a height
of 9.5 cm, and the latter tester has a 2.4 cm diameter column
and an 18.5 cm height. Details concerning the operation of
the testers are given in Hedden et al.
25
and ASTM D-6941.
26
A schematic of the experimental set up is shown in Figure
1. The experiments were carried out in both columns. A sin-
tered metal plate with an average pore diameter of 40 lm
was used as a gas distributor for both the columns. The air
enters the column from the bottom, and its flow rate is con-
trolled by a mass flow controller. The pressure drop across
the entire setup was measured with a pressure transducer.
The instantaneous pressure drop and velocity data were
recorded on a computer.
Experiments were performed with glass and polystyrene
particles of different sizes, ranging from 100 to 550 lm.
These particles are in the Geldart B class. The particles were
carefully sieved, then dried in an oven for 12 h, and finally
subject to an antistatic bar (Takk industries) to eliminate
accumulated electrostatic charge. The particles were stored
in a desiccator so that cohesive forces due to moisture were
minimized. To further reduce electrostatic charges that de-
velop during fluidization, a very small amount ($5 mg) of
antistatic powder (LarostatVR
HTS 905 S, manufactured by
BASF Corporation) was mixed with the particles, preceding
the experiments. Table 1 summarizes the experimental mate-
rials and their properties. The notation used to represent the
particle type is: G for glass and P for polystyrene, followed
by the average particle size in microns. The values for the
dynamic friction coefficients of glass and polystyrene on
acrylic (the column wall material) are not readily available
and so values of friction coefficients, for glass on glass and
polystyrene on polystyrene, obtained from Persson and
Tosatti
27
were used in the model calculations.
Before running an experiment, air was passed through the
empty column to get a background pressure drop due to the
column, diffuser, and the filter sections. Next, the particulate
material was weighed and loaded into the column from the
top. The height, H, to which the column was filled was
Figure 1. Schematic of the experimental set up.
Table 1. Experimental Materials and Properties
Material Diameter (lm) Density (kg/m
3
) Sphericity Coefficient of Friction Notation
Glass 105–125 2500 0.9 0.4 G116
Glass 210–250 2500 0.9 0.4 G231
Glass 250–300 2500 0.9 0.4 G275
Glass 350–420 2500 0.9 0.4 G385
Glass 420–500 2500 0.9 0.4 G462
Glass 500–600 2500 0.9 0.4 G550
Polystyrene 250–300 1250 0.9 0.5 P275
Polystyrene 300–354 1250 0.9 0.5 P328
AIChE Journal September 2010 Vol. 56, No. 9 Published on behalf of the AIChE DOI 10.1002/aic 2305
recorded. The antistatic powder was mixed with the par-
ticles. The air velocity was slowly increased beyond the
point of fluidization and then decreased to zero to get the
entire pressure drop profile. The point of intersection of the
pressure drop line for the fixed bed and the horizontal line
for the fully fluidized bed is typically defined as the mini-
mum fluidization velocity. However, just before fluidization,
the pressure drop across the bed overshoots the expected
value and then decreases to a constant value. Such behavior
is expected in columns with a small diameter. This over-
shoot was not observed during defluidization (Figure 2).
Thus, the defluidization pressure drop curve was used to
determine the point of fluidization.
7,28
A minimal of three
cycles of fluidization and defluidization were carried out to
determine an average minimum fluidization velocity. Each
experiment was repeated twice and the procedure was
repeated for all particle types. The maximum fluctuation
observed in the value of U
mf
was $7% (for G550 in D ¼
1.6 cm and H ¼ 4.5 cm; U
mf
¼ 24.5 Æ 0.9 cm/s). The fairly
consistent value of U
mf
observed over the three cycles and
two repetitions show that the trends in U
mf
are not due to
poor gas distribution but rather due to the wall effects. Addi-
tionally, gas channeling and maldistribution were not
observed through the transparent columns during any of the
experiments.
Figures 3–6 present the experimental data for the mini-
mum fluidization velocity (U
mf
) for the glass and polystyrene
particles, obtained from both the columns. In these figures,
the Reynolds number at U
mf
is plotted against the aspect ra-
tio of the bed (H/D). In addition to the experimental data,
the plots also include theoretical curves that will be
described in the following section. The error bars represent
the difference in the reading between two repetitions and not
amongst different cycles.
The experimental data (Figures 3–6) show that minimum
fluidization velocity monotonically increases as the height of
the bed is increased. Further, wall effects on the minimum
fluidization velocity are prominent when column diameter, D
¼ 1.6 cm (Figures 3 and 5). While, wall effects on minimum
fluidization velocity are much less pronounced for column
diameter, D ¼ 2.4 cm (Figures 4 and 6). Although the model
proposed by Di Felice and Gibilaro
23
predicts this trend,
their model is limited to column to particle diameter ratios
of less than 15. In the experiments performed here, this ratio
Figure 2. Example of a pressure drop profile (fluidiza-
tion and defluidization) using G231 particles
in the 1.6 cm diameter column.
The minimum fluidization velocity is also shown in the
figure.
Figure 3. Reynolds number, Re, based on the minimum
fluidization velocity as a function of aspect
ratio, H/D, for different glass particles in the
1.6 cm diameter column.
Figure 4. Reynolds number, Re, based on the minimum
fluidization velocity as a function of aspect
ratio, H/D, for different glass particles in the
2.4 cm diameter column.
2306 DOI 10.1002/aic Published on behalf of the AIChE September 2010 Vol. 56, No. 9 AIChE Journal
can be as large as 150, well outside the range of validity in
Di Felice and Gibilaro’s model.
As the height of the bed increases, the minimum fluidization
velocity also increases. Delebarre’s
24
model accounting for
changes in the fluidizing gas density over the bed height predicts
at most a minimum fluidization velocity increase of $0.5%,
whereas the observed minimum fluidization velocity increase is
greater than 20% in certain cases for the experiments examined
here. Thus, this model does not predict the measured data.
Theory
As existing correlations do not include the influence of
column diameter or bed height, and because the recent mod-
els by Di Felice and Gibilaro
23
and Delebarre
24
do not pre-
dict the measured data, a new model is developed in this
section to account for the observed trends. As will be pre-
sented, a key component of the model is the stress between
the column wall and the bed.
The pressure drop, DP, for a fixed bed of height, H, is
given by the semiempirical correlation of Ergun
16
as:
À
DP
H
¼ bU þaU
2
(4)
where
a ¼ 1:75
ð1À 2Þ
2
3
q
g
#d
_ _
and (5)
b ¼ 150
ð1À 2Þ
2
2
3
l
#
2
d
2
_ _
: (6)
The parameter U is the superficial fluid velocity and [ is
the bed voidage. The bed void fraction and fluid density are
assumed to remain constant throughout the bed.
The pressure drop expression (Eq. 4) can be used to pre-
dict the minimum fluidization velocity of a mono-sized ma-
terial by balancing the pressure force acting on the bed by
the effective bed weight.
1:75
ð1À 2Þ
2
3
_ _
Re
2
þ 150
ð1À 2Þ
2
2
3
_ _
Re ¼ ð1À 2ÞAr ½ Š (7)
Note that in the previous analysis, the stress between the
wall and the bed has not been included.
In the case of a fixed bed with no interstitial fluid, a frac-
tion of the bed weight is supported by the walls of the col-
umn. This phenomenon is known as Janssen’s wall effect.
To determine the wall force acting on the bed, consider a
thin horizontal slice of the particle bed of thickness dz (Fig-
ure 7). The vertical stress, r
v
, acting on the top face and the
weight of the element are balanced by the vertical stress act-
ing on the bottom face and the frictional forces applied by
the wall. The resulting differential equation is,
dr
v
dz
þc
1
r
v
¼ c
2
(8)
where
c
1
¼
4tanuk
D
and (9)
c
2
¼ ð1À 2Þq
s
g: (10)
The parameter u is the friction angle between the wall
and particles, and D is the column diameter. A horizontal
frictional stress, r
H
, arises from the vertical stress, r
v
, pro-
ducing a wall friction force of r
H
tanu.
29
It is assumed here
Figure 5. Reynolds number, Re, based on the minimum
fluidization velocity as a function of aspect
ratio, H/D, for different polystyrene particles
in the 1.6 cm diameter column.
Figure 6. Reynolds number, Re, based on the minimum
fluidization velocity as a function of aspect
ratio, H/D, for different polystyrene particles
in the 2.4 cm column.
AIChE Journal September 2010 Vol. 56, No. 9 Published on behalf of the AIChE DOI 10.1002/aic 2307
that this horizontal stress is directly proportional to the verti-
cal stress with k as the proportionality constant
29
:
r
H
¼ kr
v
: (11)
Solving Eq. 8 subject to the boundary condition that the
stress at the top of the bed is zero gives,
r
v
¼
c
2
c
1
ð1 Àe
Àc
1
z
Þ: (12)
The stress increases nearly linearly near the free surface,
but asymptotes to a constant value deeper into the bed.
These wall induced stresses will affect the minimum fluid-
ization velocity in narrow columns.
Now consider the situation when fluid is flowing through
a fixed bed of solids with the same coordinate system as
shown in Figure 7. At equilibrium, the force due to stress on
the top face of the element, F
T
, and the weight of the ele-
ment, F
W
, will be balanced by the force due to the stress on
the bottom face, F
B
, the wall friction, F
F
(Janssen’s effect),
the drag force on the solids, F
D
, and the buoyancy force,
F
Bu
. Thus, a force balance on the solids yields:
F
B
þF
F
þF
D
þF
Bu
¼ F
T
þF
W
(13)
which, when expanded is
p
4
D
2
ðr
v
þdr
v
Þ þðtanur
H
ÞpDdz þ
p
4
D
2
dp ¼
p
4
D
2
r
v
þðð1À 2Þðq
s
Àq
g
ÞgÞ
p
4
D
2
dz: (14)
In the above expression, F
w
and F
Bu
have been com-
bined to give the last term on the right hand side of Eq.
14. Simplifying and rearranging the previous equation
gives,
dr
v
dz
þ
4tanur
H
D
¼ ð1À 2Þðq
s
Àq
g
Þg À
dp
dz
_ _
: (15)
If r
H
¼ kr
v
is assumed, as in Eq. 11,
dr
v
dz
þ
4tanukr
v
D
¼ ð1À 2Þðq
s
Àq
g
Þg À
dp
dz
_ _
(16)
which may be written as
dr
v
dz
þc
1
r
v
¼ c
2
(17)
where
c
1
¼
4tanuk
D
and (18)
c
2
¼ ð1À 2Þðq
s
Àq
g
Þg À
dp
dz
_ _
: (19)
In Eq. 19 there are two additional terms as compared with
Eq. 10: the drag term, represented by dp/dz, and the buoy-
ancy term.
The solution to Eq. 17 is equivalent to that of Eq. 8
r
v
¼
c
2
c
1
ð1 Àe
Àc
1
z
Þ (20)
but with different c
1
and c
2
.
Figure 8 presents Eq. 20 in graphical form. When the ve-
locity is zero (i.e., there is no gas flowing through the col-
umn), then Eq. 20 simplifies to Janssen’s equation. As the
velocity increases, the vertical contact stresses in the bed are
smaller because the drag on the particles supports a fraction
of the bed’s weight. At and above the point of fluidization,
the stresses in the column are zero (as contact does not exist
Figure 7. Sketch showing the fluidized bed coordinate
system.
Figure 8. A schematic showing how the vertical stress
in the bed varies with bed depth at different
fluid velocity U
1
, U
2
, and U
3
.
2308 DOI 10.1002/aic Published on behalf of the AIChE September 2010 Vol. 56, No. 9 AIChE Journal
amongst the particles), and thus dr
v
/dz is also zero. This sit-
uation is only possible if c
2
¼ 0,
c
2
¼ ð1À 2Þðq
s
Àq
g
Þg À
dp
dz
_ _
¼ 0 (21)
In Eq. 21, if Ergun’s pressure drop expression is substi-
tuted (Eq. 4), the resultant expression will be the same as
Eq. 7, and the wall effects will not be included in the analy-
sis.
It is now assumed that the horizontal stress is not only a
function of the downward vertical stress but also a function
the upward drag forces caused by the gas flowing through
the column. As Ergun’s pressure drop equation is used to
calculate the pressure drop, it is assumed that the structure
of these new terms on which the horizontal stress depends is
similar in form to those given by Ergun’s equation, but the
velocity terms are being scaled differently. These scales are
chosen as they give a favorable fit to the experimental data.
r
H
¼ kr
v
Àk
1
H
l
D#d
_ _
U Àk
2
H
q
g
D
_ _
U
2
(22)
where k
1
and k
2
are constants.
Substituting Eq. 22 back into Eq. 15, yields:
dr
v
dz
þ
4tanukr
v
D
¼
4tanuk
2
ðq
g
Þð
H
D
Þ
D
_ _
U
2
þ
4tanuk
1
l
#d
_ _
ð
H
D
Þ
D
_ _
U þ ð1À 2Þðq
s
Àq
g
Þg
_ ¸
À
dp
dz
: (23)
Comparing Eq. 23 to Eq. 16, there are two additional
terms, proportional to U
2
and U, appearing on the right hand
side of Eq. 23. These terms have come about due to the hor-
izontal stress from the wall and the flow of the fluid through
the bed.
Substituting Ergun’s equation for pressure drop (Eq. 4)
into Eq. 23:
dr
v
dz
þ
4tanukr
v
D
¼
4tanuk
2
ðq
g
Þð
H
D
Þ
D
Àa
_ _
U
2
þ
4tanuk
2
l
#d
_ _
ð
H
D
Þ
D
Àb
_ _
U þ ð1À 2Þðq
s
Àq
g
Þg
_ ¸
: (24)
This equation is again of the form:
dr
v
dz
þc
1
r
v
¼ c
2
(25)
where:
c
1
¼
4tanuk
D
and (26)
Àc
2
¼ a À
4tanuk
2
ðq
g
Þð
H
D
Þ
D
_ _
U
2
þ b À
4tanuk
2
l
#d
_ _
ð
H
D
Þ
D
_ _
U
À ð1À 2Þðq
s
Àq
g
Þg
_ ¸
: (27)
Setting c
2
¼ 0, the condition which is necessary for fluid-
ization, and integrating over the length of the column:
a À
4tanuk
0
2
ðq
g
Þð
H
D
Þ
D
_ _
U
2
þ b À
4tanuk
0
1
l
#d
_ _
ð
H
D
Þ
D
_ _
U
À ð1À 2Þðq
s
Àq
g
Þg
_ ¸
¼ 0 (28)
where k
0
1
and k
0
2
are lumped parameters which also include
integration coefficients. Substituting for a and b, which are
Ergun’s constants, yields;
1:75
ð1À 2Þ
2
3
À4tanuk
0
2
#d
D
_ _
H
D
_ _ _ _
q
g
U
2
#d
þ
_
150
ð1À 2Þ
2
2
3
À4tanuk
0
1
#d
D
_ _
H
D
_ _
_
lU
ð#dÞ
2
¼ ð1À 2Þðq
s
Àq
g
Þg
_ ¸
(29)
Multiplying throughout by the factor
q
g
ð#dÞ
3
l
2
to make the
equation dimensionless gives:
1:75
ð1À 2Þ
2
3
À4tanuk
0
2
#d
D
_ _
ð
H
D
Þ
_ _
Re
2
þ
_
150
ð1À 2Þ
2
2
3
À4tanuk
0
1
#d
D
_ _
H
D
_ _
_
Re ¼ ð1À 2ÞAr ½ Š (30)
Equation 30 is a quadratic equation that can be easily
solved for Re, with k
0
1
and k
0
2
remaining constant. Rewriting
Eq. 30 results in
a
00
2; u;
#d
D
_ _
;
H
D
_ _ _ _
Re
2
þb
00
2; u;
#d
D
_ _
;
H
D
_ _ _ _
Re
Àð1À 2ÞAr ¼ 0 (31)
where a
00
and b
00
are functions not only of [ and u but also of
(Wd/D) and (H/D). Hence, both the ratio of particle size to
column diameter (Wd/D) and the bed aspect ratio (H/D) play a
role in determining U
mf
.
Results and Discussion
The values of the universal constants k
0
1
and k
0
2
which give
the minimum error for the experimental data set presented in
this study were found out to be 610 and 30.1, respectively. In
Figures 3–6 and Figures 9 and 10, the values of the universal
constants k
0
1
and k
0
2
are kept at these values for all of the par-
ticle types, bed heights, and column diameters.
Figures 3–6 show that including the wall effects in the
prediction of U
mf
is successful in describing the effects of
height vs. column diameter (H/D) and the effect of particle
size to column diameter (d/D). The model only slightly
under-predicts the increase in U
mf
for very small sized par-
ticles (d \120 lm) in the column with D ¼ 1.6 cm, which
is likely due electrostatic cohesive and adhesive forces.
For a fixed column diameter, as the H/D ratio increases,
U
mf
increases (Figures 3–6). Hence, as the column becomes
taller in comparison to its diameter, the wall effects are
more prominent, making it more difficult to fluidize the
AIChE Journal September 2010 Vol. 56, No. 9 Published on behalf of the AIChE DOI 10.1002/aic 2309
particles. Also, for a constant particle size, as d/D ratio
increases (due to changes in column diameter), U
mf
increases
(compare Figures 3 and 4 or Figures 5 and 6). As the col-
umn diameter decreases in comparison to the particle diame-
ter, the ratio of wall contact surface area to the bulk volume
increases, leading to a more significant wall effect. This wall
effect reduces as the column increases in diameter relative to
the particles, and eventually becomes negligible as d/D
becomes very small. The effect of H/D and d/D on U
mf
are
very similar; in fact, the increase in U
mf
can be characterized
by the product of H/D and d/D (Eq. 30). Additionally, as
particle size increases and approaches the same order of
magnitude as the column diameter (d/D $ O(1)), then fluid-
ization becomes difficult. Above a certain d/D ratio, the par-
ticle mixture is impossible to fluidize. This situation is pre-
dicted by the model, as the quadratic equation yields com-
plex roots. Similarly, Eq. 30 breaks down for large H/D
ratio predicting that at large H/D the gas may not fluidize
the bed. For example, the model predicts that if glass par-
ticles of size 250 lm are fluidized in a 760 lm diameter col-
umn (D/d $ 3) with H/D ¼ 1, then the particles will not flu-
idize. Similarly, the model also predicts that if these par-
ticles are loaded into a 100 cm diameter column to a bed
height of about 13 m (H/D $ 13), the gas may not fluidize
the bed.
Figure 9 compares the experimental data from Liu et al.
7
to the values predicted by the model. The data are plotted in
the same dimensional form as given in the orignal work of
Liu et al.
7
The model does not predict the increase in U
mf
for the smaller particles (d ¼ 96.4 lm), but it does give
good predictions for the larger particle sizes. In this case
also, the values of k
0
1
and k
0
2
are maintained at 610 and
30.1, respectively. For the largest particles (d ¼ 460 lm),
the prediction for the smallest diameter column is not that
good, but this may be due to experimental error, which is
not reported in Liu et al.
7
Also, the values of voidage were
not clearly stated; only values of bulk density were given. In
the model, the values of bulk density were kept constant
with respect to the column diameter, but it has been
observed that this value changes depending on the column
diameter,
28
particularly for small column diameters.
Figure 10 compares the model results to the average pre-
dictions of the Reynolds number at U
mf
based on a large
number of existing correlations.
8–22
The scatter bars on the
correlation line represent the deviation in minimum fluidiza-
tion velocity predictions using the different correlations. The
experimental data obtained when the wall effects are mini-
mum (at H ¼ 2 cm and D ¼ 2.4 cm) are consistent with the
model prediction and with the averaged correlation curve.
On the other hand, only the new model is able to predict ex-
perimental data obtained when the wall effects are signifi-
cant (at H ¼ 10 cm, D ¼ 1.6 cm), as the existing correla-
tions for U
mf
do not take wall effects into account.
Conclusions
Experiments show that wall effects influence the minimum
fluidization velocity. Existing correlations fail to incorporate
these wall effects and, hence, there is a need for a new
model that can take these effects into consideration. The
model presented in this study attempts to include the wall
influence by introducing Janssen’s wall effect in the force
balance during fluidization. It is assumed that the horizontal
stresses acting at the wall are not only a function of the local
vertical stress but also are a quadratic function of velocity.
These new terms have the same structure as that of the drag
term, i.e., the pressure drop, as given by the widely accepted
Ergun equation. This assumption leads to a modified Ergun’s
equation incorporating two universal constants. The new
model fits experimental data reasonably well for the mini-
mum fluidization velocity over a range of particles sizes, bed
heights, and column diameters. Some deficiencies in the
model predictions are noted for smaller particles (around
100 lm and smaller), but these deficiencies may be due to
Figure 10. Comparison of the curves from Eq. 30 to the
experimental data and existing correlations.
Figure 9. The minimum fluidization velocity as a func-
tion of the column diameter using the experi-
mental data from Liu et al.
7
The curves are from Eq. 30.
2310 DOI 10.1002/aic Published on behalf of the AIChE September 2010 Vol. 56, No. 9 AIChE Journal
the significance of cohesive and adhesive forces at this scale.
This new model should greatly facilitate scaling results from
MFBs to more traditional fluid bed sizes.
Acknowledgments
Jim Prescott and colleagues at Jenike and Johanson are thanked for
their support of this research and for providing the segregation testing
equipment.
Notation
[ ¼ voidage
W ¼ sphericity
l ¼ viscosity
q
g
¼ density of gas
q
s
¼ density of solid
r
H
¼ stress in horizontal direction
r
v
¼ stress in vertical direction
u ¼ friction angle
Dp ¼ pressure drop
a, b ¼ Ergun constants
A ¼ cross-sectional area
a
0
,b
0
,c
0
¼ various constants of the quadratic
Ar ¼ Archimedes number
D ¼ column diameter
d ¼ particle diameter
F ¼ force
g ¼ Gravity
H ¼ height of fixed bed
k, k
1
, k
2
, k
1
0
, k
2
0
,c
1
,c
2
¼ various constants
Re ¼ Reynolds number
U ¼ fluid velocity
U
mf
¼ Minimum Fluidization velocity
z ¼ instantaneous height measured from the top of
the bed
Literature Cited
1. Potic B, Kerstn S, Ye M, Van der Hoef M, van Swaaij W. Fluidiza-
tion of hot compressed water in micro reactors. Chem Eng Sci.
2005;60:5982–5990.
2. Chiba S, Nienow A, Chiba T, Kobayashi H. Fluidized binary mix-
tures in which denser component may be flotsam. Powder Technol.
1980;26:1–10.
3. Garcia F, Romero A, Villar J, Bello A. A study of segregation in a
gas solid fluidized bed: particles of different density. Powder Tech-
nol. 1989;58:169–174.
4. Huilin L, Yurong D, Gidaspow D, Lidan Y, Yukun Q. Size segrega-
tion of binary mixture of solid in bubbling fluidized bed. Powder
Technol. 2003;134:86–97.
5. Joseph G, Leboreiro J, Hrenya C, Stevens A. Experimental segrega-
tion profiles in bubbling gas fluidized bed. AIChE J. 2007;53:2804–
2813.
6. Geldart D, Baeyens J, Pope D, Van De Wijer P. Segregation in beds of
large particles at high velocity. Powder Technol. 1981;30:195–205.
7. Liu X, Xu G, Gao S. Micro fluidized beds: wall effects and oper-
ability. Chem Eng J. 2008;137:302–307.
8. Babu S, Shah B, Talwalker A. Fluidization correlation for coal gasi-
fication materials—minimum fluidization velocity and fluidized bed
expansion. AIChE Symp Ser. 1978;74:176–186.
9. Baerg A, Klassen J, Gishler P. Heat transfer in a fluidized solid bed.
Can J Res Sect F. 1950;28:287.
10. Broadhurst T, Becker H. Onset of fluidization and slugging in beds
of uniform particles. AIChE J. 1975;21:238–247.
11. Davies L, Richardson J. Gas interchange between bubbles and the
continuous phase in a fluidized bed. Trans Inst Chem Eng.
1966;44:293–305.
12. Goroshko V, Rozenbaum R, Todes O. Hydrodynamics and Heat
Transfer in fluidized beds. Cambridge: M.I.T. Press, 1966.
13. Grewal N, Saxena S. Comparision of commonly used correlations
for minimum fluidization velocityof small solid particles. Powder
Technol. 1980;26:229–234.
14. Johnson E. Institute of gas engineers report, Publication No. 318.
London, 1949.
15. Kumar P, Sen Gupta P. Prediction of minimum fluidization velocity
for multicomponent mixtures (short communication). Indian J Tech-
nol. 1974;12:225.
16. Kunii D, Levenspiel O. Fluidization Engineering, 2nd ed. Newton:
Butterworth-Heinemann, 1991.
17. Leva M. Fluidization. New York: McGraw-Hill, 1959.
18. Miller C, Longwinuk A. Fluidization studies of solid particles. Ind
Eng Chem. 1951;43:1220–1226.
19. Pillai B, Raja Rao M. Pressure drop and minimum fluidization
velocities of air-fluidized bed. Indian J Technol. 1971;9:77.
20. Saxena S, Vogel G. The measurement of incipient fluidization veloc-
ities in a bed of coarse dolomite at temperature and pressure. Trans
Inst Chem Eng. 1977;55:184–189.
21. van Heerdeen C, Nobel A, Krevelen v. Studies on fluidization ll—
heat transfer. Chem Eng Sci. 1951;1:51–66.
22. Wen C, Yu Y. Mechanics of fluidization. Chem Eng Prog Symp.
1966;62:100–111.
23. Di Felice R, Gibilaro L. Wall effects for the pressure drop in fixed
beds. Chem Eng Sci. 2004;59:3037–3040.
24. Delebarre A. Does the minimum fluidization exist? J Fluid Eng.
2002;124:595–600.
25. Hedden D, Brone D, Clement S, McCall M, Olsofsy A, Patel P, Pre-
scott J, Hancock B. Development of an improved fluidization segre-
gation tester for use with pharmaceutical powders. Pharm Technol.
2006;30:54–64.
26. ASTM International. Standard practice for measuring fluidization
segregation tendencies of powders, ASTM International, D6941–03,
2003.
27. Persson B, Tosatti E. Physics of Sliding Friction (NATO Science Se-
ries E). New York: Springer Publishing Company, 1966.
28. Loezos P, Costamagna P, Sundaresan S. The role of contact stresses
and wall friction on fluidization. Chem Eng Sci. 2002;57:5123–5141.
29. Janssen HA. Versuche uber getreidedruck in silozellen. Z Ver
Deutsch Ing. 1895;39:1045–1049.
Manuscript received Aug. 14, 2009, and revision received Dec. 1, 2009.
AIChE Journal September 2010 Vol. 56, No. 9 Published on behalf of the AIChE DOI 10.1002/aic 2311

4 0.9 0.4 0.9 0. the minimum fluidization velocity. and g is the gravitational acceleration. air was passed through the empty column to get a background pressure drop due to the column.9 0. Next. No. The particles were stored in a desiccator so that cohesive forces due to moisture were minimized.6 cm and a height of 9. A sintered metal plate with an average pore diameter of 40 lm was used as a gas distributor for both the columns. to which the column was filled was Figure 1. however. then dried in an oven for 12 h. Neither Di Felice and Gibilaro’s23 or Delebarre’s24 models are able to capture the experimental results. where the columns are short. where the voidage varies because of the presence of the wall. the particulate material was weighed and loaded into the column from the top. The experiments were carried out in both columns. l is gas dynamic viscosity. As there is a difference in the voidage over the cross section of the column. this effect does not play much of a role. The air enters the column from the bottom. qs is solid density.4 0. Details concerning the operation of the testers are given in Hedden et al. ranging from 100 to 550 lm. Experimental Materials and Properties Material Glass Glass Glass Glass Glass Glass Polystyrene Polystyrene Diameter (lm) 105–125 210–250 250–300 350–420 420–500 500–600 250–300 300–354 Density (kg/m3) 2500 2500 2500 2500 2500 2500 1250 1250 Sphericity 0. where the voidage remains nearly constant. and the filter sections. 9 Published on behalf of the AIChE DOI 10. 56. This effect only occurs if the column is very tall. The notation used to represent the particle type is: G for glass and P for polystyrene. and the latter tester has a 2. which in turn influences the minimum fluidizing velocity. The height.25 and ASTM D-6941.tion velocity.9 0. Before running an experiment. D/d \ 15. Table 1 summarizes the experimental materials and their properties. taking into account the effect of column diameter. The remainder of this article describes the theory and experiments in which a new mechanism is proposed for accounting for bed height and column wall effects. factors which are likely to be of interest when MFBs are considered. The values for the dynamic friction coefficients of glass and polystyrene on acrylic (the column wall material) are not readily available and so values of friction coefficients. diffuser.4 cm diameter column and an 18. Experimental Setup Two fluidization segregation units (Jenike and Johanson. i. The particles were carefully sieved.9 Coefficient of Friction 0. These particles are in the Geldart B class. for glass on glass and polystyrene on polystyrene. but this effect is only observed at very small column diameter (D) to particle diameter (d) ratios. and an outer annular section.9 0. The instantaneous pressure drop and velocity data were recorded on a computer. Two recent investigations. The pressure drop across the entire setup was measured with a pressure transducer.4 0.5 cm height. the velocity also varies across the column’s cross section.1002/aic 2305 . and its flow rate is controlled by a mass flow controller.e. and finally subject to an antistatic bar (Takk industries) to eliminate accumulated electrostatic charge.9 0. They considered the column to be comprised of two sections: an inner core. As will be shown in the following section.5 0.5 cm. None of the correlations include the effect of the bed height or diameter of the column.26 A schematic of the experimental set up is shown in Figure 1. preceding the experiments. obtained from Persson and Tosatti27 were used in the model calculations. have examined the influence of column diameter and bed height.4 0. This fact was used to develop a modified Ergun’s equation. were used in the experiments. experiments in which the bed height and column diameter are varied show that both parameters influence the bed’s minimum fluidiza- Table 1. Di Felice and Gibilaro23 described a method for predicting the pressure drop across a particle bed.9 0. Schematic of the experimental set up. To further reduce electrostatic charges that develop during fluidization. For MFBs.5 Notation G116 G231 G275 G385 G462 G550 P275 P328 AIChE Journal September 2010 Vol. Fluidization Material Sparing—Segregation Tester and Fluidization Segregation Tester). followed by the average particle size in microns. Experiments were performed with glass and polystyrene particles of different sizes.4 0. Delebarre24 assumed that the fluidizing gas density is a function of the instantanious bed height. The first tester has a column diameter of 1. H. a very small amount ($5 mg) of R antistatic powder (LarostatV HTS 905 S. manufactured by BASF Corporation) was mixed with the particles.

The air velocity was slowly increased beyond the point of fluidization and then decreased to zero to get the entire pressure drop profile. The fairly consistent value of Umf observed over the three cycles and two repetitions show that the trends in Umf are not due to poor gas distribution but rather due to the wall effects. The maximum fluctuation observed in the value of Umf was $7% (for G550 in D ¼ 1. Umf ¼ 24. 56. Additionally. While.6 cm diameter column. No. In these figures. September 2010 Vol. In addition to the experimental data. their model is limited to column to particle diameter ratios of less than 15. Figure 3. the defluidization pressure drop curve was used to determine the point of fluidization. D ¼ 2. Example of a pressure drop profile (fluidization and defluidization) using G231 particles in the 1. just before fluidization.7.recorded. Reynolds number. Each experiment was repeated twice and the procedure was repeated for all particle types. Further. the Reynolds number at Umf is plotted against the aspect ratio of the bed (H/D). The point of intersection of the pressure drop line for the fixed bed and the horizontal line for the fully fluidized bed is typically defined as the minimum fluidization velocity.28 A minimal of three cycles of fluidization and defluidization were carried out to determine an average minimum fluidization velocity. gas channeling and maldistribution were not observed through the transparent columns during any of the experiments. Figure 4. Re. based on the minimum fluidization velocity as a function of aspect ratio. Although the model proposed by Di Felice and Gibilaro23 predicts this trend.6 cm (Figures 3 and 5). Reynolds number. this ratio Figure 2. However. The experimental data (Figures 3–6) show that minimum fluidization velocity monotonically increases as the height of the bed is increased. H/D. The error bars represent the difference in the reading between two repetitions and not amongst different cycles. Re. Figures 3–6 present the experimental data for the minimum fluidization velocity (Umf) for the glass and polystyrene particles. the pressure drop across the bed overshoots the expected value and then decreases to a constant value. wall effects on minimum fluidization velocity are much less pronounced for column diameter.5 Æ 0. for different glass particles in the 1.5 cm. wall effects on the minimum fluidization velocity are prominent when column diameter. The antistatic powder was mixed with the particles.4 cm diameter column. based on the minimum fluidization velocity as a function of aspect ratio.4 cm (Figures 4 and 6).9 cm/s). Such behavior is expected in columns with a small diameter. D ¼ 1. In the experiments performed here. the plots also include theoretical curves that will be described in the following section. H/D. 9 AIChE Journal 2306 DOI 10.1002/aic Published on behalf of the AIChE . for different glass particles in the 2. Thus.6 cm and H ¼ 4.6 cm diameter column. This overshoot was not observed during defluidization (Figure 2). obtained from both the columns. The minimum fluidization velocity is also shown in the figure.

rv.1002/aic 2307 DP ¼ bU þ aU2 H (4) b ¼ 150 ð1À 2Þ2  l  : 23 #2 d 2 The parameter U is the superficial fluid velocity and [ is the bed voidage. The pressure drop. and because the recent models by Di Felice and Gibilaro23 and Delebarre24 do not predict the measured data. Re. the minimum fluidization velocity also increases. producing a wall friction force of rHtanu. " # ! ð1À 2Þ ð1À 2Þ2 2 Re þ 150 Re ¼ ½ð1À 2ÞArŠ 1:75 23 23 (7) Figure 5. for a fixed bed of height. for different polystyrene particles in the 1. Note that in the previous analysis. rH. DOI 10. In the case of a fixed bed with no interstitial fluid.The pressure drop expression (Eq. whereas the observed minimum fluidization velocity increase is greater than 20% in certain cases for the experiments examined here. a key component of the model is the stress between the column wall and the bed. a new model is developed in this section to account for the observed trends. No. DP. Thus. based on the minimum fluidization velocity as a function of aspect ratio. 9 Published on behalf of the AIChE . is given by the semiempirical correlation of Ergun16 as: À where a ¼ 1:75 ð1À 2Þ  qg  #d 23 and (5) (6) Figure 6. Reynolds number. the stress between the wall and the bed has not been included. The resulting differential equation is. A horizontal frictional stress. well outside the range of validity in Di Felice and Gibilaro’s model. The vertical stress. Delebarre’s24 model accounting for changes in the fluidizing gas density over the bed height predicts at most a minimum fluidization velocity increase of $0. based on the minimum fluidization velocity as a function of aspect ratio. arises from the vertical stress.4 cm column. consider a thin horizontal slice of the particle bed of thickness dz (Figure 7). This phenomenon is known as Janssen’s wall effect. As will be presented. this model does not predict the measured data. and D is the column diameter. Reynolds number. 56. H/D.5%.29 It is assumed here Theory As existing correlations do not include the influence of column diameter or bed height. for different polystyrene particles in the 2. The bed void fraction and fluid density are assumed to remain constant throughout the bed. H/D.6 cm diameter column. As the height of the bed increases. c2 ¼ ð1À 2Þqs g: The parameter u is the friction angle between the wall and particles. AIChE Journal September 2010 Vol. a fraction of the bed weight is supported by the walls of the column. acting on the top face and the weight of the element are balanced by the vertical stress acting on the bottom face and the frictional forces applied by the wall. 4) can be used to predict the minimum fluidization velocity of a mono-sized material by balancing the pressure force acting on the bed by the effective bed weight. H. To determine the wall force acting on the bed. rv. Re. drv þ c 1 r v ¼ c2 dz where c1 ¼ 4tanuk and D (9) (10) (8) can be as large as 150.

2308 DOI 10. September 2010 Vol. At equilibrium. will be balanced by the force due to the stress on the bottom face. Simplifying and rearranging the previous equation gives.1002/aic Figure 8. The solution to Eq. FT. 56. Now consider the situation when fluid is flowing through a fixed bed of solids with the same coordinate system as shown in Figure 7. there is no gas flowing through the column).. FW. FD.e. but asymptotes to a constant value deeper into the bed. the force due to stress on the top face of the element. FB. 8 subject to the boundary condition that the stress at the top of the bed is zero gives. Fw and FBu have been combined to give the last term on the right hand side of Eq. the vertical contact stresses in the bed are smaller because the drag on the particles supports a fraction of the bed’s weight. the stresses in the column are zero (as contact does not exist (14) In the above expression. When the velocity is zero (i. No.! drv 4tanurH dp þ ¼ ð1À 2Þðqs À qg Þg À : dz D dz If rH ¼ krv is assumed. These wall induced stresses will affect the minimum fluidization velocity in narrow columns. then Eq. as in Eq. the drag force on the solids. FBu. As the velocity increases. 9 AIChE Journal Published on behalf of the AIChE . and the buoyancy force. and the weight of the element. 20 simplifies to Janssen’s equation. and the buoyancy term. At and above the point of fluidization. 19 there are two additional terms as compared with Eq. rv ¼ c2 ð1 À eÀc1 z Þ: c1 (12) The stress increases nearly linearly near the free surface. represented by dp/dz. Sketch showing the fluidized bed coordinate system. 11. (18) (19) c2 ¼ ð1À 2Þðqs À qg Þg À that this horizontal stress is directly proportional to the vertical stress with k as the proportionality constant29: rH ¼ krv : (11) In Eq. 10: the drag term. A schematic showing how the vertical stress in the bed varies with bed depth at different fluid velocity U1. 17 is equivalent to that of Eq. when expanded is p 2 p p D ðrv þ drv Þ þ ðtanurH ÞpDdz þ D2 dp ¼ D2 rv 4 4 4 p þ ðð1À 2Þðqs À qg ÞgÞ D2 dz: 4 (13) but with different c1 and c2. Figure 8 presents Eq. 20 in graphical form. drv 4tanukrv dp þ ¼ ð1À 2Þðqs À qg Þg À dz dz D which may be written as drv þ c 1 r v ¼ c2 dz where c1 ¼ 4tanuk D and ! dp : dz ! (15) (16) (17) Figure 7. FF (Janssen’s effect). U2. Thus. the wall friction. 8 rv ¼ c2 ð1 À eÀc1 z Þ c1 (20) Solving Eq. a force balance on the solids yields: FB þ FF þ FD þ FBu ¼ FT þ FW which. and U3. 14.

if Ergun’s pressure drop expression is substituted (Eq. making it more difficult to fluidize the DOI 10. with k0 1 and k0 2 remaining constant. the wall effects are more prominent. For a fixed column diameter. 4). respectively. and integrating over the length of the column: À Á ! ! 0 0 l 4tanuk2 ðqg ÞðHÞ 2 4tanuk1 #d ðHÞ D D U þ bÀ U aÀ D D  à À ð1À 2Þðqs À qg Þg ¼ 0 (28) where k0 1 and k0 2 are lumped parameters which also include integration coefficients. Hence. both the ratio of particle size to column diameter (Wd/D) and the bed aspect ratio (H/D) play a role in determining Umf. It is now assumed that the horizontal stress is not only a function of the downward vertical stress but also a function the upward drag forces caused by the gas flowing through the column.1002/aic 2309 (25) ÀlÁ ! ! 4tanuk2 ðqg ÞðHÞ 2 4tanuk2 #d ðHÞ D D Àc2 ¼ a À U þ bÀ U D D  à À ð1À 2Þðqs À qg Þg : (27) AIChE Journal September 2010 Vol. u. This situation is only possible if c2 ¼ 0.6 cm. as the column becomes taller in comparison to its diameter. which are Ergun’s constants. the resultant expression will be the same as Eq. "   ! ð1À 2Þ H qg U 2 ð1À 2Þ2 0 #d 1:75 þ 150 À 4tanuk2 23 D D #d 23   #  à H lU 0 #d À 4tanuk1 ¼ ð1À 2Þðqs À qg Þg (29) D D ð#dÞ2 Multiplying throughout by the factor equation dimensionless gives: qg ð#dÞ3 l2 to make the "   ! ð1À 2Þ H ð1À 2Þ2 0 #d 2 1:75 ð Þ Re þ 150 À 4tanuk2 23 D D 23 #    H 0 #d Re ¼ ½ð1À 2ÞAr Š (30) À 4tanuk1 D D (23) Equation 30 is a quadratic equation that can be easily solved for Re. 23: ! 4tanuk2 ðqg ÞðHÞ drv 4tanukrv D À a U2 þ ¼ D dz D ÀlÁ !  à 4tanuk2 #d ðHÞ D þ À b U þ ð1À 2Þðqs À qg Þg : D This equation is again of the form: drv þ c1 r v ¼ c2 dz where: c1 ¼ 4tanuk D and (26) À ð1À 2ÞAr ¼ 0 (31) where a00 and b00 are functions not only of [ and u but also of (Wd/D) and (H/D). These terms have come about due to the horizontal stress from the wall and the flow of the fluid through the bed. as the H/D ratio increases. 22 back into Eq. (24) Results and Discussion The values of the universal constants k0 1 and k0 2 which give the minimum error for the experimental data set presented in this study were found out to be 610 and 30. 30 results in           #d H #d H 2 00 a 2. which is likely due electrostatic cohesive and adhesive forces. 21. 16. there are two additional terms. and thus drv/dz is also zero. Rewriting Eq. Substituting Ergun’s equation for pressure drop (Eq. 56. q   l  g rH ¼ krv À k1 H U2 U À k2 H D D#d where k1 and k2 are constants. 15. Re D D D D 00 Comparing Eq. 7. u. proportional to U2 and U. These scales are chosen as they give a favorable fit to the experimental data. Figures 3–6 show that including the wall effects in the prediction of Umf is successful in describing the effects of height vs. . the values of the universal constants k0 1 and k0 2 are kept at these values for all of the particle types. Umf increases (Figures 3–6). Substituting for a and b. yields. No.1. yields: ! 4tanuk2 ðqg ÞðHÞ 2 drv 4tanukrv D U þ ¼ D dz D ÀlÁ H !  à dp 4tanuk1 #d ðDÞ þ U þ ð1À 2Þðqs À qg Þg À : dz D (22) Setting c2 ¼ 0. but the velocity terms are being scaled differently. In Figures 3–6 and Figures 9 and 10. Substituting Eq. column diameter (H/D) and the effect of particle size to column diameter (d/D). ! dp ¼0 (21) c2 ¼ ð1À 2Þðqs À qg Þg À dz In Eq. bed heights. 23. . 4) into Eq. Re þ b 2. 23 to Eq. it is assumed that the structure of these new terms on which the horizontal stress depends is similar in form to those given by Ergun’s equation. The model only slightly under-predicts the increase in Umf for very small sized particles (d \ 120 lm) in the column with D ¼ 1. the condition which is necessary for fluidization.amongst the particles). and the wall effects will not be included in the analysis. 9 Published on behalf of the AIChE . Hence. As Ergun’s pressure drop equation is used to calculate the pressure drop. and column diameters. appearing on the right hand side of Eq.

there is a need for a new model that can take these effects into consideration. The model presented in this study attempts to include the wall influence by introducing Janssen’s wall effect in the force balance during fluidization. the values of k0 1 and k0 2 are maintained at 610 and 30. Similarly. Eq. the values of bulk density were kept constant with respect to the column diameter. Similarly. Comparison of the curves from Eq. respectively.7 The model does not predict the increase in Umf for the smaller particles (d ¼ 96.4 lm). 2310 DOI 10. These new terms have the same structure as that of the drag term. For example. as d/D ratio increases (due to changes in column diameter). not clearly stated. and column diameters. as given by the widely accepted Ergun equation.7 Also. the prediction for the smallest diameter column is not that good. 30.6 cm). It is assumed that the horizontal stresses acting at the wall are not only a function of the local vertical stress but also are a quadratic function of velocity. 9 AIChE Journal Figure 9. 30 breaks down for large H/D ratio predicting that at large H/D the gas may not fluidize the bed. The minimum fluidization velocity as a function of the column diameter using the experimental data from Liu et al. 30). the increase in Umf can be characterized by the product of H/D and d/D (Eq. in fact. but this may be due to experimental error. the particle mixture is impossible to fluidize. as the existing correlations for Umf do not take wall effects into account. No. hence. 56. Conclusions Experiments show that wall effects influence the minimum fluidization velocity. Above a certain d/D ratio. but it has been observed that this value changes depending on the column diameter. as the quadratic equation yields complex roots. bed heights. the gas may not fluidize the bed. but these deficiencies may be due to September 2010 Vol. The new model fits experimental data reasonably well for the minimum fluidization velocity over a range of particles sizes.7 The curves are from Eq.e. the model also predicts that if these particles are loaded into a 100 cm diameter column to a bed height of about 13 m (H/D $ 13). which is not reported in Liu et al. the model predicts that if glass particles of size 250 lm are fluidized in a 760 lm diameter column (D/d $ 3) with H/D ¼ 1. Figure 10 compares the model results to the average predictions of the Reynolds number at Umf based on a large number of existing correlations. On the other hand. and eventually becomes negligible as d/D becomes very small. then the particles will not fluidize. The experimental data obtained when the wall effects are minimum (at H ¼ 2 cm and D ¼ 2. This situation is predicted by the model.particles. This wall effect reduces as the column increases in diameter relative to the particles. the pressure drop. Additionally. In the model. In this case also. Also.7 to the values predicted by the model. The effect of H/D and d/D on Umf are very similar. D ¼ 1. only the new model is able to predict experimental data obtained when the wall effects are significant (at H ¼ 10 cm. leading to a more significant wall effect. then fluidization becomes difficult.1002/aic Published on behalf of the AIChE .. the ratio of wall contact surface area to the bulk volume increases. Figure 9 compares the experimental data from Liu et al. Existing correlations fail to incorporate these wall effects and.4 cm) are consistent with the model prediction and with the averaged correlation curve. as particle size increases and approaches the same order of magnitude as the column diameter (d/D $ O(1)). The data are plotted in the same dimensional form as given in the orignal work of Liu et al. 30 to the experimental data and existing correlations. As the column diameter decreases in comparison to the particle diameter. Umf increases (compare Figures 3 and 4 or Figures 5 and 6). i.28 particularly for small column diameters. Some deficiencies in the model predictions are noted for smaller particles (around 100 lm and smaller). only values of bulk density were given.8–22 The scatter bars on the correlation line represent the deviation in minimum fluidization velocity predictions using the different correlations. but it does give good predictions for the larger particle sizes.1. the values of voidage were Figure 10. For the largest particles (d ¼ 460 lm). for a constant particle size. This assumption leads to a modified Ergun’s equation incorporating two universal constants.

Persson B. Potic B. Nienow A. 1989.137:302–307. Does the minimum fluidization exist? J Fluid Eng. 9 Published on behalf of the AIChE DOI 10. London. Pope D. 24. Experimental segregation profiles in bubbling gas fluidized bed. 1959. Cambridge: M. 1977. 1978. 2.55:184–189. Versuche uber getreidedruck in silozellen. 1966. Size segregation of binary mixture of solid in bubbling fluidized bed. 8. van Swaaij W. 28. Chem Eng Sci. Chem Eng J. Geldart D. 1971. 2006. 5. Kerstn S.28:287. 26. Fluidization studies of solid particles. Johnson E. Trans Inst Chem Eng. 14. Longwinuk A. Prescott J. Kobayashi H. Baerg A. Gishler P. Trans Inst Chem Eng. Richardson J. Gidaspow D. 2008. and revision received Dec. 1949.62:100–111.39:1045–1049. Raja Rao M. 2nd ed. Babu S. 6. Fluidization correlation for coal gasification materials—minimum fluidization velocity and fluidized bed expansion. Wall effects for the pressure drop in fixed beds. 1966. 2002. Van De Wijer P.44:293–305. 2009. Saxena S. Chem Eng Sci. k10 . 17. Fluidization. Garcia F. Bello A. Yukun Q. Clement S. A study of segregation in a gas solid fluidized bed: particles of different density. Studies on fluidization ll— heat transfer. Publication No.c1.26:1–10. 15.30:54–64. Indian J Technol. Standard practice for measuring fluidization segregation tendencies of powders. Comparision of commonly used correlations for minimum fluidization velocityof small solid particles. Rozenbaum R. k20 . Joseph G. Powder Technol. van Heerdeen C. Olsofsy A. The role of contact stresses and wall friction on fluidization. Newton: Butterworth-Heinemann. Xu G.I. Yurong D. 13. 7. Micro fluidized beds: wall effects and operability. Leva M. Chem Eng Sci. Becker H. 56.57:5123–5141.12:225. AIChE J. Hrenya C. Sen Gupta P. 1951. 27.134:86–97. 1981. 1980. Acknowledgments Jim Prescott and colleagues at Jenike and Johanson are thanked for their support of this research and for providing the segregation testing equipment. Delebarre A. Broadhurst T. Loezos P. Fluidized binary mixtures in which denser component may be flotsam. D6941–03.T. The measurement of incipient fluidization velocities in a bed of coarse dolomite at temperature and pressure. Fluidization Engineering. 2002. 11. 318. 23. Grewal N. Nobel A.43:1220–1226. b A a0 .c0 Ar D d F g H k. Powder Technol. Heat transfer in a fluidized solid bed.1:51–66. Chem Eng Prog Symp. 1951. Janssen HA.21:238–247.9:77. Mechanics of fluidization. 22. Di Felice R. 18. Yu Y.59:3037–3040. ASTM International. AIChE Journal September 2010 Vol. Brone D. Powder Technol. Pillai B.53:2804– 2813. Pressure drop and minimum fluidization velocities of air-fluidized bed. 4. 10. Villar J. 9. Baeyens J.1002/aic 2311 .60:5982–5990. Wen C. Goroshko V. Ind Eng Chem. Stevens A. Development of an improved fluidization segregation tester for use with pharmaceutical powders. 2007.124:595–600. Romero A. Notation [ W l qg qs rH rv u Dp a. Levenspiel O.26:229–234. McCall M. Chem Eng Sci. Davies L. Lidan Y. Costamagna P. Can J Res Sect F. 2009. 12. Pharm Technol. Ye M. 25. 29. Prediction of minimum fluidization velocity for multicomponent mixtures (short communication). 20. 1950.30:195–205. Miller C. AIChE J. Patel P. 1980. Kumar P. AIChE Symp Ser. 3. 1991. Liu X. 1966. 19. k1. Sundaresan S. Hancock B. 1975.the significance of cohesive and adhesive forces at this scale. Leboreiro J. 1966. Gas interchange between bubbles and the continuous phase in a fluidized bed.74:176–186. 1. Z Ver Deutsch Ing. Indian J Technol. Saxena S. Tosatti E. Chiba S. Van der Hoef M. 16. 2004. Todes O. No. 21. Powder Technol. Manuscript received Aug. Gibilaro L. Press. Chiba T. 2003. Shah B. 1895. 14. Klassen J. Vogel G. Powder Technol. ASTM International. Talwalker A.58:169–174. Onset of fluidization and slugging in beds of uniform particles. Segregation in beds of large particles at high velocity. New York: Springer Publishing Company. Hedden D. Institute of gas engineers report.c2 Re U Umf z ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ ¼ voidage sphericity viscosity density of gas density of solid stress in horizontal direction stress in vertical direction friction angle pressure drop Ergun constants cross-sectional area various constants of the quadratic Archimedes number column diameter particle diameter force Gravity height of fixed bed various constants Reynolds number fluid velocity Minimum Fluidization velocity instantaneous height measured from the top of the bed Literature Cited 1. Kunii D. 2003. k2. Gao S. Physics of Sliding Friction (NATO Science Series E). Hydrodynamics and Heat Transfer in fluidized beds.b0 . New York: McGraw-Hill. This new model should greatly facilitate scaling results from MFBs to more traditional fluid bed sizes. Krevelen v. 2005. 1974. Huilin L. Fluidization of hot compressed water in micro reactors.