Gas-Solid Mass Transfer in a Rotating Drum

RITIK4 JA UHARI, MURRAY R. GRA I"'and JACOB H . MASLIYAH

Department of Chemical and Materials Engineering, University of Alberta, Edmonton, AB T6G 2G6.
Gassolid volumetric mass transfer coefficients (k.8)were measured in a rotary drum by evaporation of n-decane into

dry air from the surface of porous solids. The effect of drum rotational speed, N (0.09 to 2.0 m i d ) , solids volume fraction,

q (0.043 to 0.25) and the presence of baffles on k/ were investigated. In the presence of baffles, kd was independent
of q and higher than in the case of a rolling bed where no baMes were present. For the rolling bed case, k.8 increased
with increasing q. Mass transfer in the rolling bed was modeled based on the particle motion.

On a mesure les coefficients de transfert de matiere volumetrique gaz-solides (kd)dans un tambour rotatif par
evaporation de n-decane dans I'air sec de la surface de solides poreux. On a etudie I'effet surIc8 de la vitesse du tambour,
N (de 0,09 a 2,OO m i d ) , de la fraction de volume des solides, q (de 0,043 a 0,25) et de la presence de chicanes. En
presence de chicanes, kd est independant de q et plus grand que dans le cas d'un lit roulant non muni de chicanes. Dans
le cas du lit roulant, k/ augmente avec I'augmentation de q. Le transfert de matiere dans le lit roulant a etC modelise a
partir du deplacement des particules.

Keywords: rotary drum, gassolid mass transfer, rolling bed.

otary drums and kilns are widely used for processing of

granular materials in a variety of applications such as
drying, incineration, mixing, heating, humidification, calcining, reducing, sintering, and gas-solid reactions. Most of
the reaction processes operate at high temperature, hence the
main functions of the rotary kiln are to heat the solids to the
reaction temperature and provide heat for endothermic reactions. In the drying and reaction zones of the rotary kiln,
heat transfer may be the rate limiting process, therefore literature studies have concentrated on the various modes of heat
transfer (Ban et a1.,1989a,b). Mixing of solids plays an
important role in the performance of rotary drums, particularly
in heat transfer between gases and solids (Rao et al., 1991).
The solids bed motion in a rotating drum is dependent on
the drum rotational speed and the geometry of its internals.
At moderate rotation speeds without baffles, the bed motion
is in the rolling mode. An active layer of particles moves
downward along the surface of the bed, as particles are carried upward by drum rotation (Figure 1). The active upper
zone tends to be a crescent shape, with the shallowest depths
at the top and bottom of the upper surface of the bed (Henein
et al., 1983). The behaviour of the upper surface layer is
expected to be particularly significant for gas-solid mass
transfer. The rotary drums Henein et al. used for observing
bed behaviour were 0.4 m and 1.0 m in diameter.
The rolling bed at low rotational speeds has a constant
angle of inclination and a flat planar surface. The inclination
of the rolling bed is at the dynamic angle of repose, 8. At
constant bed depth, the thickness of the active layer (ha) was
found to be thinner for smaller sized particles. Increasing
bed depth and rotational speed, increased the active layer
thickness. For shallow beds at rotational speeds less than 2
rpm, the active layer may occupy up to one-third of the bed
depth, but for deeper beds (h/Dp> 50) the fraction declined
to about 8%.
Tscheng (1978) derived the following equation to predict
the surface particle velocity in the active layer region:

V, = 120ha L:

- 8Rh,

cos(

$)] . . . . . . . . . . . . . . .

'Author to whom correspondence should be addressed. E-mail address:

rnurray.gray@ualberta.ca
224

er
Figure 1 -Particle motion in the absence of baffles (rolling bed)
and velocity profiles in the active and bulk regions of the rolling bed.

The velocity gradient in the active layer was assumed to

be linear (Figure 1). Experimental data suggest a weaker
Data from Lebas et al.
dependence of V, on rotation 0).
(1995) suggest V, cc
while the data of Mu and
Perlmutter (1980a,b) indicated Vs oc
The analogy between heat and mass transfer can be used
to predict mass transfer coefficients.Tscheng and Watkinson
(1979) conducted an experimental study of convective heat
transfer from hot air to sand in a counter-current, non-fired
rotary kiln. They measured the gassolid convective heat
transfer coefficient (h ), as a function of N (0.4-6.0 m i d )
and 7 (0.065 to 0.17).%he solids motion in the drum was in
the rolling mode except at 0.4 min-'. They found that
increasing the gas flow rate increased h ,but increasing the
drum rotational speed gave only a smafincrease in h , . The
velocity of the particles along the bed surface was found to
be proportional to the square root of the rotational speed
similar to Mu and Perlmutter. Tscheng and Watkinson
(1979) obtained the following dimensionless expression for
the Nusselt number from their experimental data (based on
equivalent diameter, 0,):

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, APRIL, 1998

1600<Reg< 8000.........................

(2)

Mu and Perlmutter (198 1) developed a model to describe

the performance of rotary reactors which included a detailed
description of particle motion and heat transfer coefficients.
In the model, they used the following correlation from the
literature to evaluate the gas-solid convective heat transfer
coefficient:
0.2

hgs =o.023[$r8[$]

[u2 +b2]o.4. . . . . . . . (3)

Due to the more complex particle motion in drums with

baffles, general correlations are not available. Friedman and
Marshall (1949) measured volumetric heat transfer coeficients (Uu)in a flighted rotary dryer as a function of air rate,
feed rate, rotation rate, slope and number of flights. By
decreasing the slope of the dryer, q was increased. The volumetric heat transfer coefficient, Uu increased with decreasing slope until critical solids volume fraction of 0.030 was
reached. Further decrease in slope did not change Uu as the
solids holdup in the drum was above the critical value of
0.030. As holdup in the dryer increased, flightsbecame l l l y
loaded until finally a point was reached where any increase
in holdup caused no further increase in flight loading and a
bed of material formed at the bottom of the dryer. This bed
did not effectively contact the air. They expected this critical
solids volume fraction to change with dryer size or flight
design, however, they did not conduct any experiments to
verify their hypothesis. The value of Uu increased rapidly
from zero flights to two flights. As the number of flights
were fbrther increased to four and eight, Uu did not increase
rapidly. This indicated that Uu was little influenced by further
increase in the number of flights.
Any attempt to define a heat or mass transfer coefficient
for beds of solids in rotary drums requires a surface area for
gassolid contacting. A common assumption in rolling beds
is to use the chord length (L,) multiplied by the length of the
drum (L), even though this area underestimates the true
value under all conditions (Tscheng and Watkinson, 1979).
Drums with baMes give a higher area due to showering of
particles through the gas. This particle motion has been used
to calculate contact area (Hirosue and Shinohara, 1976),
although the effectiveness of this contact will decrease when
the number of falling particles is increased (Hirosue, 1989).
In some other rotary drum applications the gassolid
mass transfer is important. Examples include production of
some cefamics (wei, 1983) and solid-state fermentation
(SSF) processes. Rotary drums are attractive for SSF because
the drum rotation enhances mixing and exposure of particles
to the gas (Lonsane et al., 1985). The performance of rotary
drums has been improved by including internals like baMes
which further enhance the mixing of solids and improves the
oxygen transfer (Fung and Mitchell, 1995). Although gassolid
mass transfer is important for solid-state fermentationsystems,
information about oxygen transfer or gassolid mass transfer
in rotary drums is unavailable.
The objective of this project was to study the effect of
drum rotational speed, solids volume fraction, and baffles
on the volumetric gas-solid mass transfer coefficient, k / .
Mass transfer coefficients were determined from measurements of the evaporation of n-decane from porous alumina
particles. The experimental mass transfer data were compared

with data from analogous heat transfer studies in the literature. Mass transfer in a rotating drum was then modeled
based on the particle motion.

Theory
If the flowing gas phase is well mixed, then the steadystate rate of mass transfer of a compound from the solid to
the gas will equal the total amount of compound leaving the
drum:

P)= QC' .......................

k/(@"-

(4)

If the gas phase concentration of the compound is measured by a sensor that gives an output proportional to the
concentration, Equation (4) can be rewritten in terms of the
sensor signal at a given steady-state condition:

,/(Fur

- Sy)

= QSy ........................

(5)

If the flow rate of gas is changed at otherwise constant

experimental conditions, then a new steady-state will be
established:

,/(Fur - SF) = Q2SY

.......................

From Equations ( 5 ) and (6), the following expression for

evaluating the volumetric mass transfer coefficient was
obtained:

ksA=[

ss"Q -sya ......................

sy-r

(7)

The above derivation assumes that gas velocities through

the drum are too low to affect k/. This approximation will
be satisfied if the particle motion is greater than the axial gas
velocity. For every experimental condition, the ratio of
drum wall velocity, Vw to the maximum superficial axial
gas velocity has to be greater than unity in order to satisfy
the approximation mentioned above. By evaluating the data
in a piecewise manner, the sensor signals at two steadystates can be used to obtain k/ without detailed calibration
or measurementof the response to the saturated concentration.
Equation ( 5 ) can also be rearranged to obtain the following expression:

If data from a series of steady-states are plotted as l/Sss

versus Q for a given experimental condition (i.e. fixed N,
fured q and presence or absence of baffles) then linear
regression will give an intercept, b and a slope, m. The volumetric mass transfer coefficient, k / can then be evaluated
from the intercept and the slope:

b
ksA=m

. . . . . ...........................

(9)

Experimental

MATERIALS
The porous solid phase was Alumina S-201, 5-8 mesh
catalyst carriers (LaRoche Chemical Industries Inc., Baton

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76,APRIL, 1998

225

Rouge). The particle size was between 2.4 and 4.0 mm in

diameter. The pore volume was 0.46 cm3/g(based on vacuumdried solids). The solid phase gave low dusting and high
mechanical strength. The liquid hydrocarbon used to saturate
the porous solids was n-decane (Sigma, St. Louis). This
hydrocarbon has a low vapor pressure at ambient temperature (Wilkins and Thodos, 1969) so there is no wet-bulb
temperature effect during the mass transfer experiments.
The porous solid phase saturated with n-decane was loaded
in the drum, The preparation of the saturated solids was as
follows: supplied solids (- 700 g) were vacuum dried at
250C and 6.5 kPa for eight hours. The vacuum dried solids
were cooled overnight in a dessicator and then saturated in
700 cm3 of n-decane for twenty-four hours at room temperature (21-22OC). The saturated solids were recovered and
wiped dry to remove the free surface liquid.
METHOD

To avoid any internal resistances to mass transfer, the

experiments were conducted during the constant-ratedrying
period of the saturated solids. Experiments in a tray dryer
showed that the rate of n-decane evaporation was constant
up to 1549% loss of hydrocarbon (Jauhari, 1997). Consequently, the evaporation of n-decane in rotary drum experiments was kept below 10% to ensure constant rate drying.
A schematic diagram of the equipment used in the rotating drum experiments is shown in Figure 2. The 20 L drum
was made of Plexiglas. The drum length was 0.308 m and
the drum diameter was 0.29 m. For measurement of mass
transfer coefficients in the presence of baffles, a stainless
steel cage with eight baffles was inserted in the drum. The
width of the baffles was 10% of the drum diameter. For
measurement of mass transfer coefficients in the absence of
baffles, a stainless steel cage with 16 thin rods (2.3 mm
diameter) was inserted to prevent slippage of the solids
against the wall.
The drum was placed on two rollers which were rotated,
through a chain and gears, by means of a variable speed
motor. The air was passed through a coalescing filter to
remove any entrained hydrocarbons, then a Drierite Gas
Purifier (W.A. Hammond Drierite Co., Xenia) to remove
moisture. Flow rate of dry air into the drum was regulated
by a flow meter and controller. The gas stream exiting the
drum was sent to a mass spectrometer. The mass spectrum
of n-decane (Stenhagen et al., 1969) indicated that the
strongest signal was at mass to charge ratio of 43. The ndecane signal from the mass spectrometer was recorded on
a computer. Details of the equipment have been reported
previously (Jauhari, 1997).
Absolute calibration of the mass spectrometer signal
could not be done due to overlap between the signals for ndecane and possible internal standards. Drift in the mass
spectrometer signal precluded calibration with a n-decane
standard. The analysis of the data by Equation (7) compensated for the drift in the signal by using successive steadystate readings. The drum operating conditions were: N (0.09
to 2.0 min-') and h (0.043 to 0.25). Dry air flow rates (at
room temperature and pressure) were in the range 1-14
L/min. This range of rotation speed was selected with bioreactor operation in mind. Low rotation speeds prevent excessive
shear forces, which can damage the mycelia of fungi, while
at the same time providing sufficient aeration. The volume
fraction of solids covered the typical range of operation of
226

pl,

-Db

....

i"L, ;
...

1. A a supply valve
2. Preswe regulator
3 coalcoolog filter
4 Gnshifisr

5. Flow meter
6 . Controller

7 Rotruy drum
8. Glass trap

9 Mass spectrometer
10. Computer
I I . Motor

12. Roller

Figure 2 -Schematic diagram of the equipment for rotating drum

experiments.

rotating drums and kilns. The air flow rate was selected to
obtain accurate measurements of n-decane concentration,
and hence flux from the solid bed to the vapor phase.
The experimental procedure was as follows: the weight
of the saturated solids was measured and the solids were
loaded in the drum. The solids bed temperature at the start
of the experiment was recorded. After the air was introduced
in the drum,the n-decane signal from the mass spectrometer
was recorded at 30 s intervals. For the same experimental
conditions (i.e. fixed N, fvred q and presence or absence of
baffles ) after a steady-state was reached, the air flow rate
was changed. At every experimental condition, a series of
step up or down in flow rate was done and the corresponding steady-state signals from the mass spectrometer were
recorded. Measurements of bed temperature at the end of
each experiment showed no significant change, within f IC".
Drum rotational speed was measured by placing a mark
on the drum and measuring the time required for one revolution. The quantities that were measured while the drum
was in motion are shown in Figure 1. The dynamic angle of
repose, 8, was measured with a long arm protractor by viewing the bed through the Plexiglas end of the drum. The chord
length of the solids bed, L,, was measured by a metric scale.
To measure the active layer, h,, and total bed depth, h, a millimeter scale was pasted across the center of the Plexiglas
end of the drum. As the drum rotated, h, and h were measured at the point of maximum thickness of the rolling bed.
Qualitative observations about the surface particle velocity,
V,, were also made.

Results and discussion

PARTICLE MOTION OBSERVATIONS

In the presence of baffles, showering patterndfrequencies,

types of particle motion and surface particle velocities were
observed for drum rotational speeds in the range 0.09 to
2.0 min-*. The particle motion was studied at q values of
0.043 and 0.086. Figure 3(a) shows the particle motion
observed at q = 0.043. From the figure it can be observed
that baffles A, B and C were not overloaded with solids. As
the dnun rotated, the top layer of solids on baffle A showered onto baffle B and top layers of particles from baffle B
showered onto baffle C. Particle mixing was rapid in the
presence of baffles. At rotational speeds less than 0.5 min-',
the solids stayed on the baffles for a longer period of time.
The frequency of showering from baffle A to B and baffle B
to C was low, therefore, the bed renewal was slow. The
showering pattern was like flow over a weir. At rotational

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, APRIL, 1998

TABLE1
Measurements of Characteristics of Rolling Bed at Different
Rotational Speeds (q = 0.043)
N (mid)
0.291
0.576
1.213
1.657
1.967

B
Figure 3 - Particle motion in the presence of eight baMes and
varying drum rotational speed (a) q = 0.043 (b) q = 0.086.

speeds larger than 0.5 m i d , the showering frequency

markedly increased with increasing N. Therefore the bed
renewal was faster.
Figure 3(b) shows the particle motion that was observed
at q = 0.086. The baffles were overloaded and baffle B was
completely buried. The solids bed between baffles A, B and
C was inclined. Showering occurred from baffle H to A,
while rolling of particles was observed from baffle A
towards baffle C. The showering pattern was like flow over
a weir. Both the showering frequency and the particle velocity
along the inclined surface were visually observed to increase
with increasing N.
In the presence of baffles the particle velocities were high
due to the free fall that occurred during showering. Due to
the complex particle motion, area available for mass transfer,
A, could not be estimated.
Since the particle motion in the absence of baffles is relatively less complex than with baffles, detailed observations
(qualitative and quantitative) were made. The particle
motion observations were done in the following range of N:
0.29 to 2.0 min-' and q: 0.043 to 0.25. Figure I is a
schematic diagram of the particle motion that was observed,
consistent with the prior work of Henein et al. (1983). As the

h, (m)
6.17 x l t 3
6.77 x I t 3
8.00 x lo3
1.00 x 1 t 2
1.18 x 1W2

h(m)
2.62 x
2.67 x
2.65 x
2.67 x
2.63 x

1t2
1t2
1t2
lt2

It2

hdh(%)

L,(m)

23.5
25.4
30.2
37.5
44.9

0.160
0.160
0.160
0.155
0.160

Q(O)

28.0
27.5
28.5
28.0
28.5

drum rotated, a thii top layer of particles in the active layer

region rolled down the inclined bed surface. The surface
particle velocity, V, was visually observed to increase with
increasing N. Particles that rolled down often collided with
other particles. These collisions promoted mixing in the
active layer region. The particles in the bottom bulk flow
region had no motion relative to the drum wall, hence there
was no mixing in this region of the solids bed.
Table 1 lists the results that were obtained at q = 0.043
and varying rotational speed. The solids bed was shallow at
this solids volume fraction. At a fixed q and varying N, the
total bed depth, chord length and dynamic angle of repose
were almost constant. The thickness of the active layer
region, h, increased with N such that the fraction of bed in
active motion (hdh) increased from 23.5% to 44.9%. This
result indicated that overall solids mixing increased with
rotational speed as both the surface particle velocity and
fraction of bed in active motion increased.
Table 2 shows the results that were obtained at a fixed N
value of 1.1 m i d and varying q from 0.043 to 0.25. From
the measurements shown in Table 2, it can be concluded that
0 was almost constant. The chord length increased with
increasing q. The active layer depth increased with q, however the fraction of bed in active motion (hdh) decreased
from 30.2% to 19.4%. The surface particle velocity did not
visually seem to change with h. The bed behavior was more
defied at solids volume fractions greater than 0.043 as the
solids bed was not shallow. The rolling motion of the particles
was also smoother (less collisions) as the solids volume
fraction was increased.
For the rolling bed, the experimentally measured values
of the chord length were close to the geometric values
(Jauhari, 1997) which indicated that the solids bed had not
loosened. Also particle segregation was not observed when
the solids bed was viewed through the Plexiglas end. The
current experimental results are in agreement with the trends
observed by Henein et al. (1983).
The principle of mass conservation suggests that in a
rolling bed, the volumetic flow rate of solids to the active
layer region is equal to the volumetric flow rate of solids to
the bulk flow region due to drum rotation. Expressions for
volumetic flow rate of solids to the active and bulk regions
were derived by assuming linear velocity profiles as shown
in Figure 1. The following equation for evaluating the surface
particle velocity was obtained (Jauhari, 1997) by equating
the active and bulk volumetric flow rates of solids:

v, =m[L-iJ..
30 h,
. . . .. . . . .. . . . ..

From the experimental measurements of h, and h for q

varying from 0.043 to 0.25 and N varying from 0.29 to 2.0
min-I, the surface particle velocity, V, was estimated using

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, APRIL, 1998

227

TABLE2
Measurements of Characteristics of Rolling Bed at Different Solids Volume Fractions ( N = 1 . 1 m i d )

("/.I

L, (m)

L, (m)
geometric
chord length

30.2
27.4
19.4
19.8

0.160
0.195
0.240
0.262

0.164
0.201
0.243
0.265

h,lh

h, (m)

rl

0.043
0.086
0.172
0.250

8.00 x
1.163 x
1.130 x
1.719 x

lV3
1C2
1W2
1V2

h (m)
2.65 x
4.23 x
6.69 x
8.65 x

1V2
lV2
1V2
1W2

0 ("1
28.5
27.5
28.0
27.5

Equations (1) and (10). The following regression equation

was obtained when V, was estimated using Equation (1):
V, = 0.057No.55

(R2 = 0.89)

...............

(1 1)

2.0e-6

25%

25%(2)

1.4 L/min

1.4 L/min

Similarly, the following regression equation was obtained

when V, was estimated using Equation (10):
V, = 0.031No.54

(R2= 0.87). . . . . . . . . . . . . . . . (12)

From Equations (1 1) and (12) it is evident that the dependence of V, on N was of the same order as observed by
Tscheng and Watkinson (1979), Mu and Perlmutter
(1980a,b) and Lebas et al.( 1995). Using experimental measurements at N = 1.1 m i d , surface particle velocities calculated from Equations ( l ) and (10) were used to obtain the
following regression equations:
V, = 0.14q0.21 (V, from Equation (l), R2 = 0.72)- . (13)

2
3
(I)

20

40

60

80
t

V, = 0.10q0.29 (V, from Equation (lo), R2 = 0.84). . (14)

From Equations (13) and (14) it can be concluded that V,
should increase with q . However, in the current experimental
study, V, as observed visually did not seem to increase with
increasing q. A possible reason for this discrepancy is that
V, was observed only qualitatively. Also surface velocity
calculations depend on the accuracy of h and ha measurements, which were measured to a precision of approximately
one particle diameter which is 2.4 to 4.0 mm.

MEASUREMENT
OF VOLUMETRIC MASS TRANSFER COEFFICIENT,
kd-4

A sample of the raw data is given in Figure 4, showing a

series of steady-states in response to steps up and down in
the air flow rate. The notation 100%(2),60%(2) and 25%(2)
means that the gas flow rate was set to loo%, 60% and 25%
(based on the maximum flow rate supplied by the flow
meter) for the second time within the same experiment to
confirm the steady-states that were obtained. The data in
Table 3 show calculated values of k/ from the piecewise
method (Equation (7)). Six segments were considered for
the piecewise method. From sensor signals and air flow
rates at 100% flow and 60% flow, the value of k / using
Equation (7) was 20.58 L/min. The average of the k/ values
from the six piecewise steady-statesegments was 2 1.76L/min.
The value of k/ obtained by the slope method, Equation
(9), was 2 1.72 L/min. In the absence of drift in the signal,
the two methods were equivalent. Where drift occurred, k/
was calculated by the piecewise method. The ratio of the
drum wall velocity to the maximum superficial gas velocity
for the mass transfer experiments varied from 1.8 to 20.4.
228

100

120

140

160

(min)

Figure 4 - Sensor signal for rolling bed experiment at N = 1.21

m i d and q = 0.043.

Thus the value of k/ evaluated from the raw data was

essentially independent of the gas flow rate, Q.
GAS-SOLID
VOLUMETRIC MASS TRANSFER COEFFICIENT IN THE
PRESENCE OF BAFFLES

Figure 5 shows the mass transfer coefficient plotted as a

function of dnun rotational speed at a solids volume fraction
of 0.043. The volumetric mass transfer coefficient increased
linearly with increasing dnun rotational speed consistent
with an increase in the rate of particle mixing. The scatter in
the data is comparable with the scatter evident in equivalent
mass transfer data for packed beds (Petrovic and Thodos,
1968). The data of Figure 5 also show the volumetric mass
transfer coefficient, k/ as a function of drum rotational
speed in the presence of baffles when the solids volume
fraction was doubled to 0.086. For some mass transfer
experiments, only point estimates of k/ were available. The
value of k& increased linearly with increasing drum rotational speed. The observed particle motion supported this
trend as the mixing increased with increasing N.
The volumetric mass transfer coefficient, k/ in the presence of baffles was independent of q within experimental
error. Although the particle motion observed at these two
solids volume fractions was different, mass transfer coefficients were the same, consistent with the results of Friedman
and Marshall (1949). At q = 0.086, the baffles were overloaded and excess solids formed a rolling bed. Only the particles that were showering effectively contacted the air.
Therefore, by doubling the solids volume fraction, the number

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, APRIL, 1998

TABLE3
Example Values of kd for the Rolling Bed Experiment at
N = 1.21 m i d and q = 0.043

(L/min)
5.700
3.414
1.413
5.697
1.413
3.413

(Pa)
1.5716 x 10-6
1.7213 x 10-6
1.8764 x 10-6
1.5887 x 10-6
1.8688 x 10-6
1.7292 x 10-6

Flow

("/I

100

60
25
1OO(2)
25(2)
W2)

100

I
0
0
0

k p (L/min)
Piece-wise
Method

Step change in
flow rate
100%to 60%
60% to 25%
25% to 100%(2)
100%(2) to 25%(2)
25%(2) to 60%(2)
100% to 25%

20.58
20.80
22.24
22.87
23.35
20.69
Mean. 21.76

rT-G&zq

q = 0.043 and baffles

q = 0.086 and baffles
qxO.086 and baffles (point estimates)

-Linear trend line (baffles)

+ q4.043 and mlling bed
-- Linear trend line (rollingbad)

.f

3.<

80

70

80

a40-

30

20

lot

0.00

0.05

0.10

0.15

0.20

01

1
0.25

0.30

11
0 '
0.0

0.2

0.4

0.6 0.8

1.0

1.2

1.4

1.6

1.8

2.0

N (min")

Figure 5 - Volumetric mass transfer coefficient in the presence

and absence of baffles for varying drum rotational speed.

of particles that effectively contacted the air did not increase

by a large amount. Consequently, k/ was independent of q.

GAS-SOLID MASS TRANSFER COEFFICIENT, k/

IN THE ABSENCE

OF BAFFLES

The volumetric mass transfer coefficients, k/ obtained

from the rolling bed experiments at constant solids volume
fraction of 0.043 and varying drum rotational speed are
shown in Figure 5 . The volumetric mass transfer coefficient
increased linearly with increasing drum rotational speed.
The error bars indicate that the variation in the k/ value is
up to 36% which is comparable to the variation in mass
transfer data for packed beds (Petrovic and Thodos, 1968).
Observed particle motions indicated that the solids mixing
increased with increasing drum rotational speed, which suggests that the kd value should increase at higher rotational
speeds due to better solids mixing.
The data of Figure 6 show the volumetric mass transfer
coeficients obtained at a constant rotational speed of 1.1
m i d and varying solids volume fraction. The value of k /
increased with increasing 7. Mass transfer would depend on
the velocity of solids with respect to the gas phase, so that

Figure 6 -Volumetric mass transfer coefficient at N = 1.1 m i d

and varying solids volume fraction. The curve shows the trend of
the data.

constant velocity would suggest a constant value of k,. As

the surface particle velocity is little affected by q, this suggests that there may be an increase in the mass transfer area
with increasing q which resulted in the overall increase of
the k/ value. The mass transfer area in a rolling bed could
not be estimated directly.
The data of Figure 5 show that the volumetric mass transfer
coefficients in the presence of baffles are higher than for the
rolling bed case. These results can be explained through the
observed particle motions. In the presence of baffles, the
particle velocity was high relative to the air due to the free
fall involved during showering from the baffles. As a result,
the solids mixing was high and the particles effectively contacted the air. The particle velocity in the rolling bed was
lower, resulting in comparatively low solids mixing and a
lower velocity of the surface particles relative to the gas.
Similar trend for overall heat transfer coefficients (Va)was
observed by Friedman and Marshall (1949) when the number
of baflles in the dnun was increased from zero to eight.

NORMALIZED
MASS TRANSFER COEFFICIENT, k,'

FOR THE

ROLLING BED

The volumetric mass transfer coefficients, kd,for the

rolling bed were normalized on the basis of the volume of
bed renewed by drum rotation following the surface renewal

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, APRIL, 1998

229

0.8

0.6

0.5

'm

0.4

-m

r:

R'=096

Figure 7 - Schematic diagram of the rolling bed with the crosssectional area of the active bed, .S,

1 '

0.1
0.25 0.3

approach to gas-liquid mass transfer. The volume of bed

renewed was assumed to be equal to the cross-sectional area
of the active bed, S, times the drum length, L, where SA was
calculated from an active layer of constant thickness h,
across the top of the bed as illustrated in Figure 7 (Jauhari,
1997). The normalized mass transfer coefficient, k,' for the
rolling bed was calculated using the following equation:

q=OO43

0.6

0.8

0.4

N (min-')

Figure 8 - Normalized mass transfer coefficient for the rolling

bed as a function of drum rotational speed at q = 0.043.
1.2
0 Experimental data at 1.1 min"
95% confidence limits
Mean value = 0.63 s-'

1 .o

In Figure 8, the normalized mass transfer coefficient, k,' at

q = 0.043 is plotted as a function of drum rotational speed.
Regression of the data on a logarithmic scale gave the following regression equation:
k,' = 0.73No.56. . . . . . . . . . . . . . . . . . . . . . . . . . . .

'b 0.8

(16)

These results are consistent with the increase in particle

mixing with increasing rotational speed, due to the increase
in both the surface particle velocity (V,) and the fraction of
bed in active motion (hdh).
The normalized mass transfer coefficient, k,' at rotational
speed of 1.1 m i d are plotted versus q in Figure 9. The
value of k,' decreased sharply as q was increased from 0.043
to 0.086. Increasing q from 0.086 to 0.25 did not significantly change the k,' value. The mean value of k,' for q in the
range 0.086 to 0.25 was 0.63 s-l. The 95% confidence limits
of the mean value are also shown in Figure 9. Since the
experimental data (except one point) lie within the confidence interval, it can be concluded that statistically the ki
value did not change for q in the range 0.086 to 0.25. The
initial decrease in k,' value could be due to the difference in
the bed motion as q was increased. At q = 0.043, the solids
bed was quite shallow and approximately 1/3 of the total
bed was in active motion. As q was increased to 0.086, the
fraction of bed in active motion had declined. Thus the k,'
value was higher at the lower solids volume fraction due to
better solids mixing.
COMPARISON OF ROLLING BED MASS TRANSFER RESULTS WITH
HEAT TRANSFER STUDIES IN THE LITERATURE

Tscheng and Watkinson (1979) found that the effect of

drum rotational speed on hgs was lower than the effect of gas
230

0.4 I
0.00

0.05

0.10

0.15

0.20

0.25

Figure 9 -Normalized mass transfer coefficient at N = 1.1 min-'

and varying solids volume fraction.

velocity (Equation (2)), because their experiments were

conducted at air velocities about 2 to 3 orders of magnitude
larger than in the current experiments. They also observed a
slight negative effect of increasing q on h s, consistent with
the present finding that the value of k,' cfecreased when q
was increased from 0.043 to 0.086.
Mu and Perlmutter (1981) used Equation (3) to evaluate
h s. For the case where the particle velocity is much larger
& the gas velocity then:

h,

Q:

Vs0.8

...............................

(17)

The normalized mass transfer coefficient, k,' for the

rolling bed at q = 0.043 and varying drum rotational speed
were plotted as a function of V,. From the logarithmic plot

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, APRIL, 1998

of k, versus V, calculated using Equation (l), the following

regression equation was obtained:
kA!= 1 1 .9V,0.97

(R2=0.96).

. . . . . . . . . . . . . . . (18)

From the logarithmic plot of k, versus Vs calculated using

Equdtion (lo), the following regression equation was obtained:
k.,! = 20.7V.:.97

( R 2 = 0.95). . . . . . . . . . . . . . . . . (19)

Comparing Equations (1 8) and (19) with Equation (1 7)

suggests that the dependency of the heat or mass transfer
coefficient on V,s is roughly of the same order.

IMPIKATIONSFOR SOLID-STATE FERMENTATION (SSF)

This study provided mass transfer coefficients as a function of the major operating parameters, drum rotation rate
and volume fraction of solids. The Sherwood numbers were
not calculated because the gas-solid contact area could not
be estimated. Nevertheless, these data provide insight into
mass transfer for processes such as SSF.
In the current experiments, the volumetric mass transfer
coefficients (k,A) were found using n-decane as a model
compound. The gas-solid mass transfer coefficients for oxygen
in a rotating drum were calculated from the a-decane mass
transfer coefficients (Jauhari, 1997). This was done by correcting the k,,A values by using the Schmidt number (Sc) to
correct for the difference in the gas phase diffusivities of ndecane and oxygen. The gas-solid oxygen mass transfer
coefficients in the presence and absence of baffles were
approximately one to two orders of magnitude higher than
the gas-liquid oxygen mass transfer coefficients for a typical
sparged and agitated liquid fermentor (Yegneswaran et al.,
1990).The maximum oxygen flux to the microorganisms is
proportional to the oxygen mass transfer coefficients.
Therefore, oxygen transfer to the microorganisms is more
efficient in SSF systems than liquid fermentation systems.
The large increase in oxygen transfer in the rotating drum
suggests that oxygen flux would not be limiting in such a
fermentor, unlike submerged liquid fermentation. Instead,
other factors such as shear and heat transfer would likely
limit performance.

Nomenclature
total mass transfer area, m2
intercept, Equation (8), Pa-
gas phase saturation concentration of the diffusing
compound, rnol/L
C\\ = steady-state gas phase concentration of the diffusing
compound, rnol/L
DAB = diffusivity, m2/s
De = equivalent diameter,
[0.5D(2x - + sing)]/[n: - pi2 + sin(b/2)], ni
f,, = particle diameter, m
= total bed depth, m
h , = active layer thickness, m
h,, = gas-solid heat transfer coefficient, W/m2.K
k, = gas-particle mass transfer coefficient, L/minm2
k, = gas-solid normalized mass transfer coefficient, s-I
L
= drum length, m
L , = chord length, m
rn
= slope in Equation (8), rnin/L.Pa
N
= drum rotational weed in revolution Der minute. min-
= Nusselt number based on De,hK,s D J h , Q = gas flow rate, Umin
R
= radium of the drum, rn
R2 = correlation coefficient, ReK = gas phase Reynolds number, up DJp, Re, = transverse Reynolds number, De2op/p, S ( t ) = n-decane mass spectrometer signal, Pa
s, = cross-sectional area of the active bed, rn2
S = gas phase saturation signal of the diffusing compound, Pa
ss
= gas phase steady-state signal of the diffusing compound.
Pa
sc = Schmidt number, @p DAB,I
= time, min
u
= gas velocity, m/s
v, = surface particle velocity, m/s
v, = drum wall velocity, nRNI30, m/s
Ua = volumetric heat transfer coefficient, Wim3.K

=
=
CUf=

Greek letters

b
q
0
h
p

central angle, rad

solids volume fraction, dynamic angle of repose, degree
= thermal conductivity, W1m.K
= viscosity, Pa.s
= fluid density, kg/m3
= drum angular velocity, radis
=

=
=

Conclusions

The volumetric mass transfer coefficient, k,,A in the presence

of baffles increased linearly with the drum rotational speed.
These results are consistent with the observed mixing of the
solids. The same trend was observed in the absence of baffles,
but the values of k,,A were approximately half as large due
to the less effective gas-solid contacting. The volumetric mass
transfer coefficient in the presence of baffles was independent
of the solids volume fraction although two types of solids
motion were observed. In the absence of baffles, the volumetric mass transfer coefficient increased with increasing q.
The normalized mass transfer coefficient, k,\ for the
rolling bed at a solids volume fraction of 0.043 were correlated with the drum rotational speed. The normalized mass
transfer coefficient for the rolling bed at a fixed rotational
speed of 1.1 min- and solids volume varying from 0.086 to
0.25 was essentially constant at a mean value of 0.63 s-l.

Subscripts

Acknowledgement
The authors wish to thank the Natural Sciences and Engineering
Research Council (NSERC) of Canada for the financial support.

first steady-state
second steady-state

References
Barr, P. V., J. K. Brimacornbe and A. P. Watkinson, A Heat
Transfer Model for the Rotary . Kiln: Part I: Pilot Kiln Trials.,
Metall. Trans., 20B, 391-402( 1989a).
Barr, P. V., J. K. Brimacombe and A. P. Watkinson, A. P., A Hcat
Transfer Model for the Rotary Kiln: Part 11: Development of the
Cross-Section Model, Metall. Trans., 20B, 403-419( 1989b).
Friedman, S. J. and W. R. Marshall, Studies in Rotary Drying.
Part 11- Heat and Mass Transfer, Chem. Eng. Prog., 45(9),
573-588( 1949).
Fung, C. J. and D. A. Mitchell, Baffles Increase Performance of
Solid-state Fermentation in Rotating Drum Bioreactors,
Biotechnol. Tech., 9(4), 295-298( 1995).
Henein, H., J. K. Brimacombe and A. P. Watkinson, Experimental
Study of Transverse Bed Motion in Rotary Kilns, Metall.
Trans., 14B, 191-205( 1983).

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, APRIL, 1998

23 I

Hirosue, H., lnfluence of Particles Falling from Flights on

Volumetric Heat Transfer Coefficient in Rotary Dryers and
Coolers, Powder Technol., 59, 125-128( 1989).
Hirosue, H. and J. Shinohara, Volumetric Heat Transfer Coefficient
in Rotary Dryers and Coolers, Heat transfer-Jp. Res., 5(4), 59-66
(1 976).
Jauhari, R., Gassolid Mass Transfer in a Rotary Drum,
University
of Alberta, Edmonton, AB (1997), pp. 1-216.
Lebas, E., F. Hanrot, D. Ablitzer and J. L. Houzelot, Experimental
Study of Residence Time, Particle Movement and Bed Depth
Profile in Rotary Kilns, Can. J. Chem. Eng., 73, 173-1 79( 1995).
Lonsane, B. K., N. P. Ghildyal, S. Budiatman and S. V.
Ramakrishna, S. V., Engineering Aspects of Solid State
Fermentation, Enzyme Microb. Technol., 7,258-265( 1985).
Mu, J. and D. D. Perlmutter, A Simulation Study of Rotary Reactor
Performance, Chem. Eng. Commun., 9, 101-120 (1981).
Mu, J. and D.D. Perlmutter, The Mixing of Granular Solids in a
Rotary Cylinder, AIChE J., 26(6), 92&934( 1980a).
Mu, J. and D. D. Perlmutter, The Mixing of Granular Solids in a
Rotary Cylinder. Part 2. Experiment Results and Parameter
Estimation,Inst. Chem. Eng. Symp. Ser., 59, 5:7/1-5:7/20
(1 980b).
Petrovic, L. J. and G. Thodos, Mass Transfer in the Flow of Gases
through Packed Beds, Ind. Eng. Chem. Fundam., 7(2),
274-280 (1968).

232

Rao, S. J., S. K. Bhatia and D. V. Khakar, Axial Transport of

Granular Solids in Rotating Cylinders. Part 2: Experiments in a
Non-Flow system, Powder Technol., 67, 153-1 62( 199 1).
Stenhagen, E., S. Abrahamsson and F. W. McLafferty, Atlas of
Mass Spectral Data, vol. 2, Interscience Publishers, New
York(1969), pp. 780.
Tscheng, S.H.,
Convective Heat Transfer in a Rotary Kiln, Ph.D.
Thesis, University of British Columbia, Vancouver, BC (1 978),
pp. 226-232.
Tscheng, S. H. and A. P. Watkinson, Convective Heat Transfer in
a Rotary Kiln, Can. J. Chem. Eng., 57,433-443 (1979).
Wei, G. C., Beta S i c Powders Produced by Carbothermic
Reduction of Silica in a High-Temperature Rotary Furnace,
Commun. Am. Ceram. SOC.,66, 1 1 1-1 13 (1983).
Wilkins, G. S. and G. Thodos, Mass Transfer Driving Forces in
Packed and Fluidized Beds, AIChE J, 15(1), 47-50 (1969).
Yegneswaran, P. K., M. R. Gray and 8. G. Thompson, Kinetics
of CO, Hydration in Fermentors: pH and Pressure Effects,
Biotechnol. Bioeng., 36,92-96( 1990).

Manuscript received April 16, 1997; revised manuscript

received January 12, 1998; accepted for publication January 19,
1998.

THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING, VOLUME 76, APRIL, I998