Computer Simulation of a Rotary Dryer

Part I: Retention Time

A computer model of a single-pass, rotary-drum dryer with or without a center- F. A. KAMKE and J. 6. WILSON
fill flighting section, describes the drying behavior of wood particles within the
drum. This part of a two-part study examines retention time; the second part Department of Forest Products
incorporates heat and mass transfer for the complete drying model. Simulation College of Forestry
results with retention time data from a large-scale rotary dryer showed a favor- Oregon State University
able comparison, with a root mean square error of 14.2%. Corvallis, OR 97331

SCOPE
Rotary dryers can process a variety of materials, including board manufacturing process. Over the past few years, how-
wood particles, bark, grain, coal, fertilizer, sand, and other ever, these dryers have increasingly been put to use in conjunc-
aggregates. For these numerous applications, a wide assort- tion with exhaust stack gases and wood-fired boilers for drying
ment of dryer configurations has evolved. The drum may con- hogged-wood residue. In such applications the operation and
sist of a single- or multiple-pass arrangement, operate under design of rotary dryers has largely been an art.
cocurrent or countercurrent flow, be fired directly or indi- With the coming of computer process control, microproces-
rectly, and have various interior lifting-flight configurations. sors, and improvements in metering and sensing devices, there
In any case, the principle of operation is the same. The wet is opportunity for improving rotary dryer control strategies,
particles enter the rotating drum and are then continuously providing that the desired product and raw material character-
cascaded through the moving gas stream. Particles caught on istics can be linked with the necessary operating conditions.
lifting flights rotate with the drum. With cocurrent flow, a This paper comprises the first part of a two-part study di-
particle falling from a flight is moved along the length of the rected toward developing a computer model of the drying be-
drum as a result of a drag force imparted by the gas stream. havior of wood particles in a rotary dryer. It analyzes those
With countercurrent flow, kiln action and a sloped drum work factors affecting the retention time of particles within a rotary
against the gas flow to move the particles through the dryer. drum, providing information on particle location and behavior
Drying occurs primarily by convective heat and mass transfer. with the gas stream. Part I1 reports on simultaneous heat and
Rotary dryer performance is heavily influenced by the rate mass transfer for a complete drying analysis. The model is
of heat transfer from gas to particles and by the retention time based on first principles and empirical correlations, which
of particles in the drum. The dryingcharacteristicsof the solids were derived independently of rotary dryers. In addition, we
often dictate the required retention time in the drum; there- conducted a series of experiments measuring retention time
fore, an analysis of heat transfer and factors affecting retention and drying characteristics in a large-scale rotary dryer, and
time is important for designing rotary dryers. compared the results to predictions from the computer model.
In the forest products industry, rotary dryers have been used Retention time measurements were made by means ofa radio-
almost exclusively for drying wood particles in the particle- active tracer technique.

CONCLUSIONS AND SIGNIFICANCE

Comparison of measured retention time with the results in the gas stream. Measured retention times of individual par-
from the computer simulation yielded a root mean square ticles ranging in screen size from 1.53 to 5.14 mm showed no
error of 14.2%, averaged over all test runs. The simulation significant difference, indicating that cascading particles do
results compared most favorably when the characteristic parti- not behave independently but are influenced by bulk particle
cle dimension was represented by the mean size of all particles flow. Within the size range examined, interaction between
rather than discrete size of individual particles. particles completely masked effects of individual particle size.
The simulation predicted a much larger effect of individual Computer simulations predicted that retention time de-
particle size than was measured experimentally, a discrepancy creases nonlinearly with an increase in gas velocity, while at
due to the assumption that the particles behave independently higher velocities this effect diminishes. Retention time de-
creases nonlinearly with increasing drum speed. Increasing
the drum diameter caused a nearly proportional decline in the
Correspondence concerning this paper should be addressed to F. A. M e , Depart-
retention time. Results of the retention time simulation may
ment of Forest Products,Virginia Polytechnic Institute and State University,Blacksburg. be used directly for modeling heat and mass transfer character-
VA 24061. istics in a rotary dryer.

AlChE Journal (Vol. 32, No. 2) February, 1986 Page 263

BACKGROUND operating conditions. All work reported thus far has dealt only
with a single-pass, open-center drum. Experimental work has
Retention time in rotary dryers is dependent on the particle been limited to small-scale laboratory drums.
flow path, which consists of flow in a gas-particle stream and This study considers single-pass drums with an allowance for
travel on lifting flights. The arrangement and shape of the lift- centerfill flights, as shown on the right in Figure 1. A computer
ing flights influences the flow path. Two possible flighting sys- program was developed to predict the average retention time
tems for a single-pass rotary dryer are shown in Figure 1. The at any point along the drum length. The model does not require
broken lines indicate possible paths for a particle along the empirical constants or correlations derived from experiments
length of the drum. Other factors that influence retention time with rotary dryers. We compared data collected from a large-
are gas flow rate, particle feed rate, particle characteristics, scale rotary dryer to simulation results from the computer pro-
drum diameter, drum length, drum slope, and rate of drum gram.
rotation.
Early research on retention time concentrated on developing
empirical correlations from experiments with small-scale ro- MODEL DEVELOPMENT
tary drums. The work of Friedman and Marshall (1949) and
their summary of the work of Prutton et al. (1942) and Smith Incorporated into the model of rotary drum retention time
(1942) on retention time and holdup in rotary drums is the are these assumptions:
broadest data base available, covering avariety of materials and 1. Particles fall independently of one another with no parti-
operating conditions. Friedman and Marshall’s correlation for cle-to-particle interaction.
retention time in a single-pass, open-center drum, with cocur- 2. Irregularly shaped wood particles may be approximated
rent or countercurrent flow, is the most often cited empirical as spheres for purposes of analyzing the gas and particle inter-
approach (Perry and Chilton, 1975): action.
3. The particle lifting flights are rectangular in cross section
13.8 L LG
t= + 0.59 - or may be approximated as such (Figure 2).
-
(tan a)NO.gDd SD:.5 4. The drum is single-pass, with or without centerfill flights.
Saeman and Mitchell (1954) used a graphical analysis of the A force balance equation for a spherical particle falling in
drum and lifting flight cross section to examine particle trajec- crossflow to a moving gas stream was presented by Schofield
tories and thus to predict retention time distributions. Miskell and Glikin (1962) as:
andMarshall (1956) studiedretention time in adrum of0.14 m
diameter with radioactive tracers. Results showed an optimum
holdup condition at which deviation from the average retention
time was minimized. From left to right, the terms represent inertial force, force due
Theoretical retention time models have been developed by to gravity, and drag force on a particle, all in the longitudinal
Schofield and Glikin (1962), Kelly and O’Donnell (1968, direction. This relation may be integrated twice to yield the
1977), Glikin (1978), Kisakiirek (1982), and Platin et al. longitudinal advance of a particle per cascade:
(1982). The concept of a cascade cycle was incorporated into
all of these models, in which a fluid dynamic analysis of the
gas-particle interaction was critical. Glikin’s model predicts re-
tention time in a single-pass, open-center drum as follows: In analyzing rotary dryers, the drag coefficient, C,, has been
t= Le

ij [sin a
[f+(yq (2)
estimated under assumption 1 using Eq. 5 (Glikin, 1978; Kelly
and O’Donnell, 1968, 1977; Thorne, 1979), which was origi-
nally developed by Schiller and Naumann (1933) for spherical
bodies in motion under a gravitational force:
The drag factor, J , and the relative particle velocity, u,, are FD = 3zDpu,p(l + 0.15 Re0.687) (5)
calculated with empirical correlations relating particle charac-
teristics to gas velocity, and with the assumption that particles Combining the last term in Eq. 3 with Eq. 5, the drag coefficient
behave independently when they fall through the gas stream. may be evaluated as:
The effective drum length, L,, is derived through experiment 24
with rotary drums. Kelly and O’Donnell’s model is similar, but C D= - (1
Re
+ 0.15 Re0.887) (6)
includes the effects of kiln action and particle bouncing.
Because of the diversity in rotary dryer design, it is difficult Since the drag coefficient is a function of the relative particle
for one model to handle all possible dryer configurations and velocity, Eq. 4 and 6 must be solved iteratively.
In a rotating drum with both outer and centerfill flights, the

-Ah-
Q = Flight ongle to horizontal
Figure 1. Rotary-drumcross sections showing possible parti- +=Kinetic angle of repose
cle cascade patterns in an open-centerdrum (left) without and
(right) with centerfill flights. Figure 2. Cross section view of particles in a rectangular flight.

Page 264 February, 1986 AlChE Journal (Vol. 32,No. 2)

time of travel on lifting flights consists of that on both outer and
centerfill flights (between points D and A and points B and C in
Figure 1).Likewise, the time of fall through the gas stream
consists of fall from both outer and centerfill flights (between The procedure for calculating the total retention time and
points A and B and points C and D in Figure 1).Points A, B, C, drum holdup is repeated, and successive iterations are per-
and D represent a possible cascade in the travel of a particle. formed until convergence is achieved.
The average positions of these points may be defined by a tech- It is common practice to approximate the fluid dynamicprop-
nique presented by Glikin (1978)for outer flights and modified erties of irregularly shaped particles to another shape, such as a
by Kamke (1984) for centertill flights. For point A, the average sphere, and to calculate an equivalent particle diameter (Toro-
outer flight angle from which a particle falls is given as: bin and Gauvin, 1960; Heywood, 1962; Mason, 1980). A
method proposed by Levenspiel (1984), which is based on a
sieve screen analysis, is given by:
(7)
D p = a *D, (14)
For point C, the average centerfill flight angle from which a The angle of repose (Figure 2) for particles being carried in
particle falls is given as: lifting flights is simply the angle of the particle bed surface
relative to the horizontal. When the particle bed is in motion,
this is known as the kinetic angle of repose.
Kamke (1984) used a photographic technique to study the
To integrate Eqs. 7 and 8, a relationship is needed between the kinetic angle of repose for sawdust, examining the effects of
flight angle and the flight holdup. This is derived from a knowl- particle moisture content, drum speed, and drum diameter.
edge of the flight geometry and the characteristic kinetic angle The kinetic angle of repose was found to be independent of the
of repose of the particles. Glikin (1978) gives this relationship Froude number within the range of 200 < Fr < 100,000.
for the outer rectangular flight shown in Figure 2. A similar There was a slight dependence on the angle ofthe flight relative
relationship has been developed for centerfill flights (Kamke, to the horizontal and a significant dependence on moisture
1984). content in the range of 0.10 < X < 1.46. The mean kinetic
The average angle of entry onto the centerfill flights, ve, and angle of repose was approximately 82.6 degrees, which is
somewhat higher than the 71 degrees previously reported by
the average angle of entry onto the outer flights, Oe, are repre-
sented by points B and D,respectively, in Figure 1. If a com- Friedman and Marshall (1949) for sawdust of unknown mois-
pletely vertical fall is assumed, these angles, and the distance of ture content.
particle fall, y, may be determined by plane geometry.
The time of fall, tr, is then approximated as: EXPERIMENTAL
+
t f = (2gy)o.5 (2gyp.5 (9)
A commercially designed and manufactured rotary dryer was used in
and the time of travel on the flights is given by: the experimental portion ofthis work; the drum is 1 . 2 m in diameter and
+ +
t = [(360 8 - a=) (F- F,)]/(360 N) (10)
5.5 m in length, with six centerfill and twelve outer flights (Figure 1,
right). Wood particles entering the drum through a screw-feed con-
The total retention time is calculated by Eq.11,in which the veyor come in contact with the conveying gas at the entrance in acocur-
number of cascades, C, is determined by dividing the length rent flow (Figure 3). Wood particles and gas exit the drum into a drop-
of the drum by the longitudinal advance per cascade, x , from out hopper, at which point approximately 90% of the particles are
collected. The remaining fines and exhaust gas travel to the cyclone
Eq. 4.
t T = C(t + tf) (11)
separation system, where most of the fines are removed. Other minute
particulates and the exhaust gas pass through an induction fan and out
The analysis has so far assumed that the drum is fully loaded. the stack. For the retention time experiment, the drum was operated at
This means that at 8 = 0 the flight is filled to capacity and cas- ambient conditions with wood particles that were previously dried.
cading begins. Overloading the drum, such that a portion of the Retention time was measured by a radioactive tracer technique in
holdup bypasses the lifting flights, is undesirable because mate- which representative test particles were tagged with a radioactive iso-
tope. The signals from two NaI gamma ray detectors inside the drop-out
rial accumulating on the bottom of the drum will not participate hopper were individually processed through a separate preamplifier
in continual cascading. Particle feed rates are often limited by and amplifier circuit (Figure 3). The two signals were then joined and
burner capacity and drying rates. Thcefore overloading is not routed through a single rate meter, and the output was transmitted to a
likely to occur under normal operating conditions. strip-chart recorder. A remote switch at the particle inlet manually
For underloading of the drum, which is the most likely case, controlled the strip-chart recorder and initiated the starting time for
cascading begins at some peripheral flight angle greater than each run. Air flow was monitored by a venturi in the exhaust gas stream,
zero; therefore Eq. 7 must be modified. An iterative procedure and the particle feed rate was continuously metered through a conveyor
is proposed. First, retention time and holdup are calculated as belt balance in front of the inlet screw-feed conveyor.
The wood particles were commercially prepared. Their size distribu-
we have outlined. The calculated drum holdup is then com-
tion, with acalculatedmeansphericityof0.75, isshown inFigure4. The
pared to the known design holdup for the drum and the mate- relative frequency is the weight fraction divided by the incremental
rial being processed. A fractional drum holdup is determined as screen opening. Representative test particles were selected from three
m = H/H". If the fractional drum holdup is less than one, an size classes defined by a screen analysis; total dry weight was 1.2 g per
iteration is required. If drum holdup is assumed to be propor- size class. Enough particles for six test runs were prepared.
tional to flight holdup, a new value for the flight holdup when Test particles were tagged with a predetermined amount of an
cascading begins may be calculated as: aqueous solution of Na*', a neutron-rich isotope of NaS3.This nuclide
was selected because of its relatively energetic gamma rays at 1.37 and
h(8,)= mh"(0) (12) 2.75 MeV per disintegration and because the test site location and
travel time were in keeping with the 15 h half-life.
This value is then compared to successive values of h"(8)as 8 is Before the test runs, the tagged particles and the bulk particles to be
increased until h(8,)just exceeds h"(8),at which point the pe- fed into the drum were dried to approximate equilibrium in ambient
ripheral flight angle when cascading begins will be identified. conditions. At the start ofeach test run, the tagged particles were simul-
Equation 7 then becomes: taneously injected through an access port at the particle inlet immedi-

AlChE Journal (Vol. 32, No. 2) February, 1986 Page 265

 Exhaust
fines to
,tides

1 Porticles

PreamPli-
fier
J .
Ampl,fler
- Rate
meter

Scaler
- Preampti-,
fier Amplifier

Strip-
chart
I recorder I

Figure 3. Experimental setup for measuring retention time by means of a radioactive tracer detection system.

ately ahead of the rotating drum. Tagged particles were mixed with the The effect of drum speed is readily apparent in Figure 6.
bulk particle feed just before entering the drum. Increasing the speed decreased the average retention time. Of
particular interest is the effect of particle size, which the com-
puter model predicted to be more significant than was shown
RESULTS by the data. The model assumes the particles act indepen-
dently; experimentation showed that this was not the case. Test
Frequency histograms were prepared from the output ofthe particles ofdiscrete size were injectedinto the rotary drum and
strip-chart recorder (Figure 5 ) , and means and standard devia- mixed immediately with the bulk particle flow. During a cas-
tions were plotted with mean retention time predictions from cade, the particles fall in curtains, separated by relatively parti-
the computer program (Figure 6). cle-free areas. Within a curtain, particle contact and shielding

0.4
t
0
2
w
03
[L
LL
W
>
-
0.2
-I
w
[L

0.1

U
1.0 2 .o 3.0 4.0 5.0 6.0 7.0 8.0
SCREEN OPENING (mm)

Figure 4. Wood particle size distributionused in retention time study. Relativefrequency = weight fraction + incremental screen
opening (Sweco).

Page 266 February, 1986 AlChE Journal (Vol. 32, No. 2)

 7.2 rpm 3.0rprn of the gas flow by other particles can affect flow. The denser the
curtain, the more interaction. As a result, individual particle
PARTICLE SIZE(rnrn4
flow is influenced by bulk particle flow.
If particles behave as a group in gas-particle interaction, a
1.144
representative particle dimension on which to base a drag coef-
ficient calculation would be mean particle size (Figure 4). We
plotted retention time with the mean particle size (Figure 6)
against measured retention time averaged over each drum ro-
tation rate. The combined root mean square error for all test
runs based on the discrete test particle size was 109.6%;based
on the mean particle size, it was 14.2%.
Kelly and O’Donnell (1977) recorded a root mean square
0 5 10 15 20 25 30 error of 23.2%for their predictions of retention time ofpumice
particles in a drum 0.31 m in diameter with no centerfill. Their
retention time data were taken in a test section of the drum that
was less than 0.5 m long. Both the test particles and the bulk
particles were a discrete size. Kelly and O’Donnell’s study in-
corporated the effects of kiln motion and particle bouncing,
both of which result from a sloped drum. The rotary drum of
our study was not sloped.
The accuracy of Kelly and O’Donnell’s model declined as air
velocity increased. At 1.61 m/s the root mean square error was
34.1%.The air velocity in our study was approximately the
same at 1.58 m/s; the root mean square error was 14.2%.
We conducted a series of computer simulation trials to exam-
ine the effects of various process parameters on retention time.
Figure 7 (left) shows a plot of retention time as a function of gas
velocity. The three lines represent different drum holdups as a
fraction of the design (fully loaded) drum holdup. As expected,
retention time decreases with increasing gas velocity; the effect
diminishes at the higher gas velocities examined. A plot of re-

k
tention time as a function of drum rotation rate showed that
increasing the speed decreases the retention time (Figure 7,
center). The effect appears to be nonlinear over the entire
range of drum speeds examined. Increasing the drum diameter
0 5 10 15 20 25 30 decreases the retention time (Figure 7, right) because the
longer distance of particle fall with the larger diameters allows
a longer time for gas and particle interaction.
RETENTION TIME ( r n i n )
In all instances, the fractional drum holdup was inversely
Figure 5. Measured retentiontime distributionswith gas veloc- related to retention time. Fractional drum holdup was varied
ity 1.58 m/s and feed rate 0.334 dry kg/s. by changing the particle feed rate. Holdup, feed rate, and re-
tention time are related as follows:

2500
c Changing the particle feed rate resulted in a less than propor-
tional change in the drum holdup, which indicates that reten-
5 2000
I- tion time is more sensitive to such changes than is drum holdup.
z This result is confirmed by the data of Saeman and Mitchell
b
w
1500
(1954) and Kelly and O’Donnell (1977).
0:

ACKNOWLEDGMENT

The authors are pleased to acknowledge the interest and

I L I support of the Forest Research Laboratory and Radiation
0 500 I000 1500 2000 2500 3300
Center at Oregon State University and the Weyerhaeuser
MEASURED RETENTION TIME ( 5 ) Company.
PARTICLE SlZE(mml DRUM SPEED
3 0 r p m 7.2rpm
-1813 + I 5 3 0 . NOTATION
-335 + I 8 8 A A
-514 +335 o m
MEAN 206 0 .

Figure 6. Predictedvs. measured retentiontime for wood parti- C = number of cascades

cles in the rotary drum with gas velocity 1.58 m/s and feed rate c, = drag coefficient
0.334 dry kg/s. D = diameter, m

AlChE Journal (Vol. 32, No. 2) February, 1986 Page 267

 1000

1 I 1 1 1
0

0,= mean aperture size of two screens defining a particle Superscripts

size, m
FD = drag force, N * = design condition
Fr = Froude number - = average
g = acceleration due to gravity, m/se
G = dry gas mass flow rate, kg/s
h = flight holdup, m3 LITERATURE CITED
H = drum holdup, m3
J = drag factor Friedman, S. J., and W. K. Marshall, “Studies in Rotary Drying. I:
Holdup and Dusting,” Chem. Eng. Prog., 45, 482 (1949).
Glikin, P. G., “Transport of Solids Through Flighted Rotating Drums,”
Trans. Inst. Chem. Eng., 56, 120 (1978).
L = drum length, m Heywood, H., “Uniform and Nonuniform Motion of Particles in
L, = effective drum length, m Fluids,” Proc. Symp. Interaction between Fluids and Particles. ICE
m = ratio of actual holdup to design holdup Symp. Ser., 9 (1962).
N = drum rotation rate, rev/min Kamke, F. A., “Engineering Analysis of aRotary Dryer: Drying ofWood
Re = Reynolds number Particles,” Ph.D. Thesis, Dept. Forest Prod., Oregon State Univ.,
S = dry solids feed rate, kg/s Corvdlis (1984).
t = time, s Kelly, J. J., and J. P. O’Donnell, “Dynamics of Granular Material in
Rotary Dryers and Coolers,” ICE Symp. Ser., 29, 33 (1968).
u = velocity, m/s
-, “Residence Time Model for Rotary Drums,’’ Trans. Inst. Chem.
u, = uc - u, = relative particle velocity, m/s Eng., 55, 243 (1977).
x = longitudinal advance of a particle per cascade, m Kisakurek, B., “Retention Time in a Rotary Dryer,” Proc. 3rd Int. Dry-
X = moisture content of the wood, dry basis ing Symp., 2 (Univ. Birmingham), J. C. Ashworth, ed., Drying Re-
y = vertical distance of particle fall, m search Ltd., Wolverhampton, England (1982).
Levenspiel, O., Engineering Flow and Heat Exchange, Plenum Press,
Greek Letters New York, 120 (1984).
Mason, M. A., “Drag on Freely-falling Wood Chips and Other Irregu-
a = drum slope to horizontal, degrees larly Shaped Bodies,” M. S. Thesis, Univ. Wisconsin, Madison (1980).
6 = peripheral flight angle, degrees Miskell, F., and W. R.Marshall, “A Study ofRetention Time in aRotary
0, = peripheral flight angle of entry, degrees Dryer,” Chem. Eng. Prog. 52(1), 35 (1956).
6, = peripheral flight angle when cascading begins, degrees Perry, R. H., andC. H. Chilton, ChemicalEngineers’Handbook, Sthed.,
p = viscosity, Pa * s McGraw-Hill, New York, Ch. 20, 34-35 (1975).
n = 3.1416 Platin, B. E., A. Erden, and 0. L. Gulder, “Modelling and Design of
p = density, kg/m3 Rotary Dryers,” Proc. 3rd Int. Drying Symp., 2 (Univ. Birmingham), J.
u = sphericity C. Ashworth, ed., Drying Research Ltd., Wolverhampton, England
4 = kinetic angle of repose, degrees (1982).
Prutton, C. F., C. 0. Miller, and W. H. Schuette, “Factors Influencing
Y = centerfill flight angle, degrees the Performance ofRotary Dryers,” Trans. AIChE, 38, 123 (1942).
Ye= centerfill flight angle of entry, degrees Saeman, W. C., and T. R. Mitchell, “Analysis of Rotary Dryer and
Y, = centerfill flight angle when cascading begins, degrees Cooler Performance,” Chem. Eng. Prog. 50(9), 467 (1954).
Schiller, L., and A. Naumann, “Concerning the Fundamental Calcula-
Subscripts tion of Gravitational Effects,” Ver. Dtsch. Ing. 77, 12, 318 (1933).
B = bulk Schofield, F.R.,and P. G. Glikin, “Rotary Dryers and Coolers for Gran-
ular Fertilizers,” Trans. Inst. C h a . Eng., 40,183 (1962).
c = centerfill flight
Smith, B. A., “Factors Influencing Rotary dryer Performance,” Trans.
d = drum AZChE, 38,251 (1942).
e = peripheral flight Thorne, B.,“The Computer Simulation of the Rotary Drying Process,”
f= particle fall Ph.D. Thesis, Univ. Coll., Dublin, Ireland (1979).
G = gas Torobin, L. B., and W. H.Gauvin, “Fundamental Aspects ofSolids-Gas
p = particle or solids Flow,” Cun.]. Chem. Eng., 38(5), 142 (1960).
T = total
x = directional coordinate Manuscrfpt received July 27.1984, and revision received Apr. 24,1085.

Page 268 February, 1986 AlChE Journal (Vol. 32, No. 2)