Mixing Effects in a Simple Air Classifier
M. W. BIDDULPH
INTRODUCTION
The term air classifier is applied to a number of different
types of apparatus used in the field of waste recycling. The
common feature is the fact that the differing behavior of dif-
ferent species of solid particles in a moving stream of air is
used to effect a separation.
This paper follows earlier papers (Biddulph, 1983, 1984) con-
cerned with producing a mathematical model, based on dif-
fusion, of a simple vertical water elutriator. The intention here
is to apply the model to the much higher fluid velocity situation
generally found in air classifiers, with the objective of improv-
ing the design basis of these units.
Models of air classifiers have been developed and presented
by Senden and Tels (1976, 1978), Leschonski (1981), Fan (1975),
Vesilind et al. (1982), and Vesilind (1984). Although models
including the use of dispersion coefficients have been men-
tioned, no values of such coefficients appropriate to waste re-
cycling operations have been published.
The feedstock to an air classifier in a waste recycling plant,
for example comminuted municipal solid waste, is very com-
plicated in nature. There are many particles having irregular
shapes, often composed of different materials combined to-
gether. For this reason the design of air classifiers has been
largely based on trial-and-error methods. Bccause of the com-
plicated and largely unpredictable nature of the solid species
to be separated, it probably will not be possible to apply any
rigorous model involving trajectories of individual particles.
However, it may be possible to establish the basic operating
characteristics by using mixtures of particles of uniform and
predictable behavior.
Mathematical Model
The equations which result from the eddy diffusion model
have been presented previously (Biddulph, 1983). Equations
7-10 given there enable direct calculation of the rate of loss
of each species from the top and bottom of the column, assum-
ing that a value for the effective eddy diffusion coefficient is
known. The model assumes no interaction of solid particles,
and thus is likely to be a better representation of the behavior
at low particle concentrations, that is, at high aidsolids ratios.
During the earlier study on the water elutriator (Biddulph,
1983), it was observed that so-called wall flow of particles was
occurring. This refers to the tendency of the solid particles to
prefer the slower moving fluid near the wall rather than the
faster fluid near the axis of the tube. The order of fluid velocity
AlChE Journal (Vol. 32,No. 2)
Department of Chemical Engineerlng
Unlverslty of Nottlngham
Nottingham, England
used in the water elutriator was 0.3-0.6 mIs, whereas the air
velocities used in this study are in the range 10-14 m/s. As a
result of this large difference it would be anticipated that wall
flow would be less of a problem in the air classifier, and it will
be seen later that this is indeed the case.
Measures of Performance
There have been various ways proposed for defining the
separating performance of air classifiers (Worrell and Vesilind,
1979). An attempt has been made recently to standardize the
method of defining the efficiency (Vesilind, 1984).
Compositions and recovery fractions are important. If both
top and bottom products are to be optimized, the criterion
which has been used before (Biddulph 1983) has been the sum
of the loss fractions of the species (C). A zero value for C
represents perfect separation, while a value of unity represents
no separation. This will again be used here for comparison with
earlier studies. Other definitions are available (Vesilind, 1984).
In air classifier operations the top product composition is
often called the light fraction quality, and this, together with
the recovery fraction will also be used.
EXPERIMENTAL
A diagram of the equipment used in this investigation is shown in
Figure 1. The main vertical duct where separation occurs is constructed
from a transparent polymer sheet to make observation of the behavior
possible. The dimensions of the duct are 3.60 m by 30 cm by 15 cm.
The main centrifugal blower, driven by a 15 HP motor via a variable
speed gearbox, provides the air flow. The solids feed comes from a
hopper and a screw feeder driven by a 1/2 HP motor and a variable
speed gearbox.
RESULTS
Three different mixtures of solid particles were studied, all
being made u p of precision polymer spheres. The three binary
mixtures were made up from particles whose properties are
shown in Table 1. The top and bottom products were collected
and analyzed by sieving. The values of the terminal velocities
( U t ) in air were predicted using the standard equations, and
these agreed well with the air velocity required to just hold a
species steady in the column.
Three binary mixtures were studied: AIB, RIC, and CID.
Each mixture consisted of approximately equal numbers of
February, 1986 Page 317
 . 1.01 1

0 50 100
AlRlSOLlDS RATIO 4
Figure 2. Typical dependence of performance on
airlsolids ratio.
S
TOP PRODUCT light species to penetrate downward. It is reasonable that the
heavy species should require a diffusivity slightly lower than
the free stream value, since presumably only zero mass par-
ticles could exhibit a value equal to the free stream value in
this situation.

S
Y
d
COMPLETE
/ MIXING
'
BOTTOM PRODUCT LOWER

Figure 1. Experimental air classifier.

g 0.7-
particles of each species, and the runs were made at high air/
QI- 1 '(D~L = 1.27 DeFs
\ D s . 0 . 6 3 DeF5
solids ratios, Figure 2 (Vesilind and Rimer, 1981). 2 0.6 - I \
U
The experimental separation results from the three mixtures LL
studied are presented as light fraction quality, light species $ 0.5-
recovery fraction,. and the sum of the loss fractions of both
I? I '\
species (C). 8 0.4 -
I
Figure 3 shows a comparison between the experimental sep-
arations, expressed as C values, and the predictions from the
model for mixture AIB. Values of effective diffusivity for each
species were adjusted to obtain the best fit, these diffusivities
I
0.3- I
I
'\

+
\
\
being related to the appropriate free stream dispersion coef- 0.2 -
I 4.A-4 #' I
I
ficients based on the data of Levenspiel (1972).The slight skew
to the right shown by the curve indicates greater mixing of the
t i l
-
light species downwards than of the heavy species upwards, O.l 1two
bearing in mind that the feed point is above the middle of the 0 9.5 10.0 10.5 4
column, a fact that would give a curve skewed to the left if the
diffusivities were equal. Both effective diffusivities are quite AIR VELOCHITY (m/s)
close to the free stream value (Defs),the light species requiring
1.27 Defs while the heavy species requires 0.63 Defs.This is Figure 3. Comparison of experimental A and pre-
a similar but less pronounced effect to that observed in the dicted separation for mixture N B ; airlsolids ratio
water elutriator, and is due to the slight wall flow assisting the 124.

TABLE1. PARTICLE
PROPERTIES
Ut Density Dia.
Species Material Color mls kdm3 mm
A Nylon White 9.55 1,134 3.175
B Polypropylene, Green 10.50 875 4.763
C Polystyrene Red 12.30 987 6.00
D Polypropylene Green 13.25 875 7.94

Page 318 February, 1986 AlChE Journal (Vol. 32, No. 2)

 I!
n
I-
u)

I
P
1.o

0.9

0.8 -
-
I

I
/
,x 0--9
\

A\
\

/I '
-1
LL
0 0.7 - PREDICTION \
/
> OF QUALITY
U
I MODEL ,
0.6 i PREDICTION\+
W
>
0
0.6 - I
DeL=1.25 DeFS \
DeH:0.60 D q S \
I OFQUALITY \ 0 I \
DeL ~ 1 . 2 7DeFS\\
W
LI 0.5 -
' I \
\
U I DeH.0.63 DeFs \ 0: I
0
I
I f 0
> 0.4 I
\
\
I '\ t
-1
I \
\ MODEL PREDICTION 3
i
MODEL PREDICTION
\,
a
zz 0.3 -
I
I
I

/
OFRECOVERY
De,. 1.27 DeFS
o_ 0.2 -
I-
I
OFRECOVERY
DeL=1.25DeFS
0
a I De,*0.60 D e ~ s
0 De~zO.63DeFs
I e
I-
0.1 - I
I
L?g
-1 0 9.5 10.0 10.5 s
I
P
-1 0
I 10.0
I i
11.0
I 1
12.0
AIR VELOCITY ( m / d + AIR VELOCITY (m/s)+

Figure 4. Comparison of experimental A 0 and pre-
dicted quality and recovery for N B ; airlsolids ratio
124.
Figure 4 shows the comparisons of top product composition
(weight fraction) or light fraction quality, and the recovery
fraction of the light species as measured experimentally and
predicted by the model. It can be seen that quite a good match
is achieved. Figures 5 , 6, 7 and 8 show the equivalent com-
parisons for mixtures BIC and C D .
The absolute values of the effective difisivity deduced by
Figure 6. Comparison of experimental A 0 and pre-
dicted quality and recovery for B E ; air/solids ratio
120.
this method are shown in Table 2, each at optimum separation
conditions. In this range of air velocities, commonly used in
waste recycling operations, and under the conditions employed
in these experiments, the values of the effective diffusivities
remain fairly constant. With increasing velocity there may be
a tendency for the diffusivity relative to the free stream value
to decrease slightly.

0.9 -
+ I

I
1
I
I
I
1 .o

0.9 \ -
\
I
I
I I I
I 9I 0.8 - AI I
I I
I
I I
I 0.7-' \ II
v)
0 MODEL PREDICTIONS I 2 /

,
p M O D E L PREDICTION
I- 0.6
0
a
Y D e , = 1.25 DeFS
I DeH=0.60DeFs
II
I
0
F 0.6
0
- De,. 1.05 DeFs /
I
a De,=0.54 DeFS
I [r I
0.5-
0 \ AI
\ I
3 0.4 -
u)

LL
0
\
\
tiI'
5 0.3 - \ /
2
v) \
A //
0.2 - A\
h d A

I I
12.5 13.0
AIR VELOCITY ( m / d AIR VELOCITY ( m / s ) d

Figure 5. Comparison of experimental A and pre- Figure 7. Comparison of experimental A and pre-
dicted separation for mixture B/C; air/solids ratio dicted separation for mixture C/D; airkolids ratio
120. 130.

AlChE Joumal (Vol. 32,No. 2) February, 1986 Page 319

 The results obtained here apply to ideal mixtures under ideal
conditions of no particle interaction. Real mixtures contain par-
ticles with a range of sizes and shapes, and which are likely to
interfere with each other. The model can be used to study a
p 0.8
;-
J I mixture having a range of terminal velocities by dividing the
mixture into notional slices and summing the separation achieved
for each slice (Biddulph, 1984). It is better to operate air clas-
0.7 sifiers of this type at high air/solids ratios if possible, giving
better, and more predictable, separation. The estimated per-
0.6 formance at lower airkolids ratios might be obtainable by using
0 the model to effectively locate the curve of the type shown in
w
0.5
Figure 2.

’i
0
t
3
0 0.3
0.4

/ OF QUALITY
De,= 1.05 De,,
I
ACKNOWLEDGMENT

The author wishes to thank the Science and Engineering

Research Council for support for this study.
De, = 0.54 DeFS
U
a NOTATION

$ 1
-I
0 12.0
,
12.5
,
13.0
J C
Defs
= sum of the loss fractions of species
= free stream fluid diffusivity, m2/s
AIR VELOCITY ( m l s ) 4 Deh = effective eddy diffusivity for heavy species, m2/s
Del = effective eddy diffusivity for light species, mYs
Figure 8. Comparison of experimental A 0 and pre- Ut = terminal falling velocity in air, m/s
dicted quality and recovery for C/D; airkolids ratio
130.
LITERATURE CITED
TABLE2. EFFECTIVE
DIFFUSIVITY
VALUES
Biddulph, M. W., “Separating Etficiency of a Water Ehtriator,” AlChE
Air J., 29, 956 (1983).
Veloc. Del Deh , “Water Elutriators in Materials Recycling,” Can. 1. Chem.
Mixture mls m21s m21s Eng., 62, 357 (1984).
AIB 10.1 0.51 0.26 Fan, D. N., “On the Air Classified Light Fraction of Shredded Mu-
BIC 11.4 0.57 0.28 nicipal Solid Waste,” Resour. Recovery and Conseru., 1, 141 (1975).
CID 12.7 0.54 0.28 Leschonski, K . , “Recent Developments in the Theory and Practice of
Particle Classification,” Znst. Chem. Eng., Symp. Ser., 63, D2-MI
(1981).
Levenspiel, O., Chemical Reaction Engineering, 2nd. ed., Wiley, New
York (1972).
Senden, M. M. G., and M. Tels, “Some Principles of Air Classifica-
IMPLICATIONS FOR DESIGN tion,” Proc. 1st. Znt. Symp. Materials and Energy from Refuse, 109,
Antwerp (Oct. 21, 1976).
As in the case of the water elutriator, it is beneficial to locate , “Mathematical Model of Vertical Air Classifiers, Resour. Re-
the solids feed point noticeably above the middle of the column. cooery Consert;., 3, 129 (1978).
The exact location could be established by using the model Worrell, W. A., and P. A. Vesilind, “Evaluation of Air Classifier Per-
under the required conditions. This is a feature that has emerged formance,” Resour. Recooery and Consero., 4, 247 (1979).
by trial and error in many air classifier designs, but this model Vesilind, P. A., J. J. Peirce, and M. McNabb, “Predicting Particle
provides a means of locating the best feed point. Behavior in Air Classifiers,” Consertj. and Recyc.. 5, 209 (1982).
Vesilind, P. A., “Air Classification of Shredded Refuse,” Consero. and
The best total length of column can be found for a given set Recyc., in press (1984).
of conditions-the optimum length, which can be established
in the same way as before (Biddulph, 1984). Manuscript receited Mar 12, 1985.

Page 320 February, 1986 AlChE Journal (Vol. 32, No. 2)