United States
Environmental Protection
Agency
Municipal Environmental Research
Laboratory
Cincinnati OH 45268
Research and Development
EPA-600/S2-81-221   Dec. 1981
Project Summary
Comparative  Study of
Air  Classifiers
G. M. Savage, L. F. Diaz, G. J. Trezek, V. Hopkins, B. Slmister, D. Fiscus, S. C.
James, and D. Brunner
  This paper presents the results of an
extensive air classifier  field test
program conducted for the U.S.
Environmental Protection  Agency
(EPA). Methods of testing, criteria for
evaluation, operating conditions,  and
assessment of air classifier perform-
ance are described. Topics germane to
the design and operation of air classi-
fiers are also covered. Test results
show that air classifiers performances
can  be compared under equitable
conditions. The comparisons presented
here permit judgements  about  the
relative performances of air classifiers
and the tradeoffs that must be con-
sidered when describing air classifier
performance.
  This Project Summary was devel-
oped by EPA's Municipal Environmen-
tal Research  Laboratory, Cincinnati,
OH, to announce key findings of the
research  project  that is fully docu-
mented in a  separate report of the
same title (see Project Report ordering
information at back).

Introduction
  Seven air classification systems with
nominal throughputs ranging from  4 to
91 Mg/hr have been field  tested  and
evaluated under a 2-year program
sponsored by  the U.S. Environmental
Protection Agency (EPA).1 The purpose
of the testing  program was to charac-
terize and compare the operation  and
performance of air classifiers located in
the field. During the course of the work,
characterization criteria were developed
that enabled the comparison  of all air
classifiers  on an equivalent basis.
Because the best air classifier perform-
ance existed only for sufficiently dilute
air/solids mixtures within the air
classifier column, the air/solids ratio
was chosen as the means for establish-
ing the  performance parameters. For
each set of operating conditions, the
performance parameters varied rela-
tively little when the air/solids ratio was
greater than a critical value, that value
being defined  by the point where the
constant light fraction split fell off by 1
percent.  (Constant light fraction split is
defined as that value of the light fraction
split that does not change at a given air
flow and is independent of the air/solids
ratio.) The unvarying nature of material
characteristics above the critical air/
solids ratio has been documented
previously.2
  The seven air classifiers tested in this
study, their locations, and their general
descriptions are given in  Table 1.
(Further details concerning the geomet-
rical configuration of the air classifiers
are available in the final report to EPA.)
  An examination of  some of the key
characteristics of the  solid waste
encountered during the air classifier
testing program shows the importance
of normalizing the performance param-
eters in terms of the air classifier infeed
composition. Wide variations exist in
the waste characteristics from site to
site. For  example, the paper and plastic
content of the infeed to the Los Angeles
air classifier  average 30.2 percent,
whereas it exceeded 50 percent at the
other six air classifiers.

-------
Table 1.    Summary of Air Classifiers Tested
Site
Tacoma, WA
Baltimore Co., MD
Richmond, CA
Ames, IA
Los Angeles, CA
Akron, OH
Pompano Beach, FL
Type of
Air Classifier
Horizontal
Vertical
Vertical
Vertical
Vertical
Vibratory,
inclined
Vibratory,
inclined
Design
Throughput
(Mg/hrJ
73
91
4
45
4
64
6
Cross
Sectional Shape
of Column
Rectangular
Circular
Rectangular
Rectangular
Rectangular
Rectangular
Rectangular
  Two of  the  air classifiers were fed
shredded  and screened solid waste
(Ames and Pompano Beach), and the
other  five air  classifiers received
processed waste that had been shredded
only.  Consequently, normalizing the
performance parameters in terms of the
infeed composition was essential to
allow  comparisons among the air
classifiers. Note that the ash content of
the screened air classifier feed material
at Ames and Pompano  Beach ranged
from 13 to 14 percent, whereas it was
31  to 40 percent at the five sites not
employing screening before  air classifi-
cation.
  Abbreviations for terms specific to air
classifier performance  and  operation
are listed in Table 2.
Table 2.
Air Classification
Abbreviations
Abbreviation
A/S ratio
LF
HF
Fe
NonFe
PP
Term
Air /sol ids ratio
Light fraction
Heavy fraction
Ferrous metals
Nonferrous metals
Paper and plastic
Test Procedures and Methods
of Analysis
  For  each  air classifier, the test
program consisted of collecting samples
of heavy  and light fractions under a
specified  set of operating  conditions.
Samples were collected at three discrete
air  flow settings (high, medium, and
low) over as  wide  a  range  of material
flowrates as  was  possible  within the
physical constraints  imposed by the
facility or the air classifier system. One
of the three air flow settings was the
operating mode used by the plant under
normal  processing conditions. The
reason for selecting three air flows and
a full range of throughputs is that  no
generally accepted definition exists  for
an  optimum air  classifier operating
condition or setpoint in the industry. In
addition, manufacturers' data or recom-
mendations  as  to an air classifier
setpoint are not typically available.
  Five samples of heavy and  light
fractions were collected simultaneously
for each  air flow  setting. Before each
actual test run, a  dry run was used to
determine the time intervals needed to
collect simultaneous samples of heavies
and lights. Particle size of the heavy and
light fractions was  used to determine
the quantity of material to be used  for
the size  distribution and composition
analyses. Size distribution and composi-
tion analyses were performed with
subsamples from  the heavy and light
samples collected for mass flowrate
measurements. The latter samples
generally  ranged  from  10 to 50  kg.
Samples for  the size distribution and
composition analyses ranged from 2 to
10kg.
  The  general procedure involved
setting a particular airflow through the
classifier, by varying the rpm of the air
classifier fans or  adjusting a damper.
Subsequently, samples of  heavy and
light fractions were collected simul-
taneously for a  number of different
material feed rates to the air classifier.
Samples of heavy and light fractions
were collected from conveyors down-
stream of the classifier. These samples
served the dual purpose  of  allowing
calculation of the flowrate of the heavy
and  light  fractions and providing the
material from  which  representative
samples were chosen for later labora-
tory  analyses. All  material was com-
pletely removed from a given length ol
conveyor belting, thus alleviating the
problem of stratification of components
in the flow stream (which might have
skewed the results if only grab samples
had been collected).
  Laboratory  analysis consisted of air
drying, screening, and manually sorting
the light and heavy  fractions—except
for the Ames air classifier testing,  in
which  it was only possible to collect
infeed and heavy samples. In addition,
heating value determinations and ash
analyses were carried out on light and
heavy fractions collected at each site
following the procedures proposed  by
the ASTM for refuse  derived fuel No. 3
(RDF-3). Both manual and mechanical
screening were used to determine the
size  distribution of the samples.  Size
distribution analyses were carried out to
determine the amount of fines (minus
14 mesh) in  the heavy and  light
fractions and also to provide a quanti-
tative means  of describing the particle
size  of the shredded  air classifier feed
material. Manual sorting of the samples
consisted  of separating ferrous, papen
and plastic, and nonferrous components
from the  heavy and light fractions.
Composition and size data were used
subsequently to develop the character-
ization parameters.
  Also as part of the test program, both
air flow and system power requirements
were measured for each operating
setpoint (i.e. air flow  setting).

Results and  Discussion
   Operating parameters determined for
each air classifier are summarized in
Table 3. Data are shown for each of the
three air flow settings used during the
test program at each site. The volumetric
air flows corresponding to the high,
medium, and low air-flow settings are
reported in the final report.
   Critical  throughputs  are reported in
Table 3 instead of the maximum tested
feedrates because the air classifiers
were operating in a choked condition for
feedrates greater than  the  critical
throughput value.  Consequently,  they
were not exhibiting their best perform-
ance for each air flow setting.
  The air classifier performance param-
eters were calculated from  data for
which the air/solids ratios exceeded the
critical value. For  reasons previously

-------
 Table 3.    Summary of Operating Parameters

Tacoma


Baltimore


Richmond


Ames


Los Angeles


Akron


Pompano Beach


Air Flow
Setting
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
Critical
Specific Energy
KWh/mg (KWh/t)
460
372
36.2
98
5.9
3.9
8.8
4.9
2.0
>186
>176
>27.4
8.8
78
49
167
21 6
20.6
118
88
88
2/4.7;
(3.8)
(37)
(1.0)
(06)
104)
10.9)
(0.5)
(02)
T31.9)
S/1.8)
2J28I
(0.9)
(0.8)
(0.5)
(1.7)
(2.2)
(21)
(12)
(09)
(0.9)
Critical
A/S Ratio
<16
1 7
1.7
3.5
55
75
57
6.8
as
£2.6
£2.5
120
99
95
36
21
13
33
1 7
0.9
S210
3205
237.8
3.2
2.8
2.0
19.9
26.7
249
44
3.3
3.4
2732
109
104
40
23
14
36
19
10
323 /
3225
335.0
35
3.1
20
21.9
294
27.4
48
36
3.7
 ' ( = ton, tph = ton per hour
 b Air velocity in the vicinity of the heavy fraction discharge point of the Triple S air classifiers
 " NA- not applicable.
 a Re-adfustment of the air bleed of the air classifier caused an increase in the column velocity despite a decrease in the air flow control setting
 discussed, the normalized  values  are
 the significant parameters.  Because of
 the differing characteristics of the air
 classifier infeed material among  the
 /arious sites, the absolute values for the
 performance  parameters are only of
 value  when they  are  used to judge
 individual air  classifier  performance
 and not to  compare air classifiers
 among different sites. In addition,  the
 use of the absolute values for evaluating
 air classifier performance at a particular
 site presupposes that the waste compo-
 sition does not vary for the duration of
 the tests—a questionable assumption.
   Many comparisons and conclusions
 can be drawn from the  data. For
 example, the data  show that approxi-
 mately 70 to 99 percent of the  input
 energy content can be recovered in the
 light fraction. In addition, for most air
 classifiers near their normal operating
 points,  recovered energy ranges from
 90 to 99 percent. As a second example,
 the percentage of paper  and plastic in
 the heavy fraction  ranged from 0.8 to
 42.8 percent, excluding  the Pompano
 Beach data, which was skewed because
 of the significant amount of preprocess-
 ing of the waste before air classification.
  The light fractions obtained in Ames
and Pompano  Beach have relatively
high heating values. This result maybe
due to the composition of the raw waste
and/or the effect of screening before air
   ssif ication. The light fraction obtained
 in Los Angeles also has a relatively high
 heating value, but it cannot be attributed
 to screening, since  that unit process
 was not employed at the site. Instead it
 is partly  a  consequence of the  air
 classifier  feed at  Los Angeles  which
 also has a relatively high heating value.
  To establish criteria for air classifier
 evaluation, two key ratios can be used:
 recovered  paper and plastic/retained
 fines, and recovered energy/retained
 ash. The  first ratio (recovered  PP/re-
 tained fines) could be used to determine
 optimum air  classifier performance for
 selectively separating paper and plastic
 components  from  fine  inorganic con-
 taminants (for  instance, when  fiber
 recovery  is the intent  of a resource
 recovery process).  On the other hand,
 for recovery of a refuse derived fuel, the
 ratio of  recovered  energy  (RE) to
 retained ash  (RA) is significant and has
 been chosen  for use in this study as the
 key  criterion  for establishing the per-
formance  characterization of the seven
 air classifiers that were  tested.
  The selection  of the RE/RA ratio as
the key criterion is based on the fact that
 both recovered energy and retained ash
 in the light fraction are normalized on an
 infeed  basis, thus eliminating  the
effects of  variation in refuse composi-
tion. Large values of this  ratio imply that
an air classifier is recovering in the light
fraction a  significant percentage of the
energy available in the infeed material
and simultaneously dropping out into
the  heavy fraction  a significant per-
centage of the ash-carrying components.
  If the maximum values of RE/RA for
each air  classifier are  chosen as the
criterion (or  premise) for determining
the optimum operating point for recovery
of a  high-quality  RDF, then other
performance  parameters  can also be
chosen for comparison based on the air
flow setting that results in the maximum
value of RE/RA for each air classifier.
  A comprehensive assessment can be
made of the various air classifiers under
their respective optimum air setting for
RDF recovery (Table 4). Included are the
judging criteria and the key parameters
used to quantify each judging criteria.
The  first  three parameters (specific
energy, column loading,  and RE/RA)
were determined from field measure-
ments. The last two parameters (system
complexity and  1980 costs) are of a
qualitative nature.  The judgment of
system complexity was based on visual
observation  of  the total air classifier
system (including number of conveyors,
blowers,  cyclones,  mechanical equip-
ment required) the engineering and
structural  complexity, and an engineer-
ing judgement of the degree and type of
control necessary  to maintain  the air
classifier in operating order. Costs were
determined  by escalating  reported
capital costs (10 percent per annum) to
1980 levels and dividing them by the

-------
Table 4.    Operating and Performance Parameters of Seven Air Classifiers for RF Recovery

                  	Judging criteria and key parameter
                  Energy Requirement     Unit Size
                     RDF Quality

                Avg. Recovered Energy
System Complexity
Cost*
Air Classifier
1. Tacoma
2. Baltimore
3. Richmond
4. Ames
5. LA.
6. Akron
7. Pompano Beach
Specific Energy
(KWh/Mg)
(0,91
15.9)
(5.1 >
(7.6)
(11.4)
(11.4)
(11.2)
Column Loading
(Mg/hr) m
(36.2)
(9.8)
(8.8)
(18.6)
(4.9)
(20.6)
(1 1.8)
Avg. Retained Ash
(Ratio)
(1.5)
(1.8)
(2.3)
(2.0)
(1.9)
(1.3)
(1.5)
Design Simplicity
(Ranking)
6h
3
1
4
2
6
5
1980 Cost Basis
($Mg/hr)
$ 5,800
12,600
4,300
14.OOO
8,200
6,600
6,200
 a Estimated capital cost (equipment and engineering) for air classifier system; costs exclude operating and maintenance expenses.
  Mg/hr - throughput at the critical A/S ratio.
 b  / = simplest system; 6 = most complex system.
critical throughput for each air classifier
at its optimum air flow setting.


Conclusions
  The testing and performance charac-
terization  of  seven  air classifiers has
shown the ranges of operating  condi-
tions and  performance  that can be
expected  for each air classifier.  In
addition, methods have been presented
for comparing different types  of air
classifiers operating under different air
and material flows and handling
shredded  refuse of differing composi-
tion.  No absolute  means  exists for
comparing air classifier  performance.
As the data in Table 4 show, positive and
negative points exist for all air classifiers.
But  it is now impossible to judge air
classifier  performance  on a relative
basis if the  judging  parameters are
judiciously chosen so as to allow an
equitable  comparison (i.e.,  normaliza-
tion of the system outputs to eliminate
the effect of varying infeed composition).
The methods and data presented  here
should prove  useful in the evaluation of
other air classifiers.
   Since the field tests covered a number
of different air classifiers,  analysis of
the test data can establish quantities of
waste constituents that  should be of
interest to  the resource recovery
industry— for example, the percentages
of ferrous  metals, nonferrous metals,
retained fines, and paper and plastic in
the light fraction. Results of all seven
field tests taken collectively enable an
overall  (or average) value of certain
parameters  to  be calculated. Such
average values  subsequently can be
used for calculation of mass and energy
balances on an air classifier system. For
example,  the  mass  fraction of non-
ferrous metals in the light fraction can
be computed from the mass fraction of
nonferrous metals in the air classifier
feed and the percentage of nonferrous
metals retained in the light fraction.
  The testing  of  these  air classifiers
makes apparent the need for a standard
method of testing and evaluation for the
resource recovery industry. Such stan-
dards would not only allow performance
comparisons among different air classi-
fication  systems,  but they would  also
establish the correct operating settings
for producing a  specified output.
Presently little control is exercised over
the air classification process.  Conse-
quently, the quality of the RDF output
often suffers. As with most resource
recovery processes,  air classification
has  yet to progress  from an art to a
controlled  process.
  The  full report was  submitted in
fulfillment of Contract No. 68-03-2730
by Midwest Research Institute under
sponsorship of the U.S. Environmental
Protection Agency.

References
1.  Hopkins, V., Simister, B., and Savage,
   G., Comparative Study of Air Classi-
   fiers,  EPA-600/2-81-221, NTIS
   PB82  106121,  U.S. Environmental
   Protection Agency, Cincinnati, Ohio,
   September 1981.
2.  Savage, G., Diaz, L and Trezek, G.,
   "Performance Characterization of
   Air Classifiers in Resource Recovery
   Processing," Proceedings of the
   1980 National Waste Processing
   Conference: Resource Recovery
   Today  and Tomorrow,  American
    Society of Mechanical Engineers,
    New York, May 1980.

-------
G. M. Savage, L F. Diaz, and G. J. Trezek are with Cat Recovery Systems, Inc.,
  Richmond, CA 948O4; V. Hopkins, B. Simister, andD. Fiscus are with Midwest
  Research Institute. Kansas City,  MO 64110;  S. C. James (also the EPA
  Project Officer,  see below) is with the Municipal Environmental Research
  Laboratory, Cincinnati, OH 45268; (Don Brunner, with the Naval Civil Engi-
  neering Laboratory, Port Hueneme, CA 93043, directed two small-scale sys-
  tem studies.).
 The complete report, entitled "Comparative Study of Air Classifiers," (Order
   No. PB 82-106 121; Cost: $22.50, subject to change) will be available only
   from:
        National Technical Information Service
        5285 Port Royal Road
        Springfield, VA 22161
         Telephone: 703-487-4650

 The EPA Project Officer can be contacted at:
        Municipal Environmental Research Laboratory
        U.S. Environmental Protection Agency
        Cincinnati, OH 45268
                                                                              5
                                                                        •fr U.S. GOVERNMENT PRINTING OFFICE:1981--559-092/3358

-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Postage and
Fees Paid
Environmental
Protection
Agency
EPA 335
Official Business
Penalty for Private Use $300

RETURN POSTAGE GUARANTEED
                                 PS
                                  -   0000329
                                U S  6NVIR  PROTECTION  AGENCY
                                REGION  5 LIBRARY
                                230  S DEARBORN  STREE)
                                CHICAGO  It  60604

-------