Resources, Conservation and Recycling, 2 (1989) 261-273

261

Elsevier Science Publishers B.V./Pergamon Press plc - - Printed in The Netherlands

An Empirical Approach to the Design of Trommel

Screens for Fine Screening of Domestic Refuse
P.A. WHEELER, J.R. BARTON and R. NEW

Warren Spring Laboratory, Stevenage, Hertfordshire (U.K.)

(Received January 6, 1989; accepted in revised form March 13, 1989)

ABSTRACT

Wheeler, P.A., Barton, J.R. and New, R., 1989. An empirical approach to the design of trommel
screens for fine screening of domestic refuse. Resour. Conserv. Recycl., 2: 261-273.
This paper describes an empirical method of designing trommel screens for processing domestic
refuse. It details the method of design and uses a case study to explain the method fully. The
results obtained show that the method is robust enough to ensure effective trommel design within
defined limits. A greater understanding of the fundamentals of trommel screening is still required
to refine the method and to broaden its application. However, some extensive use of the proposed
method will facilitate identification of which variables (feed and machine) warrant more detailed
attention.

INTRODUCTION

Trommel screening is one of the most important primary unit operations in

advanced refuse processing systems for separating combustible and non-combustible components. A typical trommel is shown diagrammatically in Fig. 1;
it is essentially a rotating drum with holes in the surface through which the
undersize material passes. However, despite this prominent position, attempts
to predict performance of trommels have generally been unsatisfactory. This

FEED

- - - -

,Vr

.o

~,

:.

UNDERSIZE
Fig. 1. Typical trommel.

0921-3449/89/$03.50

1989 Elsevier Science Publishers B.V./Pergamon Press plc

262
is due in part to the fact that refuse is a complex and variable mixture, but
more importantly to a lack of understanding of the screening process when
applied to refuse.
Not surprisingly this has led to the over-design, or more commonly underdesign, of trommels which can cause severe process problems for equipment
downstream of the screen. An example of this is the trommel initially installed
at the Byker reclamation plant [1], which was so ineffective at removal of
small sized, high ash and moisture content particles from the pulverised refuse
that a secondary screen was installed to remove these components from the
fuel fraction. Thus, in order to avoid similar mistakes and to make trommel
design more reliable, some predictive method of modelling trommel operation
is needed.
Attempts have been made to establish theoretical models of trommel operation [2,3]; one theory postulated by Alter et al. was the basis of extensive
studies undertaken by the National Centre for Resource Recovery (NCRR)
and sponsored by the U.S. Department of Energy. However, it was found that,
in general, theoretical models did not simulate the screening process sufficiently accurately to permit their use for trommel design. The main areas of
difficulty encountered with the theoretical models is the inability to predict
the following:
( 1 ) breakage of friable components and the liberation of entrained undersize
particles;
(2) particle/particle interactions interfering with particle presentations to
the screen;
(3) the probability of a particle passing through a given aperture size;
(4) the heterogeneous nature of the material;
(5) the simulation of the effect of lifters.
The use of empirical data as the basis of a design guide for trommel screens
avoids many of the above problems, because many of these factors are taken
into account by the experimental results.
The purpose of this paper is to propose an empirically developed method of
extrapolating experimental data from existing trommels processing domestic
refuse, so that design guidelines for different applications can be established.
The basic structure is presented in the hope that other theoretical factors can
be added in due course to reduce the dependence on experimental data collection, allow extrapolation to conditions not covered by experimental data, and
enhance the fidelity of the resultant designs.
THE APPROACH
Specific work at Warren Spring Laboratory (WSL) on trommel screening
has been sponsored by NCRR [4,5], the European Community, and the Department of Trade and Industry [6]. In addition, a significant number of data

263

has been obtained from various refuse processing plants in the U.K. and Europe in the course of plant development and monitoring work. From these
studies an empirical method of trommel screen design has been developed.
Past experience has shown that screening performance of trommels operating on domestic refuse is generally related to the amount of material which
is passing over the screen plate in unit time rather than either the quantity of
material removed to undersize or the overall feedrate to the screen. This is
shown by the results obtained from the Doncaster refuse processing plant [ 7 ]
(Fig. 2a), where the relationship between feedrate and recovery had some outlying points. These tended to be associated with periods where the feed size
distribution differed markedly from the norm. Plotting the same results against
the flowrate of true oversize material tended to bring these outlying points into
line; this is more clearly shown in Fig. 2b and is confirmed by the improved
correlation coefficients shown in Table 1.
This seems to imply that the performance of the screen is directly related
1 O0

90
olOmm
"I-

8o
z

~-,F'' ~ ~ ~

13

70

-20+10ram

~. 6O
rim
0 5O

ll-10mm
+ -20+10mm
-40+20ram

l,,

010

12

14

16

f'11
18

I
20

FEEDRATE, t h -1

100
a~
N
re
UJ
a
Z
O
I--

90
80
70

~"
,'r
U.I

60

50

-10mm

7+" "~*--**...~
_

-20+10ram

~ r n m

'~1
0

1
4

I
6

I
8

FLOWRATE OF T R U E OVERSIZE,

-lOmm
+-20+10ram
{3 - 4 0 + 2 0 r a m
I
10
t h -1

Fig. 2. (a) Recovery to undersize vs feedrate of refuse - Doncaster trommel. (b) Recovery to
undersize vs flowrateof true oversize - Doncastertrommel.

264
TABLE 1
Doncaster trommel fine screeningsection; correlation coefficients(R)
Size range
(mm)
10
-20 +10
-10 +20
-

Correlation coefficientsof size range vs.

feedrate

flowrateof true oversize

0.600
0.701
0.536

0.858
0.854
0.849

TABLE 2
Independentvariables affectingtrommel screening
Machine variables
Rotational speed
Screen aperture size
Screening length
Trommel diameter
Open fractionalarea
Angle of inclination
Presence of liftersa

Feed variables
Feedrate
Particle size distribution
Particle shape

aA lifter is a plate mountedperpendicularto the screen surface lifting material higher than would
normallybe achievedby the rotational speed alone.
not to feedrate but to some other parameter of the feed material. The most
likely is the volume of the material passing over the screen. However, this is
difficult to quantify and measure reliably. A useful guide to the volumetric
feedrate might be the flowrate of true oversize ( + 40 mm material for the results reported in this paper). This is because the coarser components contribute significantly to the overall volume whilst fine particles generally do not
and simply occupy the voids between the larger components.
The independent variables t h a t govern trommel screen operation are listed
in Table 2. Ideally any model of trommel operation will include all of these
variables. However, in empirical models some or all of these variables can be
incorporated by using relevant experimental data rather t h a n a mathematical
relationship.
To compare data from different trommels, some method of normalising the
results must be used, which gives factors relating to the feedrate and resultant
recovery to undersize. At the present time not all of the variables are incorporated into the design guide. However, two of these factors, trommel length
and trommel diameter, can be used to create a "recovery index" and a "feedrate
index" respectively.
Variables not presently used are rotational speed, presence of lifters, the

265
angle of inclination, and the screen aperture size. The rotational speed is generally pre-determined by the requirement for good screening and, for refuse,
this requires a cascade mode of operation. This is achieved by operating the
screen at or above 50% of the critical speed with lifters or 80% of the critical
speed without lifters (critical speed is the rotational speed at which material
centrifuges and is a function of trommel diameter). Generally, the results from
tests carried out in either of these regimes are comparable.
Information on aperture size and angle of inclination is supplied by the experimental data curves.
The trommel length can be used in a recovery index by assuming that the
screening rate is constant along the length of the screen. This assumption is
not entirely correct, and results have shown that the screening rate decreases
asymptotically along the length of the screen. However, for most practical applications the error is not large, although care must be exercised when using
data from very long or very short trommels. The reason for this reduction in
effectiveness has been explained elsewhere [6,7] but is mainly due to the depletion of fine material at the screen surface and then to the subsequent rate
at which additional fines become available for screening due to liberation or
breakage effects. However, there is little variation in unit recovery along the
length of the screen beyond the first metre or so of the trommel. Thus, a recovery per unit length can be calculated from the total recovery to undersize,
and the screening length. A more general term could possibly be established if
the particle-to-aperture size ratio were used as opposed to data based on discrete size ranges with particular screen aperture sizes. However, results from
trommels with significantly different aperture sizes do not compare well. This
is not surprising, since the nature of the oversize material can vary substantially with screen aperture size. For example, the oversize fraction of refuse
screened at 200 mm consists mainly of paper, textiles and plastic film whilst
the oversize fraction of refuse screened at 50 mm will contain metals, dense
plastics, and putrescibles in addition to the paper and film plastics. However,
if screen sizes are grouped together then there is some agreement between results and this may further improve the design guide. Invariably, performances
with screen aperture sizes between 40 and 60 mm compare well whilst performances with coarser aperture sizes do not.
Trommel diameter is incorporated into a feedrate index by dividing the flowrate of true oversize by the cross-sectional area of the trommel. Ideally, an
additional factor should be included to reflect the particle transport velocity
through the trommel. This would make the feedrate index a dimensionless
factor related to the dynamic loading of material within the trommel. Use of
the tangent of the inclination angle of the trommel, which is theoretically proportional to the material transport rate (derived from Alter [3] ), does not
completely correlate results from different angled trommels, although the results are closer than would otherwise be the case. Differences are probably due
Q

266
to the effect of particle/particle interactions at higher feedrate densities, and
so it may in any case be very difficult to establish empirical or theoretical
factors for different trommel inclination, angles.
Thus, it is possible to calculate the dimensions of a trommel by simply extrapolating existing experimental data assuming that the results have been
obtained with the trommel operating in an efficient mode. The effect of different aperture size has not been fully quantified, and so the method is limited to
trommel aperture sizes that have experimental results available. Generally, the
recoveries of size fractions smaller t h a n 50% of the aperture size have been
found to be fairly insensitive to small changes in aperture size, and problems
are encountered only with "near size" particles.
METHOD
To obtain a "feedrate index" (flowrate density of true oversize) the flowrate
of true oversize is divided by the trommel cross-sectional area. The recovery
index is expressed as a recovery per unit length of screen and can be calculated
using the following equation*:
/'ln ( 1 - R t ) ~
Ru = 1 - e x p ~
n
]

(1)

where Ru is the recovery to undersize per metre; Rt is the total recovery to

undersize; n is the length of the trommel screen in metres.
To use this information, the recovery per unit length is plotted against the
flowrate density of true oversize. This provides a curve which describes the
recovery of a particular refuse size fraction with a particular screen size. Figure
3 shows the response of the - 40 + 20 mm, - 20 + 10 m m and - 10 m m size
fractions of raw refuse being screened at 40 m m apertures. These curves consist
of results obtained from the Doncaster refuse processing plant, WSL pilot plant,
and Herten (FRG) refuse-derived fuel plant. It is evident that there are good
correlations considering the wide differences between these three installations.
Therefore, by using this type of curve the design of a trommel can be estimated
by following the procedure shown in the flow diagram (Fig. 4 ).
The most important step in the design of a trommel is for the user to determine the required recoveries of the particular size fractions to undersize and
the allowable loss of required material to undersize. From the feedstock size
distribution and feedrate, the oversize flowrate is calculated. This allows the
diameter of the trommel to be selected to give an optimal flowrate density of
true oversize. From the relevant screening response curve the unit recovery is
*This equation is a simple derivative from a power law expression. For those not familiar with
this type of mathematical progression, it is equivalent to the common equation for compound
interest, but with negative interest, i.e., a fixed proportion being extracted each increment.

267

-1Omm

-20lOmm....._.
8

"- ~--~ +2"+"-

50

~
":F " - - .~..

~=~,~

(,9
c
z
o
>c

e~

-40+20mm-----

="~--

31.7

I ~
25

tr

....

D O N C A S T E R W S L HERTEN
-lOmm

-20+10mm
-40+20mm

0
0

O
+

14.8

x
I
1

I
2

2.18

FEEDRATE OF TRUE O V E R S I Z E / C R O S S S E C T I O N A L AREA OF T R O M M E L

Mg h "1 m "2

Fig. 3. Comparison of doncaster, W S L and Herten trommels with 40 mm apertures - raw waste.
Herten 4.5 , W S L 5 inclination, Doncaster scroll transport.

read off for the selected flowrate density, and so the total required screening
length is calculated from the following equation:
ln(1-Rt)
n=ln(l_Ru)"

(2)

This result is then compared with the initial specification and then adjustments are made as necessary.
it must be pointed out that there is no one design that will "fit" an application, but a number are contained within an envelope of design and operation
conditions. Thus, selection of a particular set of dimensions may well be determined by the building or engineering constraints rather than by the specific
requirements of the screening process.
This procedure can be more easily understood by the use of an example. One
such example of applying this design procedure was for the retro-fit of a front
end trommel at the Byker plant. The design specification and constraints were
as follows:
(1) feedstock raw refuse - 4 0 mm content 30% by weight,
(2) feedrate 30 t h - 1,
(3) oversize product; minimal loss of + 40 mm material,

268
FOR EXAMPLE:

'~

~ . INITIATEDESIGN ~

~OND,T'ON.PEEDRATEAND |
PHYSICALLIMITATIONS

/ / ~

MuST CONSIDER IF ANOTHER _

- -_I
SCREEN TYPEWOULD BE BETTER
SUITED TO THE PROPOSED DU'E]%RI . . . . .

SET APERTURESIZE. ANGLE

OF INCLINATION,ROTATIONAL
SPEED AS % OF CRITICAL
PRESENCE OR NOT OF-LIFTERS

r
/
J
|

....

TOGIEFEEDRATE

DENSITY OF 1 Mg h4 m"2

LS..A..,~~
/ ~ _TROMMEL
" ............
~
~EF(~!~TH~--~

----

' I
FROMSPECIFICATION
DECIDEOPERATING
PARAMETERS

USE ANOTHER __

TYPEOF SCREEN ~/

=
I
t
I

y
~

J
N

/ DES,GN
/ POSSIBLEAT
~ANY CONDITIONS/;
~
TO MEET

'
I
_[ CALCULATED,AMETER I
ROM FEED SPECIFICATION

CHANGE DIAMETERTO
WITHIN SPECIFICATION
LIMIT
LIMITS/

Y[
NOTE FLOWRATEOF TRUE
OVERSIZEAND CROSS
SECTIONALAREAUSED
FROM EXPERIMENTALDATA
OR EXTRAPOLATIONSOF
THE EXPERIMENTALCURVES

CALCU'
T

-- - - - -

FEEDRATEDENSITY'

/
/

------7
/'

FROMFEEDRATE

DENSITY OBTAIN
/
RECOVERIESPER UNIT / ~
LENIGTH
Y

CALCULATE LENGTH OF/

TROMMEL TO MEET I
I
SPECIFI?ATION }

HIS LENGTH
WITHIN /

Fig. 4. Flow diagram of trommel design procedure.

- ( x A PRACTICAL

...~,M,TAT,O
7

1
,

ADJUST ANGLE OF
INCLINATION OR
DIAMETER

JS DESIGN
SSIBLE AT THIS "~
/pO
OTATIONALSPEED TO
MEET SPECIFICATION/

269
(4) undersize product; 90% removal of - 1 0 mm material and significant
( + 50% ) removal of - 40 + 10 mm material to reduce ash and moisture
contents of fuel fraction,
(5) trommel constraints diameter < 3.5 m (limited head room) length < 5.0
m of screen (total length limited to 6 m),
(6) other factors: (a) glass breakage and removal is deemed beneficial, (b)
low power consumption is beneficial.
The above specifications would indicate that the basis for the trommel design operating conditions should be 40 mm apertures (3), trommel rotation
limited to 50% of critical rotational speed (6b), provision of lifters (6a and b ),
and a minimum diameter of 2.5 m (6a). As a first estimate, an angle of inclination of 5 o can be taken.
From this, the feedrate of true oversize can be calculated as the feedrate
multiplied by the fraction of feed larger than aperture size:
30 t h-~ >< ( 1 - 0 . 3 ) = 2 1 t h -~.
Thus, at the maximum diameter of 3.5 m, this gives a flowrate density of
2 1 / [ ( 3 . 5 / 2 )2n)] =2.18 t h - ' m -2.
The unit recoveries for a flowrate density of 2.18 t h -1 m -2 in a trommel
with inclination of 5 and 40 mm apertures, are given in Table 3 (these are
extracted from Fig. 3 by reading from the curves the unit recovery corresponding to a feedrate density of 2.18). From these data it can be calculated (using
eq. 2) that to meet the specification of 90% of - 10 mm material reporting to
undersize a length of 5.6 m is required. This does not fall within the specification. To improve the recovery per unit length the angle of inclination (Fig.
TABLE 3
Unit recovery data taken from Figs. 2 and 4
Parameter

Feedrate density (t h -1 m -2)

Unit recoveries (% m- 1)
- 10 mm
- 20 + 10 mm
-40 +20 mm
Trommel length to
meet
90%
- 10 mm
recovery (m)
nd: not determined.

Angle of inclination
5

Calculated 4 o

2.18

2.18

2.18

33.7
31.7
14.8

55.5
50.5
35.0

40.9
38.0
21.5

5.6

nd

4.38

270

4) must be decreased to increase residence time of material in the screen. Fig. 5

shows the results for 2 inclination on the WSL pilot scale trommel. Although
these data are less precise than Fig. 3, because there is only one source of data,
it does provide a rough guide to the unit recoveries at this trommel inclination
angle. Thus assuming a linear relationship between the two angles, the unit
recoveries at an inclination of 4 can be interpolated (Table 3). From these
data the length required to meet the specification is 4.38 m (using eq. 2 and
the 90% required recovery to undersize of - 10 mm material) which is within
the required length of 5 m. The estimates of recovery to undersize for each size
0.6
Jr

~-,

E0 . 5

nO

,....~p ~
-I-

/55.5

Jr
~

0.4

I
I

I
.~,

i.z 35.0

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

="

-40+20ram

I
I

0.2

nZ
D

O 0.3
tyfr
O

il
-20+lOmm

uJ
N

-1Omm

I " / 50.5

I
I
I
I

0.1

i
1

1.4

Ii

1.8
FLOWRATE

2.2

2.6

DENSITY OF TRUE OVERSIZE Mg h-~m -2

Fig. 5. Unit recovery curves from W S L pilotscaletrommel with 12 inclination.

TABLE 4
Recommended and installeddimensions and conditions of operation for the Byker trommel
Parameter

Initiallyrecommended

Installed

Diameter (m)
Screening length (m)
Aperture size (mm)
Inclinationangle (deg)
Rotational speed (rpm)
Lifters

3.5
4.4
40
4
11
two off
250 m m high
full length
50

3.5
4
50
variable 3-7
variable 11-13
two off
250 m m high
full length
53

Aperture open area (%)

26.9

30.7

24.8

30

A
Model prediction
at same conditions

B
Model prediction
at same conditions

C
Model prediction
at same conditions

Design specifications

nd: not determined.

Feedrate
(t/h)

Test No.

21.0

16.0
16.0

19.4
19.4

16.9
16.9

O/S
flowrate
(t/h)

Byker t r o m m e l screen efficiencies a n d predicted efficiencies

TABLE 5

Inclination
angle
(deg)

4.4

Screening
length
(m)

71.8
66.6

77.8
81.4

12.4
nd

9.4
nd

67.9

77.3
73.9

16.3
nd

nd

-40 +20
mm

-80 +40
mm

Recovery to undersize ( % )

89.4

89.5
93.1

81.6
83.0

82.2
87.7

-20 +10
mm

90.2

91.5
95.3

83.5
85.9

85.2
91.2

-10mm

~=~

272
fraction on a trommel 4.4 m in length are 90.1%, 87.8% and 65.5% for the - 10
mm, - 20 + 10 mm and - 40 + 20 mm size fractions, respectively.
Based on these calculations the recommended and installed dimensions and
conditions of operation for the Byker trommel are given in Table 4.
These changes to the design were made following discussions with the engineers for the trommel and the plant operators. Although the reduced screen
length adversely affects predicted screening efficiency, the increased aperture
size and built-in flexibility provided by the variable angle and speed facilities
ensured that the specification could still be met within a broader design and
operating envelope.
Tests were performed on the Byker front end trommel shortly after its installation [8]. The predicted recoveries to undersize for the Byker trommel at
an oversize feedrate of 21 t h - ' are given in Table 5. The screening efficiency
results obtained during the test series are given along with predictions under
the same conditions using the design procedure in Table 5. The test B predictions (the closest to the design specification) are reasonably close to the actual
values; however, the predicted values at the lower feedrate are much higher
than the actual values measured. This is probably due to the reduced efficiency
of the liberation device in front of the trommel at lower feedrates, which would
almost certainly have the effect of lowering the measured screening efficiency.
The predictions for test C (3 inclination) are also moderately high, but this
may well be due to the inaccuracy involved with interpolation between the 2
and 5 response curves. The - 4 0 + 20 mm recovery results and predictions
are not so well correlated as the results for the smaller size ranges, and this is
due to the difference in aperture size between the response curves (40 mm)
and the Byker trommel (50 mm ) affecting the recovery of near-sized particles.
These differences apart, the results show that the technique used in the design
procedure described is sufficiently accurate to give a good initial estimate of
trommel dimensions and, provided there is some flexibility built into the trommel design (e.g. variable speed/angle), confidence that the plant will meet the
process specifications.
CONCLUSIONS
The procedure described can, with a fair degree of accuracy, predict the response of a trommel used in the screening of domestic refuse given a sufficiently accurate unit recovery response curve. This procedure can also assist
in selecting operating parameters to improve performance in existing trommels; for example, reducing the angle of inclination thereby increasing separation efficiency.
The procedure is only as good as the data for the response curves, and so has
substantial limitations as a practical design tool at this stage. However, if'factors which can account for the other machine variables can be established either

273

by theoretical or empirical means then the usefulness of the design guide will
be increased. The main areas where these factors are needed are the angle of
inclination and the aperture size. Further consideration of the waste characteristics is also required because feedstock differences affect the screening performance. This effect will manifest itself as a displacement of the response
curve. The main distinction at present is between raw and pulverised waste
and a separate data set is used by WSL for designing screens for shredded
feedstocks. However, with better understanding of the screening process the
detailed effects of waste composition and prior processing will be quantifiable.
When added into the design guide together with the results already available,
this would provide a sound basis for designing trommels for a much wider range
of screening requirements.
Thus within limits the efficiency of the design method has been demonstrated. The results illustrate that, despite the complex nature of refuse, the
response characteristics are sufficiently robust to enable a modelling approach
to succeed. Although pragmatism has directed that a strongly empirical approach be taken at this stage, continued use and development of the method
can be expected to increase our understanding of the basic mechanisms of
trommel screening.

REFERENCES
1 J.R. Barton, Byker plant operation. Test Rep. No. 2, December 1984, Stevenage WSL 1985,
rep. LR 541 (MR).
2 J.C. Glaub, D.B. Jones and G.M. Savage, The design and use of trommel screens for processing
municipal solid waste. In: Proc. 10th National Waste Processing Conf. ASME, New York, May
1982, pp. 447-457.
3 M. Alter, J. Gavis and M.L. Renard, Design models oftrommels for resource recovery processing. In: Proc. 9th National Waste Processing Conf. ASME, New York, 1980, pp. 361-371.
4 J.R. Barton, Evaluation of trommels for waste to energy plants. Phase I rep. of the Doncaster
and Byker test series, Draft report. U.S. Department of Energy contract No. DE-AC03-80CS24315, December 1981.
5 J.R. Barton, Evaluation of trommel for waste to energy plants, phase 2 rep., Warren Spring
Laboratory pilot plant test series. Stevenage WSL 1983, rep. LR 462 (MR).
6 J.R. Barton, A.J. Poll and P.A. Wheeler, Investigation into the parformance of screens and air
classifiers in the processing of urban waste. Poster session paper, EEC Waste Seminar, Luxembourg, 25-27 Sept. 1984.
7 J.R. Barton and A.J. Poll, Doncaster plant operation, Test Rep. No. 2, October 1985. Stevenage
WSL 1986 rep. LR 572 (MR).
8 J.R. Barton and P.A. Wheeler, The benefits of front-end trommelling in processing municipal
solid wastes. Trials at the Byker plant - June 1987. Stevenage WSL 1988 Rep. LR 661 (MR) M.