You are on page 1of 22

Resources, Conservation and Recycling, 6 ( 1991 ) 1-22

Elsev ier Science Publishers B.V. / Pergamon Press plc

A new t r o m m e l m o d e l

Richard Ian Stessel


Department of Civil Engineering and Mechanics, Universityof South Florida, Tampa,
FL 33620-5350, USA
(Received 26 November 1990, accepted 11 June 1991 )

ABSTRACT

Stessel, R.I., 1991. A new trommel model. Resour. Conserv. RecycL, 6: 1-22.

Rotary screens are in an important unit operation in materials processing. Much of their design has
been empirical. A mathematical model was developed with three major components: particle rise on
the screen, particle trajectory through the air, and screening of the particle while in contact with the
screen. All were implemented on a computer using numerical methods, allowing the retention of all
necessary mechanisms within the theoretical expressions. Particle rise incorporated friction; particle
trajectory incorporated drag. The screening element made use of entirely new probabilistic theory
differing from previous work by incorporating consideration of the depth of the bed. Results showed
good predictive capabilities. Insights concerning the importance of bed layering were obtained. Fur-
ther recommendations aiding design were obtained and analyzed.

INTRODUCTION

Trommels, or rotary screens, are an important unit operation in materials


processing. They have a long history in the mineral dressing industry, and are
of particular interest to those concerned with solid waste processing because
of the difficulty of the material processing task. Trommels provide an elegant,
low maintenance and operating cost, screening option that often justifies their
higher initial cost.
A trommel is a rotating, cylindrical screen, lying on its side, at a small angle
from the horizontal. Material is fed in at its elevated end, and size-separation
occurs as the material spirals down the drum at an axial speed governed by
the angle of the drum to the horizontal. Behavior of the material normal to
the drum axis is governed by the rotational velocity.
An example of an important trommel application is the removal of small
particulate material ("fines") from shredded waste. Trommels are favored
for this task, but the separation is often inadequate [ 1-3 ].
Current design practice has evolved by experience. As waste processing op-
tions become more varied, roles for trommels (and other unit operations)
become more specialized. As example is a trommel used to differentiate be-

0921-3449/91/$03.50 © 1991 Elsevier Science Publishers B.V./Pergamon Press plc. All rights reserved.
2 R.I. STESSEL

tween 2 1 plastic soda bottles, and cans and bottles, and a trommel used to
separate humic material from a mined landfill. The work reported here at-
tempts to broaden the applicability of trommels by enhancing the under-
standing of the underlying physical principals of operation.
This paper presents a theoretical, descriptive or research model of trommel
operation, described by equations of the mechanics of particles moving
through the cylinder. The distinction must be made between a descriptive
model that describes, in terms of meaningful output parameters, the behavior
of a physical system based upon input parameters; and a prescriptive model,
that allows the designer to input desired output as well as input parameters,
having the model produce a finished design. Descriptive models allow a re-
searcher to ask "open-ended" questions concerning the effect of varying input
parameters, including those pertaining to design; a prescriptive model re-
quires output parameters to be set, and adjusts the design accordingly.
The descriptive model discussed here evolved from underlying physical
principals, implemented in computer code in a manner not practicable for
earlier researchers in the field. Two discrete efforts employing the model are
discussed next. The first is its use to replicate experimental data from the
literature, drawing conclusions from the required modifications of input pa-
rameters. The second is use of the model to draw conclusions by analyzing
the effect of altered input parameters.

MATHEMATICAL MODELING HISTORY AND DEVELOPMENT

As the speed of revolution of a trommel increases, it goes through different


operating modes. At the slowest speed, the trommel contents tumble, called
cascading. Mixing action is insufficient to adequately move the center of the
tumbling mass to the outside so that it is presented to the screen openings. At
the highest speed, particles are held against the outside of the drum by cen-
trifugal force. In between is a motion called cataracting, when the drum ro-
tates at a speed just slightly less than that needed for centrifugation. Particles
are held on the wall of the drum against gravity until just before the top of the
drum, where gravity overcomes the centrifugicity and the particles fall almost
along the drum diameter to the bottom [ 4 ]. Under these conditions, particles
fall the maximum distance possible, creating the greatest amount of tumbling,
leaving few to cascade. These motions are shown in Fig. 1.
Particle movement in cataracting has, most often, been divided into three
basic phases: the particle rises along the screen; it departs from the screen and
falls until it hits the screen again; while it rests on the screen, some of the
material passes through the screen. Fig. 2 shows the key variables and coor-
dinates. Each type of motion is addressed below in turn.
A NEW TROMMEL MODEL 3

BAND

@@0 (a) (b) (c)


Fig. l. (a) Cascade or kiln action; (b) cataract action showing band: (c) centrifuging.

(3

Fig. 2. Definition sketch.

Particle trajectory
Most of the past reports addressing trommel behavior were geometrically
oriented. Alter et al. [5] focused principally on departure location and an
assumed location angle of impact. The actual trajectory of the particle in flight
was of little significance to their analysis. Glaub et al. [6, 7] added more
factors to Alter's paper. All these researchers specified the initial condition
that forces on the particle are instantaneously balanced. The particle leaves
the screen when the centrifugal reaction is no longer able to hold the particle
on the screen against gravity; this gives the particle an instantaneous fixed
velocity vector.
None of these approaches fully included the effects of drag. Glaub et al. [6 ]
began a discussion, but left the drag force coefficient unspecified. In the case
of small particles, when, for example, minerals are pulverized m or in the case
of light particles, such as paper or very small particles in solid waste - - the
drag on a particle becomes an important component in the force balance. With
trommel diameters of 1 to 2 m, it is quite possible that the aerodynamically
lighter particles will have achieved terminal velocity in flight [ 8 ].
4 R.I. STESSEL

The present work uses the same initial conditions as in the Alter et al. and
Glaub et al. above, which were the velocities instantaneously upon departure
(see Fig. 2 ):
0Rcosa = z0 ( 1)
ORsina=ko
x and z are the horizontal and vertical locations of the particle.
The weight of a particle, Fe, acting through the angle of inclination fl, is
defined:
F e = mcosflg (2 )
where rn is the particle mass and g the acceleration of gravity. The inclination
of the trommel, fl, is included because it determines the movement in the
axial direction, i.e., down the trommel. When describing movement across
the trommel, the small angle approximation (with fl< n/12, cosfl.~ 1 ) justi-
fied its omission [ 5-7 ]. The drag on a body is:

FD=~CDpVIvl n -~ (3)

V=(Vp--Va)
where va is the velocity of air, and vp is the velocity of the particle. Drag is a
function of the square of the difference between air velocity and particle ve-
locity. To preserve the sense of direction, this is calculated as the net velocity
multiplied by its absolute value.
Particle trajectories come from a pair of equations balancing drag and grav-
ity and neglecting buoyancy [9 ]. Solving for acceleration in the horizontal
and vertical directions, respectively, these are:
1
J~=m[FD'x] (4)
l
2 = m [ --rng+FD,:]

using rn for mass. Calculation of the sphere Reynolds number, particularly in


the case of the addition of airflow, shows that a drag coefficient correlation
incorporating both the laminar and turbulent regimes is required; one of many
statistical fits to the drag coefficient curve is employed [ 10 ]. These equations
are solved parametrically using a fifth-order Runge-Kutta routine. Compu-
tation ends when the particle reaches the wall again, and
R2=x2+z 2 (5)

The movement clown the trommel depends on both the axial component of
A NEW TROMMEL MODEL 3

the vertical departure velocity, and the angle of declination of the trommel.
The z-direction shown in Fig. 2 is tilted into the page by the declination angle,
ft. The vertical coordinates of the point of departure and impact, ZD and z~,
respectively, are known, and the increment of axial movement during one
turn in the particle's spiral path down the drum is:
lA = [ (ZD--Z 1) + ~ot]Sinfl (6)
In addition to its importance in calculating overall trommel length, the incre-
ment of axial movement is required to determine the thickness of the bed
during screening as discussed below.
Together, these equations serve to determine the motion of the airborne
particle within the trommel.

Particle rise
The rise of particles in contact with the rotating screen gives the departure
location of the particles and the area of screen occupied by the material. The
contact are on the screen occupied by particles in one pass is defined as a band
as shown in Fig. 2. The band area is critical to determining both the amount
of screening that occurs, and the axial distance travelled, during each rota-
tion. Screening occurs only when the particle contacts the screen, and is a
function of the screen area contacted and the percentage of screen area de-
voted to holes. Ultimately, this yields the design length of the device.
The traditional approach is to first determine a point of departure, then
determine the rotational velocity, a, of the screen that would have the particle
leave the screen at that point. This results from a force balance between the
centrifugal reaction and the force of gravity [ 5-7]:
(gsino~/R )~ (7 )
This analysis does not apply if the material slips against the screen.
Glaub et al. [6] reported analyses incorporating friction. Separate force
balances were required to calculate motion without slippage, motion with
slippage, and the transition between slippage and sticking. It is necessary to
distinguish between 0, which is the location of the particle, and o~, the location
of a point on the screen. As with the simple force balance in Eq. 7, the mass
terms cancel, leading to calculation of accelerations: AR in the radial direction
and Av in the tangential direction:
AR= --R ( O)2--gsinO (8)
A-r=RO= ~./dA R - - g c o s 0 ( 9)
where/~ is the dynamic friction coefficient. Combining

b = fld 0 2 - - g [/~osin0 + cos0 ] (1o)


0 R.1. STESSEL

This equation is solved by the same fifth-order Runge-Kutta method used


in the trajectory calculations. Whether or not the particle is slipping is deter-
mined by checking the force balance between the tangential gravity force and
static friction. This is done by determing ifEq. 9 is positive, after substituting
Fts for #d. If the static force is larger, the particle velocity is that of the screen.
Similarly, departure from the screen is determined by checking Eq. 8 at each
time interval.
The screening rate is a function of the area of the screen to which the par-
ticles are exposed as they rise. The inclusion of particle slippage means that
the material rides over more screen openings than would be determined strictly
by dividing the band area by the hole frequency. The increased number of
holes is calculated by determining the time that it takes a point on the screen
to cover the arc-length of the band area, the time that it takes the feed actually
to rise through the arc-length, and taking the ratio of those times:

E= (trise)/(OI-OD)-¢ (1 1 )
0

/rise is calculate during the numerical integration.


The capacity of the screen and the processing rate is a function of depth of
material on the screen. The screen is conceptually divided into bands, as de-
scribed above. The screen area occupied by one band is determined by mul-
tiplying the width of the band (see Eq. 6) by the arc-length of the band. The
arc-length between the location of departure and the location of impact is:

lr=R [2re- ( o q - O~D)] (12)

The depth of bed, T, is then determined by dividing the flow into the trommel
occuring during one rotation, VB, by the band area:

T= VB/IAlr ( 13)
Together, calculation of the particle trajectory and particle rise describe the
movement of the particle. These are assembled into an independent com-
puter model to analyze particle motion only. Full description of trommel op-
eration requires incorporation of screening.

Screening
While the material is moving on the screen, it is being separated. The model
incorporates the screening of particles in conjunction with their rise up the
screen, in the band.
Trommel research in the literature has been probabilistic, considering in-
dividual layers, one particle thick. The central concept concerned the ratio of
the opening size to particle diameter, attributed to Gaudin's handbook [ 11 ]
A NEW TROMMEL MODEL 7

by Alter et al. [ 5 ] (also derived in [ 12 ] ). The fraction of particles passing, f,


was"

f( dh,d)=n[ d-dh/ d] (14)


where d is the diameter of the trommel screen holes, dis the particle diameter,
and H is the fraction of the screen area made up by holes. This expression was
integrated to determine the total fraction of particles of diameter less than the
hole diameter that pass, and made recursive to determine the effect of multi-
ple passes. A single layer is a difficult concept to justify when the objective of
screening is to separate a bed of material made of multiple particles sizes.
Layers above that in immediate contact with the screen were not incorpo-
rated. Vorstman and Tels [ 12 ] took the thickness of the bed into account by
determing surface concentrations arising from depletion of the bed, but were
still bound by probabilities of single particles passing. Below, field data sup-
port surface concentration as an important phenomenon.
A simple experiment was performed. A layer of granular material was placed
on a plate with an obscured hole. All the particles were considerably smaller
than the hole opening. When the hole was opened, a cone-shaped depression
was formed, as one would expect. With a range of particle sizes, the number
of particles passing is a function of the probability that a particle of any size
larger than the hole would block the hole. An average number of particles
passing through a hole is calculated by knowing the number-concentration of
particles both larger and smaller than the hole:

PN(dp) (15)
NP= fdm'XPN(dp)ddp
d dh

where Np is the number of particles passing, PN is the number particles size


distribution (PSD), dh is the diameter of the trommel openings, and dmax is
the diameter representative of the largest particles. The bottom integral gives
the number-fraction of particles larger than the hole. For each particle size,
Eq. ( 15 ) gives the average number that would pass before the hole is blocked
by any of the particles larger than the hole. To obtain the volume PSD of
particles passing:

Vr,=f, [PN(d)V(d)/ f2 o., PN(d)dd]dd (16)


rain

where Vp is the total volume of particles passing, drain is the diameter of the
smallest particles in the material, and V(d) is the volume of the particle of
diameter d. Equation 16 gives an absolute volume of particles passing through
a single opening of a given size, given input number-PSDs. Number-PSDs are
easily obtained from the more common mass-PSDs by dividing by density,
8 R.I. STESSEL

summing, and dividing each interval's value by the total number of particles.
Volume-PSDs and mass-PSDs are equal if the particle density is constant.
Often, when better information does not exist; for example for homogeneous
materials such as coal, the assumption is valid. With this model, a bulk den-
sity different from particle density may be specified to calculate bed
dimensions.
The total volume of a given feed that could pass through a hole is a function
of material properties. In the case of the hard, granular material discussed
above, a clearly-defined cone penetrates through the bed. The wall angles are
the angle of repose. In the case of flake or soft material, an angle of repose
could be greater than zc/2. Here, the wall angle is not so clearly related to the
angle of repose because of bridging. Thus, a cone angle is defined that differs
from the angle of repose. The cone angle can vary through n radians. For
materials likely to have steep angles of repose, the cone angle should be mea-
sured with the material on the screen. This can be accomplished by running
the trommel, stopping it, and introducing a slim ruler through an opening
from underneath. Upon measuring the depth of penetration, the void volume
could be taken as a cone, and the cone angle calculated. In the model, an
algorithm was developed for calculation of frustum volume depending upon
cone angle and depth of bed.
The total volume of material passing the bed is:
lIT ,(/ VpB if VpB< Vh
~ Vh otherwise

VpB= VB Pvdd (17)


min

l/h (EHVp if Vp< Vc


= ~EHVc otherwise
where V~ is the volume of an individual frustum, Vh is the volume of particles
calculated from passing through holes, Vpa is the total volume of particles in
the band, and Pv is the volume PSD. This sequence of equations set maxima
for the number of particles passing through the screen: first, the particles pass-
ing could not exceed the volume of the cones in the band; second, the particles
could not exceed those contained in the band as a whole.
Additionally, particle screening is calculated in the case of a striated bed,
with particle diameter increasing as a function of distance from the screen.
Originally, this was done to determine the effect of introducing airflow. How-
ever, the work of Vorstman and Tels [12] discussed above, and data from
Barton [13], showed that striation is evident in normal trommel opera-
tion .The thickness of separate layers is calculated for each input particle size
increment. A cone or frustum is defined as before, based solely upon the di-
A NEW TROMMEL MODEL 9

ameter of the trommel opening and the cone angle, defined as in Fig. 3. This
is subdivided into frustums based upon the thickness o f each sequential layer,
until a size increment is reached that is too large to pass through the trommel
opening. All those frustums pass through the screen in their entirety. The PSDs
of particles passing is determined by the volumes of the frustums containing
the different particle size increments. These frustums are illustrated in Fig. 3.
In determining the net volume and particle size distribution of particles,
these mechanisms do not operate in isolation. Each mechanism was sepa-
rately calculated and then, a parameter was included governing the fraction
of particles passing obeying striation, with the remainder acting as a mixed
bed.

Model operation
A strength o f this work was the initial decision to implement full differen-
tial equations without simplifications using numerical methods for integra-
tion. This made the c o m p u t e r essential to implementation.*
Input is divided into: general particle characteristics, particle size distribu-
tion, and trommel characteristics. Specifically, input variables are:
General particle characteristics:
• Mass feedrate;
• bulk and materials densities;

TOP OF BED.......... T-----v- ............


LAYERSOF PARTICLES .......... ................
\
LARGERTHAN OPENING /
iI

LAYERSOF PARTICLES / FRUSTUMS~_


SMALLERTHANOPENING j_ OF PARTICLES__~+
/ \
;NG~/--- PASSING __~
CONE
--SCREEN SCREENOPENING
Fig. 3. Pattern of sequential frustums in striated bed.

*The model is written in FORTRAN 77, and can be run on any computer with a suitable com-
piler. For this work, it was run on a variety of Intel-based microcomputers. Run time varies
with the number of increments in the specification of the particle size distribution and the num-
ber of rotations requested. On microcomputers, a typical run for this work would range from
about 8 hours for a PC to 2 rain for an Intel 80486 processor.
10 R.1. STESSEL

• static and dynamic coefficients o f friction, which can be set to 1 to model


the effect of flights; and
• cone angle, which became an important modeling variable since it was
not measured in any of the work reported in the literature.
Particle size distribution:
• N u m b e r of increments;
• whether the user desires to input a particle size distribution or have the
model calculate a R o s i n - R a m m l e r distribution;
• in the event the modeler wishes to input a distribution; for each particle
size increment:
• The representative particle diameter;
• either the raw mass of that component, or its percent of the total.
• In the event the modeler wishes the model to calculate the Rosin-Ra-
m m l e r distribution:
• Diameters representative o f the largest and smallest particle size
increments;
• shredder exponent;
• critical size.
T r o m m e l characteristics:
• T r o m m e l diameter;
• rotational velocity;
• hole size;
• n u m b e r of holes per unit area, which could be specified as a smaller num-
ber than the actual n u m b e r of holes in cases where effective screen area is
less than the band area;
• trommel incline;
• the fraction of the particles obeying the striation p h e n o m e n o n discussed
above.
One complete set of output is provided for every rotation m a d e by a hypo-
thetical particle spiraling down the trommel. For each revolution, output con-
tains: rotation number, cumulative axial distance travelled, total time elapsed,
mass of material remaining in the screen, cumulative mass in the unders, par-
ticle size distribution of the material remaining in the screen, particle size
distribution of the unders passing through the screen on that revolution, angle
of departure, and angles of impact for each particle increment.
The appendix provides a sample output for a very limited n u m b e r of revo-
lutions (limited for readability) with an idealized input.
This model is run to produce data on trommel performance. Variation of
the input parameters is combined with an understanding of how they affect
the underlying theory to draw conclusions.

USE O F T H E M O D E L

This model is used in two separate ways. In the first, published data from
trommel tests are used. Parameters available in the model are adjusted, and
A NEW TROMMEL MODEL 11

the model is run so as to attempt to duplicate laboratory results. In some cases,


only limited data are available concerning test conditions, requiring assump-
tions. Not only could model performance be demonstrated, but the settings
of the parameters, and the procedure used to eventually produce the desired
results, led to useful insights into trommel operation. The model was then
made to perform analyses independent of laboratory data.

Performance of the model with respect to published data


During the 1970s, detailed laboratory investigations on trommels were un-
dertaken by investigators in the US and the UK. It was a primary impetus
behind this work to be able to use the model to replicate the shapes of key
performance curves in the literature.
The most detailed data available in the comprehensive report of Glaub et
al. dealt with exhaustion of fines in the overs as a function of axial distance
travelled. Their data are graphed* with modeling results in Fig. 4. The trom-
mel involved was of 2.36 m diameter, 6.1 m long, with 0.12 m circular open-
ings covering 55% of the screen area. The trommel was reported to have em-

1.0~

"X
0.8

C
8 0.6
,g
L

0.4
"5

*5
O
L
u_ 0.2

o.o I I I I

0 1 2 3 4 5
Distonce down t r o m m e l (M)
Fig. 4. C o m p a r i s o n o f m o d e l results and data f r o m G l a u b et aL [ 6 ] for the retention o f fines as
a function o f axial distance.

*In this, as in subsequent graphs, lines connecting datapoints serve only to make the graphs
more readable.
12 R.I. STESSEL

ployed flights. To model this, the coefficients of friction were set to 1, allowing
no slippage, as would be the case with flights.
The rotational velocity used in the test was given as 2. l rad/s (20 rpm).
Calculations show the critical velocity to be approximately 2.9 rad/s ( 28 rpm ),
for which their velocity becomes 70% of the critical value, which is according
to standard design practice. When using the model, this produces cataracting,
but with an impact location less than 3/2n (see Fig. 2). It does not seem
reasonable that such operation would be allowed because it makes such poor
use of available screen area. Flights influence the mode of departure from the
screen in two ways. First, they would force the waste to remain on the screen
to a higher departure location angle than would be achieved without flights;
this is one of their key advantages, since it allows maintenance of a good cat-
aract with reduced rotational velocities. Second, at these higher departure an-
gles, the x-component of the screen velocity would be larger than it would be
without flights, causing the material to land at a smaller impact angle, making
better use of the screen. To model this, all that was needed was a slightly higher
rotational velocity: 2.8 rad/s (27 rpm).
A set of modeling runs shows the effect of varying rotational velocities. The
range of rotational velocities yielding good cataracting is quite small; for a 2
m diameter trommel, the model shows a range of approximately 0.5 rad/s
(4.8 rpm). The narrow range of suitable rotational velocities also shows why
flights are so common: it is difficult to maintain good cataracting behavior by
controlling rotational velocity. Constant adjustment of variable speed mo-
tors, possibly using electronic feedback systems, would be required.
Model results also show that flights have a great effect on reducing available
screen area under conditions of low speed and low feedrate. A low feedrate
(8.5 kg/s) is used in these tests, making the effects very apparent. Flights
transform the trommel into a quasi-batch operation: as each flight, carrying
its load, reaches a certain height, it unloads. This results in material being
dropped onto sequential, separate segments of screen below, as opposed to
continuously laying the material on the screen. Furthermore, as the flights are
used to raise material above the point that particles would ordinarily leave
the screen, the material comes to rest on the flight, not on the screen area
where it could still pass though the screen. Neither gravity nor centrifugal
reaction can then drive particles through the screen. The slower the trommel
turns, the less screen area is used for separation, even if the employment of
flights forces good cataracting. This was apparent in the modeling conducted
here, showing that only 3% of screen holes were used.
In the areas where screening did occur, the shallowness of the bed was re-
flected in the use of a shallow cone angle of 0.3 rad ( 17 o ). This means that
the cone always penetrated the bed, which was thin.
Glaub et al. expected that depletion would occur in a straight line, as Fig. 4
shows it does not. Glaub et al. sought to explain the lack of a straight deple-
A NEW TROMMEL MODEL 13

tion by doing further laboratory analysis, obtaining a particle size distribu-


tion of each recovered sample of unders. They showed, and discussed, the
increased recovery of smaller particles compared to larger within the unders
component.
For a slightly more revealing examination of a similar analysis, results from
Warren Spring Laboratory (WSL) are analyzed. The test analyzed for this
work involved a 6.6 m long trommel with 4.6 m of screen area length, and
1.53 m diameter. Their trommel used flights, which were again modelled by
using coefficients of friction equal to 1.
Material moved through their trommel by means of scrolls. These worked
imperfectly: they calculated a residence of time of 13.5 s, and measured resi-
dence times from 67 s for bottles to 80 s for large cards; no measurement was
made for fines remaining in the overs. Scrolls are modelled by adjusting the
incline parameter to achieve a residence time near 60 s, since it was suspected
that the smaller, more dense fines would be more likely to lie between the
scrolls, and thus have shorter residence times. The modelled incline was 0.3
rad (17 ° ). Again, a rotational velocity of 2.8 rad/s (27 rpm) was used to
achieve good cataracting behavior.
The feed rate was 10.5 kg/s. While this was not that much greater than that
of Glaub et al., the incline used in modeling WSL's work was 70% lower, the
bulk density was lower, and the diameter was 65% of Glaub et al.'s. This re-
sulted in a considerably larger bed depth. No reduction was required in the
number of holes available for screening.
The critical aspect of the WSL experimentation was that individual catch
baskets were placed under each quarter of the trommel screen length. The
material in the baskets was then analyzed to determine its particle size distri-
bution. Their laboratory screening resolution only resolved two particle sizes
completely contained within the trommel fines; the third-smallest particle size
interval was partly larger, and partly smaller, than the trommel openings. This
was modelled by further dividing WSL's third increment into two: one larger
and one smaller than the hole.
Figure 5 shows results obtained by WSL compared to the model output. In
both cases, only the two small particle size intervals that are smaller than the
holes are analyzed. Data are reported, not as retention in the material passing
over the screen, but as fractions of the material passing through the holes.
Further, the data are not cumulative. The critical result is that the smaller
particles first pass through the screen at a greater rate, then a lesser rate, than
the next larger particle size. This illustrates the effect that one would expect
from concentration of fines on the screen reported by Voorstman and Tels
[ 12 ]. Eventually, the fines, after being recovered more quickly, would dimin-
ish in concentration, and thus constitute a lower fraction of the screened
material.
To achieve the crossover behavior with the model, the capability to layer
14 R.I. STESSEL

0.6 i , i

c
0.5
~6
8

8
-~ O4
O

8
LL

0.3
0
,
1
,
'2
~
3
\~|4
Distance down trommel (M)
Fig. 5. Modeling results showing the particle size distribution of the unders: circles are the smaller
particles, squares are the larger; filled symbols indicate model results, and hollow indicate re-
suits from Warren Spring Laboratories [ 13 ].

particles is employed (see Fig. 3) together with adjustment of the cone angle
to modify the amount of each layer passing through the screen at each hole.
As one would expect with real waste and a reasonable bed depth, the cone
angle is obtuse 2.6 rad ( 150 ° ), implying significant bridging. Layering of 10%
of the particles produced the crossover as shown.
Hasselriis [ 14 ] compiled an extensive collection of data. Figure 6 presents
data obtained from a test of shredded wood, ranging from 0.4 m m to 4.75
mm, which is assumed for this modelling work to follow a Rosin-Rammler
distribution. These data are thus distinguished from the above tests by rep-
resenting a feed of uniform density, with little sheet material, breakage during
trommelling, or other difficulties posed by a solid waste feed. 25% of the feed
was smaller than the trommel hole openings, which were 2.36 mm. The Rosin-
Rammler parameters were adjusted accordingly.
No further information concerning trommel configuration was given. For
modeling, a radius of I m, a feedrate of 30 kg/s, and a particle density of 500
k g / m 3 (representative of pine) are chosen. No flights are assumed, so coeffi-
cients of friction less than 1 are required, and are taken from Glaub et al.:
static of 0.8; dynamic of 0.7. A rotational velocity of 3.2 rad/s (31 rpm)
produces a good cataract. It is to be noted that this is considerably higher than
A NEWTROMMELMODEL 15

1.0 ~.:~-;-

0.8

0.6
E

~8
>~
L
> 0.4
0

0.2

0.0 ~ ,
0 10 20 30

Number of r o t a t i o n s
Fig. 6. Recovery of small particles as a function of rotation; large circles are data from Hasselriis
( 1984); continuous line shows model results.

that used by Glaub et al., although their trommel diameter was similar. This
shows that slippage considerably increases the required rotational velocity.
Similar to the simulation of the WSL data, a cone angle of 2 radians ( 115 o )
was found suitable. No layering is necessary in the simulation; with this uni-
form feed, it is not a factor.
The model does a good job at following the data. In Hasselriis' original
graph, the data were plotted in semi-log fashion. This was not done here, be-
cause linear graphing allows greater visibility of the asymptotic behavior of
the exhaustion that results from an ever-diminishing content of fines. Asymp-
totic removal was also reported by Wheeler et al. [ 15 ].
Hasselriis divided the curve into three distinct parts: the constant flow, de-
creasing flow, and probability portion. He described each with distinct curve
fits. It is significant that the model produces the curve directly from the fun-
damental equations.
From these results, it can be seen that the model performed well in echoing
the mechanics of trommel operation. Some factors not clearly available from
laboratory data remain to be analyzed as below.

Analysing using the model


Once the basic ability of a predictive model to represent laboratory or field
data has been demonstrated, it may be used to test variables not easily ad-
16 R.I. STESSEL

dressed in the laboratory. Researching questions that might occur to design-


ers is shown below.
Trommel design, as with all others, hinges on achieving the most perform-
ance with the minimum equipment. A parameter had to be developed that
reflected the efficiency with which the trommel performed its screening job.
To save costs, it is important to minimize residence time, which is addressed
by two parameters: length and time. A recovery rate is defined:

r= I/pPBULK 7
IA (18)

where Vpp is the mass recovered and 1A/t is the velocity with which particles
have progressed down the trommel.
The usefulness of this parameter is shown by comparing Figs. 7 and 8, where
Fig. 8 presents the same data as Fig. 7, but using the recovery rate instead of
unders. Simply reporting the recovery of fines to the unders produces a curve
that is somewhat reverse-sigmoid. By contrast, recovery rate increased line-
arly. While increasing incline results in the obvious reduced recovery by the
end of a fixed-length trommel, it does so by a mixture of complex phenomena.
First, the residence time is reduced. Counteracting this is an increase in the
band area, resulting in decreased bed-depth. From Fig. 7, it can be concluded
that it would make no difference so long as one operated the trommel before
the curve began to tail down, at about 0.7 rad. Economically, though, the op-
erator also cares about throughput. Using the recovery rate to include
throughput shows that one recovers with increasing efficiency with increasing

140 I I

1.35

1.30

125

120 t I t
000 005 0 10 015 020

Fig. 7. The effectof inclineon the mass of unders recoveredafter 2 m of axial travel.
A NEW T R O M M E L MODEL 17

020 i i I

0.15

(9
klJ

< 010
>-
r~

f
LLI
>
o
o
m 005
O7

000 i I I

000 005 010 015 020

INCLINE (RAD~
Fig. 8. Effect o f incline on recovery rate at 2 m axial travel.

incline. The design conclusion is then different; given a target recovery, the
design decision is to buy the longest trommel that could be economically jus-
tified and give it the maximum incline. This would allow the maximum
throughput. Although the conclusion is intuitive, it is difficult to show it by
studying traditional parameters.
Rotational velocity is further examined. Within the narrow permissible
range of rotational velocities within which cataracting is maintained, the re-
covery rate at the end of 2 m of travel is calculated. Here in Fig. 9, a maximum
is clearly demonstrated. Examining model output shows that, at low rota-
tional velocities, the material departs the screen at very low angles, and re-
attaches at very great angles, leaving a small band area (see Figs. 1 and 2).
This reduces the recovery rate by increasing bed depth and reducing the screen
area available to the material. Conversely, at very high rotational velocities,
angles of departure are quite high, and angles of impact are quite low (above
the horizontal centerline of the trommel, n rad, see Fig. 2). As shown by Eq.
( 8 ), this reduces the axial distance travelled in every rotation, again increas-
ing the bed depth and reducing the screen area available to the material.
With a given waste, a designer might choose to examine several sequences
of unit operations, potentially involving screening to remove different size
fractions. The model is used to examine the effect of hole size on a feed of
fixed composition. The input distribution is identical in all cases; the hole size
is adjusted to allow passage of the first two, three, or four size increments into
which the distribution was divided. Results are shown in Fig. 10. Signifi-
cantly, the results follows the conclusions reached in examining WSL's parti-
18 R.I. S T E S S E L

0.14

0.12

(9

lJJ
<
rr 010
>-
rr
w
:>
O
O
LU 008
dE

006 I I

2.5 30 35 40

ROTATIONAL VELOCITY (RAD/8)

Fig. 9. The effect of rotational velocity on the recovery rate.

1.0 , /F- t. . . . . . .

0.8
>-
n-
w
>
O 0.6
o
z
Q
0.4
<o
rr
LL

0.2
........ EXH2
- - --EXH3
0.0 I I - - E X H 4

0 1 2 3

DISTANOE (M)

Fig. 10. The effect of size of hole on recovery as a function of axial distance travelled: EXH2,
EXH3, and EXH4 show recoveries with hole sizes increasing to recover the smallest two, three,
and four size increments, respectively, of a constant feed.

cle size d i s t r i b u t i o n s o f s c r e e n e d material. A l t h o u g h it is difficult to detect in


t h e graph, a n d p r o b a b l y n o t significant in t e r m s o f real o p e r a t i o n , t h e r e c o v -
ery rates are initially greater w i t h s m a l l e r t r o m m e l o p e n i n g s ; o f greater signif-
i c a n c e is the r e v e r s e d order w i t h i n c r e a s i n g progress d o w n t h e t r o m m e l .
A NEW TROMMEL MODEL ]9

Data from after 1.5 m o f t r o m m e l length is significant. I f a designer wishes


to consistently achieve the same high recovery rate, while changing trommel
hole openings so as to restrict recovery to smaller and smaller particles in a
given feed, an increasingly long trommel is required.
Thus, this model is able to give good insight into important design and
operating parameters.

SUMMARY AND CONCLUSIONS

A computer model has been developed to model trommel screen operation


and performance. Three components are. necessary: particle trajectory when
airborne, including discussion of drag; particle rise along the screen, incor-
porating friction; and the screening itself, using a three-dimensional cone con-
cept. New input parameters suggest measurement of bridging, and the degree
of striation of the bed according to particle size. Unique to this work is that
the model employs fundamental differential equations, using numerical
methods to solve them without the simplifications required for analytical
solution.
Use of the model to echo results obtained by laboratory investigations of
other investigators shows that the model is capable of demonstrating similar
parameter curves. Significant conclusions are that:
• Striation is an important p h e n o m e n o n in trommelling of material as het-
erogenous as solid waste, causing the finest of the unders to be exhausted
first; it is not relevant for a homogeneous feed.
• Even with flights employed to ensure good cataracting behavior, efficient
screening requires the retention of fairly high rotational velocities.
• At low velocities and low feedrates, flights can seriously reduce the access
of the material to screen area.
• Except in the case of extremely low feedrates, bridging is a significant fac-
tor. Thus, it is not sufficient to know the angle of repose; the cone angle has
to be directly measured.
Using the model alone, further results are derived after developing the con-
cept of a recovery rate. The model is able to produce intuitive results. Further
modelling shows that:
• Given a fixed target of recovery, the designer should seek the longest af-
fordable trommel, operating at the greatest angle possible to maximize
through-put.
• By contrast, variation of rotational velocity shows a clear m a x i m u m recov-
ery rate shy of either end of the range producing cataracting behavior.
• In reducing screen hole size to decrease the m a x i m u m particle size re-
covered, with a given feed and specified high recoveries, one would need
to increase trommel length.
20 R.I. STESSEL

This model shows itself to be a potentially useful tool for the design and study
of trommels.

LIST OF SYMBOLS

On the line DERIVATION


A acceleration L/t 2
B flowrate constant
CD coefficient of drag
d particle diameter L
E factor giving additional a m o u n t of screen to which
bed is exposed due to slippage during rise
f fraction of particles passing
g acceleration of gravity L/t 2
H fraction o f screen area taken up by holes
I intercept
l length L
m mass M
N n u m b e r (of particles)
P particle size distribution
r rate of recovery ML/t
R radius of the trommel L
Re Reynolds n u m b e r
t time t
T thickness of bed L
v horizontal velocity L/t
V volume L3
x horizontal location of particle L
z vertical location of particle L
o~ angular location o f a point on the screen rad
fl angle of inclination rad
0 angular location of a particle on the screen rad
# coefficient of friction
p density M/L 3
0 angle of action o f the friction force, rad
Subscripts
a pertaining to air
A axial
B bed
dynamic
D departure
~ gravity
h pertaining to screen holes
,at N E W T R O M M E L MODEL 21

i size interval
I impact
max largest particle size increment
rain smallest particle size increment
N number (pertaining to particle size distributions)
PB particles in bed
P
particle
r arc
R radial
s static
S on screen
T total (pertaining to volumes)
.v in the horizontal direction
in the vertical direction, normal to the trommel axis
0 pertaining to the smallest particles in a distribution.

REFERENCES

1 Glaub, J.C., Jones, D.B. and Savage, G.M., 1984. Preparing municipal solid waste for com-
posting. BioCycle: 35: 32-36,
2 Everett, J. and Peirce, J.J., 1985. Bound glass in shredded municipal solid waste. ASCE J.
Energy Eng., 111 : 91-94.
3 Barton, J.R., 1985. Fuel recovery from waste: A review of UK technology. Proc. Meeting
Refuse Derived Fuel--Prospects for the Industrial User, February 27, Institute of Energy
and Institute of Chemical Engineering, Newcastle upon Tyne, U.K. (un-numbererd
typescript ).
4 Vesilind, P. A. and Rimer, A.E., 1981. Unit Operations in Resource Recovery Engineering.
Prentice-Hall, Inc., Englewood Cliffs, N J, 143-150.
5 Alter, H., Gavis, J. and Renard, M.L.. Design models of trommels for resource recovery
processing. Resour. Conserv., 6: 223-240.
6 Glaub, J.C., Jones, D.B., Tleimat, J.U. and Savage, G.M., 1982. Trommel Screen Research
and Development for Applications in Resource Recovery. United States Department of
Energy, Washington DC: 190 pp.
7 Glaub, J.C., Jones, B. and Savage, G.M., 1982. The design and use of trommel screens for
processing municipal solid waste. In Proc. 1982 National Waste Processing Conference,
May 17-20,American Society of Mechanical Engineers, New York, 447-457.
8 McNabb, M.B. A new test for evaluation of air classifier performance. In: Fundamental
Aspects of Air Classifier Operation and Design, Durham, NC, Duke Environmental Cen-
ter, Durham, NC: CH. 4.
9 Stessel, R.I. and Peirce, J.J., 1987. Particle separation in pulsed airflow. ASCE J. Eng. Mech.,
113: 1594-1607.
10 Reynolds, T.D., 1982. Unit Operations and Processes in Environmental Engineering. PWS
Publishers, Boston, p. 74.
1 l Gaudin, A.M., 1939. Principles of Mineral Dressing. McGraw Hill, New York, Ch. 7.
12 Vorstman, M.A.G. and Tels, M., 1981. Some experiments in trommeling of particles in
multilayers. In: Materials and Energy from Refuse: Proc. 2nd Int. Symp., October 20-22,
Koninklijke Vlaamse Ingenieursvereniging, Antwerp: 1.27-1.34.
22 R.I. STESSEL

13 Barton, J.R., 1981. Evaluation of Trommels for Waste to Energy Plants Phase 2: Report of
the Warren Spring Laboratory Pilot Plant Test Series. Warren Spring Laboratories, Stev-
enage, Hertfordshire, U.K., 78 pp.
14 Hasselriis, F., 1984. Refuse-Derived Fuel Processing. Butterworth Publishers, Boston, Chap.
3 and 6.
15 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.

APPENDIX

Sample output

S i m p l e , s h o r t e x a m p l e run.
INMASRAT PART-DENS BULK-DENS CONE-ANGLE STAT-FRIC DYN-FRIC
40.00000 300.00000 150.00000 2.00000 .80000 .70000 ~
TROMR INCLINE ROTAVEL HOLERAD HOLFRQ VAIR EFFAIR
1.00000 .i0000 3.40000 .01250 i000.000 -i.000 .20000
ROTATION \DIAM/ 0 1 2 3
TIME .00000 1.46000 2.98000 4.49000
TRAVEL .00000 .30096 .60173 .90277
OVERS 40.00000 39.54058 39.04004 38.54552
UNDERS .00000 .45942 .95997 1.45449
ANGDEP .00000 .61616 .61731 .61100
PSDOVR 1 .01000 .25000 .24591 .24136 23678
PSDOVR 2 .02000 .25000 24828 24635 24435
PSDOVR 3 .03000 .25000 25290 25615 25943
PSDOVR 4 .04000 .25000 25290 25615 25943
PSDUND 1 .01000 .00000 60237 60044 59835
PSDUND 2 .02000 .00000 39763 39956 40165
PSDUND 3 .03000 .00000 00000 00000 00000
PSDUND 4 .04000 .00000 00000 00000 00000
ANGIMP 1 .01000 -1.57080 -i 8 0 4 5 4 -1.78622 -I 7 9 9 2 4
ANGIMP 2 .02000 -1.57080 -i 8 1 0 2 9 -1.78927 -i 8 0 5 1 0
ANGIMP 3 .03000 -1.57080 -i 8 1 1 8 9 -1.78941 -i 8 0 6 6 4
ANGIMP 4 .04000 -1.57080 -i 8 1 2 4 6 -1.78987 -i 8 0 7 3 1

You might also like