Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Journal of the Air & Waste Management Association

ISSN: 1096-2247 (Print) 2162-2906 (Online) Journal homepage: https://www.tandfonline.com/loi/uawm20

Laboratory Investigation of a New Trommel Model

Richard Ian Stesscl & Keith Cole

To cite this article: Richard Ian Stesscl & Keith Cole (1996) Laboratory Investigation of a New
Trommel Model, Journal of the Air & Waste Management Association, 46:6, 558-568, DOI:
10.1080/10473289.1996.10467491

To link to this article: https://doi.org/10.1080/10473289.1996.10467491

Published online: 09 Jan 2012.

Submit your article to this journal

Article views: 2041

View related articles

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=uawm20
TECHNICAL PAPER ISSN 1047-3289 /. Air & Waste Manage. Assoc. 46: 558-568
Copyright 1996 Air & Waste Management Association

Laboratory Investigation of a New Trommel Model


Richard Ian Stesscl
University of South Florida, Tampa, Florida

Keith Cole
Groundwater Technology, Marietta, Georgia

ABSTRACT Implementation of trommels within the waste industry


Rotary screens, or trommels, are an important unit operation began at a small refuse plant in Glasgow, Scotland in 1928.1
in material and fuel processing. A computer model has been However, there is very little information concerning this
developed based upon fundamental mechanics. The coeffi- trommel. The dearth of information about trommel use con-
cients and variables employed in the model thus have real tinued until 1966 when the Greater London Council in-
physical meaning; adjusting them based upon laboratory data stalled a trommel at a Barking, England refuse plant.2 The
allows the model user to draw conclusions about the behav- first testing, analysis, and evaluation of a trommel for the
ior of the trommel that can be applied to design and opera- sizing of wastes was conducted at the Barking facility. It was
tional changes. The laboratory testing program was specifically not until 1974 that the United States began a test program
designed to test the model described in this paper. aimed at optimizing the design of trommel performance
The model proved able to track the laboratory data. Impor- for use with municipal solid waste (MSW). Unfortunately,
tant phenomena that were validated or revealed included the due to lack of funding, the work went largely unfinished.
significance of particle layering, changes in bed sliding with This paper presents a laboratory evaluation of a computer
rotational velocity, and the pre-eminence of residence time. model of a trommel developed in the last few years. The
exercise of matching the laboratory results to the computer
INTRODUCTION output both confirmed the model's applicability and pro-
Trommels, or rotary screens, are an important unit opera- vided insight into the trommel design and operation. A brief
tion in materials processing. They have a long history in theoretical background is presented first, familiarizing the
the mineral dressing industry, and are of particular interest reader with the model's characteristics and uses that have
to those concerned with solid waste processing because of already been made of the model. After presenting results of
the difficulty of the material processing task. Trommels pro- the experimentation and computer analysis, conclusions are
vide an elegant screening option with low maintenance and drawn concerning trommel operating parameters.
operating costs that could often justify a higher initial cost.
A trommel is a rotating, cylindrical screen, lying on its BACKGROUND
side at a small angle from the horizontal plane. Material is An earlier era of solid waste research permitted substantial
fed at its elevated end, and size-separation occurs as the ma- research efforts in trommeling. Large laboratory studies were
terial spirals down the drum at an axial speed governed by conducted, some using research trommels that allowed con-
the angle of the drum to the horizontal plane. Behavior of siderable variation in operating parameters, but were nearly
the material normal to the drum axis is governed by the full-sized. In conjunction with laboratory testing, theory was
rotational velocity. developed and refined. Attempts were made at analytical
solutions for the underlying equations. Significant simplifi-
cations were necessary to allow any use at all of the equa-
tions; computational resources were not available to use
IMPLICATIONS equations in their differential forms.
Trommels play a key role in solid waste processing, both Most of the past analyses of trommel behavior were geo-
for RDF preparation and for removal of undesirable com- metrically oriented. Alter et al.3 focused principally on de-
ponents in the feed of mass-burn facilities. Trommel
parture location and an assumed location angle of impact.
theory may also find application in rotary-drum com-
bustors. Test results showed that the model developed The actual trajectory of the particle in flight was of little
in previous work was useful in understanding trommel significance to their analysis. Glaub, Jones, Tleimat, and Sav-
operation, and could serve as an aid to trommel de- age,4 and Glaub, Jones, and Savage5 added more factors to
sign. The addition of airflow, producing the air trommel, Alter's paper. None of these approaches fully included the
also showed potential.
effects of drag. Glaub, Jones, Tleimat, and Savage4 began a

5S8 Journal of the Air & Waste Management Association Volume 46 June 1996
Stessel and Cole

discussion, but left the drag force coefficient unspecified. In respectively, are known, and the increment of axial move-
the case of small particles, when, for example, minerals are ment during one turn in the particle's spiral path down the
pulverized—or in the case of light particles, such as paper or drum is:
very small particles in solid waste—the drag on a particle
becomes a serious component in the force balance. With
(2)
trommel diameters of lm to 2 m, it is quite possible that the
aerodynamically lighter particles will achieve terminal ve-
locity in flight.6 where the variables are as defined in Figure 1. The incre-
The development of the present model is described in an ment of axial movement is better visualized by renaming it
earlier paper.7 There are three major components of the "bed width." If a given particle is taken to represent the
trommel: the particle rise along the turning screen; passage leading edge of the bed, its progress down the trommel one
of certain particles through the screen; and the particle tra- revolution later defines the bed width. In addition to its
jectories after they leave the screen during their cataracting importance in calculating overall trommel length, the bed
motion. The right side of Figure 1 shows the trommel's ro- width is required to determine the thickness of the bed.
tation, the movement of the bed up the side of the trommel, The rise of particles in contact with the rotating screen
and the trajectory of the particles. The final results devel- gives the departure location of the particles and the area of
oped in Stessel's first paper7 are summarized below. the screen occupied by the material. The contact area on
Particle trajectories come from a pair of equations bal- the screen occupied by particles in one pass is defined as the
ancing drag and gravity and neglecting buoyancy.8 Solving bed, as shown in Figure 1. The number of bed-widths re-
for acceleration in the horizontal and vertical directions, quired to allow the feed to spiral down the trommel yields
respectively, these are: the design length of the device. Previous efforts have not
succeeded in accounting for slippage of the particle against
the screen.3S Combining accelerations in the radial and tan-
gential directions yields the following expression for accel-
eration of a particle against the screen:
(1)

e=jue2- 4fju sine + cosej (3)


where FD is the drag force, m is particle mass, and g is the
acceleration of gravity. where n is the coefficient of friction, which can be either
The movement down the trommel depends on both the static or dynamic. Whether the particle is slipping is deter-
axial component of the vertical departure velocity, and the mined by checking the force balance between the tangen-
angle of declination of the trommel. The z-direction shown tial gravity force and static friction.
in Figure 1 is tilted by the declination angle, p. The vertical While the material is moving on the screen, it is being
coordinates of the point of departure and impact, zD and zv separated. The model incorporates the screening of particles
in conjunction with their rise up the screen, in the bed. An
average number of particles passing through a hole is calcu-
lated by knowing the number-concentration of particles both
larger and smaller than the hole. To obtain the volume par-
ticle size distribution (PSD) of particles passing:

dh PN(d)V(d)
V _/ d dd
~ min (4)

where Vp is the total volume of particles passing, dmin is the


diameter of the smallest particle in the material and dmax is
the diameter representative of the largest, PN is the number
PSD, dh is the diameter of the trommel openings, and V(d)
is the volume of the particle of diameter d. The absolute vol-
ume of particles passing through a single opening of a given
Figure 1 . Sketches of the trommel; the left figures are definition
sketches viewed from the discharge end; the right are perspectives;
size, given input number-PSDs is given by Eq. 4. Number-PSDs
the bottom define the bed. are easily obtained from the more common mass-PSDs.

Volume 46 June 1996 Journal of the Air & Waste Management Association 589
Stessel and Cole

Material that has passed through the bed leaves a cone the "overs," and the particles smaller than the screen openings
or cone segment (frustum) as a void. The wall angles of the are hereafter termed the "unders.") Input was divided into
cone are a function of the material properties. In the model, general particle characteristics, particle size distribution, and
an algorithm was developed for calculation of frustum vol- trommel characteristics. Specifically, input variables include
ume depending upon cone angle and depth of bed. The the following. General particle characteristics include mass
total volume of material passing through the bed is: feedrate; bulk and material densities; static and dynamic
coefficients of friction, which can be set to 1 to model the
effect of flights; and cone angle, which became an impor-
V= IV V
ifV<V V
tant modeling variable, as it was not measured in any of the
T 1 PB * PB h
V_ otherwise work reported in the literature. For particle size distribution,
variables include the number of increments and whether
the user desires to input a particle size distribution or have
(5)
the model calculate a Rosin-Rammler distribution. In the
Vh= EHVrifVp<V event the modeler wishes to input a distribution, for each
EHV otherwise particle size increment, the variables of concern are the rep-
resentative particle diameter and either the raw mass of that
where Vc is the volume of an individual frustum, Vh is the
component, or its percent of the total. In the event the mod-
volume of particles calculated from passing through holes,
eler wishes the model to calculate the Rosin-Rammler dis-
VPB is the total volume of particles in the bed, and Pv is the
tribution, the variables to consider are the diameters
volume PSD. This sequence of equations sets maxima for
representative of the largest and smallest particle size incre-
the number of particles passing through the screen: first,
ments; the shredder exponent; and critical size. Trommel
the particles passing could not exceed the volume of the
characteristics include trommel diameter; rotational veloc-
cones in the bed; second, the particles could not exceed those
ity; hole size; number of holes per unit area (which could be
contained in the bed as a whole.
specified as a smaller number than the actual number of
Additionally, particle screening is calculated in the case
holes, in cases where effective screen area is less than the
of a striated bed, with particle diameter increasing as a func-
bed area); trommel incline; and the fraction of the particles
tion of distance from the screen. Originally, striation was
obeying the striation phenomenon discussed above.
included to determine the effect of introducing airflow. How-
The model gives output for each revolution of the
ever, the work of Voorstman and Tels,9 and data from
trommel. This output includes: rotation number; cumula-
Barton,10 showed that striation is evident in normal trommel
tive axial distance travelled; total time elapsed; mass of ma-
operation. The thickness of separate layers is calculated for
terial remaining in the screen; cumulative mass in the unders;
each input particle size increment.
particle size distribution of the material remaining in the

The Air Trommel


It was further desired to investigate the potential of an ac-
tual air trommel. Given equivalent densities, particles that
are very small tend to have larger impact angles (see Figure
1) because drag has a greater effect on smaller particles; this
phenomenon was first recognized as significant in work on
air classifiers. This phenomenon could be used to improve
trommeling by introducing airflow at the point of depar-
ture from the screen (ocD), with the air moving in the -x
direction. This would project fines further across the
trommel. With judicious selection of air velocities, small,
light particles that would have fallen on top of larger, heavier
particles with a higher trajectory could be induced to have a
still higher trajectory. Thus, for better separation, the small
particles would be layered closest to the screen.

Modeling DISTANCE DOWN TROMMEL (M)


The theory of the rise of a particle on the screen, particle
trajectory, and screening of undersized particles were inte-
Figure 2. Unders PSDs from the model (filled symbols) and WSL
grated into a computer model written in Fortran. (Note: par- (hollow symbols): circles are smaller particles, and squares are larger
ticles larger than the screen opening are hereafter termed particles.

5 6 0 Journal of the Air & Waste Management Association Volume 46 June 1996
Stessel and Cole

Tablet. variables assessed in the particle dynamics and screening tests.

Particle Dynamics Screening


Test Repetitions Rotation (rad/s) Airflow (m/s) Test Repetitions Rotation (rad/s) Airflow (m/s) Bed width (m)

A 5 5.23 0 A 5.23 14 0.508


B 5.23 14 0.635
B 5 5.02 0 C 5.23 14 0.762
D 5.23 9.4 0.508
C 3 4.81 0 E 5.23 9.4 0.635
F 5.23 9.4 0.762
D 4 5.23 9.4 G 2 5.23 0 0.508
H 2 5.23 0 0.635
E 5 5.02 9.4 I 2 5.23 0 0.762
J 5.02 14 0.508
F 6 4.81 9.4 K 5.02 14 0.635
L 5.02 14 0.762
G 4 5.23 14 M 5.02 9.4 0.508
N 5.02 9.4 0.635
H 4 5.02 14 0 5.02 9.4 0.762
P 5.02 0 0.508
1 2 4.81 14 Q 5.02 0 0.635
R 5.02 0 0.762
S 4.81 14 0.508
T 4.81 14 0.635
U 4.81 14 0.762
V 4.81 9.4 0.508
W 4.81 9.4 0.635
X 4.81 9.4 0.762
Y 4.81 0 0.508
Z 4.81 0 0.635
AA 4.81 0 0.762

screen; particle size distribution of the unders passing Experiment Variables and Parameters
through the screen on that revolution; angle of departure; Aspects of trommel operation that are typically varied in
and angles of impact for each particle increment. the field include rotational velocity and inclination. Rota-
This model was used to analyze published data. As an tional velocity would influence angles of departure and im-
example, early researchers expected depletion of unders in pact (see Figure 1), which would also change the bed width
the bed to occur in a linear fashion. Depletion, however, (see Eq. 2). Inclination ((3) would also influence bed width.
proved not to be linear; subsequent researchers gathered Bed width, in turn, would control trommel length: a given
additional data to investigate non-linear depletion. Results number of rotations, with varying bed widths, would pro-
from Warren Spring Laboratory (WSL) are analyzed.12 The duce different length trommels; conversely, the same length
results obtained by WSL compared to the model output are trommel would rotate its feed a different number of times
shown in Figure 2. In both cases, only the two particle size depending upon inclination.
intervals that are smaller than the holes are analyzed. The From its inception, the model was designed to include
critical result is that the smaller particles first pass through airflow, thus incorporating the concept of the air trommel.
the screen at a greater rate, then a lesser rate, than the next Airflow proved useful in modeling trommel studies reported
larger particle size. Initial screening of the smaller of the in the literature because one of its key effects is to layer the
undersized particles illustrates the phenomenon of bed stria- particles, as discussed above. With a very low airflow, little
tion reported by Voorstman and Tels.9 actual change occurs in the particles' trajectories, but the
In comparison with data published during the era of re- layering function in the model is implemented.
search in resource recovery, this model proved that it could Examination of these variables led to the recognition that
be used to represent test results accurately. the particle dynamics and their effect on departure and im-
pact angles had to be tested first, involving only the study
MATERIALS AND METHODS of rotational velocity and airflow. This work is shown on
Before any experimentation could take place, key output pa- the lefthand side of Table 1; the number of repetitions of
rameters had to be determined to measure trommel performance. each test is shown in the last column of that part of the

Volume 46 June 1996 Journal of the Air & Waste Management Association §61
Stessel and Cole

table. A factorial design was applied to the experiments, of- materials were evaluated, including plastic spheres, wood
fering a range from three values: low, medium, and high. cubes, paperboard cubes, and charcoal briquettes. Based
The factorial test required each variable to be tested at each upon observation of full-sized units, it appeared that the
of the values; hence the number of particle dynamics tests small holes in the laboratory trommel greatly limited the
were 3 2 or 9 tests. selection of particles that would exhibit reasonable trajec-
Screening tests involved sampling the trommel unders tories. Use of low-density materials such as plastic and card-
at intervals of five rotations. A PSD had to be determined board resulted in unders that were so small they lofted upon
for each such sample. Screening tests were thus much more leaving the screen; their fall was visibly unreasonably slow.
involved than dynamics tests. Still another variable, inclina- In solid waste practice, a trommel might be used to remove
tion, was involved. Thus, a greater number of tests was re- glass, ceramics, gravel, and dirt from less brittle materials.
quired. These are shown on the righthand side of Table 1. Thus, gravel screening represented a use of trommels for
Due to the time involved, only three of these tests were re- waste processing. Gravel particles can be described as having
peated; these are shown in the first column of that part of sphericity, like sand, allowing application of the spherical-
the table. The factorial design gave 3 3 or 27 tests. The tests shape assumption in the trajectory component of the model
were run in random order to minimize theinteraction of ex- without resorting to laboratory measurement of equivalent
perimental conditions, such as changes in volunteer helpers. diameters. The model verification undertaken made constant
density important; thus, the particles larger than the screen
Equipment Selection opening, the "overs/' had to be made of the same material as
The equipment used for the experiment consisted mainly of the the particles smaller than the screen openings, the
that in the Solid Waste Processing Laboratory. The trommel "unders." The use of gravel allows application of these re-
was handbuilt, employing a carbon steel industrial dryer drum. sults to mineral dressing and combustor ash processing.
The drum was 1.08 m long and 1.13 m in diameter. The screen The input particle size distribution was determined based
had 0.8 cm diameter apertures spaced in a staggered array to upon the test trommel's screen aperture size and experi-
yield 5320 holes per m2. Larger holes and a longer screen drum ment simplification. It was decided that four particle sizes
would have been preferable, but as this research was not funded, would be used: two unders and two overs. The trommel
custom fabrication was not possible. An aluminum housing had 0.8 cm (5/16") diameter apertures; therefore, particle
surrounded the drum to prevent particles from being projected sizes of < 0.476 cm (3/16"), 0.476 cm to 0.8 cm, 0.8 cm to
away from the screen. An aluminum chute was used under 1.27 cm, (1/2"), and >1.27 cm were chosen. For simplicity,
the drum to collect the unders for analysis. The trommel was the size denoting the difference between the two unders
powered by a variable-speed Dayton 0.5 kW motor coupled to particles, 0.476 cm, will be referred to as 0.5 cm henceforth.
a 1.75 gear reducer. For the screening tests, the trommel block, An arbitrary mass composition of 25% of each size was de-
consisting of a plywood circle, was created that fixed the length cided upon for simplicity.
at one bed width. Particle density was determined using fluid displacement
All particles were sieved and separated in a automatic to determine volume. This value was determined to be
Gilson TS-1 Sieve Shaker. The shaker included three shaker 2456.0 kg/m3. Bulk density was determined by filling a one-
screens: a 4.75 mm (#4) screen, an 8 mm screen, and a 12.5 liter container with gravel and weighing it. The average bulk
mm screen, allowing for separation of the particles into four density of the particle size distribution was determined to
size distributions. During each screen test, particles were re- be 1687 kg/m3.
sieved into their respective particle sizes using a hand oper- The particle's coefficients of static and dynamic friction
ated Soil Test Sieve Shaker with the same screen sizes. were measured using an inclined surface of the same mate-
Airflow was produced by two Dayton 0.7 kW blowers. rial as the trommel screen (perforated steel) to simulate the
These were connected in parallel and valved to control ve- trommel's surface. However, it was theorized that, as par-
locity. Airflow was measured with a Dwyer pitot/manom- ticles moved to greater elevations on the screen, the coeffi-
eter-type velometer. cient of friction would drop as particles ceased to protrude
through screen holes. To account for this situation, coeffi-
Particles cients of friction were also measured against solid steel. The
The determination of a suitable type of particle involved static coefficient of friction was 0.66 against perforated steel
meeting certain criteria that would allow for adequate com- and 0.465 against solid steel. The kinetic coefficient of fric-
parison with the analytical model and be easy enough to tion was 0.61 and 0.463 for perforated and nonperforated
work with in the laboratory. The particle had to have con- steel respectively. Coefficients of static and kinetic friction
sistent density, not be too slippery, exhibit little rebound were also measured for particle-to-particle contact using an
upon impact; be close to spherical, but not so close as to inclined surface to which gravel was glued. Values
roll on the screen; be available in different sizes; and be determined in this way were 1.35 and 0.61, respectively,
easily obtainable and inexpensive. Many different for the static and kinetic coefficients of friction.

5 6 2 Journal of the Air & Waste Management Association Volume 46 June 1996
Stessel and Cole

Trommel Operation whereupon the camera shutter was opened. The particle
Due to the trommel's small size and the method selected for made approximately 1.5 revolutions, or about 2 trajecto-
analyzing screening, batch testing was used. Batch testing did ries, before the shutter was closed. After film development,
not hinder the correlation with the analytical model, since the the particle's path was traced on each of the photographs.
model calculations followed the fate of a bed moving down From these tracings, departure and impingement angles were
the trommel. Using a feed rate of 18 tonnes/hour, assuming determined.
two seconds to deposit the pail of feed into the trommel, a In screen testing, five square 10-liter containers were at-
batch of 10 kg was chosen. The total feed was arbitrarily di- tached end-to-end, and the first was placed below the
vided equally, by mass, among the four particle sizes. trommel's unders chute. The trommel's length was set to
Proper particle motion in a trommel is usually called the predetermined position using the trommel block. The
cataracting, which refers to motion in which a trajectory can trommel speed was set to the desired rate and turned on.
be observed. Preliminary testing showed that a small range The 10 kg feed was placed into the trommel and the revolu-
of rotational speeds allowed particles to exhibit the entire tions were counted. At intervals of five revolutions, the
range of cataracting behavior. This range was between 46 "train" of 10-liter containers was pulled so as to locate a
revolutions per minute (rpm) and 50 rpm. Therefore, it was fresh container beneath the unders chute. After five con-
decided to use three rotational speeds for the experiment. tainers were filled, the trommel was turned off. A mass PSD
The low speed was 46 rpm, the medium speed was 48 rpm, was determined for each of the unders fractions, as well as
and the high speed was 50 rpm. for the overs and particles that escaped.
Inclination was the next parameter used as an operating Following the completion of the laboratory work, mod-
variable. Due to the shortness of the laboratory trommel, eling was undertaken. Model parameters were adjusted so
this parameter could not be applied directly to the trommel as to best fit the laboratory data.
as could rotation. As the direct effect of inclination is on
bed width, the trommel was used to evaluate sequential ro- RESULTS
tations of a single batch feed, but fixed in one position. Since Results of this work consisted not only of the laboratory data,
no material was being fed after the test batch, fixing the but also of the results of the modelling. The model parameter
trommel length at one bed-width approximated movement adjustment necessary to produce the best available replica-
of a given bed down the length of the trommel. Limiting tion of the laboratory data was important information for
trommel length to a single bed-width was accomplished by analysis. Results from particle dynamics, screening, and data
physically closing off the output of the trommel with a move- quality evaluations are presented below.
able disk, and leaving the machine horizontal. A combina-
tion of limitations of the trommel and the ability to feed it, PARTICLE DYNAMICS
with a desire to evaluate a wide range of inclinations, led to The dynamics testing measured departure and impingement
three values of bed width to use in the experiment: 0.508 m angles. The particles' coefficients of friction and the amount
(20"), 0.635 m (25"), and 0.762 m (30"). of drag through the trajectory were determined to be the
The final parameter used as a testing variable was air- key variables.
flow. Variations involved using no blower, using one blower, Results with no added airflow are shown in Figure 3. A
or using two blowers in parallel. Airflow was measured in- good fit was obtained. Departure angle correlation required
side the trommel with a velometer and pitot tube. The aver- manipulation of both static and dynamic coefficients of fric-
aged low, medium, and high values for airflow were 0 m/s, tion in the model. The adjustments made for the coefficient
9.4 m/s, and 14 m/s, respectively. of static friction did not seem to influence the departure
angle to a great degree. Dynamic coefficients of friction had
Testing Procedure to be reduced to match higher rotational velocity results.
The particle dynamics testing utilized a fluorescent-painted Lower effective dynamic coefficients of friction may have
rock from the median (0.8 cm to 1.27 cm) size fraction that resulted from particles spending a greater fraction of the
would not fall through the trommel's apertures. A black cur- climb with gravity working to push the particles away from
tain was placed behind the trommel to offset the fluores- the top half of the trommel; thus, the coefficient of friction
cent-painted rock. Crosshatches were drawn on the drop more closely approximated that of a smooth (rather than a
cloth at the horizontal (0 rad) and vertical (n/2 rad) axes to perforated) plate. Even though the particles moved with a
provide a way to measure particle departure and impinge- greater angular velocity, that velocity was not as great as it
ment angles in the photographs. A strobe light was placed would have been had the particles been more enmeshed in
behind the camera, and was set approximately 4.0 ft from the trommel's holes.
the discharge end of the trommel. After allowing the Further confirmation of lower departure velocities was
trommel to reach the desired test speed, the particle was provided by the need to use airflow, even when not testing
dropped in and allowed to make one complete revolution, the air trommel, to get the model to mimic particle

Volume 46 June 1896 Journal of the Air & Waste Management Association 56.3
Stessel and Cole

1.2

1.1

| 1.0

fj 0.9

~ 0.8

0.7
o Lab.: Total
0.6 A Model: <5 mm
& Laboratory • Model: > 5 mm
o Model A Lab.: < 5 mm
0.5
4.8 4.9 5.0 5.1 5.2 5.3 o Lab.: > 5 mm
4.8 4.9 5.0 5.1 5.2 5.3
Rotational Velocity (rad/s)
Rotational Velocity (rad/s)

T Figure 4. Recovery of components of the unders stream as a func-


-1.9
tion of rotational velocity comparing laboratory and modeling results.

Q -2.0
experimental results. Specifically, the particle's cone angle
2
(related to angle of repose) was increased from 2.46 to 2.56.
I "21 Also, the number of screen holes per area had to be reduced
•5 -2.2
from the actual measured amount of 5820 holes/m2 to 3300
< -2.3 - > holes/m2. The airflow effectiveness parameter was left at the
-2.4 - same value of 100%.
A Laboratory After understanding how the model reacted to changing
o Model
4.8 4.9 5.0 5.1 5.2 5.3 parameters, the following conclusions could be made. The
Rotational Velocity (rad/s) increase in cone angle served to actually increase the differ-
ence in recoveries between the two particle sizes. Hole fre-
Figure 3. Angle of departure (top) and angle of impact (bottom) as quency, on the other hand, was reduced to decrease the
a function of rotational velocity. total recovery fraction. This type of correlation allowed for
a near perfect match to the experimental data. The effect of
impingement angles. Airflow was added in the positive di- reducing the hole frequency in the model may show that
rection, so as to give a larger angle of impingement, with the entire trommel screen area was not being used during
higher rotational velocities. Thus, by the time the particle the experimentation. The trommel bed width available dur-
actually left the screen, the model then indicated a lower ing the test (0.635 m) may have been limited by the lack of
horizontal velocity component, in the negative direction. any inclination that would have spread the particle bed out
Higher departure velocities may also have been due to the to a greater degree and effectively utilized more screen area.
difficulty in representing jagged particles, such as these rocks, Also, observations showed that even though particles were
with idealized spheres. Perhaps another stage of the model in contact with the screen up to their departure point, screen-
development could provide the modeler more options in ing stopped considerably shy of it. Clearly, the net force
drag coefficient expressions. holding the particle against the screen was reduced as grav-
The addition of airflow, simulating the air trommel, ity acted to push particles away from the screen above the
showed a flaw in the experimental trommel design: the air trommel midpoint. A lower effective screening area further
pipe was located too low, near the 0 point in Figure 1. Turn- reinforced the conclusions discussed above that reduction
ing on the air actually reduced the departure angle. The con- of the coefficient of friction with higher rotational veloci-
cept would call for location of the air nozzles only slightly ties promote extended particle residence against the screen
below the point of departure, thus, by definition, not greatly near and above the trommel midpoint. As previously dis-
affecting it. Nevertheless, the impact angle was shown to cussed, the air effectiveness was left unchanged; therefore,
decrease, and by a greater amount than the departure angle no adjustment was needed.
was reduced. Thus, it was still hoped that results from screen- Recovery of the unders components was analyzed for dif-
ing tests would be of interest. ferent rotational velocities, as described in Table 1. Results
for the midpoint of the test, at a rotation number of 15, are
Screening presented in Figure 4. The overall result showed that a lower
After adjusting some of the model's input variables, how- rotational velocity, given cataracting behavior, was prob-
ever, the resultant data fit almost perfectly with the ably the most desirable, although the middle velocity showed

5 6 4 Journal of the Air & Waste Management Association Volume 46 June 1996
Stessel and Cole

an improved recovery of the larger particles. Matching the sufficient depth across the entire bed, and being concentrated
laboratory data with the model required increasing cone angle instead in the middle or at one side. Variations in bed depth
with velocity, indicating increased bridging. Layering was re- were very hard to determine by observation, as the interior of
duced and effective hole frequency increased, which would the trommel was a very confused place during operation.
indicate that presentation of the fines to the screen suffered It remained to evaluate the air trommel. Only the labora-
with higher velocities, although the trajectories achieved by tory results are presented; the model's ability to replicate
the higher velocities did expose them to more of the screen; these data had already been determined. Overall recovery
these factors counteract each other somewhat, and more in- as the bed moved down the trommel is shown in Figure 7.
vestigation would be required to ascertain their relationship. Of the greatest significance is that there was an optimum
To further analyze recovery, the two size fractions in the airflow, neither the maximum nor the minimum. It is also
unders were analyzed as a function of rotation number. These interesting that the improvement was most noted prior to
results are shown in Figure 5. The key result is how the data the 15th revolution. Separation enhancement in the early
replicated the results from Warren Springs Laboratories, stages of screening was related to the conclusion reached in
shown in Figure 2. The first rotation is somewhat at vari- conjunction with Figure 5: airflow helped by furnishing ad-
ance, but that may be due to the batch feed conditions. The ditional layering, speeding up the depletion of the small
model results shown resulted from a layering effect of 0.7 unders, which occurs first. Recovery, as a function of air ve-
and a high cone angle of 2.94, indicating significant layer- locity, attained by the 15th revolution, is shown in Figure 8.
ing and considerable bridging. Hole frequency was fixed at In Figure 7, not much difference can be observed between
the trommel's value. The crossover point could be delayed the air velocities. For total unders, Figure 7 also shows mini-
by increasing cone angle and/or decreasing layering. mal difference. The two unders components acted quite dif-
Tests of increased inclination were accomplished by mov- ferently, however, to produce this effect. The intermediate
ing the trommel block, as discussed above. Thus, the pri- airflow benefited recovery of the small unders, and hurt re-
mary effect of increasing inclination—increasing the bed covery of the large unders. Thus, particles even larger than
width—was directly tested. In this case, the number of revo- the large unders (i.e., the overs) would be even more hin-
lutions at which the analysis was undertaken had to be dif- dered in reaching the screen, actually aiding recovery of all
ferent for each bed width, so as to model a fixed-length the unders. The highest airflow aided the large unders be-
trommel. The laboratory and fitted model results for the cause they were lofted nearly as high as the small unders,
variation of incline are shown in Figure 6. As the bed width actually minimizing the difference between the two cat-
increased, corresponding recoveries decreased. Thus, the egories, given that most of the small unders had already been
trommel operator should seek the minimum inclination, removed by the 15th rotation, as shown in Figure 5. Thus,
and the longest residence time. Correlation with the model modest airflow does aid separation, as the model indicates.12
results required only, a reduction in hole frequency with
greater bed width. Thus, the wider the actual particle bed, Statistical Evaluation
the less amount of screen seemed to be available to the par- The statistical analysis for experimental particle dynamics
ticles. The reduction in effective screen area may have been testing was fairly straightforward. For each test, multiple runs
due simply to the particles' not covering the wider beds at were performed, as indicated in Table 1. The standard error

0.8

• Model: > 5mm


• Model < 5 mm
D Lab.: > 5 mm • Calculated
O Lab.: < 6 mm 0.4 o Experimental
10 15 0.5 0.6 0.7 0.8
Rotation Number Bedwidth (m)

Figure 5. Recovery fraction of small and large unders as a func- Figure 6. Recovery fraction as a function of bed width comparing
tion of rotation number showing crossover phenomenon. experimental and modeling results.

Volume 46 June 1998 Journal of the Air & Waste Management Association 565
Stessel and Cole

1.0
1.0

0.8

o 0.6

0.4

02
--- 14 m/s D Total Unders
9.4 m/s A Unders > 5 mm
0.0 0 m/s o Unders < 5 mm
0.0
5 10 15 20 25 5 10 15
Rotation Number Air Velocity (m/s)

Figure 7. Recovery as a function of rotation number for varying Figure 8. Recovery for different unders components as a function
airflows. of air velocity.

of the mean was also computed. These values are represented each test, replication of 10% of the tests was performed, as
in the error bars in Figure 3. In both cases, the error is quite shown in Table 1. Rotational speed was chosen as the repli-
high, but the first and last data points begin to show some cation variable because it was thought that it would have
significant separation. To validate the assumption that the greatest impact upon results. The results of the entire
changing variables did affect the resultant departure and replication set showed that the differences were significant
impingement angles, analysis of variance (ANOVA) was tabu- at the 90% level.
lated to determine if each tests' treatment means exhibited An analysis was performed on the experimental data to
significant differences. The conclusions were the same for determine the magnitude of the effect each variable had
each of the analyses: increasing rotation or airflow did not upon the final recovery output. The measurement of effects
significantly change the particle departure or impingement and interactions was calculated using a Pearson Correlation
angle. Again, given that the experimental conditions in- Matrix and compiled using a statistical analysis computer
volved a rock bouncing inside a spinning drum, these re- package.13 Results are shown at the top of Table 2. This
sults were expected. analysis allowed a probability to be calculated using a Bartlett
For each screen test, a mass balance was undertaken to Chi-Square statistic, which provided a measurement as to
account for error. In such testing, it is common that a sig- how strong a correlation the particular variable(s) had upon
nificant fraction of the feed is lost, due to being trapped in the results. The second matrix calculated a set of probabili-
nooks and crannies in the equipment, etc. A bar graph de- ties based upon the number of observations made to deter-
picting the results of the error analysis for all the screening mine how much confidence there was in the calculated
tests is shown in Figure 9. "Accounted-for" losses consisted correlations (first matrix).14 None of the correlations were
of the results of cleaning the floor and nearby surfaces after
each experiment. "Unaccounted-for" errors comprised that
fraction that could not be found. The percent error of par- 0.3
ticles that were unaccounted for ranges from 0.04% to 5.20%,
with only 8 out of 27 tests showing error greater than 2%.
These low errors showed that there were fairly good con- in I | i '•
trols on keeping account of all particles. The percent error 0.2 - i 111 B I
; j | ' | j 1 |
of particles not reaching the unders or overs fraction ranged ilj
c
n '• * ' ; * j '
ractio

i _ i „
from 7.15% to 19.35%, with only 6 out of 27 tests having ' i ' !

greater than 16% error. Though these percentages are not as 0.1 • -
small as the unaccounted-for error, they are fairly consis-
tent and within range of experimental error, particularly for
H Unaccounted-for
this type of testing. • Accounted-for
0.0
Experimental screening analyses provided an enormous
Test
database. However, as each specific test result of the model
comprised a sample size of one, few statistical tests lent them-
selves to proper application. To measure the precision of Figure 9. Errors in laboratory screening tests.

566 Journal of the Air & Waste Management Association Volume 46 June 1996
Stessel and Cole

Table 2. Recovery correlation matrices for the screening tests.

Pearson correlation matrix

Rotation Air flow Bed width Recovery <5mm Recovery > 5 mm Total recovery

Rotation 1.000
Air flow 0.000 1.000 — — — —
Bed width 0.000 -0.034 1.000 — — —
Recovery < 5 mm -0.209 0.014 -0.287 1.000 — —
Recovery > 5 mm 0.055 0.184 -0.245 -0.404 1.000 —
Total recovery 0.096 0.201 -0.465 0.314 0.742 1.000

Matrix of probabilities
Rotation Air flow Bed width Recovery <5mm Recovery > 5 mm Total recovery

Rotation 0.000 __
Air flow 1.000 0.000 — — — —
Bed width 1.000 0.857 0.000 — — —
Recovery < 5 mm 0.268 0.940 0.125 0.000 __ —
Recovery > 5 mm 0.774 0.331 0.192 0.027 0.000
Total recovery 0.612 0.286 0.010 0.092 0.000 0.000

Bartlet %2 probability 0.725 0.848 —

strongly either negative or positive, but strong correlations are trommel operation. The benefits of using a computer model
hard to develop in this type of laboratory testing. The follow- based upon mechanics, rather than on statistical analysis,
ing conclusions can be drawn from these statistical analyses: were that analysis of the comparison helped develop the
• The 0 probability calculated for the total matrix con- understanding of trommel behavior.
trasted with significant probabilities for the small and The model was able to replicate laboratory data quite
large unders recovery alone. In the main matrix, the well. The one point of comparison provided by the previ-
explanation became clear; the second most signifi- ously published work of other researchers showed that one
cant correlation in the entire table is the negative of the most sophisticated analyses, that of the crossover of
interaction between the small and large recoveries. the unders recoveries, was also found in this work. Analysis
If a recovery factor aided recovery of the small unders, with the model showed that bridging and layering were
it hurt the recovery of the large unders, and vice- very important phenomena in trommel operation. When
versa. feeds contained particles larger than the openings, as they
• Trommel incline (bed width), given a fixed-length should, bridges formed over the holes, occluding them. With
trommel, was relatively strongly negatively associ- a feed containing many still-less-spherical items than gravel,
ated with all recoveries. Residence time was thus such as solid waste, bridging would be expected to be even
found to be much more important than low bed more significant. To a limited degree, small particles were
depth. presented to the screen first.
• Rotational velocity hurt recovery of the small Rotational speed showed that slower was better. It must
unders, with a fairly strong negative correlation. be remembered that the range of rotational velocities man-
Even though it somewhat helped recovery of the dated proper cataracting action; slowing velocities to cas-
large unders, the penalty to the small unders over- cading (tumbling) speed greatly reduced effectiveness. Tests
shadowed it. A high rotational velocity contributed with incline similarly showed that residence time was the
to mixing, destroying the layering that helped the most important factor to good separation within a given
recovery of the fines. length trommel.
• Airflow had a strong effect on recovery of the large The question then became how to make use of a trommel
unders, with relatively modest confidence. cataracting as slowly as possible, as level as possible. Identi-
fication of layering as an important phenomenon led to
CONCLUSIONS concepts to enhance layering. The addition of air, at or
This research effort comprised a laboratory exercise designed slightly below the point of departure, showed great prom-
to assess directly the performance of a computer model of ise in enhancing separation efficiency. A side-benefit might

Volume 46 June 1996 Journal of the Air & Waste Management Association
Stessel and Cole

be clearing of the screen holes without the need for brushes. 9. Vorstman, M. A. G.; Tels, M. "Some experiments in trammeling of
particles in multilayers," in Materials and Energy from Refuse: Pro-
Analyses of data quality showed results indicative of fairly ceedings of the 2nd International Symposium, Koninkli)ke Vlaamse
high system losses. Such losses are, however, typical for in- Ingenieursvereniging, Antwerp, 1981;. 1.27-1.34.
10. Barton, J. R. Evaluation of Trommels for Waste to Energy Plants Phase
vestigator-built, laboratory-scale equipment operating for 2: Report of the Warren Spring Laboratory Pilot Plant Test Series, War-
short periods. After accounting for these errors, the model ren Spring Laboratories, Stevenage, Hertfordshire, U. K., 1981.
showed itself worthy of further development and testing. 11. Stessel, R. I.; Peirce, J. J. "Separation of solid waste with pulsed air-
flow," ASCE f. Environmental Engineering 1986, J21(6):833.
12. Stessel, R. I. "A new trommel model," Resource, Conservation, and
REFERENCES Recycling 1991, 6:1.
1. Williams, E. Review of the Literature on the Use of Trommels in Waste 13. SYSTAT Inc. SYSTAT Users Manual, Evanston, IL, 1985.
Processing and Resource Recovery, DOE Contract No. AC01-76CS20167, 14. Box, G.E.P.; Hunter, W. G.; Hunter, J. S. Statistics for Experimenters,
National Center for Resource Recovery, Washington, D. C, July, 1981. John Wiley & Sons, Inc., New York, 1978.
2. Parker, W. S. "Application of trammeling to prepared fuels," in Pro-
ceedings of the International Conference on Prepared Fuels and Resource
Recovery Technology, DOE Contract No. W-31-109-Eng-38, Argonne
National Laboratory, Argonne, IL, February, 1981.
3. Alter, H.; Gavis, J.; Renard, M.L. "Design models of trommels for
resource recovery processing," Resources and Conservation 1980,6:223.
4. Glaub, J.C.; Jones, D.B.; Tleimat, J.U.; Savage, G.M. Trommel Screen
Research and Development for Applications in Resource Recovery, United
States Department of Energy, Washington, DC, 1982. About the Authors
5. Glaub, J. C; Jones, D. B.; Savage, G. M. "The design and use of
trommel screens for processing municipal solid waste," in Proc. 1982
R.I. Stessel, Ph.D., P.E., is an Associate Professor in
National Waste Processing Conference, American Society of Mechani- the Department of Civil Engineering and Mechanics,
cal Engineers, New York, 1982; 447-457. University of South Florida, Tampa, FL, 33620-5350.
6. McNabb, M. B. "A new test for evaluation of air classifier perfor- K. Cole is a Project Engineer with Groundwater Tech-
mance," in Fundamental Aspects of Air Classifier Operation and De-
sign, Duke Environmental Center, Durham, NC, 1980, ch. 4. nology, Marietta, GA, 30066. Please address all cor-
7. Stessel, R. I.; Kranc, S. C. "Particle motion in a rotary screen," ASCE respondence to Dr. Stessel or send e-mail to
Journal of Engineering Mechanics 1992, 118(3):604. stessel@eng.usf.edu.
8. Stessel, R. I.; Peirce, J. J. "Particle separation in pulsed airflow," ASCE
J. of Engineering Mechanics 1987, 213:1594.

5 6 8 Journal of the Air & Waste Management Association Volume 46 June 1996

You might also like