Powder Technology 193 (2009) 274288

Contents lists available at ScienceDirect

Powder Technology
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / p ow t e c

Prediction of screw conveyor performance using the Discrete Element Method (DEM)
P.J. Owen , P.W. Cleary
CSIRO Mathematical and Information Sciences, Clayton, Vic 3168, Australia

a r t i c l e

i n f o

Available online 17 March 2009

Keywords:
Screw conveyor
Discrete Element Method
Mass ow rate
Energy dissipation
Power consumption

a b s t r a c t
Screw conveyors are used extensively in agriculture and industry for transporting and/or elevating bulk
materials over short to medium distances. They are very effective conveying devices for dry particulate solids,
giving good control over the throughput. Despite their apparent simplicity, the mechanics of the
transportation action is very complex and designers have tended to rely heavily on empirical performance
data. The performance of a screw conveyor is affected by the operating conditions, such as: the rotational
speed of the screw; the inclination of the screw conveyor; and the volumetric ll level of the bulk material. In
this paper we examine how these operating conditions inuence the performance of a screw conveyor by
applying the Discrete Element Method (DEM) to simulate a single-pitch screw conveyor with periodic
boundary conditions. The DEM modelling gives predictions of screw conveyor performance in terms of
variations of: particle speeds, mass ow rate, energy dissipation and power consumption, due to changes in
the operating conditions.
Crown Copyright 2009 Published by Elsevier B.V. All rights reserved.

1. Introduction
Screw conveyors are widely used for transporting and/or elevating
particulates at controlled and steady rates. They are used in many bulk
material applications in industries ranging from industrial minerals,
agriculture (grains), pharmaceuticals, chemicals, pigments, plastics,
cement, sand, salt and food processing. They are also used for
metering (measuring the ow rate) from storage bins and adding
small controlled amounts of trace materials (dosing) such as pigments
to granular materials or powders. If not designed properly for the
transported material, problems experienced include: surging and
unsteady ow rates, inaccurate metering and dosing, inhomogeneity
of the product, product degradation, excessive power consumption,
high start-up torques, high equipment wear and variable residence
time and segregation.
The basic design of a typical screw conveyor has three major
components:
(1) a hopper or bin;
(2) a stationary screw casing (tubular, open or covered trough);
and
(3) a rotating screw.
Fig. 1 shows a 45 inclined screw conveyor with a tubular screw
casing. The rotating screw draws down the bulk material from the
hopper and transports it along the cylindrical tube to the discharge
opening. A summary of current design methods and problems

Corresponding author.
E-mail address: phil.owen@csiro.au (P.J. Owen).

experienced for screw conveyors can be found in Bortolamasi and

Fottner [1]. The description of the theoretical behaviour of screw
conveyors can be found in articles by Yu and Arnold [2], and Roberts [3].
In this paper, we will use the Discrete Element Method (DEM) to
analyse the performance of a screw conveyor in terms of: particle
speed, mass ow rate, energy dissipation and power consumption. We
also analyse the change in performance due to changes in the
operating conditions. DEM modelling of particulate ow in a screw
conveyor was rst reported by Shimizu and Cundall [4]. They
examined the performance of horizontal and vertical screw conveyors
and compared their results with previous work and empirical
equations. Owen et al. [5] introduced the use of a periodic slice
model to explore the performance of a long screw conveyor. Cleary [6]
used DEM to study draw down patterns from a hopper by a 45
inclined screw conveyor (as shown in Fig. 1). Cleary [7] extended this
work to examine the effect of particle shape on the draw down ow
from the hopper and on the transport characteristics of the screw
conveyor. Here we extend the original work of Owen et al. [5] to look
at the detailed operational performance of a screw conveyor including
quantitative variation of ow characteristics with ll level, angle of
inclination and rotational speed.

2. Model description
DEM simulations involve following the motion of every particle
(coarser than some cut-off size) and modelling each collision between
the particles and between the particles and their environment (e.g.
the internal surface of the screw casing and the surface of the rotating
screw). The boundary geometry is built using a CAD package and
imported as a triangular surface mesh into the DEM package. This

0032-5910/$ see front matter. Crown Copyright 2009 Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.powtec.2009.03.012

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

275

Fig. 1. Inclined screw conveyor: (left) external view showing hopper and screw casing, (right) internal view showing screw and particulate material.

provides unlimited exibility in specifying the three dimensional

geometries with which the particles interact. Here the particles are
modelled as spheres. The DEM code used here is described in more
detail in Cleary [6,8,9].
The screw conveyor used in this study was a standard pitch, single
ight screw conveyor with dimensions similar to the one used by
Roberts and Willis [10] in their experiments. The pitch of the screw is
dened as the length, along the drive shaft, of one turn of the helical
blade, as shown in Fig. 2. A standard pitch screw has its pitch equal to
the outer diameter of the helical blade.
The DEM model was simplied by applying periodic boundary
conditions to a single pitch of the screw as shown in Fig. 2. The
diameter and the pitch of the screw were 38 mm, the diameter of the
screw shaft was 13 mm, and the blade thickness was approximately
1 mm. The internal diameter for tubular case was 40 mm, giving a gap
of about 1 mm between the outer edge of screw blade and the internal
surface of the casing.
Roberts and Willis [10] used millet (a grain that is very close to
spherical in shape) for the dry particulates in their experiments. The
diameters of the spherical particles used here ranged from 2 mm to
3 mm, were uniformly distributed on a mass weighted basis, and had a
density of 700 kg/m3 to best match the millet used in these
experiments. The particleparticle and particleboundary frictions
used for these DEM simulations were 0.7 and 0.5, respectively; and the
particleparticle and particleboundary coefcients of restitution
were 0.1 and 0.3, respectively. The maximum overlap between
particles is determined by the normal spring stiffness. Typically,
average overlaps of 0.10.5% are desirable, requiring a spring constant
of 1000 N/m for this type of simulation.
A series of DEM simulations was carried out for a range of screw
conveyor operating conditions. Three different rotational speeds and
three different volumetric ll levels were used, and the inclination of
the screw conveyor was varied from 0 to 90 in steps of 10. Table 1
summaries the series of operating conditions simulated in this study.
The DEM modelling gave predictions of the changes in the screw
conveyor performance due to changes in the operating conditions in

terms of variations of: particle speed, mass ow rate, energy

dissipation and power consumption.
3. Particle distributions within the screw conveyor
Fig. 3 shows the distribution of the particles inside the screw
conveyor for a 30% by volume ll level and with the screw rotating at
1000 rpm for angles ranging from horizontal to vertical. The particles
are coloured according to their diameter, with the smaller particles
(~2 mm diameter) being light grey and the larger particles (~3 mm
diameter) coloured dark grey. Note that the change inclination was
obtained by changing the orientation of the gravity vector. The screw
therefore remains in the same position and orientation in the pictures
allowing easier comparison of the ow patterns. The rst frame, Fig. 3
(a), shows the particle distribution inside a horizontal screw conveyor.
In each subsequent frame the inclination is increased until the screw
conveyor is vertical in the last frame, Fig. 3(f).
For the horizontal screw conveyor (Fig. 3(a)) the particles form a
heap against the leading face of the screw. After reaching the top of
the screw most of the particles tumble down the surface of the heap
and a few particles fall behind the screw shaft as pictured. There is a
slightly higher concentration of smaller (lighter grey) particles in
contact with the leading face of the screw and a higher concentration
of larger (darker grey) particles in the outer parts of the heap. This
indicates that the particles have become mildly size segregated due to
shear generated by the face of the screw. The extent of the size
segregation reported by Owen et al., [5] for a larger (125 mm) screw
conveyor rotating at 60 rpm with particle sizes ranging from 2 mm to
5 mm, was more pronounced than it is for this screw conveyor.
The next three frames of Fig. 3 show the changes in the distribution
of the particles as the inclination of the screw conveyor is increased to
30 in 10 steps. Fig. 3(b), (c) and (d) shows the heap in front of the
leading face of the screw blade progressively redistributing as the
inclination of the screw increases, with less material at the bottom
near the casing and more material near the top. The free surface of the
bed of particles becomes increasingly aligned with the angle of the
screw blade. As the inclination of the screw increases more particles

Table 1
Screw conveyor operating conditions.

Fig. 2. Standard pitch single ight screw conveyor (pitch equal to diameter).

rpm

Fill level (%)

Number of particles

Screw conveyor inclination to horizontal

600
1000
1400
1000
1000

30
30
30
50
70

907
907
907
1440
2012

0
0
0
0
0

to 90
to 90
to 90
to 90
to 90

in
in
in
in
in

steps of
steps of
steps of
steps of
steps of

10
10
10
10
10

276

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

Fig. 3. Particle distributions within the screw conveyor inclined at angles from horizontal to vertical. Particles are coloured by diameter: smaller ones are light grey and larger ones are
dark grey.

are in contact with the leading surface of the screw. At an inclination

of about 30, most of the leading face of the screw is covered with
particles. For higher angles, as shown in Fig. 3(e) and (f) the layer of
particles above the screw surface becomes more even leading to a bed
of uniform depth on top of the conveying screw at high angles.
In Fig. 4, the particles from Fig. 3 are re-coloured according to their
speed with the slower particles (0.4 m/s) being light grey and the
faster particles (0.9 m/s) coloured dark grey. For the horizontal
screw conveyor, in Fig. 4(a), the particles tumbling down the free
surface on the heap clearly gather speed as they go. This gure also
shows that a few particles have fallen over the shaft. The next two
frames of Fig. 4 show that increasing inclination of the screw conveyor
results in increasing numbers of particles falling backwards over the

shaft. Fig. 4(d) shows a shear layer starting to develop. The layer of
particles in contact with the leading surface of the screw is owing
faster than the rest of the particles in the bed above, so the shear is
concentrated in a thin basal layer just above the screw. The last two
frames, Fig. 4(e) and (f), show that the shear layer thickens with
continuing inclination increase of the screw with all particles
experiencing shear in the vertical screw.
In Fig. 5, the particles from Fig. 3 are re-coloured according to their
axial speed with the slower particles (0.2 m/s) being light grey and
the faster particles (0.8 m/s) coloured dark grey. There are a
signicant number of particles avalanching down along the surface
layers of the heap for the horizontal screw conveyor with axial speeds
exceeding 0.8 m/s. In contrast material adjacent to the screw is barely

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

277

Fig. 4. Particle ow within the screw conveyor inclined at angles from horizontal to vertical. Particles are coloured by their speed: from light grey to dark grey for 0.4 to 0.9 m/s.

moving in the axial direction. These particles were moving with

overall high speeds (Fig. 4a) so we can deduce that the particles are
owing vertically upward along the face of the screw. This means that
there is a strong recirculatory ow within the heap. The screw blade
acts much like a bull dozer pushing a heap in front of it. In a frame
moving with the screw blade motion, the particles within this heap
ow recirculate with ow up along the screw blade, avalanching down
the free surface and then moving back along the bottom towards the
screw blade. There is also a weak swirling (in the direction of rotation
around the shaft) component of the recirculation which leads to the
transport of some particles over the top of the shaft from the top of the
heap and a return ow from under the shaft at the base of the heap.
This extent of the avalanching observed at the free surface decreases
with increasing inclination of the screw conveyor until about 30, at

which stage the free surface of the particles becomes close to parallel to
the angle of the screw blade. Consequently, the layer of particles above
the screw surface becomes more evenly spread out to produce a bed of
uniform depth on top of the conveying screw. Fig. 5 shows that the axial
velocity of the particles is fairly constant through the full thickness of the
bed. The ow pattern at high angles is therefore quite different to that
observed at low angles. To help understand the high angle ow it is
helpful to consider the motion if the particles were bonded together. The
uniform depth particle bed would then be lifted upwards and would
rotate around the shaft as a rigid body. If the screw were stationary then
the mobile particles in a granular bed would ow uniformly down along
the spiral screw. So the ow for the rotating screw can be characterised
as an upward and swirling rigid body increment due to the screw motion
with a smaller downward ow of the particles relative to the screw

278

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

Fig. 5. Particles within the screw conveyor coloured by their axial speed for inclinations from horizontal to vertical (light grey corresponds to 0.2 m/s and dark grey to 0.8 m/s).

blade. The result is an upward spiral ow that is slower than would occur
for a rigid body due to the ability of the bed to shear and ow partially
back down the spiral.
At intermediate angles, the ow is a super-position of these two
basic ow patterns from the horizontal and vertical cases. Below we
quantitatively study this transition between the two basic ow
patterns and how it changes with ll level and screw rotation
speed.

3.1. Particle distributions at various screw speeds

Fig. 6 shows the particle distributions inside the screw conveyor
for a ll level of 30% by volume and with the screw rotating at three
different rotational speeds (600, 1000 and 1400 rpm) for various
screw conveyor angles ranging from horizontal to vertical. Here the

particles are coloured by their speed using the same greyscale range as
used in Fig. 4.
It is quite clear that for each inclination, increasing the rotational
speed of the screw increases the speed of the particles. There is an
apparent increase in the ll level with increasing speed and the angle of
the top surface of the heap becomes steeper. This is most evident on the
horizontal screw conveyor where the height of the heap for 600 rpm
case does not reach the top of the screw. At 1000 rpm, the height of the
heap has reached the top of the screw and at 1400 rpm there is a thick
layer of particles at the top of the screw. For the vertical screw
(inclination 90) the layer of particles above the screw surface has
spread out to produce a bed of uniform depth on top of the conveying
screw. The depth of this uniform bed clearly increases with increasing
speed.
In Fig. 7 the particle distributions shown in Fig. 6 are redrawn from
a viewpoint looking along the axial directional of the screw conveyor.

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

279

Fig. 6. Particle ow within the screw conveyor inclined at various angles and operating at various screw speeds. Particles are coloured by their speed: from light grey to dark grey for 0.4 to 0.9 m/s.

The particles that were in the foreground of Fig. 6 appear on the left
hand side of each image in Fig. 7. In this frame of reference the screw is
rotating clockwise.
For the horizontal screw conveyor, the ow pictures in Fig. 7 look
very similar to those obtained for rotating drums and mill, for example
see Cleary [11]. Using the same terminology used for mills, at 600 rpm
the heap has a well dened toe and shoulder. The toe is located at
about the 4 o'clock position and the shoulder is located at about the 10
o'clock position. Material is clearly transported at low speed by the
rotation of the screw up to the shoulder and it then avalanches down
the free surface leaving a cavity behind the core of the screw. Particles
pile up at a toe position.
This recirculation ow in the heap, also noted above, is similar to
the ow found inside rotating mill where particles are lifted along the
surface of the mill from the toe to the shoulder. At the shoulder slower
particles will cascade or avalanche down the free surface, and faster
particles are thrown away from the free surface. In mills this
behaviour is called cataracting.

The sequence of images in Fig. 7, for the horizontal screw conveyor,

shows the change in ow behaviour with increasing rotational speed of
the screw. The rst thing to note is that the toe and shoulder move
further clockwise with increasing screw speed. The second thing to note
is that with increasing screw speed the speed of the particles increases
and as a consequence more particles are thrown or fall from the shoulder
region over the top of the shaft of the screw. This three dimensional ow
behaviour re-distributes the particles from the back to the front of the
screw and makes the heap of particles appear larger when viewed from
the front as shown in Fig. 6.
At an inclination of 30, Fig. 7 shows that there are more particles
in the cataracting stream and that the toe and shoulder disappear.
However there is still some asymmetry in the shape of the cavity (the
region devoid of particles below the shaft of the screw). There are two
different changes in the void shape with increasing screw speed. The
rst is that the angular position of the centre of the void is located at
the 4 o'clock, 4:30 and 5 o'clock positions for speeds of 600, 1000 and
1400 rpm respectively. The void moves further around the casing in

280

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

Fig. 7. Particle ows from Fig. 6 when viewed along the axis of the screw conveyor. The particles are coloured by their speed: from light grey to dark grey for 0.4 to 0.9 m/s.

the screw rotation direction with increasing screw speed. The second
change is in the shape of the void. It becomes larger in circumferential
direction and to a lesser extent in the radial direction with increasing
screw speed. This shows that the centrifugal component is increasing
with increasing screw rotation rate.
For a horizontal screw conveyor the critical speed for centrifugal
motion for a particle in contact with the screw casing is about 0.43 m/s
in the circumferential (swirl) direction. So ows with swirl speeds above
0.43 m/s would have centrifugal components. The broad range of swirl

speeds found for the current operating range is 0.3 to 0.9 m/s, which
means that there is almost always some centrifugal force inuencing the
ow pattern. At the lower speeds this has only a minor inuence on the
ow pattern, but at higher speeds it becomes a signicant factor in
controlling the ow of particles.
At inclinations above 30, Fig. 6 shows the effect of gravity
attening the particle bed, whereas, Fig. 7 shows the void in the
particle bed growing larger with increasing speed. The section below
on average particle swirl speed shows that for all simulations with the

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

screw conveyor inclined at a high angle, above about 60, the average
swirl speed is above the critical centrifugal speed.
So, for the vertical screw conveyor, Figs. 6 and 7 show that the
screw ow forms a bed on the screw blade due to both gravity and the
outward centrifugal force. The centrifugal component increases with
increasing screw speed. This is clearly shown by the particle bed
moving further away from the shaft with increasing screw speed and
corresponding increasing particle swirl speeds. The ow patterns for
the horizontal screw conveyor are quite different to ow patterns for
the vertical case being best characterised as a bull dozer pushing a
circulating heap in front of it. While the particle movement in the
vertical screw conveyor is an upward spiral ow.
At intermediate angles, the ow is a superposition of these two
basic ow patterns from the horizontal and vertical cases. For
example, for the screw inclined at 30 and rotating at 600 rpm,
Fig. 7 shows slower particles on the left as they are being lifted up and
faster particle on the right as they cataract and accelerate under

281

gravity. This is consistent with the bull dozer circulating heap like
ow. On the other hand, the bed on particles is in contact with the
screw casing all the way around the turn of the screw. This is
consistent with the upward centrifugal spiral ow.
3.2. Particle distributions at various ll levels
Fig. 8 shows the particle distributions inside the screw conveyor
operating at 1000 rpm for various ll levels (30%, 50% and 70% by
volume) for different screw conveyor angles ranging from horizontal
to vertical. Again the particles are coloured by their speed using the
same greyscale as used in Fig. 4.
For the horizontal screw conveyor (inclination 0), at the 30% ll
level avalanching of particles along the free surface of the heap is very
strong with particles starting to ow from the screw surface at the top
of the pile and just reaching the back of the previous screw pitch. For a
50% ll level the extent of avalanching is much reduced. The free

Fig. 8. Particle distributions within the screw conveyor inclined at various angles and operating with different ll levels. Particles are coloured by their speed: from light grey to dark
grey for 0.4 to 0.9 m/s.

282

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

will be explored in future work. Overall, the match to the

experimental results is very encouraging.
5. Particle transport velocities
5.1. Variation of average particle speeds

Fig. 9. Standard pitch single ight screw conveyor (pitch equal to diameter).

surface is much steeper but also much shorter in length with the pile
making contact with the back of the previous screw pitch at a much
higher level. The start of the free surface has also moved well to the
right of the screw. At the 70%, the heap on the front of the shaft almost
lls the entire region between leading face of the screw and the
trailing face of the next turn of the screw. This leaves no free space for
avalanching to occur. The top row in Fig. 8 also shows that the speed of
the particles in contact with the leading face of the screw increases
with increasing ll level. This suggests that a shear layer develops near
the leading screw face for the higher ll levels. The shear layers do not
appear to vary much when the angle of the screw conveyor is
increased. For screw inclinations from 30 to 90, the slope of free
surface doesn't seem to change with changing ll level the depth of
the bed just increases with the increasing ll level.
Increasing ll level clearly changes the balance between the two
basic ow patterns. As the free space available between the turns of
the screw progressively declines the recirculatory ow driven by
surface avalanching weakens and the shearing swirling ow along the
screw blade increases. The transition between the two ow patterns is
very strongly inuenced by the ll level. So for ll levels lower than
30% one would expect that the avalanching recirculatory ow in the
bulldozed heap will dominate for most inclinations and screw speeds.

Fig. 10 shows that the predicted average particle speed increases

only marginally with increasing inclination of the screw conveyor and
that this is consistent for a range of screw rotational rates and ll
levels. At very steep angles the overall average speed of the particles
becomes quite insensitive to further orientation change. More
specically, Fig. 10(a) shows the average particle speed versus screw
conveyor inclination for a 30% ll level and with the screw rotating at
three different speeds. For each of the three rotation rates, the particle
speed is insensitive to angle when close to horizontal or vertical. In
between there is gentle linear increase in particle speed with angle
ranging for 6% for 1400 rpm to 23% for 600 rpm.
The variation of the average particle speed with inclination is much
smaller than the change due to increasing rotation rate (as seen by the
vertical translation of the speed versus angle curves shown in Fig.10(a)).
Clearly, the average particle speed increases with increasing screw
speed, however the average speed of the particles does not scale with
the speed of the conveyor.
Fig. 10(b) shows the variation in the average particle speed with
inclination for three ll levels. For all three ll levels there is only a

4. Comparison with experimental results

Tracking each particle in a DEM simulation enables collection of
information to measure the performance of a screw conveyor, for
example: average speed of the particles, average mass ow rates,
power draw and energy dissipation. Fig. 9 shows mass ow rates,
obtained from these DEM simulations and from the experimental
measurements of Roberts and Willis [10], plotted against rotational
speed for various inclinations of the screw conveyor. Fig. 9 also shows
the theoretical maximum mass throughput of the screw conveyor
which is a linear function of the rotational speed of the screw. There is
excellent agreement between DEM predictions and the experimentally measured throughput for the horizontal (0) and the vertical
screw conveyor (90) cases across the full range of screw rotation
speeds. The DEM predictions for the screw conveyor inclined at 30
and 60 follow the correct qualitative trend but modestly underestimate the mass ow rates by 16% and 24% respectively. These
underestimations may result from differences between particlewall
and particleparticle friction which were not explored in this work
and/or particle shape effects. Cleary [7] has previously found that the
drawdown pattern in a screw conveyor hopper can be quite sensitive
to the particle shape, so this is likely to be a contributing factor. This

Fig. 10. Average particle speed versus inclination for: (a) 3 rotational speeds at a 30% ll
level; and (b) 3 ll levels at 1000 rpm.

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

283

weak change in the particle speed. The three curves are very close
showing that the average particle speed is almost independent of the
ll level. The maximum difference occurs for the 30% when the screw
is vertical with speed that is 11% higher than for the 70% ll level.

5.2. Average particle axial speed

Fig. 11 shows the average axial velocity of the particles for various
operating conditions. Fig. 11(a) shows changes in axial velocity with
angle for the three rotation rates of the screw. In each case it decreases
strongly and linearly with increasing screw conveyor inclination until
about 60. For steeper angles, where the blade of the screw is covered
with a uniform bed of particles (as shown in Fig. 3), the average axial
velocity asymptotes to a constant value.
The curves for each of the different speeds which are very close to
parallel are well separated. The variation due to rotation rate changes
is much stronger than for the screw inclination changes. As expected,
the average axial velocity of the particles increases with increasing
screw speed, however no consistent power law scaling could be found
to describe the relativity between these three axial particle velocity
curves and the rotational speed of the screw.
Fig. 11(b) shows that the average axial speed of the particles is
almost invariant to changes in the volumetric ll level for the full

Fig. 12. Average swirl speed versus inclination for: (a) 3 rotation rates at a ll level of
30%; and (b) 3 ll levels at a rotation rate of 1000 rpm.

range of inclinations. For a horizontal screw conveyor, (the most

extreme case) the average axial speed of the particles decreases by
only 9.5% when the ll level is increased from 30% to 70%. The ll level
changes the depth of the bed on top of the ights of the screw but does
not change the overall distribution within the screw and therefore
does not change the average axial transport speeds.

5.3. Average particle swirl speed

Fig. 11. Average particle axial speed versus inclination: (a) ll level 30% for 3 speeds; and
(b) 1000 rpm for 3 ll levels.

Fig. 12 shows the predicted average swirling speed of the particles

(the motion component around the axis of the screw conveyor) for
various operating conditions. More specically, Fig. 12(a) shows the
average swirling speed versus inclination for a 30% ll level and with
the screw rotating at three different speeds. At rst, it increases
strongly and linearly with increasing inclination of the screw
conveyor. For angles above about 40 the rate of increase of the
swirling speed starts to decline. Below 40 the average swirl speed is
below the critical speed for centrifugal motion. So, increasing the
angle up to this inclination would see an increase in the centrifugal
component of the ow. Above this angle the centrifugal component
dominates the ow. By about 80 it has saturated and there is little
further change up to 90. The three curves for the different rotation
rates are close to parallel but unlike for the earlier axial speed, the
variation due to the inclination has a stronger effect than rotation rate.

284

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

Similar to the earlier axial speed, the swirl speed does not scale with
screw speed.
Fig. 12(b) shows that increasing the ll level of the screw tends to
increase the swirling speed for smaller inclinations of the screw, and
decrease it for inclinations above 20. For a screw conveyor that is 30%
full by volume, the swirl speed for a vertical screw conveyor is about
2.8 times faster than the swirl speed for a horizontal one. Whereas, for
a ll level of 70% the swirl speed for the vertical screw is only about 1.4
times faster than the horizontal one. So the low ll levels have the
highest sensitivity to inclination for the swirling motion. This can be
consistent with ow observations made for Fig. 8 where increasing ll
level led to reduction of the recirculatory heap avalanche component
of the ow and an increase in the rotating bed ow component. This
explains the increase observed here in average swirling velocity at
higher ll levels for lower screw inclinations. As the ll level increases
the swirling component of the ow becomes decreasingly sensitive to
the changing orientation because the ow pattern has already become
dominated by the rotating bed ow pattern. At an angle of around 25
the swirl speed is independent of ll level.
5.4. Average particle speeds summary
In summary, as the inclination of the screw conveyor increases
from the horizontal position to the vertical position:
The average swirling speed of the particles increases.
The average axial velocity of the particles decreases.
These two trends broadly cancel each other to give an overall
average particle speed which is fairly insensitive to inclination. The
insensitivity of the average speed masks strong structural changes in
the motion between the axial and swirling velocity components
driven by the change in ow pattern from a recirculating avalanching heap to a rotating bed of constant depth owing along the screw.
6. Mass ow rates
The ow behaviour inside the screw conveyor can also be
examined quantitatively by measuring the mass ow rates of the
particles as they are transported along the screw conveyor. The mass
ow rate was determined by recording the mass of each particle that
has passed through a plane perpendicular to the axis of the screw. This
plane was located half-way between the two periodic boundaries.
Fig. 13(a) shows the average mass ow rate of the particles versus
the inclination of a screw conveyor for a 30% ll level for the screw
rotating at three different speeds. Exhibiting the same behaviour as
the average axial speed, the average mass ow rate decreases strongly
but linearly with increasing screw conveyor inclination until about
60. For steeper angles, where the blade of the screw is covered by a
uniform depth bed of particles, the average mass ow rate approaches
a constant value. The volumetric transport rate is reduced by 47%, 31%
and 23% when operating the screw conveyor vertically compared to its
optimal transport when operated horizontally and with screw speeds
600, 1000 and 1400 rpm, respectively.
The curves for each of the different speeds which are very close to
parallel are well separated, just like the ones for average axial particle
speeds shown in Fig. 11(a). Again, the variation due to rotation rate
changes is much stronger than for the screw inclination changes.
Again, no consistent power law scaling could be found to describe the
relativity between these three mass ow rate curves.
Fig. 13(b) shows the average mass ow rate of the particles versus
the inclination of a screw conveyor for three volumetric ll levels at a
screw rotation rate of 1000 rpm. The average mass ow rate decreases
linearly with increasing screw conveyor inclination until about 60.
For steeper angles, where the blade of the screw is covered with a
uniform bed of particles, the average mass ow rate again approaches
a constant value.

Fig. 13. Average mass ow rate versus screw conveyor inclination for: (a) 3 rotation
rates at a ll level of 30%; and (b) 3 ll levels at a rotation rate of 1000 rpm.

It was noted above that the average axial speed of the particles is
almost invariant to changes in the volumetric ll level for the full
range of inclinations. Consequently, it is reasonable to expect that the
mass ow rate would increase linearly with increasing volumetric ll
level across the full range of inclinations.
7. Power draw
Power draw is determined from DEM predictions of the forces
exerted by the millet particles on the rotating screw. For each
operating condition, the screw conveyor reaches steady state
within 23 turns of the screw and the power draw is then quite
steady. Fig. 14(a) shows the steady state power draw for the screw
conveyor operating at 30% ll level, inclined at various angles, with the
screw rotating at 3 different speeds. This clearly shows that with
increasing inclination angle the screw conveyor draws more power.
Initially, the increase is linear up to about 50. The rate of change then
declines leading to the power draw being independent of angle for
inclinations above 80. The rate of increase in power draw reects the
signicant energy input required to increase and maintain the high
swirl speeds observed at the higher screw angles. So as the inclination
increases the transport rates achieved decline gently but the cost of
maintaining the swirling motion in the uniform thickness bed owing
on top of the screw blade rises strongly.
Increasing the rotational speed of the screw also draws more
power. However, no consistent power law scaling could be found to

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

285

energy supplied. That is, it should be the same as the power draw
discussed in the previous section.
8.1. Shear energy dissipation
Fig. 15(a) shows the percentage of shear energy dissipation for all
collisions in the screw conveyor operating at 30% ll level, inclined at
various angles, with the screw rotating at three different speeds. At
lower inclinations up to about 50, it shows that shear energy
dissipation increases with increasing angle of inclination and
increases with increasing rotational speed of the screw. For inclinations above 50 the shear energy dissipation fraction becomes
invariant to changes in inclination or rotation rate of the screw. This
means that the recirculatory avalanching heap ow pattern observed
at low inclinations dissipates energy less by shear which is reasonable
considering the impacts that occur in the avalanching region.
Conversely, the swirling bed ow supported by the screw underneath
has less opportunity for direct impact dissipation and more energy
dissipation by shear as layers of particles slide over each other.
Fig. 15(b) shows the shear energy dissipation for all collisions in
the screw conveyor rotating at 1000 rpm and inclined at various
angles for three different ll levels. For the 30% ll level the shear
energy dissipated increases linearly for inclinations up to about 30.
The rate of change then starts to decline for angles above 40. For the
other two ll levels, the fraction of the total energy dissipated as shear
energy increases steadily with increasing inclination. At each screw

Fig. 14. Power versus angle of inclination for: (a) 3 rotation rates at a ll level of 30%;
and (b) 3 ll levels at a rotation rate of 1000 rpm.

describe the relativity between these three power curves. The power
relationship between the 1000 rpm and 600 rpm curves scales with
rotational speed to the power 1.61. However, the power between the
1000 rpm and 1400 rpm curves scales with rotational speed to the
power 1.80. This indicates that the relationship between power draw,
rotational speed and inclination is non-linear and non-trivial.
Fig. 14(b) shows the variation of steady state power draw with
inclination for three different volumetric ll levels with the screw
operating at 1000 rpm. For all three ll levels the power draw
increases linearly for inclinations up to about 50. Again the rates of
change then start to decline and for angles above 80 there is no
further change in the power with angle.
Increasing the volumetric ll level of the screw conveyor also
increases the power draw. Again, a simple consistent power law scaling
could not be found to describe the relativity between these curves.
Between the 30% and 50% ll level curves the power scales with ll level
to the power 1.16, and between the 30% and 70% ll level curves it scales
with ll level to the power 1.40. This indicates that the relationship
between power draw, ll level and inclination is also non-linear.

8. Energy dissipation
DEM simulations provide detailed information about the energy
dissipated in every collision over the period simulated. In accurate
DEM simulations the total energy dissipated should be the same as the

Fig. 15. Shear energy dissipation versus screw conveyor inclination for: (a) 3 rotational
speeds at a ll level of 30%; and (b) 3 ll levels at 1000 rpm.

286

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

conveyor inclination, increasing the volumetric ll level decreases the

percentage of the energy that is dissipated through the shear
interactions. Conversely, a larger fraction of the energy is dissipated
via normal impacts with increasing ll level as the weight of the bed
increases.

8.2. Energy dissipation in particleboundary collisions

Fig. 16(a) shows the fraction of the total energy supplied that is
dissipated during particleboundary collisions in the screw conveyor
operating at 30% ll level, for various inclination angles and at three
different speeds. At the lowest rotation rate of 600 rpm, the fraction of
the supplied energy dissipated during particleboundary collisions
starts at around 44% for a horizontal screw and increases approximately linearly, with increasing inclination of the screw, up to about
49% at 70 and then remains at that level for steeper screw angles. At
1000 rpm, the fraction of the supplied energy dissipated during
particleboundary collisions increases linearly with increasing inclination, up to 49% at about 30. Above this inclination the curve levels
off at about the same level as the 600 rpm curve. The increase in the
particleboundary energy fraction reects the decreasing slope length
and decreasing dissipation in the avalanching ow as the ow
transitions from the avalanching mode to the shearing bed mode
where more energy is dissipated by the screw and shell.

Fig. 17. Energy spectra for all collisions in a horizontal screw conveyor with 30% ll level
with screw rotation rate at 1000 rpm, with tted curves added.

At the highest screw speed simulated of 1400 rpm, the fraction of

the supplied energy dissipated during particleboundary collisions
starts above the asymptotic limit at around 50.6% and then decreases
slightly, with increasing inclination of the screw, down to around 50%
at about 50. So, at higher inclinations about 50% of the supplied
energy is consumed in particleboundary collisions irrespective of the
screw rotational speed.
Fig. 16(b) shows the fraction of the total energy supplied that is
dissipated during particleboundary collisions in the screw conveyor
rotating at 1000 rpm for various inclinations at three ll levels. For the
30% ll level, the fraction of the total energy dissipated as shear energy
increases linearly for inclinations up to about 30. The rate of change
then starts to decline for angles above 40. For the 50% and 70% ll
levels, the shear energy fraction increases steadily with increasing
inclination.
At each screw conveyor inclination, increasing the volumetric ll
level decreases the percentage of the energy that is dissipated during
particleboundary collisions. Again, this also means that a larger
fraction of the supplied energy is dissipated as normal impacts with
increasing ll level. Increasing from 30% to 50% lled increases the
fraction by around 2.5%, while an increase to 70% lled increases the
fraction by another 2%. This reects the increasing amount of
boundary exposed to impacts with particles as the ll level rises.
8.3. Energy spectra

Fig. 16. Energy dissipation in particleboundary collisions versus screw conveyor

inclination: (a) ll level 30% for 3 rotational speeds; and (b) 1000 rpm for three ll
levels.

To analyse the nature of the collisional environment the collision

energy spectra are constructed. These are representations of the
frequency distribution of different energy levels for collisions
occurring in the screw conveyor. These provide a means to help
understand where and how the energy in the screw conveyor is
consumed and the extent to which the structure of these collisions
varies with changes in operating conditions.
The distribution of impact and shear energies inside the screw
conveyor can inuence the wear rates on the screw and screw casing
and the degradation of the particles. High energy impacts may cause
particles (millet) to fracture and high shear energies may cause
abrasive erosion to the particles and/or the equipment. Energy spectra
can therefore be used to assess the relative degradation risk for the
material being transported.
Fig. 17 shows the energy spectra for all collisions occurring in a
horizontal screw conveyor with a 30% by volume ll level and a screw
rotation rate at 1000 rpm. This gure shows: normal energy loss
(which is relevant to brittle impact fracture of the particles), shear
energy loss (which is related to abrasive wear) and total energy loss.
Each energy loss curve is a histogram of collision frequency (or the
number of collisions per second) versus the collision energy
dissipation in those collisions. All three curves have the same shape:

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

each is a skewed bell shape curve with a long low energy tail, a
rounded central (or modal) peak declining to a maximum energy.
Fig. 17 also shows tted fourth order polynomial curves that can be
used to characterise the essential features of the energy spectra. The
rst feature extracted is the location of the modal peak which gives
the most frequent collisional energy. The second feature is a
dimensionless measure of the high energy spectral width from the
central peak to the maximum energy. It is dened by log10(Emax /
Emodal), where Emodal is the energy at the modal peak and Emax is the
maximum energy dissipated in a collision. Emax is obtained from the
tted curve at a collision frequency of 1 collision/s. The characteristic
features obtained from the tted curves are:
For the total energy the modal peak is about 25.4 nJ/collision and
the collision frequency at this peak is about 10,000 collisions/s. The
maximum energy is about 16.5 J which gives a high energy spectral
width of about 2.8.
For the shear energy the modal peak is about 14.6 nJ/collision with a
collision frequency also of about 10,000 collisions/s. The high
energy spectral width is about 3.0.
For the normal energy the modal peak is also about 10,000 collisions/s
located at about 7.03 nJ/collision. The high energy spectral width is
about 2.7.

287

The spectra width was found to be essentially constant over the full
range of operating conditions. They were 2.7, 2.9 and 2.5 for the total,
shear and normal spectra respectively. The relative positions of the
three types of modal peaks were broadly invariant to changes in
operating conditions. The shear peak was approximately 53% of the
energy of the total energy peak and the normal energy peak was about
37% of the energy of the total energy modal peak. This conrms the
consistent pattern observed of much more energy being dissipated in
shear interactions than in normal ones.
The only spectra characteristic that was found to vary was the
location of the modal peak for the total energy dissipation in each
collision. Fig. 18(a) shows the modal peak energy for the screw
conveyor operating at 30% ll level, inclined at various angles for three
screw rotation rates. It shows a consistent upward trend in the average
collision energy with increasing inclination up to about 5060 after
which it becomes constant. These curves are very similar in shape to
the power draw curves in Fig. 14. Increasing the rotational speed of the
screw, which draws more power, results in a shift of the energy
spectra (to the right) to higher collisional energies. As for the power
draw no consistent power law scaling could be found to describe the
relativity of these curves.
Fig. 18(b) shows the variation of the modal peak energy with
inclination for three different volumetric ll levels with the screw
operating at 1000 rpm. As before the modal peak increases essentially
linearly over the range of inclination with asymptotic behaviour when
near vertical. As the angle increases and the particle swirl speeds
increase the energy available to be dissipated increases so that the
average energy per collision nearly doubles for a vertical screw
compared to a horizontal one. The three curves for the different ll
levels are very close together indicating that the modal peak energy is
broadly insensitive to changes in ll level.
9. Conclusion
DEM simulation shows that the nature of the particle ow in an
inclined screw conveyor is surprisingly complex and is reasonably
sensitive to the operating conditions. The predicted mass ow rate for
transport by the screw conveyor was in excellent agreement with
experimentally measured values for the horizontal and the vertical
congurations across the full range of screw rotation rates. The
throughput predictions for the screw conveyor inclined at 30 and 60
followed the correct qualitative trend but modestly under-estimated
the mass ow rates by 16% and 24% respectively. These underestimations will be explored in future work by looking at variations of
particlewall and particleparticle friction and/or particle shape
effects.
The screw conveyor exhibited two principle types of ow patterns:
1. A recirculatory ow in a heap of particles being bull-dozed along by
the screw with avalanching down along the free surface of the heap
and vertical transport up the face of the screw. When viewed along the
axial direction the ow exhibits behaviour observed in rotating drums
and mills with cascading, cataracting and centrifuging behaviour.
2. A shearing ow in a bed of uniform depth owing, driven by gravity
to be spread evenly across the screw surface and a centrifugal
component pressing the bed against the screw casing and away
from the screw shaft.

Fig. 18. Energy at modal peak versus screw conveyor inclination: (a) ll level 30%for 3
rotational speeds; and (b) 1000 rpm for three ll levels.

The rst ow type occurs at low angles and at low ll levels while
the second ow type occurs at higher inclinations and at higher ll
levels. For intermediate screw inclinations a weighted superposition
of these two ow patterns were observed. Operating performance
changes can all be related to this change in the ow pattern type. The
change from the circulating heap ow to the bed type ow pattern is
strongly inuenced by the transition to a centrifuging axial particle
distribution as the swirl speeds increase above the critical speed.

288

P.J. Owen, P.W. Cleary / Powder Technology 193 (2009) 274288

The ow behaviour was quantitatively characterised in terms of

particle speeds and mass ow rates for the various angles and
operating conditions. For increasing screw inclination:
The average axial velocity of the particles decreases linearly at low to
medium angles. The rate of decline reduces at mid-range inclinations, and the axial velocity becomes insensitive to inclination at
high angles.
The average swirling speed of the particles increases linearly at low
to medium angles asymptoting to a constant level at high angles.
These two trends are caused by the progressive transition from heap
ow at lower angles to the swirling bed ow at higher angles.
The trends in the axial and swirling velocities broadly cancel each
other to give an overall average particle speed that is fairly
insensitive to inclinations. This masks the strong structural changes
in the type of ow and is therefore not a good indicator of ow
behaviour.
The trends in the mass ow rates mirrored the trends in the average
axial velocity of the particles over the full range of the screw
conveyor operating conditions.
Increasing power was needed to move the particles leading to
increased energy being dissipated in particleparticle and particle
wall collisions, particularly as the swirl velocity increases with the
transition from heap ow to swirling bed ow.
So with increasing inclination of the screw, the screw has to work
harder and put in more energy, but this leads to a progressively lower
axial transport with the extra energy being used in the particle
swirling motion around the screw. As the angle approaches vertical all
ow characteristics become insensitive to further angle increases.
For a given ll level and a given angle of inclination, increasing the
rotational speed of the screw conveyor:
Increases the power needed to move the particles. The relationship
between the power draw and rotational speed is non-linear.
Increases the average axial, average swirl speed and average overall
speed of the particles, however, no consistent power laws could be
found between these speeds and the rotational speed of the screw.

For a given rotational speed of the screw conveyor and a given

angle of inclination, increasing the ll level:
Increases the power needed to maintain the particle ow. The
relationship between the power draw and ll level is non-linear.
The average axial speed of the particles is almost invariant to ll
level.
The average swirl speed of the particles increases for low screw
inclination angles but decreases for inclinations above 25.
The overall average particle speed is insensitive to changes in ll
level.

References
[1] M. Bortolamasi, J. Fottner, Design and sizing of screw feeders, Proc. Partec 2001, Int.
Congress for Particle Technology, Nuremberg, Germany, March 2001, pp. 2729,
Paper 69.
[2] Y. Yu, P.C. Arnold, Theoretical modelling of torque requirements for single screw
feeders, Powder Technology 93 (1997) 151162.
[3] A.W. Roberts, The inuence of granular vortex motion on the volumetric
performance of enclosed screw conveyors, Powder Technology 104 (1999) 5667.
[4] Y. Shimizu, P.A. Cundall, Three-dimensional DEM simulation of bulk handling
screw conveyors, Journal of Engineering Mechanics (September 2001) 864872.
[5] P.J. Owen, P.W. Cleary, B. McBride, Simulated granular ow in screw feeders using
3D Discrete Element Method (DEM), CHEMECA 2003, 31st Australasian Chemical
Engineering Conference, 2003, ISBN: 0-86396-829-5, Paper No. 203.
[6] P.W. Cleary, Large scale industrial DEM modelling, Engineering Computation 21
(2004) 169204.
[7] P.W. Cleary, DEM modelling of particulate ow in a screw feeder, Progress in
Computational Fluid Dynamics 7 (Nos. 2/3/4) (2007) 128138.
[8] P.W. Cleary, Predicting charge motion, power draw, segregation, wear and particle
breakage in ball mills using discrete element methods, Minerals Engineering 11
(1998) 10611080.
[9] P.W. Cleary, Discrete Element Modelling of industrial granular ow applications,
TASK, Quarterly Scientic Bulletin 2 (1998) 385416.
[10] A.W. Roberts, A.H. Willis, Performance of grain augers, Proceedings of the
Institution of Mechanical Engineers 176 (No 8) (1962) 165194.
[11] P.W. Cleary, Charge behaviour and power consumption in ball mills: sensitivity to
mill operating conditions, liner geometry and charge composition, International
Journal of Mineral Processing 63 (2001) 79114.