The Numerical Modelling of Excavator Bucket Filling Using DEM

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Available online at www.sciencedirect.

com
Journal
of
Terramechanics
Journal of Terramechanics 46 (2009) 217–227
www.elsevier.com/locate/jterra

The numerical modelling of excavator bucket filling using DEM


C.J. Coetzee *, D.N.J. Els
Department of Mechanical and Mechatronic Engineering, University of Stellenbosch, Private Bag X1, Matieland 7602, South Africa

Received 15 February 2007; received in revised form 25 February 2009; accepted 28 May 2009
Available online 25 June 2009

Abstract

The filling of an excavator bucket is a complex granular flow problem. In order to optimize the filling process, it is important to under-
stand the different mechanisms involved. The discrete element method (DEM) is a promising approach to model soil-implement inter-
actions and it was used in this study to model the filling process of an excavator bucket. Model validation was based on the accuracy with
which the model predicted the bucket drag force and the development of the different flow regions. Compared to experimental measure-
ments, DEM predicted lower bucket drag forces, but the general trend was accurately modelled. At the end of the filling process the error
in predicted drag force was 20%. Qualitatively, there was a good agreement between the observed and the modelled flow regions in terms
of position and transition from one stage to the other. During all stages of filling, DEM was able to predict the volume of material inside
the bucket accurately to within ±6%.
Ó 2009 ISTVS. Published by Elsevier Ltd. All rights reserved.

1. Introduction are also found on hydraulic excavators, loaders and shovel


excavators.
Earthmoving equipment plays an important role in the The filling of a bucket is a complex granular flow prob-
agricultural, earthmoving and mining industries. The lem. Instrumentation of field equipment for measuring
equipment is highly diverse in shape and function, but most bucket filling is difficult and expensive. It is possible to
of the soil cutting machines can be categorised into one of use small-scale (usually 1/10th scale) experimental rigs to
three principal classes, namely blades, rippers and buckets evaluate bucket designs [1,2] but they are expensive and
(shovels). This paper focuses on the numerical modelling of there are questions regarding the validity of scaling [3,4].
excavator bucket filling using the discrete element method To scale-up results from model experiments is problematic
(DEM). since there are no general scaling laws for granular flows as
Buckets are found on a number of earthmoving machin- there are for fluid dynamics [5].
ery. Draglines are used to remove blasted overburden from According to Cleary [5] the filling of buckets, in the
open cut mines. Its removal exposes the coal deposits absence of very large rocks, is observed to be relatively
beneath for mining. A dragline is a crane-like structure two-dimensional with little motion in the transverse direc-
with a huge bucket of up to 100 m3 in volume suspended tion. The flow pattern along a cross-section of the bucket in
by steel ropes. Draglines are an expensive and essential part the drag direction is the most important aspect of filling
of mine operations and play an important role in the com- and can be analysed satisfactorily using two-dimensional
petitiveness of South African mines. In the coal mining models. Rowlands [2] made similar observations based on
industry it is generally accepted that a 1% improvement dragline bucket filling experiments.
in the efficiency of a dragline will result in an R1 million According to Maciejewski et al. [6], in practical cases
increase in annual production per dragline [1]. Buckets when the motion of a bucket or bulldozer blade is dis-
cussed, plane strain conditions apply only in some defor-
*
Corresponding author. Tel.: +27 21 808 4239; fax: +27 21 808 4958.
mation regions. The plane strain solution for such tools
E-mail address: ccoetzee@sun.ac.za (C.J. Coetzee). can be assumed only with limited accuracy. Maciejewski

0022-4898/$36.00 Ó 2009 ISTVS. Published by Elsevier Ltd. All rights reserved.


doi:10.1016/j.jterra.2009.05.003
218 C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227

et al. [6] also investigated the assumption of plane strain The main objective of this study was to demonstrate the
conditions in soil bins where the soil and tool motion is ability of DEM to predict the drag force on the bucket and
constrained between two transparent walls. For measure- the material flow patterns that develop as the bucket fills
ments in such a bin, the force acting on the tool due to up. The DEM results were compared to experiments per-
the friction between the soil and the sidewalls has to be esti- formed in a soil bin.
mated or neglected. They have shown that for a high num-
ber of teeth on the bucket, the teeth do not act as separate 2. The discrete element method
three-dimensional objects but as one wide tool built up
from several modules. The deformation pattern in front Discrete element methods are based on the simulation of
of such an assembly of teeth was found to be plane strain the motion of granular material as separate particles. DEM
deformation. The authors, however, concluded that this was first applied to rock mechanics by Cundall and Strack
was true for the particular cohesive soil (sandy clay) and [16]. In this study, all the simulations were two-dimensional
may not apply to other (especially rocky and brittle) mate- and performed using commercial DEM software PFC2D [17].
rials. In this study the bucket had a full-width lip with no A linear contact model was used with a spring stiffness kn
teeth and based on the findings by Maciejewski et al. [6], in the normal direction and a spring stiffness ks in the shear
the assumption of plane strain was made and two-dimen- direction (Fig. 1). Frictional slip is allowed in the tangential
sional DEM models were used. direction with a friction coefficient l. The damping force acts
Analytical methods [7–11] used to model soil–tool inter- on a particle in the opposite direction to the particle velocity
action are limited to infinitesimal motion of the tool and and is proportional to the resultant force acting on the par-
the given geometry of the problem. These methods were ticle with a proportionality constant (damping coefficient)
not expected to be valid for the analysis of the subsequent C [17]. For a detailed description of DEM, the reader is
stages of advanced earth digging problems [12]. The analyt- referred to Cleary and Sawley [18], Cundall and Strack
ical methods are based on Terzaghi’s passive earth pressure [16], Hogue [19] and Zhang and Whiten [20].
theory and assumptions of a preliminary soil failure pattern
[13]. Complicated tool geometry (such as buckets) and large 3. Experimental
deformations cannot be modelled using these methods [14].
The discrete element method is a promising approach to Two parallel glass panels were fixed 200 mm apart to
model soil-implement interaction and can be used to over- form the soil bin. The bucket profile was fixed to a trolley
come some of the difficulties encountered by analytical which was driven by a ball screw and stepper motor. The
methods [15]. In DEM, the failure patterns and material
deformation are not needed in advance. The tools are mod-
elled using a number of flat walls and the complexity of the
tool geometry does not complicate the DEM model. Large
deformation in the granular material and the development
of the granular material free surface are automatically han-
dled by the method.
Cleary [5] modelled dragline bucket filling using DEM.
Trends were shown and qualitative comparisons made, but
no experimental results were presented. The process of
hydraulic excavator bucket filling was investigated experi-
kn
mentally by Maciejewski and Jarzebowski [12]. The aim of
their research was optimization of the digging process and
bucket trajectories. It is shown that the most energy efficient Friction µ ks
bucket is the one where the pushing effect of the back wall is
minimized. Owen et al. [21] modelled 3D dragline bucket fill-
ing. In there approach, the bucket was modelled with the
finite element method and the soil with DEM. Ellipsoids
and clumped spheres were used to approximate the particle
angularity. The bucket followed a prescribed path.
Esterhuyse [1] and Rowlands [2] investigated the filling
behaviour of scaled dragline buckets experimentally with
the focus on rigging configuration, bucket shape and teeth
spacing. They have shown that the aspect ratio of the
bucket (width to depth) plays and important role in the
drag distance needed to fill a bucket. The bucket with the
shortest fill distance was found to produce the highest peak
drag force. Fig. 1. DEM contact model.
C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227 219

complete rig could be set at an angle h to the horizontal as calibration checked regularly to avoid drift in the measure-
shown in Fig. 2a. The first arm was then rotated and fixed ments. For rig angles other than zero, the force transducer
such that both arms remained vertical. The second arm was zeroed before the drag commenced. This compensated
remained free to move in the vertical direction. First, coun- for the component of the bucket weight that acted in the drag
terweights were added at position A (Fig. 2a) to balance direction. The vertical displacement of the bucket was mea-
the combined weight of the bucket profile and the second sured with a linear variable differential transformer (LVDT)
arm assembly. This resulted in a ‘‘weightless” bucket. and used as input to the DEM simulation. In both the exper-
Counterweights were then added at position B to set the iments and the DEM simulations the bucket was given a drag
‘‘effective” bucket weight. Since arm 2 was always vertical velocity of 10 mm s 1. The dimensions of the bucket profile
even for rig angles other then zero, the effective bucket are shown in Fig. 2b.
weight always acted vertically downwards (Fig. 2c). Bucket In this study corn grains were used. Although the corn
weights of 49.1 N, 93.2 N, 138.3 N and 202.1 N were used. grains are not real soil, Rowlands [2] observed that seed
When the bucket was dragged in the direction as indi- grains are suitable for experimental testing and closely
cated, it was also free to move in the vertical direction as resemble natural soil flow into dragline buckets.
a result of the effective bucket weight and the force of the
grains acting on it. The bottom edge of the bucket was 4. DEM parameters and numerical model
always set to be parallel to the drag direction and the mate-
rial free surface. This type of motion resembles that of a Fig. 3 shows the range of measured grain dimensions
dragline bucket which is dragged in the drag direction by and the equivalent DEM grain. A normal distribution
a set of ropes, but with freedom of motion in all other within the range of dimensions given was used to create
directions [2]. the clumped particles. Clumps can be formed by adding
Spring loaded Teflon wipers were used to seal the small two or more particles (discs in 2D and spheres in 3D)
opening between the bucket profile and the glass panels. A together to form one rigid particle, i.e. particles included
force transducer was designed and built to measure the drag in the clump remain at a fixed distance from each other
force on the bucket. A set of strain gauges was bonded to a [17]. Particles within a clump can overlap to any extent
steel beam of which the position is shown in Fig. 2a. The and contact forces are not generated between these parti-
set of four strain gauges was used to measure the force in cles. Clumps cannot break up during simulations regardless
the drag direction. Other force components were not of the forces acting upon them. In the model 20,000–30,000
measured. The force transducer was calibrated and the clumped particles were used.
A calibration process, presented in another paper, was
developed for cohesionless material. The particle size, shape
a Counter weights
and density were determined from physical measurements.
A The laboratory shear tests and compressions tests were used
2nd Arm B to determine the material internal friction angle and stiffness
1st Arm
respectively. These tests were repeated numerically using
DEM models with different sets of particle friction coeffi-
Direction of Direction of drag cients and particle stiffness values. The combination of shear
vertical motion test and compression test results could be used to determine a
θ unique set of particle friction and particle stiffness values,
Table 1.

a 5-9 b
Force transducer
R 2.5 - 4.5

5-6
8 - 12

3.0 - 5.0

b c
150 mm

Max volume
Wb θ Wb⋅cosθ

35 mm 4-5 R 1.5 - 3.0


100 mm
45° θ 3-6
200 mm
Fig. 3. (a) Physical grain dimensions and (b) DEM grain model.
Fig. 2. Experimental setup. Dimensions in (mm).
220 C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227

In the software used, PFC2D, so-called walls are used to 120


build structures. The test rig and the bucket, with the same Wb = 202.1 N
dimensions as in the experiment, were built from walls. The 100 138.3 N
walls are rigid and move according to prescribed transla-

Vertical displacement [mm]


93.2 N
tional and rotational velocities. The forces and moments
acting on the walls do not influence the motion of the wall. 80
During the experiments a constant drag velocity of 49.1 N
10 mm s 1 was applied while the vertical displacement 60
was measured. The vertical displacement was influenced
by both the rig angle and the effective bucket weight. A typ- 40
ical result is shown in Fig. 4. Except for the initial transi-
tion, the vertical velocity was nearly constant, for a given
setup, and increased with an increase in bucket weight. In 20
the DEM model, the drag velocity was set to 10 mm s 1
and the measured vertical displacement was read from a
0 100 200 300 400 500 600 700
data file and applied to the bucket.
Drag displacement [mm]
Standard functions build into PFC2D were used to
obtain the forces and moments acting on individual walls Fig. 4. Measured vertical displacement of the bucket with h = 10° and
and on the bucket as a whole. For rig angles other than four values of effective bucket weight Wb.
zero, the rig was kept horizontal but the gravity compo-
nents were set accordingly.
The values and the positions where they were measured
5. Results and discussion are indicated in the figures. Taking the nominal particle
size as 10 mm, DEM predicted the heap height accurately
It is difficult to make quantitative comparisons regard- within 1.5–4.5 particle diameters.
ing flow patterns. When comparing the material free Fig. 7 shows typical drag forces obtained from experi-
surface, some comparisons could however be made. Figs. ments and simulations. The large jump in the drag force
5 and 6 show how the material flowed into the bucket for at the beginning of the experiment was observed in most
rig angles of h = 0° and h = 20°, respectively. When com- of the runs and could not be explained and needs further
paring the shape of the material free surface, the simula- investigation. From this result, it is clear that the DEM
tions were able to predict the general shape during the model captured the general trend in drag force, but it pre-
initial stages of filling. The simulations however failed to dicted lower values compared to the measured values. Over
accurately predict the material free surface during the final the complete drag of 800 mm, the model predicted a force
stages of filling. which was 15–50 N lower than the measured force. At the
Curves were fitted to the experimental free surface and end of the drag the error was 20%. The frictional force
overlaid on the numerical results in Figs. 5 and 6. The max- between the Teflon wipers and the glass panels was mea-
imum difference between the two free surfaces (heap sured in a run without grains. This frictional force was sub-
height) was measured along the direction perpendicular tracted from the measured drag force. Frictional forces
to the drag direction. Two measurements were made, one between the grains and the side panels would also have
where DEM predicted a higher heap height, and one an influence on the measured results. These frictional forces
measurement where DEM predicted a lower heap height. could not be measured or included in the 2D DEM model
and might be the reason why the model predicts lower drag
Table 1 forces [6].
Summary of corn properties and DEM parameters used. The drag energy was defined as the area under the drag
Macro property Measured DEM force–displacement curve. Making use of different rig
Internal friction angle 23° 24° angles h and effective bucket weights Wb, the drag energy
Angle of repose 25 ± 2° 24 ± 1° E700 up to a displacement of 700 mm is compared in Fig. 8.
3 3
Bulk density 778 kg m 778 kg m The first observation that could me made was that with
Confined bulk modulus 1.60 MPa 1.52 MPa an increase in effective bucket weight, for a given rig angle
Material-steel friction 14° 14°
h, there was a linear increase in required drag energy. A
Calibrated DEM properties closer investigation showed that with an increase in bucket
Particle stiffness, kn = ks 450 kN/m weight, the bucket was forced deeper into the material
Particle density, qp 855 kg/m3 which caused a higher drag force when compared to a
Particle friction coefficient, l 0.12 bucket with less weight.
Other properties The second observation that can be made is that with an
Damping, C 0.2 increase in the rig angle, there is a decrease in drag energy.
Model width 0.2 m
The effective bucket weight Wb always acted vertically
C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227 221

Fig. 5. Bucket filling results with rig angle h = 0°.

downward (Fig. 2c) so that the normal force pushing the Using the simulation results it was possible to identify
bucket into the material is given by Wb  cos (h). Thus, with how much of the total force was exerted on each of the
an increase in rig angle, there is a decrease in the normal bucket sections. In Fig. 9 the bucket was divided into six
force pushing the bucket into the material. This caused a sections. The graphs show, as a ratio of the total drag
reduction in the drag force, and hence a reduction in the force, the force on each of the sections. From the start
drag energy, when compared to results using a lower rig up to a displacement of 200 mm (25% of total displace-
angle. The DEM simulations were able to capture the gen- ment) the total force acted mainly on the lip and the bot-
eral trends, but it predicted drag energies lower than the tom section. As material started to flow into the bucket,
measured. The reason for this is that the predicted drag the other sections came into play, first the inner curve
forces were too low due to the exclusion of the friction and finally the front section. Less than 5% of the force
forces between the grains and the glass panels. It would, acted on the top section. This was far less than the bottom
however, still be possible to use the simulation results for section (30%). The reason for this is that the material inside
qualitative optimization of bucket filling. the bucket showed little movement relative to the bucket
222 C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227

Fig. 6. Bucket filling results with rig angle h = 20°.

and the pressure on the top section was only due to the Theory. He observed that definite planes of shear (rupture)
weight of the material inside the bucket. On the bottom formed between distinct moving material regimes. These
section, the pressure was due to the combined weight of shear planes changed orientation and location depending
the material inside the bucket and the weight of the bucket on initial setup and during different stages of the filling pro-
itself. During the complete filling process, 20–30% of the cess itself. The generalised theory is shown in Fig. 10. The
drag force acted on the lip. This shows that the design of different flow regions, as named by Rowlands [2], are indi-
the lip and teeth is important. It is well known that the cated on the figure. The movements of the material relative
length of the lip/teeth and the angle of attack are important to the bucket are indicated by the arrows.
factors influencing bucket filling [2] . The virgin material remains largely undisturbed until the
Rowlands [2] made use of mixtures of millet, peas and final third of the drag during which ‘‘bulldozing” occurs.
corn in his 2D test rig. The observation of the filling behav- The initial laminar layer flows into the bucket during the
iour led to the development of a theory that describes the first third of the drag (Fig. 10a). After entering to a certain
flow characteristics and patterns of material entering the distance, this layer fails at the bucket lip and subsequently
bucket. Rowlands [2] named this concept the Shear Zone becomes stationary with respect to the bucket for the
C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227 223

250
Front
0.5 Front
Experiment Inner curve
Top Inner curve
200 Lip
0.4 Top
Bottom Lip
Outer curve

Drag force ratio


Outer curve Bottom
Drag force [N]

150 0.3

Simulation
0.2
100

0.1
50

0
0
100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Displacement in drag direction [mm] Displacement [mm]

Fig. 7. Typical bucket drag forces with rig angle h = 10° and a bucket Fig. 9. Bucket drag force distribution with h = 10°.
weight Wb = 138.3 N.

remainder of the drag (Fig. 10b and c). At steeper drag


a
angles, the material flows more rapidly towards the rear
because of the added gravitational assistance. This effect
can be seen by comparing Figs. 5 and 6. Initial laminar layer Active dig zone Virgin material
With the laminar layer becoming stationary, a new zone,
the active flow zone, develops (Fig. 10). In this zone, the
material displacement is predominantly in the vertical
direction. The active dig zone is located above the teeth Shear line
and bucket lip. This area develops as material starts to
enter the bucket and increases in size after failure of the ini- b
tial laminar layer. In this zone, the virgin material fails and Initial laminar layer Active flow Active dig Virgin
either flows into the bucket as part of the laminar layer zone zone material
during the first part of filling or moves into the active flow
zone during the latter part of filling.
The dead load that has resulted from ‘‘live” material in
the active flow zone ramps up and over the initial laminar
layer. Some of the material in the initial laminar layer fails Shear line

120 c
Dead load Active flow Active dig Virgin
110 zone zone material

100

90
E700 [J]

80
Dead load Initial laminar Shear line
70 shear line layer

60 Fig. 10. The Shear Zone Theory according to Rowlands [2].


θ = 0° θ = 10°
θ = 20°
50 Experiment
Simulation
40 and starts to form part of the dead load (Fig. 10c). During
40 60 80 100 120 140 160 180 200 220 experiments and while the material was flowing, a definite
Wb [N] rupture or shear line could be observed here. With an
Fig. 8. Bucket drag energy E700 as a function of the bucket weight Wb for increase in drag angle, the active dig zone and active flow
different rig angles h. zone tended to join into one continuous band.
224 C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227

Fig. 11. Flow regions using the particle–bucket displacement ratio.

It should be noted that Fig. 10 only shows three stages 100 mm. The displacement vector of each particle was then
of the filling process, but in reality there is a gradual tran- set to be zero after which the bucket was given a further
sition from one stage to the next. It should also be noted displacement of 10–15 mm (1–3 particle lengths). The par-
that this is a generalised theory and there will be variations ticle displacement ratio PDR was defined as the ratio of the
in the results when different materials and bucket geome- magnitude of the particle absolute displacement vector to
tries are used. During experiments two definite shear lines the magnitude of the bucket absolute displacement vector.
could be observed. The one extended from the tip of the The particles were then coloured according to their individ-
lip up to the free surface. This is named the cutting shear ual PDR values. A PDR equal to unity means that the par-
line. The second line is the one between the initial laminar ticle is moving with the bucket. The result is shown in
layer and the dead load, called the dead load shear line. Fig. 11. This is in effect the average velocity ratio over a
Making use of DEM and investigating the flow regions short period.
further, the following procedure was devised. The bucket The flow regimes as predicted by the Shear Zone Theory
was moved through the material and ‘‘paused” after each are indicated on the figure. The three pictures correspond
C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227 225

to the three pictures given in Fig. 10. After a displacement 60


of 100 mm, the active dig zone is clearly visible with + 6%
0.40 6 PDR < 0.65. The initial laminar layer moves into 50
- 6%
the bucket with 0.10 6 PDR < 0.25. This corresponds well
to the flow zones shown in Fig. 10a.
After 500 mm, the characteristic ‘‘V” shape of the active 40

Simulation %
flow zone can be seen with 0.10 6 PDR < 0.25. Although
the PDR is relatively low, the displacement is predomi- 30
nantly in the vertical direction. The active dig zone is still
present and in the back of the bucket, the initial laminar 20
layer starts to become stationary relative to the bucket. θ = 0°, Wb = 49.1 N
θ = 10°, Wb = 49.1 N
This is visible by the PDR values that increase towards θ = 20°, Wb = 49.1 N
the back of the bucket. This corresponds well to the flow 10 θ = 0°, Wb = 138.3 N
θ = 10°, Wb = 138.3 N
zones shown in Fig. 10b. θ = 20°, Wb = 138.3 N
After 800 mm the existence of the dead load shear line is 0
clearly visible. When compared to Fig. 10c, the active flow 10 20 30 40 50 60
Experimental %
zone and active dig zone cannot be distinguished from the
dead load. The reason for this is that at a bucket displace- Fig. 13. Comparison between experimental and DEM fill percentages.
ment of 800 mm, the bulldozing effect is large and over-
shadows the other flow zones.
Dragline bucket optimization is very important in terms filled in 2–3 bucket-lengths. With an increase of the rig
of force and energy requirements and cycle time. In some angle from 0° to 10°, there is a slight increase in fill percent-
applications it would be advantageous to fill the bucket age towards the latter stages of filling. This is due to the
using the minimum amount of energy. In other applica- fact that when material is disturbed, it flows more easily
tions, it would be advantageous to fill the bucket as quickly into the bucket. When the rig angle is further increased
as possible to decrease cycle time [1]. To investigate fill to 20° there is, however, a decrease in fill percentage. A fur-
rates, images from the experiment were taken at different ther investigation showed that with an increase in rig angle,
stages of filling, the outline of the material digitized, and the bucket displacement into the material is less. It has
the volume of material inside the bucket calculated and been shown that the force perpendicular to the material
expressed as a percentage of the maximum bucket volume. surface is given by Wb  cos (h). Hence, with an increase
The maximum bucket volume of 0.0146 m3 is defined in in the rig angle, the force component forcing the bucket
Fig. 2b. Using the DEM results, the same procedure was to dig in, decreases. When this force component decreases,
followed and the results compared. the penetration depth of the bucket into the material is
Fig. 12 shows the experimental results using three differ- reduced and the bucket scoops up less material. When
ent rig angles. The bucket fill percentage is plotted against the bucket scoops up less material, there is a decrease in fill
bucket displacement in terms of bucket-lengths. In the percentage.
dragline industry, the target is to get the bucket completely The comparison between experimental and DEM fill
percentages is summarised in Fig. 13. Using three rig
100 angles h = 0°, 10° and 30° and two effective bucket
weights Wb = 49.1 N and 138.3 N, the fill percentage
90
was calculated at displacements of 100, 200, 300, 400,
80 500, 600 and 700 mm. The 42 data points were plotted
70
and the two lines indicate that in all cases, except for
two, the DEM results were within ±6% of the experi-
Bucket fill %

60 mental results.
50 In practice, the bucket is rotated to prevent the majority
of the material to fall out when the bucket is disengaged.
40
This principle is depicted in Fig. 14 where, at the end of
30 its displacement, the bucket was lifted out of the material
20
and kept at the rig angle. The effect of bucket orientation
θ = 0° is clear on the amount of material that the bucket could
10 θ = 10° hold. Again, the experimental free surface outline is shown
θ = 20°
0 on the DEM results with good agreement for h = 0°. For
0.5 1 1.5 2 2.5
h = 20°, the DEM model predicts additional material in
Displacement [bucket length] the back of the bucket which can be explained by the differ-
Fig. 12. Bucket fill percentage as a function of bucket displacement for ence in the final fill state as seen in Fig. 6 at a displacement
different rig angles. of 800 mm.
226 C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227

Fig. 14. The effect of bucket orientation on the maximum holding capacity.

6. Conclusions 6. The DEM models, using different rig angles and bucket
weights, are able to predict the volume of material inside
The main objective of this paper was to demonstrate the bucket accurately within ±6%.
how accurately the discrete element method can predict
the process of excavator bucket filling. The flow patterns
of material entering the bucket, drag force acting on bucket References
due to material interaction, energy requirements and the
bucket fill rates were compared to experimental observa- [1] Esterhuyse SWP. The influence of geometry on dragline bucket filling
tions and measurements. The study was limited to cohe- performance. M.Sc. Thesis Mechanical Engineering, University of
Stellenbosch, Stellenbosch, South Africa; 1997.
sionless granular material and two-dimensional models. [2] Rowlands JC. Dragline bucket filling. Ph.D Thesis, University of
The conclusions of the paper are: Queensland, Queensland, Australia; 1991.
[3] Poschel T, Saluena C, Schwager T. Scaling properties of granular
1. Comparing the material free surface, DEM can accu- materials. Phys Rev E 2001:64.
[4] Poschel T, Saluena C, Schwager T. Scaling properties of granular
rately model the flow of material into the bucket during
materials. In: Vermeer PA, Diebels S, Ehlers W, Herrmann HJ,
the initial stages of filling. During the latter stages of fill- Luding S, Ramm E, editors. Continuous and discontinuous modelling
ing DEM, however, fails to accurately predict the mate- of cohesive frictional materials. Berlin: Springer; 2001. p. 173–84.
rial free surface. [5] Cleary PW. The filling of dragline buckets. Math Eng Ind 1998;
2. DEM can accurately predict the general trend in bucket 7(1):1–24.
drag force. Over the complete drag of 800 mm DEM [6] Maciejewski J, Jarzebowski A, Trampczynski W. Study on the
efficiency of the digging process using the model of excavator bucket.
predicts a drag force 15–50 N lower than the measured J Terramech 2004;40:221–33.
values. The maximum measured drag force is 250 N [7] Dechao Z, Yusuf Y. A dynamic model for soil cutting by blade and
while DEM predicts a maximum drag force of 200 N. tine. J Terramech 1992;29(3):317–27.
3. DEM fails to accurately predict the drag energy. The [8] Ibarra SY, McKeys E, Broughton RS. A model of stress distribution
and cracking in cohesive soils produced by simple tillage implements.
general trends are however correct and it is shown that
J Terramech 2005;42:115–39.
the drag energy increases linearly with an increase in [9] Onwualu AP, Watts KC. Draught and vertical forces obtained from
bucket weight. dynamic soil cutting by plane tillage tools. Soil Till Res 1998;48:
4. Based on the DEM results, between 20% and 30% of the 239–53.
total bucket force acts on the lip. With the current [10] Osman MS. The mechanics of soil cutting blades. J Agric Eng Res
1964;9(4):313–28.
experimental setup this cannot be validated.
[11] Swick WC, Perumpral JV. A model for predicting soil–tool interac-
5. The DEM results show good agreement with the Shear tion. J Terramech 1998;25(1):43–56.
Zone Theory. Based on qualitative comparisons, DEM [12] Maciejewski J, Jarzebowski A. Laboratory optimization of the soil
could predict the position of the initial laminar layer, digging process. J Terramech 2002;39:161–79.
the active dig zone, the active flow zone and the dead load. [13] Terzaghi K. Theoretical soil mechanics. New York: Wiley; 1943.
C.J. Coetzee, D.N.J. Els / Journal of Terramechanics 46 (2009) 217–227 227

[14] Karmakar S, Kushwaha R. Dynamic modelling of soil–tool interac- [19] Hogue C. Shape representation and contact detection for discrete
tion: an overview from a fluid flow perspective. J Terramech element simulations of arbitrary geometries. Eng Comput 1998;15(3):
2006;43(4):411–25. 374–90.
[15] Asaf Z, Rubinstein D, Shmulevich I. Evaluation of link-track [20] Zhang D, Whiten WJ. The calculation of contact forces between
performances using DEM. J Terramech 2006;43:141–61. particles using spring and damping models. Powder Technol 1996;88:
[16] Cundall PA, Strack ODL. A discrete numerical method for granular 59–64.
assemblies. Geotechnique 1979;29:47–65. [21] Owen DRJ, Feng YT, de Souza Neto EA, Cottrell M, Wang F,
[17] Itasca. PFC2D Theory and background manual 1999: version 2.0. Andrade Pires FM, et al. The modelling of multi-fracturing solids and
<www.itascacg.com>. particulate media. In: Mang HA, Rammerstorfer FG, Eberhardste-
[18] Cleary PW, Sawley ML. Three-dimensional modelling of industrial iner J, editors. Fifth conference on computational mechanics, July 7–
granular flows. In: Second international conference on CFD in the 12, 2002, Vienna, Austria.
minerals and process industries; 1999. p. 95–100.

You might also like