Mathematical and Computer Modelling 44 (2006) 816–833

www.elsevier.com/locate/mcm

Comparison of measured and simulated motion of a scaled dragline

excavation system
Jason Kyle, Mark Costello ∗
Department of Mechanical Engineering, Oregon State University, Corvallis, OR. 97331, United States

Received 8 August 2005; received in revised form 13 February 2006; accepted 21 February 2006

Abstract

Motion measurements obtained from a maneuvering 1/16-scale dragline excavation system using a six camera motion capture
system are compared against motion predictions of the same scenario using a complex dragline excavation system simulation.
The dynamic model includes motion of the bucket, spreader bar, hoist cluster, and drag cluster, along with flexible rope and
chain segments including the hoist rope, drag rope, dump rope, hoist chains, and drag chains. In general, measured and simulated
motions agree quite favorably. However, a consistent difference between the measurements and the simulation model is damping
of oscillations, where it is noticed that measurement data tend to damp more rapidly than simulation model results.
c 2006 Elsevier Ltd. All rights reserved.

Keywords: Dragline; Excavator; Simulation; Validation; Dynamic model; Motion capture

1. Introduction

Surface mining extracts underground resources that reside close to the Earth’s surface and involves a process
of moving soil covering the resource to a storage location and subsequently removing the resource. Overburden,
defined as soil covering a resource, is most efficiently moved by a dragline excavation system in many surface mining
scenarios. As shown in Figs. 1 and 2, a dragline excavation system is fitted with an open bucket supported above by
a cable hanging from the end of a long boom. The bucket is dragged along the ground by a cable until it is filled
with earth, which is then dumped elsewhere. Draglines are used primarily in high-volume overburden removal and
excavation of deep holes. The complete system contains many dynamic components, including the excavator, boom,
bucket, spreader bar, drag coupler, hoist coupler, hoist chains, drag chains, hoist rope, drag rope, and dump rope. The
bucket is indirectly controlled by an operator in the excavator through rotation of the boom about a vertical axis and
payout and windup of the hoist and drag ropes. Lowering and lifting of the bucket is accomplished mainly with the
hoist rope, while longitudinal motion of the bucket is accomplished mainly through the drag rope.
Physically, dragline excavator systems are typically very large-scale devices. Common medium-sized dragline
excavation systems weigh 500 metric tons with a bucket capable of moving 23 m3 of material in a single cycle. An
average-sized boom is 91 m long. Not only are dragline excavation systems physically big, they are equally large from

∗ Corresponding author.
E-mail address: [email protected] (M. Costello).

c 2006 Elsevier Ltd. All rights reserved.

0895-7177/$ - see front matter
doi:10.1016/j.mcm.2006.02.015
 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833 817

Fig. 1. Dragline excavation system schematic.

Fig. 2. Dragline bucket schematic.

an economic perspective. Surface mining at large sites is big business, with high revenue opportunities as well as high
operating costs. In many surface mining operations, removal of overburden dominates the cost side of operating a
mine. Relatively small improvements in dragline excavator system performance such as cycle time, bucket capacity,
working envelope, component life, etc. can map into sizeable economic gains for the mine operator.
For slowly operating dragline excavation systems, static system models provide a satisfactory estimate of
component loads that is utilized in structural sizing of system components. Static system models also provide bucket
position and orientation information for various hoist and drag rope lengths and various bucket riggings that are useful
for site excavation planning. In cases where relatively rapid maneuvering of the dragline system is encountered, the
use of a dynamic dragline excavation model is more appropriate for estimating bucket orientation excursions and
maximum loads on different components.
818 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833

Several research groups have developed dragline excavation system models at various levels of sophistication
for different analytical purposes. McCoy and Crowgey [1] investigated static and dynamic dragline bucket tightline
control. Using simple geometric analysis, they constructed both static and dynamic tightline limits based on the
length of the drag and hoist ropes, and subsequently used these limits as part of an anti-tightline control system.
Haneman et al. [2] reported on dragline performance evaluations using physical modeling. Rigging geometry and
bucket size were optimized using physical modeling and compared against field data. Rowlands and Just [3]
documented experimental results on a dragline bucket digging and filling capability. They concluded that design
changes to conventional dragline buckets can be made to improve performance by reducing filling times and dig energy
consumption. Knight and Shanks [4] investigated dragline productivity improvements through short term monitoring.
A productivity benefit of 2% was realized using a monitoring system that computes an optimized bucket rigging.
Pathak et al. [5] generated a method to compute the working zone of a dragline bucket. A graphical approach for the
determination of the bucket path profile of a standard dragline was indicated, considering the simple three-period duty
cycle in each operation. Hainsworth et al. [6] used machine vision techniques to measure the location of a dragline
bucket in space remotely. Using only a single camera, an image segmentation process is used to classify the bucket
and to identify its position in the scene. Cleary, [7] investigated dragline bucket filling using a discrete element model
to simulate the filling process. Ridley and Corke [8] developed a technique to estimate dragline bucket pose under
gravity loading. Their model is based on static equilibrium of the loads on the different components of the system.
The intricate geometric arrangement of different dragline excavation system components, coupled to the fact that
each component can significantly alter motion characteristics, leads to the necessity of employing a fairly complex
dynamic simulation model to replicate physical behavior. In the work reported here, measured and simulation motion
for an example 1/16-scale dragline excavation system are compared. Measurements are recorded with an optical
motion capture system. The dynamic model includes three-dimensional (3-D) rigid body motion of the bucket,
spreader bar, drag cluster, and hoist cluster, as well as flexible motion of all chains and ropes. A typical dig maneuver,
without a bucket load, is presented in the comparison of results.

2. Dragline excavation system dynamic model

The dynamic event considered consists of the motion of a complete dragline excavation system. As shown in Figs. 1
and 2, the total system contains four rigid bodies and nine ropes. The rigid bodies include the bucket, spreader bar,
hoist coupler, and drag coupler. The ropes include the drag rope, dump rope, hoist rope, left drag chain, left lower
hoist chain, left upper hoist chain, right drag chain, right lower hoist chain, and right upper hoist chain. The surface
of the ground is assumed to be a satisfactory inertial reference. As shown in Fig. 1, the inertial reference frame is a
right-handed coordinate system defined with its origin on the ground surface located below the excavator at the point
of azimuth rotation of the excavator. The IEI and JEI axes form the ground plane, while the KE I axis completes the
right-handed triad and points into the ground.
The ropes (drag rope, dump rope, and hoist rope) and chains (left drag chain, left lower hoist chain, left upper
hoist chain, right drag chain, right lower hoist chain, and right upper hoist chain) are dynamically modeled in the
same manner. In this paper, all ropes and chains are called “rope segments”. Differences in dynamic characteristics
are realized through appropriate physical properties for a particular rope or chain. The rope segments of the dragline
excavator system couple the various system components together. The motion of the end-points of a rope segment
largely influences the dynamic behavior of the rope segment as a whole. For a rope segment, each edge is associated
with either the beginning or end of the rope segment. For example, the left lower hoist chain is connected to the bucket
and the left side of the spreader bar. The connection with the bucket is defined as the beginning connection point, while
the connection with the left lower hoist chain is defined as the ending connection. Some rope segments interact with
a pulley in the middle of the rope segment. For example, the dump rope is connected to the hoist coupler pulley.
The bodies that constitute the dragline excavator system (bucket, spreader bar, hoist coupler, drag coupler) are
linked together by rope segments. Generally, each body has several rope segments attached to it. The weight of the
body and the forces and moments generated by the rope connections dictate the motion of the individual bodies. For
example, the bucket has five rope segments connected to it, namely the left drag chain, the right drag chain, the left
lower hoist chain, the right lower hoist chain, and the dump rope.
Each rope segment is dynamically modeled as a set of particles connected by springs and dampers. Thus, all the
mass elements of the system, including all bodies (bucket, spreader bar, hoist coupler, and drag coupler) and all
 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833 819

Fig. 3. Body dynamic model schematic.

particles on each rope segment (drag rope, dump rope, hoist rope, left drag chain, left lower hoist chain, left upper
hoist chain, right drag chain, right lower hoist chain, and right upper hoist chain), are connected to other system mass
elements by springs and dampers. Connection forces are simply treated as external forces acting on the body. Thus,
derivatives of the state vector can be directly computed from the state vector in a straightforward manner.

2.1. Generic body equations of motion

The dynamic equations for all body elements assume the same structural form. This section develops the basic
dynamic equations of motion used for any body element. Each body is modeled as being rigid and undergoes three-
dimensional motion. Hence, 6 degrees of freedom or 12 state variables are required to describe the position and
orientation of each rigid body at a given instant in time. The degrees of freedom for each body include three position
components of the mass center of the body as well as three Euler orientation angles of the body. A schematic of a
general rigid body of the system is shown as Fig. 3. The translational dynamics of the mass center of an arbitrary body
are given by Eqs. (1) and (2):
   
ẋ B   u B 
ẏ B = v B (1)
ż B wB
   

 u̇ B   X B /m B 
   

v̇ B = Y B /m B (2)
ẇ B Z B /m B + g
   

In Eq. (2), m B is the mass of the body and X B , Y B , Z B are the components of the external rope coupling forces on
the body expressed in the inertial reference frame. The orientation of a body element is defined by a sequence of
three body fixed Euler angle rotations [9] shown in Fig. 4. Starting from the inertial reference frame, a rotation of θ B
is executed about the JEI axis. The resulting rotated reference frame is called the O frame. Next, the O frame is rotated
about the IEO axis by the angle φ B . The resulting reference frame is denoted the T frame. The T frame is subsequently
rotated about the KE T axis by the angle ψ B , yielding the body reference frame. The angles φ B , θ B , and ψ B are the
Euler angles associated with the body. The kinematic relationship between time derivatives of the Euler angles and
body frame angular velocity components is shown in Eq. (3):

 φ̇ B 
    
cψ B −sψ B 0  pB 
θ̇ = sψ B /cφ B cψ B /cφ B 0 q B (3)
 B
ψ̇ B sψ B t φ B cψ B tφ B 1 rB
 
820 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833

Fig. 4. Bucket Euler angle diagrams (each rotation shown with other angles zero).

a. Continuous rope segment.

b. Discrete rope segment.

Fig. 5. Rope dynamic model schematic.

Eq. (3) uses the following shorthand notation for trigonometric sine, cosine, and tangent functions: sα ≡ sin α,
cα ≡ cos α, and tα ≡ tan α. The rotational dynamic equation for a body is given in Eq. (4):
       
 ṗ B  LB  0 −r B qB  pB 
q̇ B = [I B ]−1  M B −  r B 0 − p B  [I B ] q B  (4)
ṙ B NB −q B p B 0 rB
     

In Eq. (4), I B denotes the mass moment of inertia matrix of the body about its own mass center and L B , M B , N B
represent the components of the total externally applied moment vector on the body about its own mass center
expressed in its own reference frame.

2.2. Generic rope segment equations of motion

The dynamic equations for all rope segments assume the same structural form. This section develops the basic
dynamic equations of motion used for any rope segment. Fig. 5a depicts a typical rope segment consisting of a single
length of rope with both ends attached to a moving body. Each rope segment is split into nr particles and nr + 1
elastic elements using truss finite elements. A standard linear solid visco-elastic element is used to connect particles.
 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833 821

Collectively, the motion of the particles defines the motion of the complete rope segment. Each particle on a rope
segment is a point mass possessing three translation degrees of freedom. Forces that drive the motion of the particles
include particle weight and adjacent visco-elastic element line forces. The dynamic equations for one particle on a
rope segment are structurally the same for all particles, so the formulas to follow are shown only for the ith particle
on an arbitrary rope segment:
   
ẋ Ri   u Ri 
ẏ R = v R (5)
 i  i 
ż Ri w Ri
     
 u̇ Ri  X − X Ti+1   0 
1  Ti
v̇ R = YT − YTi+1 + 0 (6)
 i  m Ri  i
ẇ Ri Z Ti − Z Ti+1 g
  

In Eq. (6), m Ri is the mass of the ith particle on a rope segment and g is the gravitational constant. The terms
X Ti , YTi , Z Ti and X Ti+1 , YTi+1 , Z Ti+1 represent the visco-elastic line force vector components expressed in the inertial
reference frame of the line elements adjacent to the ith particle.
The line forces are caused by strain of the rope and are directed parallel to the line. Rope line flexibility generates
resistive stiffness and damping forces caused by rope line segment extension and extension rate. In order to compute
the visco-elastic line forces, it is useful to define a rope segment position and velocity difference vector for each
rope segment. The difference vectors are formed by subtracting the position or velocity components of the i + 1
rope particle, or ending connection, from the ith rope particle component, or beginning connection. This results in
nr + 1 difference vectors for each rope. The terms 1x Ri , 1y Ri , and 1z Ri represent the components of the position
difference vectors, while the velocity difference vectors are used to construct the stretch rate, ṡ Ri . Using the rope
segment position difference and velocity difference data, an expression for the elastic line force can be formed
directly:

1x Ri 
   
 X Ri  F Ri
YR = q 1y Ri (7)
 i 1x 2Ri + 1y 2Ri + 1z 2Ri 1z Ri 
Z Ri
Eqs. (8) and (9) provide expressions for the elastic line force, FRi , in terms of the strain and strain rate of the visco-
elastic line element:

(K K v Ri K s Ri
• K v Ri v Ri + K s Ri )ṡ Ri + s Ri , s Ri − L Ri > 0
F Ri + F Ri = C v Ri (8)
C v Ri 
0, s Ri − L Ri ≤ 0
q
s Ri = 1x 2Ri + 1y 2Ri + 1z 2Ri (9)
In Eq. (8), K v Ri , Cv Ri and K s Ri are the viscous stiffness, viscous damping, and static stiffness coefficients
respectively, for an individual visco-elastic line element. Also, L Ri is the nominal unstretched line length of
an individual line element. All individual line elements for a given rope segment have the same static and
viscous stiffness coefficients, viscous damping coefficient, and unstretched line length. The first condition in Eq.
(8) represents the normal tension condition between two adjacent particles. In this case, the distance between
the two adjacent particles is greater than the unstretched elastic line element length, and a nonzero tension
force in the line persists. The second condition in Eq. (8) is the slack condition. It stipulates that, when the
unstretched elastic line element length is greater than the distance between two adjacent particles, the elastic force is
zero.
The number of state variables required to model a rope segment can be large. Consider a rope segment split into
20 particles. Since each particle requires six state variables and each segment requires an additional state variable,
this rope segment would require a total of 141 state variables to define the motion of the complete rope segment.
As the number of beads used to represent each rope and chain is increased, the simulation model converges to the
measured position. A balance between the accuracy of the model and number of states must be selected by the
analyst.
822 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833

2.3. Rope edge point position and velocity

Edge point position and velocity of a rope segment is required for visco-elastic line force computation. For edge
points of a rope segment that are fixed on a body of the system, expressions for the position and velocity are formed
from the motion of the connection point on the body. The inertial position vector components of the connection of
a rope segment and a body is constructed as the distance from the inertial reference frame to the mass center of the
body plus the distance vector from the body mass center to the connection point:
 B

 S LC − S L B 
   
xC  x B   
yC = y B + [TB→I ] B L CB − B L B (10)
zC zB B
     
W LC − W L B
 

The components xC , yC , z C in Eq. (10) represent the inertial position components of a rope segment and body
interface. The inertial location of the mass center of the body is denoted as x B , y B , z B . The second term in Eq.
(10) is the inertial position vector components of the distance from the mass center of the body to the interface
point. The matrix TB→I is an orthonormal transformation relating the body reference frame to the inertial reference
frame:
 
cψ B cθ B + sψ B sφ B sθ B −sψ B cθ B + cψ B sφ B sθ B cφ B sθ B
TB→I =  s ψ B cφ B cψ B cφ B −sφ B  (11)
−cψ B sθ B + sψ B sφ B cθ B s ψ B s θ B + cψ B s φ B cθ B cφ B cθ B

The components S L CB , B L CB , W L CB are the body frame coordinates of the interface point, while the components
S L B , B L B , W L B are the body frame coordinates of the body mass center. The velocity of the interface point with
respect to inertial space is obtained by differentiating Eq. (10).
Some rope segment edge points are not connected to bodies included in the dynamic system. For example, the
end-point of the hoist rope is connected to the hoist reel. For end-points not attached to a body element of the system,
the motion of the connection point is known as a function of time. Prior specification of an excavation maneuver
defines the inertial position and velocity of the hoist reel, drag reel, and boom points. Furthermore, definition of an
excavator maneuver also yields the length of line of the hoist and drag reels as a function of time. Computationally,
any prescribed motion variables are determined by linear interpolation of a table of data with time as the abscissa and
the prescribed motion parameter as the ordinate.

2.4. Body forces and moments

The forces and moments acting on a body of the dragline excavation system consist of body weight and elastic
line forces from rope segments connected to the body. Each rope segment is divided into nr particles and nr + 1
adjoining elastic elements. The visco-elastic forces applied to a body are generated from either the first or last
elements, depending on whether the connection is a beginning or ending connection point for the rope segment.
Beginning connections use the first line element force of a rope segment, while ending connections use the last line
element force of a rope segment:
 
 X Te j 
 
X B X nc  
YB = a j YTe j (12)
ZB
  j=1   ZT  
ej

In Eq. (12), the index e j is either 1 or nr + 1, depending on whether the jth rope segment connection is a beginning
or ending connection. When the connection is a beginning connection, a j = −1; when the connection is an ending
connection, a j = 1. The moment about the mass center of a body due to the connection forces shown above is
computed as a cross-product of the distance vector from the mass center to the connection location and the connection
force. Both the distance vector and the connection forces are expressed in the body reference frame, so the resulting
moments are also expressed in the body reference frame as required in the equations of motion:
 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833 823

Fig. 6. 1/16-scale dragline excavator rigging.

 
 X Te j 
 
LB  X c  
MB = a j [SC j ][TB→I ]T YTe j (13)
NB
  j=1  ZT 
 
ej

In Eq. (13), SC j is a matrix cross-product operator.

2.5. Pulley modeling

The dragline excavation system utilizes pulleys in rigging various ropes. The boom pulley is attached to the boom
and located at the top of the boom. It interfaces with the hoist rope and directs the hoist rope upward from the hoist
coupler and then downward to the hoist rope reel on the excavator body. The hoist coupler pulley guides the dump
rope from the bucket to the drag coupler. For the purposes of dynamic modeling, pulleys serve to change the position
of a rope segment by enforcing a constraint that part of the rope segment must lie on the pulley. Each pulley is modeled
as a small ring through which a rope segment passes. The tension in the elastic line segment exerts a force on the body
that the pulley is connected.

3. Dragline excavation system motion measurement system

A 1/16-scale dragline excavator system is the basic system examined. Motion of various bodies, ropes, and chains
are measured with a three-dimensional optical motion capture system. The finite capture volume of the optical system
covers a majority of the operating range of the 1/16-scale excavator system.

3.1. Test platform

As shown in Fig. 6, the scaled dragline excavator system consists of four rigid bodies: the bucket, spreader bar,
hoist cluster, and drag cluster. The bucket is approximately 203 mm wide by 381 mm long by 152 mm tall and weighs
just over 54 N. Weights, inertial properties and center-of-mass location for each body are listed in Table 1. The scaled
dragline excavator system also employs three ropes (drag, dump and hoist ropes), and six chains (left drag chain, left
lower hoist chain, left upper hoist chain, right drag chain, right lower hoist chain, and right upper hoist chain). Both
the drag and hoist lines are half-inch rope, the dump line is quarter-inch rope, and all chains are 9.5 mm closed-link
chain. Chain and rope properties, including mass, length, static and viscous spring constants, and viscous damping
coefficients, are listed in Table 2. Chain and rope spring constants and damping coefficients are found by directly
measuring the static spring stiffness K s ; the visco-elastic stiffness and damping coefficients are then determined by
an iterative process of comparing and matching measured and simulated rope data.
824 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833

Table 1
Physical properties of bodies

Parameter Bucket Drag cluster Hoist cluster Spreader bar

Mass, m B (kg) 5.4975 0.0876 0.3692 0.3900
C.G. location, S L B (m) 0.0933 0.0 0.0 0.0
C.G. location, B L B (m) 0.0 0.0 0.0 0.0
C.G. location, W L B (m) −0.0207 0.0 0.0 0.0
Inertia, I B X X (kg m2 ) 0.0712414 0.0000068 0.0000678 0.0059656
Inertia, I BY Y (kg m2 ) 0.0858639 0.0000298 0.0007728 0.0000502
Inertia, I B Z Z (kg m2 ) 0.0119312 0.0000244 0.0007321 0.0059385

Table 2
Physical properties of ropes

Rope segment Number of Rope Rope mass Static stiffness Viscous stiffness Viscous damping
particles length (m) (N/m) ((N/m)/m) ((N/m)/m) ((N/(m/s))/m)
Drag rope 6 1.829 0.06751 1,859 7439 74
Dump rope 5 1.088 0.07278 3,513 5269 53
Hoist rope 12 3.658 0.06751 1,859 7439 74
Left drag chain 4 0.610 0.55781 24,795 2755 14
Left lower hoist chain 2 0.183 0.55781 24,795 2755 14
Left upper hoist chain 3 0.305 0.55781 24,795 2755 14

Fig. 7. 1/16-scale dragline excavator system.

The excavator is composed of two separate structures: the boom pulley, which is mounted to the wall with a square
tube steel cantilever beam; and the excavator base, which is a 0.9 m by 0.9 m by 1.5 m steel frame vibration-isolated
and anchored to the floor opposite the boom pulley. The assembled system is shown in Fig. 7. The wall-mounted
boom pulley functions as both the boom and the boom pulley from Fig. 1. The excavator base provides a platform for
mounting drag and hoist reels. When assembled, the 1/16-scale dragline excavator system spans 4 m high by 5.7 m
long, allowing the bucket to operate in a 10.8 m2 area.
The drag and hoist reel systems are both composed of a rope spool, direct current motor, motor controller, and two
rope sheaves. The first sheave directs the rope onto the center of the reel, while the second sheave centers the rope on
the excavator base and defines the location of the reel. Drag and hoist rope windup and payout are controlled by an
operator positioned behind the excavator base. Rope speeds vary from ±1.3 m/s.
 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833 825

3.2. Instrumentation

The position of each body and a portion of the drag and hoist ropes is tracked using a 3-D optical position analysis
system from VICON Motion Systems [10]. Spherical reflective markers, 5 mm in diameter, are placed on each body
of interest. As shown in Fig. 7, six cameras strategically placed throughout the laboratory emit and collect visible
red light reflected from the markers. Using optical correlation techniques, the position of the center of each marker is
determined in three dimensions. This data is time stamped and stored in a text file, which is then used to determine
drag and hoist rope lengths, along with body position, velocity, orientation, and rotational velocities.
As shown in Fig. 6, mounting rods are fixed to each of the four system bodies to extend the markers away from the
body to reduce occlusion. Four markers are mounted to the bucket: two on either side of the lower hoist chain mount
and two on each side of the bucket nose. Three markers are mounted to the hoist cluster: one extended from the pulley
center, and one on either side of the hoist cluster base. Three markers are mounted to the spreader bar: one is directed
upwards from the cross bar, and one from both the left and right lower hoist chain mount points. The drag cluster has
only two markers extended from either side of the mass center. Three markers on a body define a plane which is used
to determine position and orientation data of the body. Two markers on a body define a single axis, which is used to
determine only position data of the body.
As shown in Fig. 7, a circle of visible red diodes is fixed around each camera lens. A filter inside the camera detects
the visible red light reflected from the markers, capturing a two-dimensional picture of the camera view. Data from
all six cameras are collected and used to determine the three-dimensional position of all visible markers. Each camera
has 1.2 Mega-pixel resolution and is capable of capturing marker position at a rate of 1000 Hz. Marker position is
determined within 1 mm at slower sample rates up to 200 Hz. The camera system is calibrated by first defining a
right-hand coordinate system with four static markers in the x–y plane. Next, two markers fixed a known distance
apart are passed throughout the entire capture area, allowing the system to determine the position of each camera
in the previously defined coordinate system. The marker position for subsequent captures is then determined in this
coordinate system.

3.3. Data processing

Data collected by the motion capture system is used to determine the position and orientation of the bucket, spreader
bar, hoist cluster, and drag cluster, along with the lengths of the drag and hoist ropes. Position and orientation data are
numerically differentiated using second-order finite differences to obtain linear and angular velocities for each body.
A running average with a 12-sample window size is implemented to reduce noise amplification.
To generate the mass center position and orientation of a body, first, marker position data are processed to generate
the transformation matrix from the local body frame to the ground frame. To illustrate this procedure, consider the
bucket that has four markers rigidly attached. These marker points are denoted as α B , β B , γ B and δ B . The unit vectors
that define the bucket body frame are defined such that the JEB axis is directed laterally. The IEB axis points forward
out of the front of the bucket and the KE B axis points down through the bottom of the bucket. The four bucket markers
shown in Fig. 6 are used to define the bucket body frame in a detailed manner using the equations below:
rEO→β B − rEO→α B
JEB = = j B X IEI + j BY JEI + j B Z KE I (14)
r O→β B − rEO→α B k
kE
(E
r O→γ B − rEO→α B ) × JEB
KE B = = k B X IEI + k BY JEI + k B Z KE I (15)
r O→γ B − rEO→α B ) × JEB k
k(E
JEB × KE B
IEB = = i B X IEI + i BY JEI + i B Z KE I (16)
k JEB × KE B k
Since the bucket frame and inertial frame unit vectors are related by

 IEB   IEI 
   
T
JEB = TB→I JE (17)
E   EI 
KB KI
equating components resolved in the inertial frame yields:
826 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833

Fig. 8. Measured and predicted hoist rope length vs. time.

   
iBX jB X kB X 1 0 0
T 0
i BY j BY k BY  = TB→I 1 0 (18)
iBZ jB Z kB Z 0 0 1
The position vector from the ground reference frame to the bucket reference frame is computed as a linear combination
of the bucket marker points. From basic geometric data, the distance vector from the bucket reference frame to the
bucket mass center is known. Thus the position vector from the ground reference frame to the bucket mass center is
computed as a superposition of these two vectors, expressing both vectors in the ground reference frame. The distance
vector from the bucket reference frame to the bucket mass center is expressed in inertial coordinates with the help of
the transformation matrix defined in Eq. (18).
The Euler orientation angles of the bucket are obtained by comparing the functional form of the transformation
matrix in Eq. (14) with the measured transformation matrix in Eq. (18), yielding

φ B = sin−1 (− j B Z ) (19)
 
iBZ
θ B = sin−1 (20)
cos(φ B )
 
jB X
ψ B = sin−1
(21)
cos(φ B )

Drag and hoist rope lengths are determined by fitting a cubic polynomial in the x–z plane to the rope marker
positions:

z i = a0 + a1 xi + a2 xi2 + a3 xi3 + e (22)

In Eq. (22), e is the residual and a0 , a1 , a2 , and a3 are unknown coefficients.

4. Dig maneuver results

A dig maneuver begins with the hoist rope set at 8.6 m and the drag rope at 4.2 m. This places the bucket 1.2 m
above the dig plane and near the end of the excavator’s travel in the longitudinal direction with a negative pitch angle
of 52◦ . At t = 0, the bucket is swept down and towards the operator by lengthening the hoist rope at 0.36 m/s while
reeling in the drag rope at 0.30 m/s. The hoist rope is then stopped for 1 s, while the drag rope pulls the bucket along
in the longitudinal direction at 0.61 m/s. While continually reeling the drag line in, the hoist line is again paid out at
0.30 m/s for another 2 s. The hoist line is then stopped, while the drag line continues for 1 s more, after which both
reels are stopped. The dig maneuver ends with the hoist rope set at 10.1 m and the drag rope at 1.4 m, placing the
bucket 0.70 m above the dig plane near the middle of the system’s longitudinal travel. Hoist and drag rope lengths are
plotted against time in Figs. 8 and 9.
 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833 827

Fig. 9. Measured and predicted drag rope length vs. time.

Fig. 10. Measured and predicted X inertial position of bucket mass center vs. time.

The bucket position begins and ends with a slight shift between measured and predicted values, while pitch differs
by two to four degrees. The position in the IEI , and KE I directions and the pitch angle for the dig maneuver are plotted
in Figs. 10 through 12. Note that predicted and measured bucket position match closely in all directions throughout
the entire dig maneuver. The predicted motion tends to oscillate slightly more in all cases, yet location and frequency
match well. The predicted values of bucket Euler pitch angle tend to oscillate more and with larger amplitude than
measured values. Bucket position in the JEI direction, bucket roll and yaw angles are all nearly zero for both measured
and simulated maneuvers.
The velocity in the IEI and KE I directions and the pitch rate of the bucket for the dig maneuver are plotted in
Figs. 13 through 15. Translational velocities correlate well throughout the entire maneuver. The maximum measured
longitudinal velocity is 0.99 m/s, occurring at 6.2 s, while the maximum predicted longitudinal velocity is 1.01 m/s
and occurs at 5.95 s. The maximum measured vertical velocity is 0.73 m/s, occurring at 6.05 s, while the maximum
predicted vertical velocity is 0.91 m/s and occurs at 6.27 s. The predicted bucket pitch rate is larger than the measured
values throughout most of the maneuver. The maximum measured pitch rate of −2.4 rad/s occurs at 6.77 s, while the
maximum predicted pitch rate of −3.2 rad/s occurs at 6.77 s.
The spreader bar and hoist cluster position in the longitudinal and vertical directions, along with Euler pitch angle,
are compared during the maneuver in Figs. 16 through 19. The positions of the drag cluster in the longitudinal and
vertical directions are compared in Fig. 20. The predicted spreader bar position in the longitudinal direction matches
the measured position through the entire maneuver with a consistent offset of approximately 51 mm. In the vertical
direction, predicted and measured motion match well until the end of the maneuver, where the measured position
settles more quickly than predicted. The spreader bar’s measured pitch also tends to settle more quickly than predicted.
828 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833

Fig. 11. Measured and predicted Z inertial position of bucket mass center vs. time.

Fig. 12. Measured and predicted bucket pitch angle vs. time.

Fig. 13. Measured and predicted X inertial velocity of bucket mass center vs. time.

Similarly to the spreader bar and bucket, the hoist and drag clusters’ predicted and measured motions match well
through the entire maneuver, yet the measured motion tends to settle faster than predicted motion in the vertical
direction and pitch angle.
 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833 829

Fig. 14. Measured and predicted Z inertial velocity of bucket mass center vs. time.

Fig. 15. Measured and predicted bucket pitch rate vs. time.

The only significant difference between predicted and measured motion throughout the dig maneuver is in
the pitch dynamics of the bucket and hoist cluster, and the effects that this motion has on the system. These
differences can mainly be attributed to un-modeled sources of damping in the system, including friction in the
chain links, torsion damping in the ropes, rope bending damping, pulley friction, and pulley rotational momentum.
Differences between predicted and measured positions and orientations are attributed to: the simple hoist cluster
pulley model, which plays a large role in the pitch dynamics of the bucket; un-modeled rigging interactions
such as rope and chain attachment moments, as well as rope and chain collisions; and system measurement
errors.

4.1. Rope discretization

As previously mentioned, each rope segment is split into nr particles and nr + 1 elastic elements using truss finite
elements. Collectively, the motion of the particles defines the motion of the complete rope segment. The effect of the
discretization level of each rope on the simulation predictions is investigated at four levels of discretization tabulated
in Table 3. Bucket pitch angle is compared against measured values for each bead discretization level in Fig. 21, where
the baseline case is used for the motion comparisons (Table 2). Simulated pitch angle approaches measured the pitch
angle as the discretization level is increased. The penalty associated with increasing the bead count is an increase in
system states.
830 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833

a. Measured and predicted X inertial position of spreader bar mass center

vs. time.

b. Measured and predicted Z inertial position of spreader bar mass center

vs. time.

Fig. 16. Measured and predicted position of spreader bar mass center vs. time.

Fig. 17. Measured and predicted spreader bar pitch angle vs. time.
J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833 831

a. Measured and predicted X inertial position of hoist cluster mass center

vs. time.

b. Measured and predicted Z inertial position of hoist cluster mass center

vs. time.

Fig. 18. Measured and predicted position of hoist cluster mass center vs. time.

Fig. 19. Measured and predicted hoist cluster pitch angle vs. time.
832 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833

a. Measured and predicted X inertial position of drag cluster mass center

vs. time.

b. Measured and predicted Z inertial position of drag cluster mass center

vs. time.

Fig. 20. Measured and predicted position of drag cluster mass center vs. time.

Fig. 21. Measured and predicted bucket pitch angle vs. time for varying bead discretization.
 J. Kyle, M. Costello / Mathematical and Computer Modelling 44 (2006) 816–833 833

Table 3
Rope bead discretization level

Discretization Drag Dump Hoist Left drag Left lower hoist Left upper hoist Total # of rope states
level rope rope rope chain chain chain
Low 4 3 8 2 1 1 133
Base line 6 5 12 4 2 3 224
Medium 8 7 16 8 3 4 322
High 14 12 28 13 4 6 539

5. Conclusions

The work reported here focused on a detailed comparison of measured longitudinal motion of a scaled dragline
excavation system with results generated by a complex dynamic simulation. Measured and predicted motions during
a dig maneuver agrees favorably. System dynamics are also predicted well, with the exception of bucket and hoist
cluster pitch dynamics. The primary difference between measured and predicted motions are relatively high-frequency,
slowly decaying oscillations exhibited by simulation results. These differences can be attributed mainly to un-modeled
sources of damping in the system, including friction in the chain links, torsion damping in the ropes, rope bending
damping, pulley friction, and pulley rotational momentum.

References

[1] C.E. McCoy, J.L. Crowgey, Anti-Tightline Control System for Draglines used in the Surface Mining Industry, IEEE, 1980, CH1575-
0/80/0000-0140.
[2] D.K. Haneman, H. Hayes, G.I. Lumley, Dragline performance evaluations for tarong coal using physical modeling, in: Third Large Open Pit
Mining Conference, Mackay, 30 August–3 September 1992.
[3] J.C. Rowlands, G.D. Just, Performance characterisitcs of dragline buckets, in: Third Large Open Pit Mining Conference, Mackay, 30 August–3
September 1992.
[4] P.F. Knights, D.H. Shanks, Dragline productivity improvements through short-term monitoring, in: Institution of Engineers, Coal Handling
and Utilization Conference, Sydney, 19–21 June 1990.
[5] K. Pathak, K. Dasgupta, A. Chattopadhyay, Determination of the working zone of a dragline bucket, Mining Technology (February) (1992).
[6] D.W. Hainsworth, P.I. Corke, G.J. Winstanley, Location of dragline bucket in space using machine vision techniques, in: Speech and Signal
Processing, Adelaide, April 1994.
[7] P.W. Cleary, The filling of dragline buckets, Math. Engng. Ind. 7 (1) (1998) 1–24.
[8] P. Ridley, P. Corke, Calculation of dragline bucket pose under gravity loading, Mechanism and Machine Theory 35 (2000) 1431–1444.
[9] J.H. Ginsberg, Advanced Engineering Dynamics, Cambridge University Press, New York, NY, 1995, pp. 119–162.
[10] VICON Motion Systems Limited, Vicon System Manual, VICON 612, 12th edition, Lake Forest, CA, July 2002.