See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/265227627

Bond Graph Modeling of Mechanical Dynamics

of an Excavator for Hydraulic System Analysis
and Design

Article · January 2009

CITATIONS READS

2 48

2 authors, including:

Onesmus Muvengei
Jomo Kenyatta University of Agriculture and Technology
10 PUBLICATIONS 173 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

optimisation of a jaw crusher View project

All content following this page was uploaded by Onesmus Muvengei on 27 January 2016.

The user has requested enhancement of the downloaded file.

 International Journal of Mechanical, Industrial and Aerospace Engineering 3:4 2009

Bond Graph Modeling of Mechanical Dynamics of

an Excavator for Hydraulic System Analysis and
Design
Mutuku Muvengei and John Kihiu,

Abstract—This paper focuses on the development of bond graph elements involved [4]. The relatively new bond graph mod-
dynamic model of the mechanical dynamics of an excavating mech- eling technique, has been proposed to successfully model
anism previously designed to be used with small tractors, which are the dynamics of manipulators and mechanisms. Since bond
fabricated in the Engineering Workshops of Jomo Kenyatta University
of Agriculture and Technology. To develop a mechanical dynamics graph method is based on the interaction of power between
model of the manipulator, forward recursive equations similar to elements, it can be used to model multi-energy domains also,
those applied in iterative Newton-Euler method were used to obtain for example the actuator system of the manipulator which
kinematic relationships between the time rates of joint variables may be electrical, pneumatic, hydraulic or mechanical. Once
and the generalized cartesian velocities for the centroids of the the bond graph model is ready, the system equations can be
links. Representing the obtained kinematic relationships in bond-
graphic form, while considering the link weights and momenta as derived from it algorithmically in a systematic manner. This
the elements led to a detailed bond graph model of the manipulator. process is usually automated using appropriate softwares such
The bond graph method was found to reduce significantly the number as ENPORT, CAMP-G, TUTSIM 20-SIM, SYMBOLS 2000,
of recursive computations performed on a 3 DOF manipulator for a etc which support bond graphs. For mechanical manipulators
mechanical dynamic model to result, hence indicating that bond graph and mechanisms, the bond graph model can be developed
method is more computationally efficient than the Newton-Euler
method in developing dynamic models of 3 DOF planar manipulators. based on kinematic relationships between the time rates of
The model was verified by comparing the joint torque expressions joint variables and the generalized cartesian velocities (transla-
of a two link planar manipulator to those obtained using Newton- tional and angular velocities). It is not necessary to have higher
Euler and Lagrangian methods as analyzed in robotic textbooks. The order time rates of variables involved, that is translational and
expressions were found to agree indicating that the model captures angular accelerations.
the aspects of rigid body dynamics of the manipulator. Based on
the model developed, actuator sizing and valve sizing methodologies The concept of bond graphs was originated by Paynter [5].
were developed and used to obtain the optimal sizes of the pistons The idea was further developed by Karnopp and Rosenberg
and spool valve ports respectively. It was found that using the pump in their textbooks [6]–[8], such that it could be used in
with the sized flow rate capacity, the engine of the tractor is able to practice. By means of the formulation by Breedveld [9] of
power the excavating mechanism in digging a sandy-loom soil. a framework based on thermodynamics, bond graph model
Keywords—Actuators, bond graphs, inverse dynamics, recursive description evolved to a systems theory. More information
equations, quintic polynomial trajectory. about bond graphs can be found in [6]–[12].
The Bond graph method can be used to obtain more intricate
I. I NTRODUCTION information such as the power required to drive each joint
actuator, or the power interaction at the interface with the
N order to design, improve performance, simulate the be-
I havior and finally control a system or plant, it is necessary
to obtain it’s dynamics. To develop the dynamics of a manip-
environment. Such information can also be used to study the
stability of the manipulator system during contact interaction
with the environment. Modifications and additions to the
ulator, a kinematic model of the manipulator is required first. system can be easily incorporated by connecting suitable bond-
The kinematic modeling is done first by attaching coordinate graphic sub-systems to its existing bond graph.
frames to every link. The usual convention applied to attach In this paper, the mechanical dynamics of the excavating
frames in the links of a manipulator is the Denavit-Hartenberg manipulator designed in [13] to be used with the small tractors
procedure [1]. The dynamics of a manipulator can be modeled (fabricated in the Engineering Workshops of Jomo Kenyatta
using various methods namely; Newton-Euler formulation, University of Agriculture and Technology) in digging medium-
Lagrangian formulation, Kane’s method, and others [2], [3]. height trenches for small scale farmers, is modeled using the
The three methods; Newton-Euler, Lagrangian and Kane’s bond graph method. The excavating manipulator is shown
methods tend to hide the physical interactions between the in Fig. 1. Inverse dynamics is performed on the developed
dynamic model for purposes of analyzing the hydraulic system
M. Muvengei is with the Department of Mechanical Engineering, Jomo
Kenyatta University of Agriculture and Technology, P.O BOX 62000-00200 design.
Kenya (Tel: +254720642441, e-mail: [email protected].) In this work, forward recursive equations for motion of
J. Kihiu is with the Department of Mechanical Engineering, Jomo Kenyatta manipulators similar to those used in Newton-Euler method
University of Agriculture and Technology, P.O BOX 62000-00200 Kenya
(email: [email protected]) are used to derive the kinematic relationships between the time
rates of joint variables, and the generalized cartesian velocities

248
 International Journal of Mechanical, Industrial and Aerospace Engineering 3:4 2009

⎛ ⎞
cos θ2 − sin θ2 0 a2 cos θ2
⎜ sin θ2 cos θ2 0 a2 sin θ2 ⎟
=⎜ ⎟
(2)
A1 ⎝ 0 ⎠ (2)
0 1 0
0 0 0 1
⎛ ⎞
cos θ3 − sin θ3 0 a3 cos θ3
⎜ sin θ3 cos θ3 0 a3 sin θ3 ⎟
(3)
A2 ⎜
=⎝ ⎟ (3)
0 0 1 0 ⎠
0 0 0 1
⎛ ⎞
cos θ4 − sin θ4 0 a4 cos θ4
⎜ sin θ4 cos θ4 0 a4 sin θ4 ⎟
(4)
A3 ⎜
=⎝ ⎟ (4)
0 0 1 0 ⎠
Fig. 1. Schematic drawing of a part of assembled excavator 0 0 0 1
Kinematic relationships between the translational velocities of
(translational and angular velocities) of mass centers of the the center of masses of the links (vGi ) to the time rates of
links. These kinematic relations are further used for graphical the joint variables (θ̇i ) (where i = 2, 3, 4) can be obtained
representation of the system dynamics using Bond graphs. by using forward recursive equations which were proposed by
i
Luh et al. [15]. The translational velocity vGi of the center of
th th
II. M ODELING THE M ANIPULATOR DYNAMICS mass of the i link as specified in the i coordinate frame
Assuming that; the inertial effects of cylinders and their is given recursively by;
pistons are negligibly small compared to those of manipulator (i) (i) (i) (i) (i)
vGi = v0i + ωi × (PGi − P0i ) (5)
links, the hydraulic cylinders transmit axial forces only, the
revolute joints have no friction, and all the links and supports Where
are rigid, a bond graph model representing the mechanical (i)
• PGi is the vector from the origin of the base coordinate
dynamics of the excavating manipulator was developed. system to the center of mass of the ith frame as expressed
in the ith coordinate system.
A. Kinematic Analysis and Forward Recursion (i)
• P0i is the vector from the origin of the base coordinate
Kinematic analysis was performed on the excavating ma- system to the origin of the ith coordinate system as
nipulator to relate the translational velocities of the center of expressed in the ith coordinate system.
masses of the links (vGi ) to the time rates of the joint variables (i)
• ωi is the rotational velocity of link i as specified in the
(θ̇i ), for i = 2, 3, 4. The choice of center of mass velocities th
i coordinate frame, and is given recursively as;
for rigid bodies leads to a highly systematic approach for
(i) (i−1) (i−1)
constructing bond graphs and is recommended [7]. ωi = Ri ωi−1 + Z̀0 q˙i (6)
The homogeneous transformation matrices (1) to (4) were
obtained by first attaching world coordinate frames to the three where
(i−1)
links as shown Fig. 2 by using Denavit-Hartenberg procedure – Ri is the rotational matrix relating two adjacent
as described in [14]. frames and is obtained from the respective homoge-
neous ⎛ ⎞
transformation matrix.
y2 0
F x2 – Z̀0 = ⎝ 0 ⎠
O2 , C
y1
I
1
G2 J
T1 T3
– q̇i = θ̇i for revolute joints.
B G3 (i)
O1 , A
T2
L
K
y3 • v0i is the translational velocity of the origin of the ith
x1
O3 , D G link coordinate frame as expressed in the ith coordinate
y0
T4
G4
frame and is given recursively as,
E
O0 x3
O4 , N
x0 (i) (i−1) (i−1) (i) (i) (i)
y4 v0i = Ri v0i−1 + ωi × (P0i − P0i−1 ) (7)
x4
where
(i)
– P0i−1 is the vector from the origin of the base
Fig. 2. Coordinate System assignment for excavator. coordinate system to the origin of the (i − 1)th
coordinate system as expressed in the ith coordinate
⎛ ⎞ system.
cos θ1 − sin θ1 0 a1 cos θ1 (i)
⎜ sin θ1 cos θ1 0 a1 sin θ1 ⎟ – P0i is the vector from the origin of the base co-
=⎜ ⎟
(1)
A0 ⎝ 0 ⎠ (1) ordinate system to the origin of the ith coordinate
0 1 0
0 0 0 1 system as expressed in the ith coordinate system.

249
 International Journal of Mechanical, Industrial and Aerospace Engineering 3:4 2009

B. Modeling the Bucket Digging Force remain constant throughout the digging process. The critical
A model that accounts for the material being retained in value of the cutting angle is given by [17],
the bucket, which was developed by Cannon [16] using force 1
αc = π − σ − sin−1 (sin σ sin ρ) (12)
equilibrium and fundamental earthmoving equation in soil 2
mechanics was applied in this study to determine the force The soil-tool force F is assumed to be applied at the cutting
F exerted by the excavator bucket to the soil. From Fig. 3, edge of the bucket. From the Newton’s third law of motion,
the force F is given in (8). the soil applies an opposite and equal reaction force at the
x
bucket, which can be resolved to a normal and tangential force
β components as shown in Fig. 4
Q cLf
z W1 W2
G
α ρ R D , O3
d φ
F
δ

caLt
y3
x3

Fig. 3. Wedge model that accounts for the material being retained in the Tb
FN
bucket.
N ,O4

2 y4
F = d wγgNw + cwdNc + Vs γgNq (8) Tb FT

Vs is the swept volume, Q is the surcharge, W1 is the weight

x4
of the material above the bucket, W2 is the weight of the rest
of material in the wedge, Lt is the length of the tool, Lf is the
length of the failure surface, R is the force of the soil resisting Fig. 4. Bucket digging force at the tip.
the moving of the wedge, F is the force exerted by the tool on
the wedge, ca is the adhesion between the soil and the blade, The horizontal and vertical components of the bucket reac-
c is the cohesiveness of the soil media, β is the failure surface tion force are given as;
angle (slip angle), α is the surface terrain slope (cutting angle), Fx =−(FT cos θb − FN sin θb ) cos(θ2 + θ3 + θ4 )
φ is the soil-soil friction angle, ρ is the rake angle of the tool
relative to the soil surface, σ is the soil-tool friction angle, d is +(FT sin θb + FN cos θb ) sin(θ2 + θ3 + θ4 ) (13)
the depth of the bucket tool perpendicular to the soil surface, Fy =−(FT cos θb − FN sin θb ) sin(θ2 + θ3 + θ4 )
w is the width of the bucket, γ is the bulk density of the soil −(FT sin θb + FN cos θb ) cos(θ2 + θ3 + θ4 ) (14)
media and g is the gravitational acceleration.
Nw , Nc Nq are N-factors which depend on: the soil’s These forces are included at the translational velocity of the
frictional strength, the bucket tool geometry and soil-to-tool origin of the 4th link coordinate frame but referenced to the
strength properties, and are given by the following equations. base coordinate frame.
 
cot β − tan α cos α + sin α cot(β + φ) C. Overall Bond Graph Model of the Manipulator
Nw =  (9)
2 cos(ρ + σ) + sin(ρ + σ) cot(β + φ) The kinematic relationship obtained from (5) for each link
was represented in bond-graphic form while considering the
1 + cot β cot(β + φ)
Nc = (10) weights and momenta of the links as the bond graph elements.
cos(ρ + σ) + sin(ρ + σ) cot(β + φ) All the bond graph sub-models were then assembled to an
cos α + sin α cot(β + φ) overall non causal bond graph of the manipulator as shown in
Nq = (11)
cos(ρ + σ) + sin(ρ + σ) cot(β + φ) Fig. 5.
Equations (8) to (11) show that the magnitude of the digging Where
force depends on many factors such as the cutting angle, • r1 = LG2 O2 sin(θ2 − σ1 ) − LO1 O2 sin θ2
specific resistance to cutting, volume of the bucket, amount • r2 = −LG2 O2 cos(θ2 − σ1 ) + LO1 O2 cos θ2
of the the material ripped into the bucket and the volume • r3 = −LO1 O2 sin θ2 + LG3 O3 sin(θ2 + θ3 − σ2 ) −
of the material surcharged. These factors are always varying LO2 O3 sin(θ2 + θ3 )
during the bucket digging operation and indicate complicated • r4 = −LG3 O3 sin(θ2 + θ3 − σ2 ) + LO2 O3 sin(θ2 + θ3 )
interactions of the bucket and the soil, hence making modeling • r5 = LO1 O2 cos θ2 − LG3 O3 cos(θ2 + θ3 − σ2 ) +
of the bucket digging force throughout the digging process a LO2 O3 cos(θ2 + θ3 )
complex and bulk process. • r6 = −LG3 O3 cos(θ2 + θ3 − σ2 ) + LO2 O3 cos(θ2 + θ3 )
In this study, a simplified model is presented by considering • r7 = −LO1 O2 sin θ2 − LO2 O3 sin(θ2 + θ3 ) −
the situation of critical force, and then assuming the force to LO3 O4 sin(θ2 + θ3 + θ4 ) + LG4 O4 sin(θ2 + θ3 + θ4 − σ3 )

250
 International Journal of Mechanical, Industrial and Aerospace Engineering 3:4 2009

MSE : Fx I : vO 4 x
0
MTF : r15
All the rotation axes at the joints are along the z-axis normal
MTF : r14
1:T 4
x
to the paper surface. Let;
MTF : r9

MTF : r7
0
I :M4
MTF : r18
LO1 O2 = LO2 O3 = l
1 : vG 4 x
MTF : r13
I :M2 MTF : r8
MTF : r12
l
I : J2
0 I : J4
I : M4
LG2 O2 = LG3 O3 =
1 : vG 2 x MTF : r1 1:T 2
x
MTF : r10 0
1 : vG 4 y 2
I :M2 1 : vG 2 y 0 SE :  M 4 g
σ 1 = σ2 = 0
MTF : r2 MTF : r3
MTF : r5
I : J3 MTF : r11

SE :  M 2 g 0 MTF : r4
x
The bond graph model of the two link planar manipulator
1:T 3
can be represented as shown in Fig. 7;
1 : vG 3 x
MTF : r6 MTF : r17
0
I : M3 SE : W 3

I :M3 1 : vG 3 y MSE : Fx 1 : vO 3 x

1 : vO 4 y MSE : Fy I : M3 1 : vG 3 y
x
MTF : r16 0 0 MTF : r6 1:T 3
SE :  M 4 g

I : M2
SE :  M 3 g MTF : r5 0 MTF : r8

Fig. 5. A non causal bond graph model representing the mechanical dynamics 1 : vG 2 y
MTF : r7 I : J3
of the manipulator. MTF : r2
SE :  M 2 g
x 0
SE : W 2 1:T 2 MTF : r10

0
• r8 = −LO2 O3 sin(θ2 + θ3 ) − LO3 O4 sin(θ2 + θ3 + θ4 ) + 1 : vG 2 x
MTF : r1
I : J2
MTF : r9

LG4 O4 sin(θ2 + θ3 + θ4 − σ3 ) I : M3
1 : vO 3 y MSE : Fy

• r9 = −LO3 O4 sin(θ2 + θ3 + θ4 ) + LG4 O4 sin(θ2 + θ3 + I : M2 MTF : r3

θ 4 − σ3 ) 1 : vG 3 x 0
MTF : r4

• r10 = LO1 O2 cos θ2 + LO2 O3 cos(θ2 + θ3 ) +

LO3 O4 cos(θ2 + θ3 + θ4 ) − LG4 O4 cos(θ2 + θ3 + θ4 − σ3 ) Fig. 7. A non causal bond graph model representing the mechanical dynamics
• r11 = LO2 O3 cos(θ2 + θ3 ) + LO3 O4 cos(θ2 + θ3 + θ4 ) − of a 2-link manipulator.
LG4 O4 cos(θ2 + θ3 + θ4 − σ3 )
• r12 = LO3 O4 cos(θ2 + θ3 + θ4 ) − LG4 O4 cos(θ2 + θ3 + Where,
θ 4 − σ3 ) • r1 = − 2 sin θ2
l

• r13 = −LO1 O2 sin θ2 − LO2 O3 sin(θ2 + θ3 ) − • r2 = 2 cos θ2

l

LO3 O4 sin(θ2 + θ3 + θ4 ) • r3 = −l sin θ2 − 2 sin(θ2 + θ3 )

l

• r14 = −LO2 O3 sin(θ2 + θ3 ) − LO3 O4 sin(θ2 + θ3 + θ4 ) • r4 = − 2 sin(θ2 + θ3 )

l

• r15 = −LO3 O4 sin(θ2 + θ3 + θ4 ) • r5 = l cos θ2 + 2 cos(θ2 + θ3 )

l

• r16 = LO1 O2 cos θ2 + LO2 O3 cos(θ2 + θ3 ) + • r6 = 2 cos(θ2 + θ3 )

l

LO3 O4 cos(θ2 + θ3 + θ4 ) • r7 = −l sin θ2 − l sin(θ2 + θ3 )

• r17 = LO2 O3 cos(θ2 + θ3 ) + LO3 O4 cos(θ2 + θ3 + θ4 ) • r8 = −l sin(θ2 + θ3 )
• r18 = LO3 O4 cos(θ2 + θ3 + θ4 ) • r9 = l cos θ2 + l cos(θ2 + θ3 )
• r10 = l cos(θ2 + θ3 )

The joint torques applied to each of the joint by the

D. Checking the Model
respective actuator can be obtained systematically from the
One way to check the bond graph model developed in Fig. bond graph using the constitutive relations, and noting that
5 is to compare results with those available in the literature. Fx = Fy = 0 since the manipulator is moving in free space.
A two link manipulator shown in Fig. 6, moving in a free The external torque applied to move link 3 can be obtained
space, and whose links are uniform and of equal lengths was from the bond graph as;
considered. Such a problem is studied using Newton-Euler, 1 1 1 1
Lagrangian, and d’Alembert’s methods in the standard robotic τ3 = m3 l2 θ̈2 + m3 l2 θ̈3 + m3 l2 θ̈2 cos θ3 + m3 l2 θ̇22
3 3 2 2
textbooks [2], [3]. 1
sin θ3 + m3 gl cos(θ2 + θ3 ) (15)
2
The external torque applied to move link 2 can be obtained
from the bond graph as;
1 4 1
τ2 = m2 l2 θ̈2 + m3 l2 θ̈2 + m3 l2 θ̈3 + m3 l2 θ̈2 cos θ3
3 3 3
1 1
+ m3 l θ̈3 cos θ3 − m3 l θ̇2 θ̇3 sin θ3 − m3 l2 θ̇22 sin θ3
2 2
2 2
1 1
+ m2 gl cos θ2 + m3 gl cos θ2 + m3 g cos(θ2 + θ3 )(16)
2 2
The equations of external torque given in (15) and (16) corre-
Fig. 6. A two link planar manipulator. spond to those obtained using Newton-Euler and Lagrangian

251
 International Journal of Mechanical, Industrial and Aerospace Engineering 3:4 2009

methods for the same planar manipulator, as illustrated in of the polynomial can be obtained and substituted in the
[2], [3]. This indicates that, the model developed captures the polynomial equation (17) to get the required displacement,
essential aspects of rigid body dynamics of the manipulator. velocity and acceleration trajectories as;
10(xmax − xmin ) 3 15(xmin − xmax ) 4
III. H YDRAULIC S YSTEM A NALYSIS AND D ESIGN U SING x(t)=xmin + t + t
t3f t4f
I NVERSE DYNAMICS
6(xmax − xmin ) 5
In inverse dynamics, the generalized joint torques are com- + t (18)
t5f
puted given the desired joint trajectories. The joint trajectories
are obtained through the trajectory planning schemes which 30(xmax − xmin ) 2 60(xmin − xmax ) 3
ẋ(t)= t + t
generally interpolate or approximate the desired manipulator t3f t4f
path by a class of polynomial functions and generates a 30(xmax − xmin ) 4
sequence of time-based set-points for the manipulator from + t (19)
t5f
the initial position and orientation to its destination [2].
60(xmax − xmin ) 180(xmin − xmax ) 2
Quintic trajectory is used to size the spool valves and the ẍ(t)= t+ t
pistons of the hydraulic cylinders, and also to check the total t3f t4f
power required when the bucket of the manipulator is digging a 120(xmax − xmin ) 3
+ t (20)
sandy-loom soil. As described in [16], all the cylinders will be t5f
considered to be extending simultaneously during the digging
operation. B. Simulink Model for the Inverse Dynamics of the 3 dof
Model parameters are needed to run simulations. Parameters System
like lengths, masses and angles were found from design In Fig. 8, the complete Simulink model of the inverse
drawings and trigonometric calculations. But parameters like dynamics of 3dof excavating manipulator is shown. This
the location of a center of mass, link moments of inertia, model is the test-bed for all inverse dynamics simulations done
products of inertia of a link could not be estimated from on the system. The whole system is run in the same time
blueprints. Auto CAD with Advanced Modeling Extension frame, and as a result the outputs must match the generated
Package was used to estimate the mass properties of all the trajectories.
links and the locations of center of masses.
torques
power

A. Trajectory Planning To Workspace5

To Workspace4

In a typical trajectory, all joints move simultaneously. For |u|

|u|
thetas |u|
the typical trajectory selected here, the boom, arm, and bucket To Workspace3
pressure_drops

To Workspace1
links move from their minimum to maximum positions and emu cylinder_forces

To Workspace2
all joints start and finish moving at the same time, although joint traj joint torque emu

different time limits can be programmed. inverse dynamics

cyl force pressure drops

piston traj flow rates flow_rates

Three common trajectories namely, trapezoidal trajectory, Hydraulics
To Workspace7

cubic polynomial trajectory, and quintic polynomial trajectory joint_traje em

have been previously applied in trajectory planning for hy- joint trajectory generator
piston traj

draulic manipulators. Sarkar [18] used the three trajectories theta traj
Jf emu

to size the valves and power requirement for an articulated joint to piston level

forestry machine. Among the three methods, the quintic poly-

nomial trajectory has advantage in that;
• the velocity trajectory is smooth unlike in trapezoidal tra- Fig. 8. The overall Simulink block for the inverse dynamics
jectory whose velocity profile has discontinuities where
the link motion starts to settle at a constant velocity, and The joint trajectory generator block produces the desired
where the link starts to decelerate. boom, arm and bucket quintic polynomial trajectories, that
• the acceleration profile has values equal to zero at starting is, the angular displacements, angular velocities and angular
and finishing times of the trajectory, unlike in the other accelerations of the links according to (18)-(20). The three sig-
trajectories where the acceleration values at the start and nals from the joint trajectory generator block each containing
final times have non zero values. the corresponding signals for the boom, arm and bucket link
Therefore the trajectory to be adopted in this work is the are fed into the inverse dynamics block to compute the joint
quintic polynomial trajectory which is given by; torques for the bucket, arm and boom links respectively. The
joint torque profiles when the bucket is digging a sandy-loom
x(t) = a0 + a1 t + a2 t2 + a3 t3 + a4 t4 + a5 t5 (17) soil is shown in Fig. 9.
The desired boundary conditions are; xt=0 = xmin ,
xt=tf = xmax , ẋt=0 = 0, ẋt=tf = 0, ẍt=0 = 0 and C. Power Profiles
ẍt=tf = 0. By taking the first and second derivatives of The dynamic model obtained also permits either sizing of
(17) and satisfying these boundary conditions, the coefficients the system power supply or checking whether the desired

252
 International Journal of Mechanical, Industrial and Aerospace Engineering 3:4 2009

(a) (b) (c)

12000 5000 3000 The power consumption as seen in (21)-(24) depends on
Boom joint Torque (Nm)

Bucket Joint Torque (Nm)

the joint torque requirements and the angular velocities of

Arm Joint Torque (Nm)

10000 4000 2000
3000
8000
2000
1000 the links. Two options are available to reduce the total power
6000
1000
0
requirement of the manipulator. These are;
4000 -1000
0 • Reduce the joint torque requirements for the manipulator,
2000 -2000
-1000
0
by reducing the mass properties of the links and/or
-2000 -3000
0 2 4
Time(secs)
6 8 0 2 4
Time(secs)
6 8 0 2 4
Time(secs)
6 8
reducing the force exerted to the ground by the bucket.
• Increase the cycle times of the link trajectories. This
Fig. 9. The torque profiles at manipulator joints, when the bucket is digging means that the pump flow rate capacity will be reduced.
a sandy-loom soil; (a) Boom joint torque (b) Arm joint torque (c) Bucket
joint torque In this work, the second option of increasing the cycle times
of the link trajectories was used. Generally, the cycle time
has a direct impact on the flow rate requirements relative
manipulator trajectory can be followed without exceeding the to the link/actuator motion requirements, and on the power
power capacity of the supply. The power required for the requirement of the manipulator. As seen in (19) increasing the
boom, arm and bucket motion is respectively given as; cycle time tf will decrease the link velocities, and this will
subsequently reduce the power requirement and also the flow
Pbo = τbo θ̇2 (21) rates requirement.
Pa = τa θ̇3 (22) The optimum cycle time necessary to ensure that the power
requirement for the manipulator motion when the bucket is
Pbu = τbu θ̇4 (23)
digging a sandy-loom soil is reduced was found to be 10s.
Where τ is the torque required to move a link at an angular Fig. 11 shows the new power requirement profiles for the
velocity of θ̇. The total power required for a given trajectory manipulator.
can be obtained by;
(a) (b)
2.5 0.4

Ptotal = |Pbo | + |Pa | + |Pbu | (24)

Boom Link Power (kW)

0.2

Arm Link Power (kW)

2
0
1.5
assuming there are no power losses. The power requirement -0.2

1 -0.4
profiles for the manipulator are given in Fig. 10 when the -0.6
bucket is digging a sandy-loom soil. 0.5
-0.8

0 -1
0 2 4 6 8 10 0 2 4 6 8 10
Time(secs) Time(secs)
(a) (b)
3 0.2
Boom link Power (kW)

2.5 0
Arm link Power (kW)

(c) (d)
-0.2 1 5
2
Bucket Link Power (kW)

-0.4
1.5 0.5 4
Total Power (kW)

-0.6
1 0
-0.8 3
0.5 -1 -0.5 2
0 -1.2
0 2 4 6 8 0 2 4 6 8 -1
Time(secs) 1
Time(secs)
-1.5 0
0 2 4 6 8 10 0 2 4 6 8 10
Time(secs) Time(secs)
(c) (d)
1.5 6
Bucket link Power (kW)

1 5 Fig. 11. (a) Boom joint power (b) Arm joint power (c) Bucket joint power
Total Power (kW)

0.5 4
(d) Total joints power ;when the manipulator is digging a sandy-loom soil
and at the optimal cycle times
0 3

-0.5 2

-1 1

-1.5 0 D. Sizing the Hydraulic Actuators and Valves based on Inverse

0 2 4 6 8 0 2 4 6 8
Time(secs) Time(secs) Dynamics
Fig. 10. (a) Boom joint power (b) Arm joint power (c) Bucket joint power
An important application of the inverse dynamic modeling
(d) Total joints power ;when the bucket is digging a sandy-loom soil and at of hydro-mechanical systems is the sizing of hydraulic compo-
the initial cycle times nents. In this section, the optimal sizes of the actuator pistons
as well as the optimal sizes of the spool valve orifice ports of
The peak value for the total power consumption when the boom, arm, and bucket cylinders are determined. For this
the bucket is digging a sandy-loom soil is shown to be purpose, the pressure drop profiles across the cylinders and
approximately 5kW (6.7hp). This value is greater than the valves need to be plotted from the inverse dynamics.
power rating of the engine (6.5hp) which is to drive the 1) Pressure Drop Profiles: The hydraulic forces required to
hydraulic pump, implying that, the digging operation cannot produce the torques necessary for manipulator motion results
be achieved under the given link trajectories without exceeding in corresponding pressure drops across the hydraulic cylinders.
the power capacity of the prime mover. The expressions for the pressure drop in the boom cylinder,

253
 International Journal of Mechanical, Industrial and Aerospace Engineering 3:4 2009

Δpbo , arm cylinder, Δpa , and bucket cylinder, Δpbu are • If Δpmax < ps , then this implies that the supply pressure
approximated using equations below; is not used maximally and hence there is no need of such
a high pump pressure since the cylinder is oversized.
Fcybo
Δpbo = (25) • If Δpmax > ps , then the pressure drop across the valve
Apbo becomes negative and this will result to a negative flow
Fcya rate through the valve. This is not possible practically and
Δpa = (26)
Apa implies that the cylinder is undersized.
Fcybu
Δpbu = (27) As seen in Fig. 12, the maximum pressure drops across
Apbu the cylinders (points with asterisks) are less than the supply
The cylinder forces are related to the joint torques by the pressure of ps = 14.5M P a. This is also shown in Fig. 13
manipulator jacobians which for the manipulator in consider- where at the point of maximum pressure drop for all the
ation are derived in [19]. The pressure drop profiles across cylinders (points with asterisks), the pressure drops across the
the cylinders when the bucket is digging a sandy-loom soil corresponding valves do not equal to zero. Therefore it can
are shown in Fig. 12. be concluded that the boom, arm and bucket cylinders are
oversized.
x 10
6
(a) x 10
6
(b) x 10
7
(c) The optimal cylinder piston sizes were determined by tuning
10 8 1
* * the piston sizes until that instant when the maximum pressure
6
Pressure drop (Pa)

Pressure drop (Pa)

Pressure drop (Pa)

4
0.5 drops equaled the supply pressure for all the cylinders. These
6
2 0
values were found to be precisely 52.2mm, 45.1mm and
4
0 40.9mm for the boom, arm and bucket cylinders respectively,
-0.5
2
-2 as shown in Fig. 14. The optimal piston diameters values of
0 -4 -1
* the cylinders were rounded to the next imperial values which
0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10
Time(secs) Time(secs) Time(secs) are available in the market as 57.15mm (2 14 in), 50.8mm (2in)
and 44.45mm (1 34 in) for the boom, arm and bucket cylinders
Fig. 12. The Pressure drop profiles across the cylinders, when the bucket is
digging a sandy-loom soil and at original piston diameters; (a) Boom cylinder respectively.
(b) Arm cylinder (c) Bucket cylinder
(a) (b) (c)
20 20 20
'p 'p 'p
Neglecting line pressure drops, the pressure drops at the MAXbo MAXa MAXbu
Pressure drop (MPa)

Pressure drop (MPa)

Pressure drop (MPa)

'p 'p 'pVbu

14.5 Vbo Va 14.5
corresponding valves are approximated as; * 14.5
* *
10 10 10

Δpbov = ps − |Δpbo | (28) 5 5 5

Δpav = ps − |Δpa | (29) 0

* 0
* 0
*
Δpbuv = ps − |Δpbu |
-5 -5 -5
(30) 5052.2 55 60 65 70 40 45.1 50 60 70 35 40.9 45 50 55 60

Piston diameter (mm) Piston diameter (mm) Piston diameter (mm)

(Boom cylinder) (Arm cylinder) (Bucket cylinder)
Where ps is the operating pressure. The pressure drop profiles
at the valves for the case when the bucket is digging a sandy- Fig. 14. The pressure drops versus piston diameters; (a) Boom cylinder
loom soil is shown in Fig. 13. (b)Arm cylinder (c) Bucket cylinder

7
x 10
6
(a) x 10
7
(b) x 10 (c)
14 1.6 1.6 3) Valve Sizing: A valve is properly sized when it can sup-
ply the demanded flow at the required pressure drop across it.
Pressure drop (Pa)
Pressure drop (Pa)
Pressure drop (Pa)

12 1.4 1.4

10 1.2 1.2 Therefore to size a valve, flow and pressure requirements must
8 1 1
be obtained as a function of time for a given task. Obviously,
6 0.8 0.8
the task becomes more demanding when the manipulator’s
* * bucket is digging a trench.
4 * 0.6 0.6
0 2 4
Time(secs)
6 8 10 0 2 4
Time(secs)
6 8 10 0 2 4
Time(secs)
6 8 10
The flow through the valves for the three actuators is
obtained from the following equations,
Fig. 13. The Pressure drop profiles across the cylinder valves, when the
bucket is digging a sandy-loom soil and at original piston diameters; (a) Qbo = Apbo ẋpbo (32)
Boom cylinder valve (b) Arm cylinder valve (c) Bucket cylinder valve
Qa = Apa ẋpa (33)
2) Actuator Sizing: As shown in (25)- (27) the cross sec- Qbu = Apbu ẋpbu (34)
tional area of the piston of a cylinder determines the pressure
drop across the cylinder during the working stroke, which is where Q is the flow to the cylinder, Ap is the average area
considered to be the extension stroke. And from (28)-(30) of the cylinder piston, and ẋp is the velocity of the cylinder
the maximum possible pressure drop (Δpmax ) across a given piston.
cylinder should be equal to the supply pressure (ps ), that is, The pressure drops across the valves is obtained from (28)-
(30). These equations can be used to plot valve flow versus
Δpmax = ps (31) valve pressure drop for the desired end-point trajectories.

254
 International Journal of Mechanical, Industrial and Aerospace Engineering 3:4 2009

-4
(a) -4
(b) -4
(c)
The resulting Q − Δp curve should lie below the valve
x 10 x 10 x 10
8 8 8
Q -'p Qv-'pv Qv-'pv
v v
pressure-flow characteristics at full valve opening, Qv − Δpv , Q-'p Q-'p Q-'p

Flow rate (m3/s)

Flow rate (m3/s)

6

Flow rate(m3/s)
6 6

typically a curve described by a relationship of the form;

4 4 4

2
Qv = Cd AO Δpv (35) 2 2 2
ρ
0
Where Qv is the flow rate through a valve, Δpv is the pressure
0 0
0 5 10 15 0 5 10 15 0 5 10 15
Pressure drop 6
x 10
Pressure drop x 106 Pressure drop 6
x 10
drop across the valve, Cd is the discharge coefficient, ρ is the (Pa) (Pa) (Pa)

fluid density and AO is the area of the orifice opening. Fig. 16. Pressure drops versus flow curves, when the bucket is digging a
If Q − Δp curve does not lie below the valve pressure-flow sandy-loom soil and at the optimal sizes of the valve’s orifices; (a) Boom
characteristic curve at full valve opening, then the pressure cylinder valve (b)Arm cylinder valve (c) Bucket cylinder valve
drop across the valve is less since the pressure drop across
the actuator is large. Therefore, the valve flow rate is not
able to provide the motion to the manipulator at the specified IV. C ONCLUSION
speed at a particular operating pressure. In this case a valve of Bond graph modeling tool has been applied to model the
larger capacity must be specified, or the value of the operating mechanical dynamics of an excavating manipulator which was
pressure increased. Fig. 15, shows the typical plots of such modeled as a 3 degree of freedom planar manipulator. This
curves when the bucket is digging a sandy-loom soil. was done by applying forward recursive equations similar to
-3
those applied in iterative Newton-Euler method only to deter-
x 10 (a) x 10
-3
(b) x 10
-3
(c)
6
Q -'p
6
Q -'p
6
Q -'p
mine the centroid velocities of the links, unlike in Newton-
v v v v v v
5
Q-'p 5 Q-'p 5 Q-'p Euler method which requires extra recursive computations to
Flow rate (m3/s)
Flow rate(m3/s)

Flow rate (m3/s)

4 4 4
determine the centroid accelerations of the links. A dynamic
3 3 3
model resulted after representing the horizontal and vertical
2 2 2
velocities of the links in bond-graphic form, while considering
1 1 1
the momenta and weights of the links as the bond graph
0 0 0
0 5
Pressure drop
10 15
6
x 10
0 5
Pressure drop
10 15
6
0 5
Pressure drop
10 15
6
elements. On the other hand, Newton-Euler method requires
x 10 x 10
(Pa) (Pa) (Pa)
backward recursion to be performed in order to obtain a
Fig. 15. Pressure drops versus flow curves, when the bucket is digging
dynamic model. This showed that the bond graph method
a sandy-loom soil and at the initial sizes of the valve’s orifices; (a) Boom reduces significantly the number of recursive computations
cylinder valve (b) Arm cylinder valve (c) Bucket cylinder valve required to be performed to a manipulator for a dynamic model
to result, and therefore it can be concluded that bond graph
As seen in Fig. 15, the Q − Δp curves for all the cylinder method is more computationally efficient than the Newton-
valves are far below the valve characteristic curves, hence it Euler method in developing dynamic models of manipulators.
can be concluded that the manipulator will be able to operate Based on the developed model, valve-sizing and actuator-
with the selected valves, although the valve orifice ports are sizing methodologies were briefly outlined and used to obtain
seen to be oversized. An optimal orifice port size should ensure the optimal sizes of the ports of the spool valves as well as
that the peak value of the Q − Δp curve is well near the valve the optimal sizes of the pistons of the hydraulic cylinders.
characteristic curve.
The optimal orifice port sizes were determined by tuning
R EFERENCES
the radii of the ports until that instant when the peak values
of the Q − Δp curves for all the valves are well below the [1] J. Denavit, and R. S. Hartenberg, “A kinematic notation for lower-pair
mechanisms based on matrices,” Journal of Applied Mechanics, pp. 215–
characteristic curves. These values were found to be precisely 221., 1955.
equal to 1.5mm for all the cylinder valves. The optimal port [2] K.S. Fu, R. C. Gonzalez, and C. S. Lee, Robotics: Control, Sensing,
diameter values of 3mm were rounded to the next imperial Vision and Intelligence. McGraw Hill Book Publishing Company, 1987.
[3] J. J. Craig, Introduction to Robotics: Mechanics and Control. Addison-
values which are available in the market as 3.175mm ( 18 in). Wesley Publishers, USA, 1986.
The resulting pressure drop versus flow curves are shown in [4] V. Anand, H. Kansal, and A. Singla, “Some aspects in bond graph
Fig. 16. modeling of robotic manipulators: Angular velocities from symbolic
manipulation of rotation matrices,” Technical Report, Department of
Mechanical Engineering, Sant Longowal Institute of Engineering and
Technology, 2003.
[5] H. M. Paynter, Analysis and Design of Engineering Systems. MIT Press
Publishers, Cambridge, 1961.
[6] D. C. Karnopp, D. L Margolis, and R. C. Rosenberg, System Dynamics:
Modelling and Simulation of Mechatronic Systems. John Wiley and Sons
Publishers, Newyork, 2000.
[7] D. C. Karnopp, D. L. Margolis, and R. C. Rosenberg, System Dynamics;
A Unified Approach. John Wiley and Sons Publishers, Newyork, 2nd ed.,
1990.
[8] D .C. Karnopp and R. C. Rosenberg, Introduction to Physical System
Dynamics. McGraw Hill Publishers, Newyork, 1983.
[9] P. Breedveld, “Bond graphs,” in Encyclopedia of Life Support Systems,
Modeling and Simulation, 2003.

255
 International Journal of Mechanical, Industrial and Aerospace Engineering 3:4 2009

[10] P. Gawthrop and L. Smith, Metamodeling: Bond Graphs and Dynamic

Systems. Prentice Hall International Publishers, UK Limited, 1996.
[11] F. Fakri, A. Rocaries, and A. Carrierre, “A simple method for con-
version of bond graph models in representation by block diagrams,”
in 1997 Proc. International Conference on Bond Graph Modeling and
Simulation.
[12] J. F. Broenink, “Introduction to Physical Systems Modeling with Bond
Graphs,” Technical Report, Department of Electrical Engineering, Uni-
versity of Twente, Netherlands, 1996.
[13] O. M. Muvengei, “Design of an excavating mechanism to be used with
juja diesel tractor jk01,” Design Innovation, Department of Mechanical
Engineering, J.K.U.A.T, 2006.
[14] A. Koivo, “Kinematics of excavator (back hoe) for transfering surface
materials,” Journal of Aerospace Engineering, vol. 7(1), pp. 7–31, 1994.
[15] J. Y. Luh, M. W. Walker, and R. P. Paul, “On-line computational
scheme for mechanical manipulators,” Journal of Dynamic Systems,
Measurement and Control, vol. 120, pp. 69–76, 1980.
[16] H. N. Cannon, “Extended earthmoving with an autonomous excavator,”
MSc. thesis, Robotics Institute, Carnegie Mellon University, Pittsburgh,
1999.
[17] H. Q. Nguyen, ”Robust low level control of robotic excavation”. PhD.
thesis, Australian Centre for Field Robotics, The University of Sydney,
2000.
[18] S. Sarkar, “Dynamic modeling of an articulated forestry machine for
simulation and control,” MSc. thesis, Department of Mechanical Engi-
neering, McGill University, Canada, 1996.
[19] O. M. Muvengei, “Simulation of the dynamic behavior of an excavator
due to interacting mechanical and hydraulic dynamics,” MSc. thesis,
Department of Mechanical Engineering, JKUAT, Kenya, 2008.

256

View publication stats