KINEMATIC ANALYSIS OF MECHANISM BY

USING BOND-GRAPH LANGUAGE

Gregorio Romero, Jesús Félez, M. Luisa Martínez and Joaquín Maroto
Engineering Graphics and Simulation Group
Department of Mechanical and Manufacturing Engineering
Universidad Politécnica de Madrid
C\José Gutierrez Abascal Nº2, CP. 28006, Madrid (Spain)
E-mail:{gromero; jfelez; muneta; jmaroto}@etsii.upm.es

KEYWORDS corresponding to the bucket movement mechanism

instead of dynamic equations. Thus, the set of equations
Bond Graph, differential, algebraic, equations, needed to obtain the movement of the whole assembly,
dynamics, kinematics, mechanism. will depend solely on the movement of the main arms
instead of on the external actions existing in all of the
implements.
ABSTRACT
In recent years, the generation of dynamic equations of
This paper presents a methodology for obtaining the systems modeled with bond graphs has been the topic of
equations corresponding to a mechanism that are considerable research. These equations have been
necessary for carrying out a kinematic simulation. A presented in a variety of forms (Bos 1985). Classical
simulation of this kind means obtaining the co-ordinates formulations express these equations in terms of a large
dependent on the system according to the movements number of momentum associated to inertances and
imposed by the degrees of freedom. Unlike a dynamic displacements associated to compliances. These
simulation, where the set of elements moves according elements can present integral or differential causality. In
to the different external forces existing, in kinematic order to obtain these equations it is necessary to
simulation the movement of the whole set depends establish the causality in the model. Karnopp &
exclusively on imposing movement on one or more of Margolis (Karnopp and Margolis 1993) contributed
the bodies according to the degrees of freedom initially with the stiff compliance approach, where high stiffness
possessed by the mechanism. After presenting an compliances are introduced in the model to eliminate
analysis of how to obtain the necessary equations for the casual loops between integral and derivative causal
several simple systems, this methodology is applied to storage ports. In this case, a set of differential equations
the particular case of a wheel loader, where in order to is obtained, including as variables the flows associated
move and tilt the bucket, various closed mechanisms are to the previous inertances, displacements associated to
integrated. the previous compliances, and the corresponding
variables associated to the new stiff elements
introduced, increasing seriously the number of
1. INTRODUCTION differential equations to be solved. Usually, this
approach needs the use of special numerical solvers for
Real time simulation is an essential requirement in stiff differential equations, requiring very small
simulations such as those of vehicle dynamics, where integration step times. For these reasons, this procedure
the driver expects an immediate response, as is the case is not appropriate for simulation in real time.
in a real situation. The implementation of each and
every part of a vehicle, together with the fact that they Another approach was the introduction of Lagrange
are more and more complicated, means that this real multipliers (Bos 1986) in the model. Some other authors
time simulation requires a thorough in-depth analysis (Gawthrop and Smith 1992) introduce residual sinks
aimed at simplifying to the maximum everything that is and sources ((Borutzky and Cellier 1996, Borutzky
either not strictly essential or requires more information 1995) in a similar solution to the Lagrange approach
than necessary. (Félez et al. 1990). When Lagrange multipliers are used
to break causal paths, a set of differential-algebraic
When it comes to simulating machinery such as equations is obtained, composed by a number of
backhoes, wheel loaders,..., there are two parts that differential equations equal to the number of inertances
come together; that corresponding to the dynamics of plus compliances, and a number or algebraic (or
the vehicle itself, and the part relative to the movement constraint) equations equal to the number of Lagrange
of implements such as bucket, arms, actuators, ... multipliers. The number of equations is the same than
with the previous method, being also not appropriate for
The aim of this article is to demonstrate the validity of real time. Nevertheless, this approach presents
implementing the relevant kinematic equations important formulation advantages that will be

Proceedings 20th European Conference on Modelling and Simulation

Wolfgang Borutzky, Alessandra Orsoni, Richard Zobel © ECMS, 2006
ISBN 0-9553018-0-7 / ISBN 0-9553018-1-5 (CD)
considered in this work. borne in mind that all of them move on a vertical plane,
it being possible to study the corresponding flat
Previous methods prevent the existence of causal loops, mechanism. Therefore, the system of co-ordinates
but it is necessary to preanalyze the model and to shown in the next figure, will be used.
modify the model in a subsequent operation (John D. et
al. 1993). Another approach to solve the problem is the
use of break variables (Félez et al. 1997, Félez et al.
2000) to open the causal loops. Causal loops always
present an algebraic character. Algebraic loops relate
their internal variables by means of algebraic
relationships. It means that these kinds of loops do not
involve integration operations. This fact leads to the
definition of Zero-order Causal Paths ZCPs (Van Dijk
and Breedveld 1991). The mathematical model obtained
from bond graphs with ZCPs and opened with break
variables is also a differential-algebraic equation set
(DAE) (Romero et al. 2005, Granda 2005). Figure 1. Planar 2D rod in global co-ordinates

This last approach reduces the number of equations to a The concept of degrees of freedom may be defined as
number equal to the number of inertances plus the number of entries that need to be provided in order
compliances plus break variables, but it is not enough to give rise to a predictable exit. Each type of entry
for real time. required will need some kind of starter or actuator,
either in the form of a solenoid engine or hydraulic
The use of Lagrange multipliers or the introduction of cylinder. Thus, a flat rod will consist of three degrees of
break variables implies the necessity of new variables freedom.
and the appearance of constraint equations, increasing
considerably the number of involved variables and the This is why a representation of a flat rod using a Bond-
number of equations to be solved. These procedures are Graph will consist of three Inertial type ports (Ix, Iy and
very useful to obtain the system equations in a IJ).
systematic way, but they have to be improved or
modified for their use in real time simulation.

2. SIMULATION OF MECHANISMS

As we know, kinematics studies the movements of

particles and stiff bodies without taking account of the
forces needed to give rise to these movements. Statics is
the study of mechanical systems where the resultant of
the system of forces is zero, and are thus balanced (at
rest or moving at constant velocity). Finally, when this
resultant is not zero, the mechanical system is charged
with accelerated movement, and these unbalanced
forces and the movements they give rise to, comprise Figure 2. Planar 2D rod in global co-ords. using a Bond-Graph
the field of study called dynamics.
In this figure we can see the name assigned to the
When it comes to studying a mechanism’s movement, it element (TF2, Ix, …), two dots (:) and the expression
is essential to choose one reference system or another. associated to it (-b·cos(_XIJ_), m, … respectively). The
Thus, a system of local co-ordinates will usually be variables associated with the different ports are P (m·V)
chosen when studying three-dimensional solids, which and X in the case of Inertances and Compliances
requires a change of reference systems in the existing respectively. If we write VI1, we think in the velocity of
joints of a mechanism of this type. When analysing the a Inertance called ‘I1’, but in the case of the mechanism
behaviour of a mechanism that only moves on a plane, we need introduce the integral of VI1 into the TF
it turns out to be much simpler to work with a system of elements, it said the angular displacement (XI1). For this
co-ordinates that are parallel to the global system, reason in the next examples I will use this notation to
which avoids having to change the reference systems in represent displacements (XName) and velocities (VName).
the joints.
By way of example, we will study the behaviour of
some simple mechanisms, with the aim of observing the
When studying the movements of the different
implements of an backhoe or wheel loader, it must be differences existing as to the process for formulating
movement equations depending on whether some
degrees of freedom are worked with rather than others, In an initial study, we will bear in mind that only the
and how these equations are influenced by whether the weight of both rods themselves act as external forces.
mechanism under study is of an open kinematic chain,
or to the contrary, contains some closed loop. If we take what is shown in figure 2 as a rod model, the
In order to analyse and study the equations system formed using a Bond-Graph, once the causality
corresponding to the use of the above Bond-Graph rod has been obtained, will be as shown in the following
model , we will apply it to an open chain mechanism, as figure.
is the case of a double pendulum, and to two closed B l2

mechanisms, a crank-slider and an articulated

l1
Y 2
1
quadrilateral. Finally, we will apply the results obtained Z1 C
A Z2
to a wheel loader. X

xC2 ??
For the different examples presented the joints are Figura 3. Double pendulum
assumed ideal.
As can be seen in the figure 4, there are only two
3. DOUBLE PENDULUM degrees of freedom, those corresponding to each rod,
(3x2) less the restrictions due to each joint (2x2).
A double pendulum comprises two rods joined together,
one of them being joined by a joint at a fixed point in The Compliances situated in the extreme of the model
space. (KptoC_x and KptoC_y) are used to see the trajectory of
the point C.

Figure 4. Double pendulum using Bond-Graph

Since the variables corresponding to the ports There are therefore 2 ports with integral causality and 4
associated with the angular inertia of each rod, that is, with differential causality, which means that the system
the angular displacements, appear as parameters within of equations corresponding to the double pendulum is
the different transformer elements, these ports must be formed by 2 differential equations and 4 algebraic ones
the ones that correspond to the degrees of freedom, and in the following way:
thus, have integral causality.
1 1
VI1_x( t )  L1 sin( XI1_J( t ) ) VI1_J( t ) (1) VI2_x( t ) L1 sin( XI1_J( t ) ) VI1_J( t ) L2 sin( XI2_J( t ) ) VI2_J( t ) (3)
2 2
1 1
VI1_y( t )  L1 cos( XI1_J( t ) ) VI1_J( t ) (2) VI2_y( t ) L1 cos( XI1_J( t ) ) VI1_J( t ) L2 cos( XI2_J( t ) ) VI2_J( t ) (4)
2 2

§ m2 § d VI2_x( t ) ·1 m1 § d VI1_x( t ) · · L1 sin( XI1_J( t ) )§ §  g 1 § d VI1_y( t ) · · m1 § § d VI2_y( t ) ·g · m2 · L1 cos( XI1_J( t ) )
d ¨¨ ¨¨ ¸¸ ¨¨ ¸¸ ¸¸ ¨¨ ¨¨ ¨¨ ¸¸ ¸¸ ¨¨ ¨¨ ¸¸ ¸¸ ¸¸
VI1_J( t )  © © dt ¹ 2 © dt ¹¹ © © 2 2 © dt ¹¹ © © dt ¹ ¹ ¹
dt J1
(5)

1 d g 1 d
m2 L2 sin( XI2_J ( t ) ) §¨¨ VI2_x( t ) ·¸¸§¨¨   §¨¨ VI2_y( t ) ·¸¸ ·¸¸ m2 L2 cos( XI2_J ( t ) )
d 2 © dt ¹ © 2 2 © dt ¹¹
dt
VI2_J( t ) 
J2
(6)

As can be seen, the equations (5)-(6) obtained depend on However, if the angular variation experienced by each
the external forces, which means that since they are not rod at a particular instant is introduced, that is, if
balanced, they will originate a movement and thus a movement is applied to as many elements as there are
dynamic simulation will be had. degrees of freedom initially existing, there will only be
inertial ports with differential causality, as can be seen
in following figure.
 Figura 5. Double pendulum Z1 and Z2 using a Bond-Graph.

Once the differential algebraic equations have been movements of the different stiff bodies, as is the case
obtained for the model (similar to eqs. (1)-(6)), it can be here.
seen that the position of the different elements no longer
depends on the external forces but only on the Thus, it may be stated that when dealing with a dynamic
movements imposed and the geometric relationships simulation where there are as many degrees of freedom
existing between the two rods. initially as impositions of movement, it is possible to
eliminate the inertial ports from the model along with
Reducing (Romero et al. 2005) the system of equations the external forces.
(1)-(6) to a differential one, we get:
Despite, if the same number of equations is to be
d 1
dt
XI1_x ( t )  L1 sin( XI1_J ( t ) ) w1( t )
2 (7) formulated, they must be substituted by zero value
compliance-type ports, obtaining a model similar to that
d
dt
XI1_J ( t ) w1( t ) (8) shown in figure 6. We need add to the model a zero
d 1
value port instead of null effort source because we need
dt
XI1_y ( t )  L1 cos( XI1_J ( t ) ) w1( t )
2 (9) variables into the model (angle produced by the the two
d 1
rods) and we cannot delete all the ports.
XI2_x ( t )  L2 sin( XI2_J ( t ) ) w2( t )L1 sin( XI1_J ( t ) ) w1( t ) (10)
dt 2
d The difference between the original model and this one,
XI2_J ( t ) w2( t ) (11)
dt is that although both systems generate the same final
d 1 equations, the former starts from differential algebraic
XI2_y ( t )  L2 cos( XI2_J ( t ) ) w2( t )L1 cos( XI1_J ( t ) ) w1( t ) (12)
dt 2 equations that have to be reduced to a system of
As previously stated, kinematic simulation does not take differential equations, whereas since the latter has all
account of the forces necessary to originate the the integral causality ports, the differential equations are
obtained directly.

Figure 6. Double pendulum equivalent using a Bond-Graph.

Since the position of all the elements of the mechanism equations (eqs. (7)-(12)) to 2 differential equations (eqs
can be obtained by simply knowing the angles of each (8)-(11)).
of the rods, the ports corresponding to the horizontal
and vertical displacements can also be eliminated, Thus, the rod to be used in mechanisms having as many
thereby passing from an initial 6 differential algebraic impositions of velocity as initial degrees of freedom,
will contain only one zero value spring-type port.
 do the same using the flat rod equivalent obtained in the
section above.

B l2

l1
Y
1
Z1 C
A 3
X

Figure 8. Crank-Slider mechanism

Figure 7. Flat rod equivalent using a Bond-Graph
Unlike the previous model, this system comprises a
closed mechanism, since we can pass from one element
4. CRANK-SLIDER MECHANISM to another without twice passing over the same element
on the ground. In this kind of element there exist fewer
A mechanism of this kind is similar to the double degrees of freedom than the number of rods.
pendulum dealt with in the above section, with the
difference that the only end of the rods that was not As can be seen in the following figure, the model only
restricted is only allowed movement in one direction has one degree of freedom, as many as correspond to
here. each rod (3x2) less the restrictions due to each joint
(2x2) and the one introduced at the very end of the rod
First, we will set out this mechanism in a Bond-Graph nº 2 (1). So, by introducing a single velocity source, we
using flat rods in traditional global co-ordinates in order can know the evolution of the rest of the model..
to obtain the corresponding equations and then we will

Figure 9. Crank-Slider mechanism using a Bond-Graph.

The introduction of this movement causes all the In the previous equation´s system Y1 and Y2 dependent
inertial-type ports to have differential causality, and variables are the flow and effort associated with the
makes it necessary to impose random causality on one intermediate graph (*) (Romero et al. 2005).
of the intermediate graphs (*), and therefore, make up a
set of differential algebraic equations (DAE): Working with the above system, we can form one made
1 up of only differential equations (ODE), which makes it
Y1  L1 cos( XI1_J( t ) ) w1( t )
2 (13) easier and quicker (sometimes) to solve:
m2 g d 1
Y2  XI1_x ( t )  L1 sin( XI1_J( t ) ) w1( t )
2 (14) dt 2 (21)
d 1 d
XI1_x ( t )  L1 sin( XI1_J( t ) ) w1( t ) XI1_J( t ) w1( t )
dt 2 (15) dt (22)
d d 1
XI1_J( t ) w1( t ) XI1_y ( t )  L1 cos( XI1_J ( t ) ) w1( t )
dt (16) dt 2 (23)
d 1 1 3
XI1_y ( t )  L1 cos( XI1_J( t ) ) w1( t )  L1 w1( t ) sin ( XI1_J ( t )  XI2_J ( t ) )  L1 w1( t ) sin ( XI1_J ( t )  XI2_J ( t ) )
dt 2 d 4 4
(17) dt
XI2_x ( t )  
cos ( XI2_J ( t ) )
d sin( XI2_J( t ) ) Y1
XI2_x ( t ) L1 sin( XI1_J( t ) ) w1( t ) d L1 cos( XI1_J( t ) ) w1( t )
dt cos( XI2_J( t ) ) (18) XI2_J( t ) 
dt L2 cos( XI2_J( t ) ) (25)
d 2 Y1
XI2_J( t )  d 1
dt L2 cos( XI2_J( t ) ) (19) XI2_y ( t )  L1 cos( XI1_J( t ) ) w1( t )
dt 2 (26)
d
XI2_y ( t ) L1 cos( XI1_J( t ) ) w1( t )Y1
dt (20) If we look closely at the initially formed DAE system
(13)-(20), we can see how the term corresponding to the
weight of the second rod appears, while in the ODE
system obtained (21)-(26) on reducing it, it does not For this reason, since no external force influences the
appear. This shows that although the external force is movement of the mechanism, it is possible to substitute
included in the differential algebraic equations, it is the initial rods by those obtained in point 3, where only
really not necessary and in no way affects the final the ports corresponding to the angular displacement of
solution, whether it appears or not. the rods appear.

Figure 10. Crank-Slider mechanism equivalent using a Bond-Graph.

As with the initial model, causality needs to be imposed This same approach will be used for the following
randomly on one of the intermediate graphs (*). To the mechanism.
contrary, all the ports that appear have integral causality
and therefore the system of equations is formed by 1 5. ARTICULATED QUADRILATERAL
algebraic equation and 2 differential equations instead
of the previous 8 differential algebraic equations. An articulated quadrilateral is a mechanism where three
1 rods are integrated by means of two joints, the
Y7  L1 cos( XI1_J ( t ) ) w1( t )
2 (27) remaining points being joined at fixed points in space.
d l2
XI1_J( t ) w1( t )
dt (28) 2

d 2 Y7 C
XI2_J( t ) 
dt L2 cos( XI2_J( t ) ) (29) B

l3
3
l1

1
Once the above equations have been operated with one Y Z1
other, it can be seen that the two differential equations
thus formed (30),(31) correspond to those obtained on A D
X
reducing the DAE system to ODE (22),(25), and therefore
it can be concluded that both Bond-Graph models d

generate the same angular movement of the rods.

Figure 11. Articulated quadrilateral.
d
XI1_J( t ) w1( t )
dt (30) As before, we will make the initial approach with flat
d L1 cos( XI1_J( t ) ) w1( t )
rods in global co-ordinates traditionally used in Bond-
XI2_J( t ) 
dt L2 cos( XI2_J( t ) ) (31) Graphs and then we will do so with the equivalent flat
rods in question..

Figure 12. Articulated quadrilateral Bond-Graph.

There are three rods and four joints in this mechanism, In order to conclude the causal analysis of the model,
which means the number of degrees of freedom will be random causality must be introduced on two of the
one (3x3 - 4x2). As can be seen in the above figure, by intermediate graphs (*). It can be observed how all the
introducing movement on one of the rods (rod 1), the ports have differential causality, for which reason, the
movement for the rest of the model may be obtained. corresponding system of equations will be of the
differential algebraic kind (DAE):

1 L1 w1( t ) cos( XI1_J ( t )XI3_J( t )XI2_J ( t ) )L1 w1( t ) cos( XI1_J( t )XI3_J ( t )XI2_J( t ) )2 L1 w1( t ) cos( XI1_J( t )XI3_J ( t )XI2_J( t ) )
Y0 
4 sin( XI3_J ( t )XI2_J( t ) ) (32)
( m3 gm2 g ) ( sin( 2 XI2_J( t )XI3_J( t ) )sin( 2 XI2_J( t )XI3_J( t ) ) )
Y1 
4 cos( 2 XI2_J( t )XI3_J( t ) )4 cos( XI3_J( t ) ) (33)
1 L1 w1( t ) sin( XI1_J( t )XI3_J( t )XI2_J( t ) )L1 w1( t ) sin( XI1_J( t )XI3_J( t )XI2_J( t ) )
Y2 
4 sin( XI3_J( t )XI2_J( t ) ) (34)
3 m3 g m2 g 1 ( m3 gm2 g ) sin( XI3_J( t )XI2_J( t ) )
Y3   
4 4 4 sin( XI3_J( t )XI2_J( t ) ) (35)
d 1
XI1_x ( t )  L1 sin( XI1_J( t ) ) w1( t )
dt 2 (36)
d
XI1_J( t ) w1( t )
dt (37)
d 1
XI1_y ( t )  L1 cos( XI1_J ( t ) ) w1( t )
dt 2 (38)
d 2 sin( XI3_J( t ) ) Y2
XI2_x ( t ) L1 sin( XI1_J( t ) ) w1( t )Y0
dt cos( XI3_J( t ) ) (39)
d 2 Y0 4 sin( XI3_J( t ) ) Y2
XI2_J( t )  
dt L2 sin( XI2_J( t ) ) cos( XI3_J( t ) ) L2 sin( XI2_J( t ) ) (40)
d cos( XI2_J( t ) ) Y0 2 sin( XI3_J( t ) ) cos( XI2_J( t ) ) Y2
XI2_y ( t ) L1 cos( XI1_J( t ) ) w1( t ) 
dt sin( XI2_J( t ) ) cos( XI3_J( t ) ) sin( XI2_J( t ) ) (41)
d sin( XI3_J( t ) ) Y2
XI3_x ( t ) 
dt cos( XI3_J ( t ) ) (42)
d 2 Y2
XI3_J( t ) 
dt L3 cos( XI3_J( t ) ) (43)
d 2 cos( XI2_J( t ) ) Y0 § 4 sin( XI3_J( t ) ) cos( XI2_J( t ) ) ·
XI3_y ( t ) L1 cos( XI1_J( t ) ) w1( t ) ¨¨ 1 ¸ Y2
dt sin( XI2_J( t ) ) © cos( XI3_J( t ) ) sin( XI2_J( t ) ) ¸¹
(44)

As can be seen, the system of equations formed (32)-(44) tedious, which makes a real time simulation
is much more complex than for the previous model (13)- unapproachable.
(20), without having introduced anything other than one
rod and one restriction. As was to be expected, as the On operating the different equations (32)-(44) with one
number of rods and restrictions increases, the system of another in order to get a system of differential
equations formed is ever more complicated and equations, we obtain:
d 1
XI1_x ( t )  L1 sin( XI1_J( t ) ) w1( t )
dt 2 (45)
d
XI1_J( t ) w1( t )
dt (46)
d 1
XI1_y ( t )  L1 cos( XI1_J( t ) ) w1( t )
dt 2 (47)
d 1 ( cos( XI1_J( t )XI3_J( t )XI2_J( t ) )cos( XI1_J( t )XI3_J( t )XI2_J( t ) )2 cos( XI1_J( t )XI3_J( t )XI2_J( t ) ) ) w1( t ) L1
XI2_x ( t ) 
dt 4 sin( XI3_J( t )XI2_J( t ) ) (48)
d L1 sin( XI1_J( t )XI3_J( t ) ) w1( t )
XI2_J( t ) 
dt L2 sin( XI3_J( t )XI2_J( t ) ) (49)
d 1 ( sin( XI1_J ( t )XI3_J ( t )XI2_J ( t ) )2 sin( XI1_J ( t )XI3_J ( t )XI2_J ( t ) ) sin( XI1_J ( t )XI3_J ( t )XI2_J ( t ) ) ) w1( t ) L1
XI2_y ( t )  
dt 4 sin( XI3_J ( t )XI2_J ( t ) ) (50)
d 1 L1 w1( t ) cos( XI1_J( t )XI3_J( t )XI2_J( t ) )L1 w1( t ) cos( XI1_J( t )XI3_J( t )XI2_J( t ) )
XI3_x ( t ) 
dt 4 sin( XI3_J( t )XI2_J( t ) ) (51)
d L1 sin( XI1_J( t )XI2_J( t ) ) w1( t )
XI3_J( t ) 
dt L3 sin( XI3_J( t )XI2_J( t ) ) (52)
d 1 L1 w1( t ) sin( XI1_J( t )XI3_J( t )XI2_J( t ) )L1 w1( t ) sin( XI1_J( t )XI3_J( t )XI2_J( t ) )
XI3_y ( t ) 
dt 4 sin( XI3_J( t )XI2_J( t ) )
(53)
If we apply the previously obtained rod equivalent, the As with the traditional case, causality must be
model formed will only comprise spring ports with introduced on two of the intermediate graphs(*), which
integral causality that will generate differential will generate algebraic equations.
equations.

Figure 13. Articulated quadrilateral equivalent Bond-Graph.

The system of equations thus formed, will comprise two algebraic equations and three differential equations:

1 L1 w1( t ) cos( XI1_J( t )XI3_J( t )XI2_J( t ) )L1 w1( t ) cos( XI1_J( t )XI3_J( t )XI2_J( t ) )2 L1 w1( t ) cos( XI1_J( t )XI3_J( t )XI2_J( t ) )
Y3 
4 sin( XI3_J( t )XI2_J( t ) ) (54)
1 L1 w1( t ) sin( XI1_J( t )XI3_J( t )XI2_J( t ) )L1 w1( t ) sin( XI1_J( t )XI3_J( t )XI2_J( t ) )
Y5 
4 sin( XI3_J( t )XI2_J( t ) ) (55)
d
XI1_J( t ) w1( t )
dt (56)
d 4 Y3 sin( XI3_J( t )XI2_J( t ) )4 Y3 sin( XI3_J( t )XI2_J( t ) )8 Y5 cos( XI3_J( t )XI2_J( t ) )8 Y5 cos( XI3_J( t )XI2_J( t ) )
XI2_J( t ) 
dt 2 L2 cos( XI3_J( t ) )L2 cos( 2 XI2_J( t )XI3_J( t ) )L2 cos( 2 XI2_J( t )XI3_J( t ) ) (57)
d 2 Y5
XI3_J ( t ) 
dt L3 cos( XI3_J( t ) )
(58)
Operating the above equations with one another, we’ve:
d
XI1_J( t ) w1( t )
dt (59)
d L1 sin( XI1_J( t )XI3_J( t ) ) w1( t )
XI2_J( t ) 
dt L2 sin( XI3_J( t )XI2_J( t ) ) (60)
d L1 sin( XI1_J( t )XI2_J( t ) ) w1( t )
XI3_J( t ) 
dt L3 sin( XI3_J( t )XI2_J( t ) ) (61)
As we have seen, in these previous mechanisms, ones
with as many velocity sources as initial degrees of
freedom, the angular variation of each rod (59)-(61) may
be obtained more quickly and effectively simply by
using the rod in question, since the complexity and
number of equations resulting from the model is
reduced.
Figure 14. Wheel loader mechanism.

6. WHEEL LOADER MECHANISM Movement of the different parts is performed by the

action of two actuators (7 and 8) whose action will be
By using this mechanism, the raising and tilting implemented by introducing the movement
movement of the bucket and the different arms that corresponding to the support arm (1) or the tilt arm (2).
make it articulate, are obtained. They will be used to
reproduce the movements of loading and unloading In order to form the Bond-Graph model for this
material, and will be formed by the arms, struts, bucket mechanism, each of the elements of which it is made up
and actuators. will be started from in isolation, and after generating the
Bond-Graph layout for each of them, they will be
grouped and assembled.
 After obtaining the differential equations corresponding
to the model, the next step is to validate it using its
numeric simulation.

The bucket movement consists of the alternate

displacement of the two main arms, the support one (1)
and the tilt one (2), so that starting from a state of rest
the bucket will be placed horizontally at a small
distance from the ground for it to be able pick up the
load, raise it, and tilt in order to unload the material.

As we are dealing with a mechanism for moving the

bucket, the movement and angle described by it will be
represented during the simulation.

The following figure shows the phases comprising the

bucket movement.
Figure 15. Support arm (element 1).
Firstly, starting from rest, the support arm (1) is initially
In order to facilitate assembly of all the elements once a lowered in order to move the bucket closer to the
Bond-Graph of one element has been made, their ground (A-B) and then this will be placed in a
complete ports and nodes should be grouped and their horizontal position (C-D) so that loading can be
entrances and exits left free so that the element can be properly performed (D-E). Once material has been
linked to the rest of the subsets (fig. 16). loaded, the bucket is tilted backwards (E-F) so that
while circulating (F-G) with the bucket full, the material
Once the different subsets of the mechanism have been loaded does not fall out. Once the point for unloading
obtained, it is essential to impose joints on the different the contents has been reached, firstly the arm is raised
nodes using restrictions, finally obtaining the (G-H) and secondly, it is tilted forward (I-J), in order to
mechanism we are dealing with. unload the material. Finally, the bucket is tilted
backwards (K-L), the arm lowered (M-N) and finally,
the starting point returned to.

Figure 16. Support arm subset

Figure 18. Bucket angle according to time

While the previous figure represents the resulting angle

on the bucket produced by the combined movement of
the actuators, the support one and the tilting one, the
following figure shows the support arm angle.

Figure 17. Wheel loader mechanism using a Bond-Graph.

In this way, the number of degrees of freedom will be

proportional to the number of rods (6x3) and the
restrictions due to joints (2x8). Thus it can be shown
that there are only two degrees of freedom (6x3 – 2x8),
as many as actuators, represented by Z1 and Z2.
 number of algebraic equations, in the way it is set out
here, no more algebraic equations are added, as no
differential causalities exist, and the number of
variables and parameters needed are reduced by
reducing the number of ports and because these are of
zero value.

Finally, when dealing with mechanisms as complex as

those of the wheel loaderl, the use of uncoupling
elements is usually resorted to, which makes real time
calculation unviable due to stiff behaviour. Should the
solution to use stiff joints be adopted, the existence of
three inertias with differential causality for each rod and
Figure 19. Support arm angle according to time the imposition of causality on intermediate graphs,
makes calculation by DAE or ODE systems
Finally, figure 20 shows the path drawn by the point unapproachable, meaning that simulation is not
situated at the end of the bucket. possible.

8. BIBLIOGRAPHY

Bos, A.M. 1986. “Modeling Multibody Systems in

terms of Multibond Graphs”. Ph. D. Thesis. Twente
University, Enschede, The Netherlands.

Félez, J. et al. 1997. “Deriving simulation models from

bond graphs with any combination of topological loop
classes”. International Conference on Bond Graph
Modeling and Simulation ICBGM’97. Phoenix. SCS
Publishing, Simulation Series. Vol.29, No.1, pp.85-93.

Figure 20. Path drawn by the end of the bucket Félez, J. et al. 2000. “Deriving simulation models from
bond graphs with algebraic loops. The extension to
This figure shows how the end descends from its initial multibond graph systems”. Journal of the Franklin
position (A) to a point located lower down (B), it is then Institute. Vol. 337 , pp.579-600.
tilted (F), partly to place the bucket horizontally, and Félez, J. et al. 1990. “Bondyn: A Bond Graph based
partly to hold the material loaded. It is then raised in simulation program”. Trans. ASME. J. Dyn. Syst.
order to unload the bucket (H) and after tilting, it is Measurement and Control. Vol.112 , pp.717-727.
moved to a point located lower down, (J). Once the
material has been unloaded, the bucket returns to a Gawthrop, P.J. and Smith, L.S. 1992. “Causal
position near to its original one after its tilting (M) and augmentation of Bond Graphs with algebraic loops”.
its descent (N). J. of the Franklin Institute. Vol. 329, No.2, pp.291-
303.

7. CONCLUSIONS Van Dijk, J. and Breedveld, P. 1991. “Simulation of

system models containing Zero-order Causal Paths- I.
Taking these results and those of the previous Classification of Zero-order Causal Paths”. J. of the
mechanisms as a basis, the advantage of setting out the Franklin Institute. Vol.328, No.5/6, pp.959-979.
elements using a Bond-Graph as presented here, stems
from being able to set out the model, and thus the Karnopp, D.C. and Margolis, D.L. 1993. “Analysis and
equations of a kinematic model, according to the simulation of planar mechanism systems using
displacements and angular velocities instead of the Bond-Graph”. J. of Mechanism Design. Vol. 101,
external forces. This makes it possible to find the No.2,pp.187-191.
equations typical of the model, not only in smaller
number, but in less computation time needed to John D. Lamb et al. 1993. “Equivalences of bond graph
calculate them. junction structures”. Intemational Conference on
Bond Graph Modeling ICBOM'93. San Diego. SCS
Although it is true that the problems of causal Publishing, Simulation Series, Vol.25, No.2, pp.79-
assignation will be the same in most cases, and it is 84.
necessary in both cases to assign random causality on
intermediate graphs, thereby originating the same
Bos, A.M. 1985. “Implicit solutions to constraints in graphics and virtual reality techniques, mainly
mechanical bond graphs”. Proceedings IMACS Ist addressed towards the development of simulators. He
World Congress on Systems Simulation and has published over 50 technical papers and has been
Scientific Computation. Oslo. Vol.4, pp.309-312. actively involved in over 25 research and development
projects. He has served as thesis advisor for 30 master’s
Romero, G. et al. 2005. “Optimised procedures for theses and four doctoral dissertations.
obtaining the symbolic equations of a dynamic system
by using the Bond-Graph technique”. International M. LUISA MARTÍNEZ received her Mechanical
Conference on Bond Graph Modeling and Simulation Engineering and Doctoral degrees from the Politechnic
ICBGM’05. New Orleans. SCS Publishing, Simulation University of Madrid (Spain) in 1990. She got her PhD
Series. Vol.37, No.1, pp.51-58. Degree in 1997 working on variational geometry. In
1990 she started to work as Associate Professor at the
Romero, G. et al. 2005. “Systematic reduction of dynamic Technical University of Madrid in Spain (UPM). Her
equations to a minimal set in systems modelled with thesis was focused on variational geometry. She usually
Bond Graphs”. International Conference on works in the field of computer graphics, virtual reality
Integrated Modeling and Analysis in Applied Control and CAD. During this time she has been involved in
and Automation IMAACA’05. Marseille. pp.59-65. different educational projects and pilot activities
promoted by the European Commission and other
Borutzky, W. and Cellier, F. 1996. “Tearing in Bond Spanish institutions. She has published over 23
Graphs with Dependent Storage Elements”. technical papers and has been actively involved in over
Proceedings. of CESA'96, pp. 1113 - 1119 16 research and development projects.

Borutzky, W. 1995. “Representing Discontinuities by JOAQUÍN MAROTO received his Control

Sinks of Fixed Causality”. International Conference Engineering and Doctoral degrees from the Madrid
on Bond Graph Modeling and Simulation ICBGM’95. Polithecnic University in 2000 and 2005. He has been
pp. 65 – 72. Assistant Professor at the Technical University of
Madrid in Spain (UPM) since year 2003. His main
Granda, J.J. 2005. “The CAMP-G Symbolic Solution to activities and research interests are mainly focused on
Algebraic Loops in Bond Graph Models”. the field of simulation, computer graphics, virtual
International Conference on Bond Graph Modeling reality and machine vision. His main contribution is in
and Simulation ICBGM’05. New Orleans. SCS the field of distributed virtual environment generation
Publishing, Simulation Series. Vol.37, No.1, and ethe generation of immersive systems. He has
published over 20 technical papers and has been
9. BIOGRAPHY actively involved in over 13 research and development
projects.
GREGORIO ROMERO received his Mechanical
Engineering and Doctoral degrees from the UNED.
(Spain) in 2000. He got his PhD Degree from the
Politechnic University of Madrid in Spain in 2005
working on simulation and virtual reality, optimizing
equations system. He has worked as Assistant Professor
at the Technical University of Madrid in Spain (UPM)
since 2001. He is developing his research in the field of
simulation and virtual reality including simulation
techniques based on bond graph methodology
integrating computer graphics and virtual reality
techniques to the simulation in real time. He has
published over 20 technical papers and has been
actively involved in over 11 research and development
projects and different educational projects.

JESÚS FÉLEZ received his Mechanical Engineering

and Doctoral degrees from the University of Zaragoza
in 1985 and 1989. He started as Associate Professor at
the Technical University of Madrid in Spain (UPM) in
1990 and became Full Professor in 1997. His main
activities and research interests are mainly focused on
the field of simulation, computer graphics and virtual
reality. His research includes simulation techniques
based on bond graph methodology integrating computer