Dynamics
Dynamics is the branch of mechanics that is concerned with the study of motion
and the relation between the forces and motion. The central focus of our study is
the dynamics of systems of rigid bodies and its application to technical problems.
Furthermore, we are basically concerned with the computer aided dynamics of rigid
bodies to give an insight into the contemporary classical dynamics from the computational point of view. This should familiarise the reader (and user of this book) with
the basic concepts of todays computational dynamics realized in various program
packages [4], [5].
The motivation for this approach stems from the fact that in the contemporary
engineering praxis a lot of dynamical problems arise but only very few of them
can be solved in the analytical form by following classical calculation by hand
approaches. For the majority of problems (large-scale problems, analytically nonsolvable differential equations, non-linear tasks, coupled problems etc.) computational methods have to be applied. This fact gives rise to many open questions
concerning the optimal use of the computational tools available within the various
program packages [17].
The experience shows that for an accurate and effective computation, the mechanical and mathematical models of the given engineering problem have to be
properly established. The computational model should contain all the necessary
pieces of information considering the mechanical phenomena under investigation.
It also should be formulated properly to suit the computational method that is intended to be utilized to obtain the final solution. On the other hand, many computational methods for the various kinds of problems are at the users disposal today.
Among them the appropriate ones for the problem at hand should be chosen and
applied.
In this chapter our main goal is to provide the basic principles of the contemporary computational dynamics of rigid body systems as well as the necessary theoretical background. The starting point is the question: What should be considered
190
CHAPTER 4. DYNAMICS
191
4.1 Introduction
4.1.1 Issues of applied dynamics
Dynamics can be classified into the several sub-domains. Each of them has its own
modelling assumptions and procedures. In most of the cases, the computational
methods are also different. According to the characteristics of the problem and the
focus of the intended dynamical analysis, the sub-domain whose approach is best
suited to the problem at hand should be chosen . In the sequel of the chapter, an
overview of the characteristics of the sub-domains and problems of the contemporary dynamics is given.
Multibody dynamics
Multibody dynamics deals with the mechanical systems of interconnected rigid bodies that undergo large displacements and rotations [4], [5], [19]. The bodies are
interconnected by kinematical constraint elements and coupling elements. Both
visco-elastic and inertia properties of the real technical system are discretised during the process of shaping of the systems mechanical model [18].
The mathematical modelling of the established discretised mechanical model
leads to the ordinary differential equations (ODE) (minimal form mathematical
models) or to the differential-algebraic equations (DAE) (mathematical models in
descriptor form) [5].
The concepts of the multibody dynamics can be successfully utilized within
the framework of the following technical applications: vehicle systems, aircraft
subsystems, robotic systems, various kind of mechanisms, biomechanical systems,
mechatronics.
Structural dynamics
Structural dynamics deals with the deformable mechanical structures whose segments generally do not undergo large displacements and rotations (not kinematical
chains). The mass and visco-elastic properties of the system are distributed along
the structure [18].
The basic mathematical modelling generally leads to partial differential equations (PDE). The discretisation of the system that is usually performed in the sequel
of the mathematical modelling procedure yields a mathematical model in the form
of ODE. By using finite element approach [1], very powerful computational proce-
CHAPTER 4. DYNAMICS
192
dures are available for tackling the problems of structural dynamics, see also chapter
5.
Typical structural dynamics applications are: plates, shells, aircraft structures,
trusses, civil engineering structures.
Flexible multibody dynamics
In the framework of flexible multibody dynamics, segments of a system are considered to be flexible [19].
Flexible multibody dynamics typically deals with non-linear structures whose
segments undergo large rigid body motion superposed by flexible deformations
[10]. Modelling and computational procedures of the multibody dynamics and
structural dynamics are being combined in order to formulate efficient procedures
for problems of this kind. The methods of the flexible multibody dynamics are
subjects of extensive ongoing research activities [5].
The applications of flexible multibody dynamics systems can be found in various multibody systems with connected rigid and flexible segments like aircraft rotary wings, flexible robots, biomechanical systems, high-speed mechanisms.
Problems of dynamics
In dynamics various classes of problems can be distinguished.
Inverse dynamics
The inverse dynamics deals with the determination of the applied and constraint forces and torques for a mechanical system whose motion is prescribed
[16].
Beside the full dynamic approach, in which all the forces of the system
are considered in the computation, the quasi-static approach of the inverse
dynamics can be applied. Within the framework of the quasi-static approach,
the inertia forces of the system are neglected.
In most of the cases, an inverse dynamics problem leads to a set of algebraic
equations.
Forward dynamics
The forward dynamics deals with the determination of the motion of the system that is subjected to prescribed applied forces and torques [4].
In most of the technical applications (systems bodies undergo large displacements and rotations, coupling elements of the system possess non-linear characteristics), a forward dynamics problem leads to solving non-linear ordinary
CHAPTER 4. DYNAMICS
193
Dynamics
Structural dynamics
Dynamics of MBS
Dynamial behaviour
(stability tests)
Inverse dynamics
Optimisation
Forward dynamics
Dynamical modelling
Lagrange approach,
Jourdains principle
etc.
NewtonEuler
approach
Inverse
dynamics
(Solving
of linear
algebraic
equations)
Minimal form
formulation
Stability criteria
Stability
analysis
Forward
dynamics
Forward
dynamics
Linear
analysis
Linearization of
the equations
Linear
forward
dynamics
DAE system
Reduction during
ODE system
integration
Integration of
DAE
Linear ODE
(vibration analysis)
Integration of
ODE
Vibrations
In most of the cases in linear domain [13] [15], solving the vibrational problem leads to the determination of the system eigenvalues and modes (Chapter
5). The system stability problem can also be mentioned in this context.
CHAPTER 4. DYNAMICS
194
In the framework of the some very important industrial applications (nonlinear vibration within the vehicle sub-systems, acoustical problems etc.) nonlinear vibrational problems have to be tackled.
Optimisation
The problem of the optimisation of the mechanical systems (weight, costs,
deformation and stresses, dynamical trajectories etc.) is very important in
engineering and lies far out of the scope of this book [2], [6].
However, it can be stated that specialised methods that allow for the optimisation of the mechanical systems according to the specified criteria may be
applied. Another possibility is to look for an improved design via repeated
simulations and variations of the systems parameters.
The first step is the mapping of the reality (technical system) into a set of simplified entities in order to establish a mechanical model [18]. The mechanical model
should include at least the effects under consideration, but not more, i.e. the models
should be as complex as necessary but as simple as possible (A. Einstein: Everything should be made as simple as possible, but not simpler). The mechanical
modelling is not an unique but an iterative process. It needs a lot of engineering
CHAPTER 4. DYNAMICS
195
experience since proper analogies between the reality and the model, dependent on
the goals of the analysis, have to be established [5].
Once the mechanical model is built, in the second step a mathematical model, i.
e. a set of the governing equations which describe the models dynamical behavior,
has to be formulated [7]. The mathematical modelling is also not an unique process.
It depends on the goals of the analysis and the computational procedures and tools
as well as the computer hardware that are intended to be used.
Mechanical modelling
Mechanical modelling is a process that is affected by the character of the problem
and focus of the intended analysis in the first place. Second, the characteristics
of the real objects are important, but only within the scope of the given task and
intended analysis.
A real object can be modelled using different mechanical elements: an aircraft
can be considered as a rigid body within the scope of flight mechanics, but it has
to be modelled as a system of elastic bodies to analyse the landing dynamics phenomena. If its space trajectory is under investigation, a large space station can be
modelled as a particle, but on the other hand, a tennis ball has to be considered to
be an elastic body in the case of its impact analysis.
The crucial modelling criterion is that the mechanical model should be able to
describe (take into account) those mechanical properties of the real system that are
under the consideration with the desired accuracy [16], [18].
Mathematical modelling
Mathematical modelling is a process of formulating a mathematical text (a set of
the equations of motion, for example), referred to the established mechanical model
by following physical laws and principles (Newtonian classical mechanics, smooth
or non-smooth theory).
A good and effective mathematical model has to reflect the type and character
of the analysis that is to be performed (linear or non-linear analysis, for example),
but also has to be properly formulated to suit the computational procedures and
algorithms that are intended to be used for the manipulation and evaluation of the
generated equations [7].
In some special cases, the solution of the established mathematical model may
be found analytically, where the obtained solution is exact under the assumptions
made during mechanical and mathematical modelling. Nevertheless, in most of the
cases computational procedures have to be utilized to find the numerical solutions
(Chapter 4.3.1).
In the past three decades numerous computational techniques and algorithms
CHAPTER 4. DYNAMICS
196
have been established to generate the governing equations for the various classes of
problems and specific kinds of the analysis (multibody systems, structural systems,
systems with the unilateral or variable constraints etc.). These algorithms are the
core of the various program packages that are offered in the market today [18], [19].
Although very often the intended mechanical analysis can be carried out by
starting initially from different mathematical models, an appropriate mathematical
modelling can influence the computational procedure itself to a great extent (reducing the computation time or gaining more accurate results).
CHAPTER 4. DYNAMICS
197
Actuators that prescribe the particular applied forces as the functions of time
(force actuators).
The motion of the system caused by this type of the actuators is generally not
known. It is a subject of the forward dynamic analysis of the system.
Actuators that prescribe the motion of the system, i.e. prescribe the particular
displacements or rotations of the systems bodies as the functions of time
(displacement actuators).
The forces imposed by the actuators of this type are generally not known.
These forces are the subject of the inverse dynamic analysis of the system
[16].
Since these actuators prescribe the systems motion (the system is constrained
to evolve in time in the specific way), the actuators of this type can be considered as the kinematical constraints. Consequently, the forces imposed by
them are classified as the constraint forces (see classification of forces and
kinematical constraints in the sequel of the chapter). The actuators of this
type are also called kinematical drivers.
CHAPTER 4. DYNAMICS
198
Passive elements
P
Coupling elements
Constraint elements
spring
rod
damper
Active elements
Coupling elements
Constraint elements
CHAPTER 4. DYNAMICS
199
The constraint forces are imposed to the system by the kinematical constraint
elements (the joints, bearings, actuators that prescribe motion of the system).
In the case of the ideal constraints, these forces are collinear to the direction
of the restricted motion (see discussion on the kinematical constraints below).
They influence the possible motion of the system.
In Figure 4.5 system forces that appear in double pendulum are analysed and classified by deriving the free body diagram of the system.
CHAPTER 4. DYNAMICS
200
and orientation of the systems bodies) uniquely [4]. In Table 4.1 typical constraint
elements and their characteristics are depicted.
Table 4.1 Types and valency of bearings
If q holonomic constraints are added to the system, its degree of freedom is reduced.
If all q constraints are independent, the degrees of freedom of the system are
f = 6p q .
F
CHAPTER 4. DYNAMICS
201
systems
boundary
Force
external
internal
applied constraint
Statical determination
If all q constraints are independent, the system is a statically determined. On
the other hand, if only r of the q constraints are independent, than n = q r
constraints are superflous. The system is statically n times overdetermined.
In this case the constraint forces can not be calculated without introducing
further modelling assumptions (elastic properties). In Figure 4.6 and Figure
4.7, the systems 4.6 d) and 4.7 d) are statically overdetermined.
CHAPTER 4. DYNAMICS
202
a)
+@+
b)
c)
+*+
+
e)
+@*+
d)
+*+
+
+@*+
+*+
(Slider)
+@*+
+*+
+
+@*+
+*+
(Pendulum)
+
Figure 4.6: Systems with various constraints and DOF: beams with various supports
Kinematical determination
A system is kinematically determined, f = 0, if the displacements and rotations of all its members are completely determined by the constraints.
If all kinematical constraints do not depend on time, the system is a statical
one. The kinematical constraints that do not depend on time are called scleronomic constraints [3]. In Figure 4.7, the systems c) and d) are kinematically
determined (statical systems).
Otherwise, if at least one constraint is dependent on time, the system does
not have a fixed configuration but evolves in time and can be considered as a
kinematical or dynamical system. The kinematical constraints that depend on
CHAPTER 4. DYNAMICS
a)
b)
c)
203
C1
+3@2*4+2
C2
+ 1 2
+4*4+0
C1
C2
+ 1
C1
+3@2*5+1
+3@2*6+0
C2
+5*5+0
+6*6+0
statically and kinematically
determined support
C3
d)
C1
C2
C3
C1
+3@3*9+0
+ 10 * 9 + 1
statically overdetermined
support
C2
Figure 4.7: Systems with various constraints and DOF: structures and mechanisms
CHAPTER 4. DYNAMICS
204
CHAPTER 4. DYNAMICS
205
operations by using operations with matrices that can be easily utilised in computer
applications.
Starting from the classical vector representation, the matrix equations will be
derived.
Dynamics of particles
f
z
r
x
Figure 4.8: Motion of particle
(4.1)
where mv is the linear momentum of a particle. The angular momentum of a particle with respect to O is
hO = r mv :
(4.2)
E QUATIONS OF MOTION OF SYSTEM OF PARTICLES
If the system of p particles shown in Figure 4.9 is considered, the Newtons law
for the i-th particle yields
f i = f ei + f ii = mi ai :
(4.3)
where f i e denotes the resultant of the external forces acting on the i-th particle and
the resultant of the internal forces f i i is given by the equation
p
X
j =1
f ij = f ii :
(4.4)
CHAPTER 4. DYNAMICS
206
f2
f1
f 2pi
f 21i
r2
f 12i
f 1pi
r1
y
f p2i
f p1i
rp
fp
f ij = f ji :
(4.5)
p
X
i=1
fi +
e
p
X
i=1
fi =
i
p
X
i=1
p
X
i=1
mi ai ;
p
X
f ei =
i=1
mi ai :
p
X
i=1
f ii = 0 ;
(4.6)
(4.7)
The equation (4.7) can be elaborated by utilizing the relations (4.8 - 4.11). The
position rC of the system mass centre C is defined by
mr C =
Pp
p
X
i=1
mi ri :
(4.8)
where m = i=1 mi is the total mass of the system. Differentiation of (4.8) with
respect to time leads to the linear momentum of the system of particles
mv C =
p
X
i=1
mi v i :
(4.9)
mi ai ;
(4.10)
maC =
p
X
i=1
CHAPTER 4. DYNAMICS
207
p
X
i=1
f ei = maC :
(4.11)
According to (4.11), the mass centre of the system of particles moves as if the
entire mass of the system were concentrated at that point and all the external forces
were applied there [20].
A NGULAR
ith particle,
r Ai
fi
vi
vA
h_ Ai = r_ Ai mivi + rA miv_ i :
(4.13)
If A coincides with the fixed point O (r_ A v i = v i v i = 0), then it can be
i
written
(4.14)
h_ Ai = lAi ;
(4.15)
or
hA =
p
X
i=1
rAi mivi :
(4.16)
CHAPTER 4. DYNAMICS
208
After differentiation of (4.16) with respect to time and some algebraic operations
and substitutions, it can be shown that equation (4.13) for the system of particles
has the form
p
h_ A = r_ AC mv C +
X
i=1
rAi f ei ;
(4.17)
where rAC is the position vector of the mass centre of the system of particles with
respect to point A.
When point A coincides with the fixed origin O (r_ AC = r_ C = v C , v C mv C =
0 ) or point A coincides with the centre of mass C, equation (4.17) reduces to
h_ A =
p
X
i=1
rAi f ei :
(4.18)
p
X
h_ C =
or
i=1
rCi f ie ;
(4.19)
h_ C = lC ;
where
lC =
p
X
i=1
(4.20)
rCi f i e
(4.21)
is the resultant moment of all external forces acting on the system of particles about
the mass centre C. It should be mentioned that dynamics of particle is uniquely described by the Newtons equation (4.3). The introduction of an angular momentum
(Eq. 4.13) does not bring any new information into account. It has been introduced
here as a pre-stage to dynamics of rigid body, where the consideration of angular
momentum leads to the essential Eulers equation.
Dynamics of rigid body
Prior to deriving governing equation of rigid body dynamics, in the next section
some basic kinematical relations will be repeated.
BASIC KINEMATICAL RELATIONS
In Figure 4.11 the following notation is used:
!
rA
(x; y; z )
(x ; y ; z )
0
CHAPTER 4. DYNAMICS
209
z
y
rC
rC
rA v A
z
O
r
x
r_
=!r :
0
(4.22)
r = rA + r
v = vA + r_
= vA + ! r :
0
(4.24)
The orientation of a body in the inertial coordinate system can be determined via
the Euler angles '; #; that specify the orientation of the fixed body system k with
respect to the inertial system k . Other possibilities to describe the orientation of
the body-fixed system include Bryant (cardan) angles, Euler parameters, Rodriguez
parameters etc. [3], [16].
The relation between ! and the derivatives of the Euler angles x_ R = ['_ #_ _ ]T
can be expressed in matrix form
0
! = HRx_ R :
L INEAR MOMENTUM
The mass of a body is given by
m=
Z
m
dm ;
(4.25)
(4.26)
CHAPTER 4. DYNAMICS
210
and the position vector of the body centre of mass C in the system k is
1
rC = m r dm :
(4.27)
rAC = rC
dm :
(4.28)
From (4.23), the linear momentum of the body can be written in the form
(v A + ! r ) dm = v A
m
or
dm + !
(v A + ! r ) dm = v A m + ! mr C
0
dm ;
(4.29)
= m(v A + ! r C )
= mv C
= mr_ C :
0
(4.30)
A NGULAR MOMENTUM
The absolute angular momentum of a body with respect to O (origin of the
inertial coordinate system) is determined by [3]
hO = r r_ dm
(4.31)
or after introducing
hO =
r = rA + r
= r A (v A + ! r C )m + r C
The term
(4.32)
(rA + r ) (v A + ! r ) dm
0
vAm + r (! r ) dm :
0
(4.33)
= IA ! ;
(4.34)
CHAPTER 4. DYNAMICS
211
IA =
E is the unit vector and r r
0
(r 0 2 E
r r ) dm ;
0
(4.35)
hO = r C v C m + I C ! :
(4.36)
d
(mv C ) = f ;
dt
(4.37)
where:
ma C = f :
(4.38)
The Eulers equation determines the dynamics of the bodys rotational motion
h_ O = lO ;
(4.39)
where:
hO
h_ O = vC v Cm + rC maC + I C !_ + ! I C ! ;
and since v C v C m = 0 ,
h_ O = rC maC + I C !_ + ! I C !:
(4.40)
(4.41)
I C !_ + ! I C ! = lO rC f = lC ;
(4.42)
CHAPTER 4. DYNAMICS
or in short
212
I C !_ + ! I C ! = lC:
(4.43)
maC = f ;
(4.44)
~ IC ! = lC :
IC + !
(4.45)
= !_
(4.46)
CHAPTER 4. DYNAMICS
213
Governing equation
(holonomic system)
p...bodies, q...constraints, f...DOF
Minimal form
f equations of motion,
f generalised coordinates y
q kinematical constraint
equations
CHAPTER 4. DYNAMICS
214
Unfortunately, in most of the cases (except for some linear models and the
simpler tasks of small dimensionality), it is not possible to find the closed
form solution and a numerical procedure has to be applied to obtain the solution of the model (the numerical procedure may be launched immediately
after the mathematical model is established, or some symbolic manipulations
and simplifications can be performed prior to the numerical calculations).
Symbolic approach
The symbolic mathematical operations consist of the manipulations with the
mathematical entities without assigning their numerical values.
If a computational tool gives possibilities for the symbolic calculations, sometimes a more efficient computational procedure can be achieved by simplifying the established mathematical model before an iterative numerical procedure is launched. Once the mathematical model using symbolic formalisms
is established, it can be used for repeated numerical calculations e.g. in numerical integration schemes [11].
However, the extent to which the efficiency may be improved using the symbolic tools is dependent on the task (mathematical model) at hand. The symbolic manipulations are often computationally more costly than the numerical
procedures and for some types of problems very efficient numerical procedures can be utilized e.g. the sparse matrices techniques).
Although symbolic procedures are much in use in todays computation, the
design and implementation of symbolic algorithms are the topics of the ongoing research activities.
Numerical approach
By using this approach, the numerical values are assigned to the symbolic
items as soon as the mathematical model is established and the whole computational process deals with the numerical values.
The majority of the computational packages on todays market are numerically oriented, especially packages and tools that are designed for a general
use [4].
CHAPTER 4. DYNAMICS
215
1
l1
f 1pi
f p1i
fp
v1
f 12i
vp
O
f 21i
x
l2
f2
2
lp
p
v2
f p2i
e
f 2pi
CHAPTER 4. DYNAMICS
216
is given by the vector of the body mass centre e.g. expressed in the inertial Cartesian
coordinate system (other coordinate systems can also be chosen)
xTi = [xi yi zi ]T ;
(4.47)
xRi = ['i #i i ]T :
(4.48)
(Note: In Eq. (4.46), (4.47) and subsequent text the index C is omitted since
centre of mass will always be reffered to describe a position of the body) By grouping equations (4.46) and (4.47) together, the body absolute position vector can be
introduced in the form
xi = [xTTi xTRi ]T :
(4.49)
Newton-Euler equations of i-th body
The Newton-Euler equations are basic equations of rigid body dynamics, see
Chapter 4.1. The Newton equation determines dynamics of the bodys translational
motion, while the bodys rotational motion is determined by the Euler equation [16].
The Newton equation is given by
mi ai = f i ;
(4.50)
mi x Ti = fi :
(4.51)
I i!_ i + !i I i!i = li ;
(4.52)
(4.53)
i = !_ i :
(4.54)
The relation between the bodys angular velocity !i and the time derivatives of Euler
angles xRi = ['i #i i ]T , by means of which the absolute orientation of the body
in the inertial coordinate system is specified, can be given in the form [3]
!i = HRix_ Ri ;
(4.55)
and the differentiation with respect to time using the chain rule yields
i = HRix Ri + i :
(4.56)
CHAPTER 4. DYNAMICS
217
In equation (4.55), all terms in which the second derivative appear linearly are ex Ri and all others are grouped in
i [18]. By taking into
pressed in the product HRi x
account equations (4.55), equation (4.52) can be written in the form
Ii HRix Ri + Ii i + !~i Ii !i = li :
(4.57)
Futhermore, the equations (4.50) and (4.56) can be grouped together to form the
Newton-Euler equations of the i-th body in the matrix form:
mi E 0
0 Ii
E 0
0 HRi
or in short
x Ti
x Ri
+
Iii + !~i Ii !i
=
fi
li
Mi Hi x i + qvi = qai :
(4.58)
(4.59)
2
m1 E 0
66 0 I1
66 0 0
M=6
66 0 0
64
0 0
0
2
E
66 0
66 0
H=6
66 0
64
0
0
0
0
0
0
0
I2
0
0
0
HR1
0
0
0
0
0
0
m2 E
0
0 0
0 0
E 0
0 HR2
0
0
0
0
3
77
77
77 ;
77
0 5
0
0
0
0
0
0
0
0
dim[M] = 6p 6p ;
(4.63)
(4.64)
mp E
0 Ip
0
0
0
0
E
0
3
77
77
77 ;
77
0 5
0
0
0
0
HRp
dim[H] = 6p 6p ;
(4.65)
CHAPTER 4. DYNAMICS
218
2
3
0
66 I11 + !~1 I1!1 77
66
77
0
v
qv = 6
66 I22 + !~2 I2!2 777 ; dim[q ] = 6p 1 ;
64
75
0
Ipp + !~p Ip! p
2 3
f1
66 l1 77
66 f2 77
a
a
q =6
66 l2 777 ; dim[q ] = 6p 1 :
64 75
fp
(4.66)
(4.67)
lp
In the case of the multibody system of free bodies, the Newton-Euler equations
(4.59) are the equations of motion of the system.
Equation (4.59) represents 6p dimensional ODE system. It can be integrated in
time for specified initial conditions x0 ; x_ 0 to determine the systems motion (vari ).
ables x ; x_ ; x
Some computational issues
The inertia matrix MH in Eq.(4.59) has non-symetrical properties which may decrese significantly the efficiency of computation.
In the framework of integration of governing equations, a non-symetric inertia
matrix prevents use of very efficient numeric procedures which require its symetrical properties (e.g. Cholesky method).
Therefore, to improve the efficiency of the procedure, it may be helpful to symmetrise the matrix MH. This can be done by premultiplying Eq.(4.59) by HT ,
which reads
HT MHx + HT qv = HT qa :
(4.68)
In many applications the integration of (4.65) requires a less computer-power than
integration of (4.59).
Constrained multibody systems
M ATHEMATICAL MODEL IN FULL DESCRIPTOR FORM
The constrained multibody system (Figure 4.13) is a mechanical system of rigid
bodies whose motion is constrained by kinematical constraints. If the system consists of p bodies whose motion is constrained by q kinematical constraints, the sys-
CHAPTER 4. DYNAMICS
tem possess f = 6p
219
q DOF [21].
ek
f1
e1
e1
f1
l2
eh
l2
f i = f ie + f ii ;
(4.69)
where the resultant force of the external applied forces, reduced to the centre of
mass Ci , is
fi =
e
k
X
j =1
f ie
(4.70)
and the resultant force of the internal applied forces (internal springs, dampers,
etc.), reduced to the centre of mass, is given by the sum of the internal applied
forces between the bodies i and j
f ii =
p
X
j =1
f iji :
(4.71)
CHAPTER 4. DYNAMICS
220
e
l1
f1
i
f 1p
c
l1
i
f p1
e
fp
f 12
f 21
l2
lp
c
c
c
f p0
f 2p
f2
lp
c
i
e
i
f p2
c
l2
The reduction of forces with different application points to one specific point implies the formulation of an equivalent couple of force and torque acting at this specific point.
The resultant torque about the centre of mass Ci of the applied forces and
torques that act on the i-th body (Figure 4.14) is
li =
h
X
j =1
li
ej
p
X
j =1
liji + lred
(4.72)
where liji is an internal applied torque (internal torsional spring, for example)
that acts between bodies i and j and lredi is the torque due to the reduction of the
forces f ie and f ii to the centre of mass Ci .
Constraint forces
The resultant constraint force that acts on the i-th body, reduced to the centre of
mass Ci (Figure 4.14), is given by the equation
fi =
c
p
X
j =0
f ijc ;
(4.73)
where f ijc is a constraint force that acts between bodies i and j (i; j = 0:::p, the
index 0 stands for the body of external world).
CHAPTER 4. DYNAMICS
221
li =
c
p
X
j =0
c
lijc + lred
(4.74)
where lijc is a constraint torque that acts between bodies i and j or a body and the
c
external world (i; j = 0:::p) and lred
is a reduction torque of the forces f ijc .
i
Newton-Euler equations of i-th body
As previously mentioned, the Newton equation determines the dynamics of a
rigid bodys translational motion, while the bodys rotational motion is determined
by the Euler equation. When a body is kinematically constrained, the constraint
forces and torques also influence the motion of a body and have to be considered in
the framework of the Newton-Euler equations together with the applied forces and
torques [19].
The Newton equation is given by
mi ai = f i + f ci ;
(4.75)
mi x Ti = fi + fic :
(4.76)
(4.77)
(4.78)
(4.79)
After introduction of the body absolute position vector (4.60), equations (4.73) and
(4.76) can be grouped together to form the Newton-Euler equations of the i-th body
mi E 0
0 Ii
E 0
0 HRi
x Ti
x Ri
+
Ii i + !~i Ii !i
=
fi
li
fic
lci
(4.80)
CHAPTER 4. DYNAMICS
or in short
222
(4.81)
@g
@g
x_ +
=0;
@x
@t
or in the short form
Q x_ =
where the matrix Q is defined as
Q(x; t) =
@g
;
@t
@g
; dim[Q] = q 6p :
@x
(4.86)
(4.87)
(4.88)
CHAPTER 4. DYNAMICS
223
(4.90)
qc = QT :
(4.91)
In the context of the equation (4.86) it can be stated here, without going to the
details, that the directions of the system constraint forces qc are expressed by the
columns of the transposed matrix QT while the magnitudes of the constraint forces
are given by the Lagrange multipliers vector .
The matrix Q, defined in (4.83), by means of which the system constrained
forces qc are expressed in (4.86), is usually called system constraint matrix [4].
The system constraint matrix is the one of the most important matrices in the domain of dynamics of constrained multibody systems. By checking its rank it can be
examined if the system is properly constrained i.e. if all constraints imposed on the
system are independent or some of them are superflous. The number of independent
constraints is equal to the rank of Q (see Chapter 4.2.3).
System governing equations
After insertion of (4.86) into (4.79), the Newton-Euler equations of the constrained multibody system can be written in the form
MHx + qv = qa + QT :
(4.92)
Unlike the Newton-Euler equations of the free multibody system (4.14), the set of
the equation (4.87) can not be solved and integrated in time directly since it contains
q additional algebraic unknowns [16].
To make a set of the equations (4.87) complete and solvable, the kinematical
constraint equations have to be added to the mathematical model and considered
simultaneously with the Newton-Euler equations.
With this aim in view, the Newton-Euler equations (4.87) and q kinematical
constraint equation (4.80) are grouped together, forming the governing equations of
the constrained multibody systems
MHx + qv = qa + QT
CHAPTER 4. DYNAMICS
224
g(x; t) = 0 :
(4.94)
MHx + qv = qa + QT
Q x = c :
(4.96)
The set of equations (4.88) as well as the set (4.89) represent the governing
equations of the constrained multibody system expressed in the full descriptor form
[5].
As it was already explained, by integrating (4.88) or (4.89) for the specified
initial conditions x0 ; x_0 the systems motion as well as the systems constraint
forces can be determined.
Because of inherent numerical instability of DAE system presented by (4.87)
and (4.88), numerical time integration of these equations is a challenging task which
has to be treated very carefully [8].
Similarlly as it was the case with Eq. (4.59), the inertia matrix MH in Eq. (4.88)
is a non-symetric one which makes time integration of (4.89) less efficient. As it
was explained, premultiplication of (4.89) by HT symmetrise the inertia matrix
MH and brings (4.87) in the form
HT MHx + HT qv = HT qa + HT QT ;
(4.97)
Basic equations: Newtons and Eulers equations, the constraint forces are
included
Absolute coordinates, 6 coordinates per body: e.g. cartesian coordinates of
the bodys mass centers and bodys Euler angles (or other parameters)
Easy-to-obtain mathematical model, straightforward universal approach [16]
CHAPTER 4. DYNAMICS
225
If it is required, the governing equations expressed in the full descriptor form can be
reduced to the mathematical model in the minimal form (that represents the equations of motion of the constrained multibody system) which will be described in the
sequel.
M ATHEMATICAL
CHAPTER 4. DYNAMICS
226
In order to shape the mathematical model in the minimal form i.e. to establish
equations of motion of the constrained multibody system, a minimal set of coordinates y, dim[y] = f (f = number of DOF), have to be chosen. By means of y, the
kinematical configuration of the system is uniquely described [18].
The relation between the system full descriptor absolute coordinates
x = f (y; t) ;
(4.99)
(an implicit formulation '(x; x_ ; t) = 0 is also possible). The corresponding equation for velocities takes the form
x_ =
@f
@f
@f
y_ +
= Jy_ +
;
@y
@t
@t
(4.100)
x = Jy + a ;
(4.101)
J=
@f
; dim[J] = 6p f :
@y
(4.102)
The matrix J is usually called Jacobian matrix. It is not unique but depends on the
chosen set of coordinates by means of which the system kinematical configuration
is described.
It can be shown [18] that the Jacobian matrix J has a property of being orthogonalcomplementary matrix to the system constraint matrix Q. The relation between
these two matrices can be expressed by
QJ = JT QT = 0 :
(4.103)
The orthogonality between the matrices J and Q stems from the fact that these
matrices span different subspaces which are mutually orthogonal. Namely, the
columns of the transposed constraint matrix QT as well as the Jacobian matrix J
are vectors which form the basis of the two vectorial subspaces: the q -dimensional
subspace of the system constraint forces is spanned by Q and the f -dimensional
subspace of the system velocities is spanned by J.
The orthogonality given by (4.93) holds only for the ideal mechanical systems
and can be briefly explained by the fact that the system velocities are always orthogonal to the system constraints. In analytical mechanics this is related to Jourdains
CHAPTER 4. DYNAMICS
227
principle which is based on virtual power. Other principles are dAlemberts principle formulated by Lagrange on the base of virtual work and Gausss principle based
on minimal constraints. The details of the introduced vectorial subspaces as well as
the strict mathematical proof are not given here and interested reader is refered to
the literature [18], [4].
By introducing equations (4.91), (4.92) and (4.93) into equation (4.86), the
Newton-Euler equations of constrained systems can be expressed via the minimal
set of coordinates y [18]
MH (Jy + a) + qv = qa + Q T
0
(4.104)
where matrices that are expressed by the new set of coordinates are denoted by.
Furthermore, if the equation (4.96) is multiplied from the left side by the transposed Jacobian matrix JT , two important effects will be achieved:
JT MH Jy + JT (MH a + qv ) = JT qa ;
(4.105)
(4.106)
Mgen = JT MH J ; dim[Mgen ] = f f ;
0
(4.107)
(4.108)
qagen = JT qa ; dim[qagen ] = f 1 :
0
(4.109)
The set of equations (4.98) represents the equations of motion of the constrained
multibody system (mathematical model in minimal form, f -dimensional ODE system). It can be integrated in time for given initial conditions y0 , y_ 0 to obtain the
systems motion.
Since the term of constraint forces QT vanishes from the governing equations
due to the left multiplication by JT , it is obvious that this term must not be formulated, if equations of motion are to be derived. Therfore, the equations of motion
CHAPTER 4. DYNAMICS
228
JT HT MH Jy + JT HT (MH a + qv ) = JT HT qa :
0
(4.110)
T=
p
X
1
i=1
1
mi viT vi + !Ti Ii !i
2
2
(4.111)
has to be determined. The expression in the bracket expresses the kinetic energy of
the i-th body in the system [3].
By introduction of absolute position vector of the i-th body
xi = [xTTi xTRi ]T
(4.112)
and vector ti of the applied forces and torques reduced to the i-th body mass centre
ti = [fiT lTi ]T ;
(4.113)
d @T
dt @ y_
@T
= qagen :
@y
(4.114)
In equation (4.104), a vector of the generalized applied forces qagen has the form
(4.115)
CHAPTER 4. DYNAMICS
229
qi =
p
X
i=1
tTi
@ xi
@ yi
(4.116)
Governing equation
(holonomic system)
p...bodies, q...constraints, f...DOF
+ q* constraints
x + fy
x + f + y
y
.
f
f
y
y
f
f
y
y
Minimal form
f equations of motion,
f generalised coordinates y
Minimal form
f * equations of motion,
f *generalised coordinates y*
t
CHAPTER 4. DYNAMICS
230
Appropriate for the forward dynamical problem. Appropriate for the inverse
dynamical problem, if only applied forces have to be determined.
Forward dynamics
The model allows for determination of the motion of the system.
Since model is based on the generalised coordinates given in minimal
form, the additional calculations are needed in order to determine position and orientation variables of the each body in the system (the additional calculation is based on the values of the generalised coordinates
and the kinematical constraints equations).
Inverse dynamics
If only the applied forces have to be determined, an utilization of the
minimal form model is plausible since computational procedure will not
be unnecessarily burdened by the superflous coordinates and constraint
forces.
Since constraint forces are not contained in the equations of motion, the
equations of the kinematical constraints have to be used, if these forces
have to be determined.
Computer algorithms
Lagrangian equations of second kind: not so appropriate
During generation of the equation of motion, differentiation of the system energy terms is needed.
This procedure is not so appropriate as to be efficiently incorporated
to the computational procedure (this holds specially for the large-scale
systems).
CHAPTER 4. DYNAMICS
231