Object-Oriented Modelling

of Hybrid Technical Systems Introduction to Bond Graphs

Jonas Eborn Invented by Henry Paynter at MIT, 1960.

Motivation:
Department of Automatic Control
Lund Institute of Technology  Energy based modelling (Lagrange)
Lund, Sweden  Power continuous, fullls rst law
Email: jonas@control.lth.se  Analogies, mechanics - electrics
 Multidisciplinary, unied approach for
modelling of mixed systems
Lecture 5 { Bond Graph Modelling  Non-causal, imposed by boundaries
 Graphical method, bonds describing
 Introduction to Bond Graphs energy interchange between systems
 Bond Graph Elements Conservation
Causality
Sources and sinks Storage
 (junction structure)

 Modelling Example
 Analysis of Models Dissipation

 Advanced Bond Graphs

Introduction to Bond Graphs

Power Conjugate Variables
Introduction to Bond Graphs
Energy interchange is a two way information Power Conjugate Variables
ow. It is described by one eort and one ow
variable.
Power conjugate, their product is power. In each engineering domain there is one pair of
Example: An electrical connection is described power conjugate variables.
by the voltage between the terminals and the Some common examples are:
current through them. The energy owing
between the two systems is
Z
Domain e, eort f, ow
E= u
i dt Electric u, voltage i, current
Mechanics F, force v, velocity
u
Rotation M, torque !, angular vel.
Voltage
source
Passive
system Hydraulic p, pressure q, volume ow
i Thermal T, temperature S_ , entropy ow
i
V u R

1
 Bond Graph Elements
Energy Storing Elements
In dynamical systems energy can be stored,
Bond Graph Elements there are two kinds of energy storage:
Basic Elements  Capacitor/Compliance ( ow storage)

 Sources/Sinks Se C
Active elements, that supplies/removes Z
eort or ow. e(t) = G( f (t)dt)
Z
 Resistance R u(t) = C1 i(t)dt
Passive dissipative elements. Removes i(t) = C du=dt
energy and relates eort to ow.
Constitutive relation:  Inductance/Inertia (eort storage)
e(t) = h(f (t)) I
f (t) = h?1 (e(t))
Z
u(t) = R i(t) f (t) = G( e(t)dt)
Z
1
v(t) = m F (t)dt
F (t) = m dv=dt

Bond Graph Elements Bond Graph Elements

Junction elements Junction examples

Elements to build model junction structure. Common eort:

Requirements are: F
power continuous, symmetrical and ability to V
be used as building blocks.
C
Se 0
R

There are two unique three-port junctions: Common ow:

 0-junction Common eort, the ows sum
to zero. Like a parallel connection in an F
electrical network. V

 1-junction Common ow, the eorts sum I

to zero. Like a closed series connection in
an electrical network. Se 1 R
Alternative notation: p- and s-junctions
Mnemonic: 0 - p - e lower than 1 - s - f C

2
 Bond Graph Elements
Energy Transducers Bond Graph Elements
Modulated Elements
Relaxing the symmetry requirement, two Some elements can be modulated by an
unique two-port junctions can be derived external variable or signal.
describing energy owing from one physical
domain into another.
 A Transformer transforms eort in one  Sources MSe
domain into eort in another, e. g., force modulated by a signal generator or a
 pressure in a hydraulic piston. control signal
F = pA MTF
F p  Transducers
v TF q nonlinear transformations, modulated by
q = vA for example angle or position
 A Gyrator relates eort in one domain Example: AC voltage source in small circuit.
with ow in another, e. g., current  sin t
torque in an electrical motor.
M = km i
u MSe R
GY M
w
u = km ! i

Causality
Element causality

Causality is shown as a bar on the bond.

Causality A
e
B A
e
B
Purpose of Modelling is Insight f f
e e
A B A B
Causality is a fundamental concept in mod- f f
elling. It shows dependencies between ele- See it as a nail, eort is imposed the at end,
ments. ow at the point.
Causality is
 NOT a property of the bondgraph
 imposed by Sources
Sources have mandatory causality. Eort or
 easily analyzed by assignment procedure ow given.
Se
Sf
Resistances are indierent to causality.

3
 Causality Causality
Integral causality Causal constraints

Storage elements are said to have desirable Constitutive equations of junctions and
causality if it is integrating. transducers give constraints or propagation
Numerical routines are designed to integrate, rules for causality.
not dierentiate. 0
 Common eort,
 Eort causality for inertias
I one eort can be imposed
f 1R 1
)f =
I e dt
 Common ow,
 Flow causality for capacitors N-1 eorts imposed
e
C 1 R  Transformer, same causality in and out
)e=
C f dt e1 ! e2
f2 ! f 1 TF
 Gyrator, switched causality
e1 ! f2
e2 ! f1 GY

Causality Causality
Assigning procedure
Example: A simple electrical circuit.
1. Assign mandatory causality to sources.
uc
i
2. Propagate causality through the junction
structure, using the causal constraints on I
junctions, transformers and gyrators.
i uc
:

Se 1 0 C

3. Assign desirable causality for one energy

storage. Return to 2. and repeat until no R R
more storage elements. 1: Sources
(2): No propagation possible
3: Desirable causality, I + prop.
3: Choose another, C + prop.

4. Remaining R-elements can be given any

causality. Choose one and propagate as in
2. until all elements have causality.

4
 Modelling example
Modeling Example
Issues in Modeling Battery vacuum cleaner, electrical dynamics of
motor neglected.
The main issues when you want to make a
model of a system are:
 Competence Willems - MPUM
Is the model adequate for our purpose?
 Simplicity Occam's razor
Is it the simplest model possible?

How can you check this?

 Causality analysis
Are there any dependencies?
 Thought experiments
Does the model behave like the real
system?

Analysis of models
Causality con icts

Analysis of models Systems with con icting/derivative causality

seem to indicate high DAE index.
Example: Two connected tanks, no ow
resistance.
What information can you extract from a bond q
A
graph model? C : g
A a
 Causality analysis shows dependencies
Sf 0: p
 Index problems and Algebraic loops V1 V2
 State equations are derived directly from a p a
q2 C : g
causal bond graph.
 Signal ow diagrams and
dV1 = q ? q
 Transfer functions can also be derived dt 2

from a bond graph using Mason's rule. dV2 = q

(Relative degree and Zero dynamics) dt 2

 Eigenvalue estimates, accurate to 10-20% p1 = p2 =) VA1 = Va2 (*)

 Controllability/Observability? Equation (*) must be dierentiated once to

(Chapter 6.7, [4]) nd q2. This is an index 2 problem.

5
 Analysis of models
Algebraic loops
Analysis of models
Algebraic loops can be found by inspection of
signal loops in the bondgraph. (Chap 6.5, [4])
Signal loop Closed causal path Deriving system equations from a causal
bondgraph.
Causal path Follows bonds with same causal
orientation, stops at storage or source 1. Choose power state variables, outputs
I from independent energy storages.

2. Write the ODE for each state from the

Sf 0 1 C constitutive relation.

R R 3. Follow the causality backwards to express

the input in 2. as outputs from other
elements.
Loop can be broken by inserting storage
elements, cf. tearing.

Analysis of models Analysis of models

System equations example (simple circuit): Eigenvalue estimation, assumption: parts of

I the bondgraph are loosely coupled.
uL Find subsystems that resemble resonance
i uc circuits. (Chapter 6.4, [4])
:

ic
:

Se 1 0 C u 1
uR1 iR2 uR
uC
i
uL I
R
R R C
1:I: i; C: uc di ; u = Ri;
i = C dudtC ; uL = L dt R
di = u =L duc = i =C
uC = u ? uL ? uR = U ? LC ddtu2C ? RC dudtC
2
2:
dt L dt c
3 : uL = V ? uR1 ? uc =
= V ? R1 i ? uc
) (LCs2 + RCs + 1)UC (s) = Ur(s)
ic = i ? iR2 = i ? uc =R2 !02 = LC1
;  = R2 CL
di = 1 (V ? R i ? u )
) dt L 1 c 0 u
i
duc = 1 (i ? uc )
dt C R2 R I
C

6
 Advanced Bond Graphs Advanced Bond Graphs
Multiports Multiports

Most energy transduction is associated with Example: A bicycle pump with blocked outlet.
energy storage. This leads to general multi-
ports, elds. Examples:
 Electrical transformer with ux energy, II
p,T

 Ideal gas in closed container, C p T T0

C IR Se :T0
 Irreversible heat transfer, IR V S S0
T1 T2
IR S_1 S_2

Q_ = T1 S_1 = T2 S_2 S_1 = h(T1 ? T2 )=T1

Q_ = h(T1 ? T2 ) S_2 = h(T1 ? T2 )=T2
) S_2 ? S_1  0
This is a nonlinear R-eld. To certify second
law, eort causality is mandatory.

Advanced Bond Graphs Advanced Bond Graphs

Multi-bonds and robotics Multi-bonds and robotics

In two- and three-dimensional mechanics you Example: (Ch 8.1 [4], Ch 23.2 [7])
need to have bonds with several eorts/ ows, Extendable robot arm.
multibonds. y
Motivation for bond graph robotics: r x

 'Mechanics is hard for humans', Steve θ

Crandle, MIT.
 Lagrangian becomes very complex alge-
braically, 6dof robot ! 60000 terms θ
ω v
 Uncoupled dynamics are easy
:

Se:τ 1 IMT IF 1 II :M
generalized uncoupled
 Handle in simple coordinate frame
{ Mechanics in uncoupled coord
{ Actuators in generalized coord

7
 Conclusions
 Bond graphs can be used to describe
simple linear & nonlinear systems.
+ Displays causality & dependencies between
parts of the system.
+ Systematic method, using physical
analogies
+ Model subsystems independently
+ Simplies nonlinear mechanics?
{ Tricky to learn, best suited for electrome-
chanics.
{ Complex technical applications?
{ Not well suited for applications with multi-
layered descriptions, simultaneous balance
equations

1. References
[1] Henry Paynter. Analysis and Design of
Engineering Systems, MIT Press, 1961.
[2] Dean Karnopp and Ronald Rosenberg,
System Dynamics: A Unied Approach,
Wiley & sons, 1975.
[3] P.E. Wellstead Introduction to Physical
System Modelling, Academic Press, 1979.
[4] Jean U. Thoma, Simulation by Bondgraphs,
Springer-Verlag, 1990.
[5] Francois Cellier, Continuous System
Modeling, Springer-Verlag, 1991.
[6] L. Ljung and T. Glad, Modellbygge och
simulering, Studentlitteratur, 1991.
[7] N. Hogan and P. Breedveld, Integrated
Modeling of Physical System Dynamics,
Internal report, Univ. Twente, 1995.
[8] Peter Gawthrop and Lorcan Smith,
Metamodelling: Bond graphs and dynamic
systems, Prentice Hall, 1995.