You are on page 1of 46

Control Engineering

Laboratory

BOND GRAPHS -
Physical systems modeling 1:
Fundamental concepts
Peter Breedveld
Cornelis J. Drebbel Institute for Mechatronics
and Control Engineering Laboratory, Electrical Engineering Department
University of Twente, Netherlands
p.c.breedveld@el.utwente.nl

7/11/2003 Summerschool Bertinoro 1


7-11 July 2003

Aim
• System theoretic approach to physical system dynamics
based on
– classification of phenomena in terms of energy
– fundamental principles of thermodynamics
• shown how, as a result:
– variables and relations describing physical systems may be
classified
– models may be organized as (‘port-
(‘port-based approach’)
• multiport elements
• interconnected in an interconnection structure corresponding to a
generalized network
– multiport elements describe basic behaviors with respect to energy
energy
and entropy
• special attention:
– role of analogies
– analogue behavior
Summerschool Bertinoro, 7-11 July 2003 (2) © Peter Breedveld

1
Introduction

• introduction & some philosophy:


• Why physical systems modeling?
• What is physical systems modeling?
• Context of explanation (focus of engineering
physics) versus justification (focus of math)
• elements and components, theory building
• synthesis between classical approaches
• choice of variables:
– mechanical versus thermodynamic
framework
Summerschool Bertinoro, 7-11 July 2003 (3) © Peter Breedveld

Three-world meta-model
• Real world (assuming existence of ‘objective’
environment’)
• Conceptual world (in our brain)
• ‘Paper’ world, including electronically stored
data

– Only in/via the paper world:


• Communication
• Support
• Systematization
– exchangeable abstractions/concepts >>>
– importance of symbols & notation (representation)
Summerschool Bertinoro, 7-11 July 2003 (4) © Peter Breedveld

2
Are physical concepts ‘real’ ?

• For example:
– energy, time, momentum, causality,…

• In a context of justification: NO!


• In a context of discovery/explanation:
YES, but rather ‘really useful’ than
just ‘real’

Summerschool Bertinoro, 7-11 July 2003 (5) © Peter Breedveld

The useless ‘quest for truth’


• A model is necessarily incomplete:
‘all models are wrong’, but ‘truth’ is not
what counts
• The issue is whether a model is
competent (to solve a problem in a
given problem context in the most
generic sense of the word)

Summerschool Bertinoro, 7-11 July 2003 (6) © Peter Breedveld

3
What is a model?
Some sort of abstraction (in the ‘paper’
paper’ world) that enables
• insight in the real world counterpart
• communication about the real world counterpart
• observation of the real world counterpart
• troubleshooting of the real world counterpart
• design of new aspects related to the real world counterpart
• modifications of the real world counterpart
• ‘explanation’
explanation’ of functionality of the real world counterpart
• measurement of the real world counterpart
but, most importantly, that is :
• competent to solve a given problem and make decisions
related to the real world

Summerschool Bertinoro, 7-11 July 2003 (7) © Peter Breedveld

Modeling
• Given a specific problem context, the decision
process to obtain a competent model to solve this
problem
• Approaches between two extremes:
– a priori knowledge
– ‘black box’
box’ (also an axiomatic concept with a priori
assumptions!!!)
• concept ‘input’
input’ implicitly contains model of imposing some
action with negligible back effect (‘(‘high input impedance’
impedance’)
• concept ‘output’
output’ implicitly contains model of measurement
with negligible effect on the system being observed (‘ (‘low
output impedance’
impedance’)
• after a competent input-
input-output relation is found, it is not
open for modifications or physical interpretation

Summerschool Bertinoro, 7-11 July 2003 (8) © Peter Breedveld

4
Model representation
• model representation:
representation:
– symbols used to represent
• the concepts being used
• (the structure of) their relations (interconnection)
• model manipulation (as opposed to modeling!) :
– transformation to different representations (including
the ‘solution’
solution’) to
• increase insight
• draw conclusions, etc.

Summerschool Bertinoro, 7-11 July 2003 (9) © Peter Breedveld

Role of representations
x1
a
Sequential: A T
x2
b
A T
x3
c
A T
x4
d
AT
x5
e
AT

t
Simultaneous:
T

Process time
versus A T x

processing time
x1 x2 x3
x5
x4

Summerschool Bertinoro, 7-11 July 2003 (10) © Peter Breedveld

5
Physical systems modeling
• all concepts used in the model are, or have a direct relation
to, physically relevant concepts (use of a priori knowledge!)
• physical relations are maintained as much as possible
• herein constrained to (for sake of ‘simplicity’
simplicity’):
– deterministic mathematical models of macroscopic systems
that
• obey basic principles of macroscopic physics:
– energy conservation
– positive entropy production
• describe the behavior in time of the common physical properties:
– mechanical (incl. hydraulic and pneumatic)
– electrical
– magnetic
– chemical
– material

Summerschool Bertinoro, 7-11 July 2003 (11) © Peter Breedveld

time & uncertainty


• time:
– derived measure for ‘regularity’ based on
counting ‘ticks’ of a ‘time-
‘time-base’ based on
repetitive behavior requiring state and change (in
order to be able to count)
– within the smallest unit used: necessarily
uncertainty
• cf. Heisenberg u.r. for displacement (= elastic state,
kinetic state of change) and momentum (= elastic state
of change, kinetic state)
– dialectic concepts!

Summerschool Bertinoro, 7-11 July 2003 (12) © Peter Breedveld

6
concepts of
‘state’ & ‘change’
• dialectic fundamental concepts
• basis for any dynamic model (more
than time!)
• within a context of discovery a shift to
space-time is understandable and
useful, like the shift from position-
momentum in Hamiltonian mechanics to
position-velocity in Lagrangian
mechanics
Summerschool Bertinoro, 7-11 July 2003 (13) © Peter Breedveld

Position: a ‘special’ state

Position
• has a dual nature:
– Energy state (related to a
conservation or symmetry principle
like all other states)
– Configuration state
• does not transform like a tensor
end of introduction

Summerschool Bertinoro, 7-11 July 2003 (14) © Peter Breedveld

7
Contents of the sequel
• Modeling pitfalls
• Port-based modeling
• Basic Concepts (ports, bonds)
• Dynamic conjugation (effort, flow)
• Multidomain modeling and the role of
energy
• (Computational) Causality

Summerschool Bertinoro, 7-11 July 2003 (15) © Peter Breedveld

Modeling pitfalls

• ‘Every model is wrong’


• Model depends on problem context
• Competent models
• Analogies are not identities
• Avoid implicit assumptions
• Avoid model extrapolation
• Confusion of components with elements

Summerschool Bertinoro, 7-11 July 2003 (16) © Peter Breedveld

8
Physical components versus
conceptual elements

physical component:
piece of rubber hose

dominant behavior:
•when falling: ideal mass
•when pulling load: ideal spring
•for vibration isolation: ideal resistor
•etc.

Summerschool Bertinoro, 7-11 July 2003 (17) © Peter Breedveld

Parasitic elements

• Next to dominant behavior:

Summerschool Bertinoro, 7-11 July 2003 (18) © Peter Breedveld

9
In engineering models:

• Avoid implicit assumptions, e.g. about


– problem context
– reference
– orientation
– coordinates
– metric
– ‘negligible’ phenomena, etc.

• Avoid model extrapolation


– danger of ignoring earlier assumptions

• Focus at competence, not ‘truth’


Summerschool Bertinoro, 7-11 July 2003 (19) © Peter Breedveld

Port-based modeling
• multidomain approach:
– ‘mechatronics’ (and beyond)
• multiple view approach: other graphical
representations: iconic diagrams, (linear
graphs), block diagrams, bond graphs, etc. and
equations
• domain independent notation using ports:
– bond graphs (& some other benefits…)
• port-based approach:
– underlying structure of 20-
20-sim,
sim, ideal tool for
demonstration
• what are ports and what are bond graphs?
Summerschool Bertinoro, 7-11 July 2003 (20) © Peter Breedveld

10
Intuitive introduction of the
‘port’ concept

J P

ideal current ideal ideal


source motor inertia potentiometer
ideal
transmission
Dominant behavior
(not necessarily competent in each context)

Summerschool Bertinoro, 7-11 July 2003 (21) © Peter Breedveld

Simple model

Summerschool Bertinoro, 7-11 July 2003 (22) © Peter Breedveld

11
Addition of relevant parasitic
behavior
Depending on
problem
context:

Polymorphic
modeling

Summerschool Bertinoro, 7-11 July 2003 (23) © Peter Breedveld

What are ports?

∆u2 ∆u3
∆T2 ∆T3
∆T1
i i J

∆u1 i i MECH

∆u4 ω ω
ω
P =∆T⋅ω
P =∆u⋅ i electrical ports (power)
(power)
mechanical ports

Summerschool Bertinoro, 7-11 July 2003 (24) © Peter Breedveld

12
Comparison with familiar
model views
Iconic diagrams
(‘ideal physical models’):
R L
i v R

Usource C F m
Fext
K=1/C

electrical mechanical

Summerschool Bertinoro, 7-11 July 2003 (25) © Peter Breedveld

Iconic diagram symbols

Port-based, but
Port-based,
domain dependent

Summerschool Bertinoro, 7-11 July 2003 (26) © Peter Breedveld

13
Iconic diagram symbols

C-type storage
I -type storage
(M)R (dissipation,
irreversible transduction)
Se (effort source)
Sf (flow source)
TF (transformer)

Summerschool Bertinoro, 7-11 July 2003 (27) © Peter Breedveld

Iconic diagrams

• Icons do not (always) represent


physical structure:
20-sim
icon:
1
i

TF (transformer)

Summerschool Bertinoro, 7-11 July 2003 (28) © Peter Breedveld

14
Common block diagram
models
R L
i v R

Usource C F m
Fext
K=1/C

Summerschool Bertinoro, 7-11 July 2003 (29) © Peter Breedveld

Common block diagram


models

Note: not all signals


are physically
meaningful variables

Summerschool Bertinoro, 7-11 July 2003 (30) © Peter Breedveld

15
Ports in iconic diagrams

R L
i v R

Usource C F m
Fext
K=1/C

Summerschool Bertinoro, 7-11 July 2003 (31) © Peter Breedveld

Ports in iconic diagrams

R L m
F
i R Fext

Usource C v

K=1/C

Summerschool Bertinoro, 7-11 July 2003 (32) © Peter Breedveld

16
Ports in block diagrams

Note: each signal is a physically


meaningful variable

Summerschool Bertinoro, 7-11 July 2003 (33) © Peter Breedveld

Structure as multiport

Summerschool Bertinoro, 7-11 July 2003 (34) © Peter Breedveld

17
Different forms

Summerschool Bertinoro, 7-11 July 2003 (35) © Peter Breedveld

Different forms

Summerschool Bertinoro, 7-11 July 2003 (36) © Peter Breedveld

18
Compact & domain independent
Domain dependent Coil Mass

Source

Source Spring

Capacitor

Resistor Damper

Summerschool Bertinoro, 7-11 July 2003 (37) © Peter Breedveld

Compact & domain independent

Domain independent I
I

Se 1 C Se 1 C

Structure explicitly R R
represented
as multiport:
JUNCTION!

Summerschool Bertinoro, 7-11 July 2003 (38) © Peter Breedveld

19
Ports in bond graph view

Se 1 C

Summerschool Bertinoro, 7-11 July 2003 (39) © Peter Breedveld

Bond graphs

Inventor (MIT, 1959): Prof. Henry M. Paynter (1923-2002)

Summerschool Bertinoro, 7-11 July 2003 (40) © Peter Breedveld

20
Bonds
Storage of kinetic energy
Storage of elastic energy

Compact notation:
C I
Energy exchange
= power

•Terminology and notation induced by chemical bonds:

Summerschool Bertinoro, 7-11 July 2003 (41) © Peter Breedveld

Ports and power bonds

• (power) bond connecting two elements


via (power) ports (Harold Wheeler,
1949)

Summerschool Bertinoro, 7-11 July 2003 (42) © Peter Breedveld

21
Dynamic conjugation

• Between signals of bilateral signal


flow of relation:
– rate of change: ‘flow’
(zero in equilibrium)
• e.g. molar rate during diffusion
– equilibrium determining variable: ‘effort’
• e.g. concentration

Summerschool Bertinoro, 7-11 July 2003 (43) © Peter Breedveld

Power conjugation

= special case of dynamic conjugation :


– ‘effort’ and ‘flow’ relate to power
– functional relation is commonly a product
– sum in case of scattering variables

dE ∂E dqi ∂E dp j
P= =∑ +∑ = ∑ ei fi + ∑ e j f j
dt i ∂qi dt j ∂p j dt i j

∂E dp j
ei = ej = 'effort'
∂qi dt
dqi ∂E
fi = fj = 'flow'
dt ∂p j

Summerschool Bertinoro, 7-11 July 2003 (44) © Peter Breedveld

22
Conjugate variables &
corresponding states

Summerschool Bertinoro, 7-11 July 2003 (45) © Peter Breedveld

mass-spring system

Summerschool Bertinoro, 7-11 July 2003 (46) © Peter Breedveld

23
Basic dynamic behaviors
& mnemonic codes
• Storage (reversible)
– C, I (q-
(q-type and p- p-type storage)
• Irreversible transformation (‘dissipation’):
– (M)R(S)
• Distribution
– 0-junction, 1-junction
• Supply and demand:
demand:
– (M)Se, (M)
(M)Sf
• Reversible transformation
– (M)TF, (M)GY
• >> 9 basic elements

Summerschool Bertinoro, 7-11 July 2003 (47) © Peter Breedveld

Orientation conventions
• 1-ports: into most elements (R, C, I), except
sources (Se, Sf)
Sf)
• 2-port transducers: 1 in, 1 out
• junction structure elements: arbitrary
• multiport generalizations: same as simple form
• MOST IMPORTANT: obeying grammar rules
minimizes sign errors!

Summerschool Bertinoro, 7-11 July 2003 (48) © Peter Breedveld

24
Basic one-port elements

Summerschool Bertinoro, 7-11 July 2003 (50) © Peter Breedveld

Basic two- and multiport


elements

Summerschool Bertinoro, 7-11 July 2003 (51) © Peter Breedveld

25
Constitutive relations not
necessarily linear!
e.g. zener diode:
dominant behavior:
irreversible
transduction
(resistor)

with nonlinear
constitutive relation

Summerschool Bertinoro, 7-11 July 2003 (52) © Peter Breedveld

Other form of non-linearity:


Modulation
e.g. modulation of a transducer: iconic diagram (IPM)

capstan
MR
Summerschool Bertinoro, 7-11 July 2003 (53) © Peter Breedveld

26
Bilateral signal flow
(computational causality)

2 possibilities:

Summerschool Bertinoro, 7-11 July 2003 (54) © Peter Breedveld

Notation

Causal stroke:

Summerschool Bertinoro, 7-11 July 2003 (55) © Peter Breedveld

27
Back to the example

Se 1 C

Summerschool Bertinoro, 7-11 July 2003 (56) © Peter Breedveld

Back to the example

Se 1 C

20-sim demo

Summerschool Bertinoro, 7-11 July 2003 (57) © Peter Breedveld

28
Causal port properties
• Fixed causality

• Preferred (integral) causality

Summerschool Bertinoro, 7-11 July 2003 (58) © Peter Breedveld

Causal constraints

Summerschool Bertinoro, 7-11 July 2003 (59) © Peter Breedveld

29
Arbitrary causality

Summerschool Bertinoro, 7-11 July 2003 (60) © Peter Breedveld

Causality assignment

Algorithmic (SCAP):
– 1) fixed causal ports with propagation via
constraints
causal ‘conflict’: ‘ill-
‘ill-posedness’
posedness’
– 2) preferred causal ports with
propagation via constraints
causal ‘conflict’: dependent state(s)
– 3) choice of arbitrary causal port with
propagation via constraints
means existence of algebraic loop(s)
loop(s)
Summerschool Bertinoro, 7-11 July 2003 (61) © Peter Breedveld

30
Example
u u u
T T
i u i u i J
ω ω
i i
•Important modeling feedback
•Automatic in 20-
20-sim
I I •Visible in bond graph causality
2nd order loop •‘Hidden’ in case of iconic diagrams

Se 1 GY 1
1st order loop
1st order loop

R R

Summerschool Bertinoro, 7-11 July 2003 (62) © Peter Breedveld

Positive orientation

orientation NOT THE SAME AS direction!

Summerschool Bertinoro, 7-11 July 2003 (63) © Peter Breedveld

31
Causality does NOT affect orientation!

Summerschool Bertinoro, 7-11 July 2003 (64) © Peter Breedveld

Mechanical versus
Thermodynamic framework

Mechanics: Thermodynamics:
Two types of storage One type of storage

Oscillatory behavior Only relaxation


(damped): C-
C-I(-
I(-R) behavior: C-
C-R
Split domains (therm
(therm.)
.) and couple by SGY
(mech.):
C-SGY-
SGY-C

Summerschool Bertinoro, 7-11 July 2003 (65) © Peter Breedveld

32
Symplectic gyrator

• Unit gyrator: C 0 GY 0 C
r=1

• SGY: C 0 SGY 0 C

• Multiport:
C11 C12 0 -1
C12 C22 C 0 SGY +1 0

20-
20-sim demo

Summerschool Bertinoro, 7-11 July 2003 (66) © Peter Breedveld

Symplectic GYrator

• Mechanical (x,p) ekin =


∂E
ekin = v ∂E
∂p e pot = F e pot =
∂x
C 0 SGY 0 C
dp dp dx
f kin = F =− f pot = v f pot =
dt dt dt
dp
– Only in inertial frames: F = − (Newton's 2nd law)
dt

• Electrical network (q,λ):


∂E ∂E
emag = emag = i eelec = u eelec = ∂q
∂λ
C 0 SGY 0 dq
C
dλ dλ felec = i f elec =
f mag = − =u dt
dt dt
– Only quasi-
quasi-stationary (non-
(non-radiating): dλ emag = i
− =u
dt

Summerschool Bertinoro, 7-11 July 2003 (67) © Peter Breedveld

33
Unit gyrator (SGY) as ‘dualizer’
Original: Port equivalent: Original: Port equivalent:

0 SGY I 0 C Se SGY 1 Sf 1

1 SGY C 1 I Sf SGY 0 Se 0

1 SGY R 1 R
1 SGY 1 SGY 0 1 0 0

1 SGY TF SGY 0 1 TF 0
SGY
0
1 SGY GY SGY 1 1 GY 1

0
1 SGY TF 1 1 GY 1
1 SGY 0 SGY 0 1 1 0
1 SGY GY 0 1 TF 0

SGY 1

0
1

Summerschool Bertinoro, 7-11 July 2003 (68) © Peter Breedveld

Mechanical framework of
variables

Summerschool Bertinoro, 7-11 July 2003 (69) © Peter Breedveld

34
Generalized thermodynamic
framework of variables
f
flow
e
effort ∫
q = f dt
generalized state
electric i u q = ∫ idt
current voltage
charge
magnetic u i λ = ∫ udt
voltage current
magnetic flux linkage
elastic/potential v F x = ∫ vdt
translation velocity force
displacement
kinetic translation F v p = ∫ Fdt
force velocity
momentum
elastic/potential ω T θ = ∫ ωdt
rotation angular torque
velocity angular displacement

kinetic rotation T ω b = ∫ Tdt


torque angular
velocity angular
momentum
elastic hydraulic ϕ p V = ∫ ϕdt
volume flow pressure
volume
kinetic hydraulic p ϕ Γ = ∫ pdt
pressure volume flow
momentum of a flow tube
thermal T fS S = ∫ fS dt
temperature entropy flow
entropy
chemical µ fN N = ∫ fN dt
molar flow
chemical number of moles
potential

Summerschool Bertinoro, 7-11 July 2003 (70) © Peter Breedveld

Co-Energy & Legendre


Transformations
A function F( x) with x = x1 ,… , x k
is homogeneous of order n if F( α x) = α n F( x)

∂F
Define y i ( x) =
∂x i
∂F( α x) α n ∂F( x) n −1
then y i ( x) or y i ( α x) = = ⋅ =α y i ( x) is homogeneous of order (n-1).
∂αx i α ∂x i
k ∂F 1 k 1
T
For a homogeneous function Euler’s theorem holds: ∑ ∂x ⋅ x i = n⋅ F( x) or F( x) = n ∑ y i ⋅ x i = n ⋅ y ⋅ x
i =1 i i =1
k ∂F k
By definition: dF = ∑ ⋅ dx i = ∑ y i dx i = y T ⋅ dx

i =1 ix i =1
1  1 1
but also: dF = d y T ⋅ x= y T ⋅dx + (dy) T ⋅ x
n  n n

Hence: (dy) T ⋅ x = ( n −1)y T ⋅dx

for n = 1: ( dy) T ⋅ x = 0
1
for n ≠ 1: dF = (dy) T ⋅ x
n -1

Summerschool Bertinoro, 7-11 July 2003 (71) © Peter Breedveld

35
Homogeneous
Energy Functions
The energy of a system with k state variables q is: E(q ) = E (q1 ,… , q k )

1
If qi is an “extensive” state variable, this means that: E( αq ) = αE(q ) = α E(q )

∂E
Hence E(q) is first order (n = 1) homogeneous, so e i (q ) =
∂q i

is zeroth order (n – 1 = 0) homogeneous, which means that ei(q) is an “intensive” variable, i.e.
e i ( αq ) = α 0 e i ( q ) = e i ( q )
∂e de
This means also that in case n=1 and k=1 e(q) is constant, i.e. = =0
∂q dq
which changes the behavior of this element into that of a source.
in order to enable storage: storage elements = multiports (k>1)
‘1-port storage element’ = n-port storage element with flows of n-1 ports kept zero
corresponding n-1 states constant and not recognized as states.
Such a state is often considered a parameter:
if E(q 1 , q 2 ,..., q n ) dq =0,∀i≠1 = E′(q1 ) then E’(q1) not necessarily first-order homogeneous in q1
i

Summerschool Bertinoro, 7-11 July 2003 (72) © Peter Breedveld

Homogeneous
Energy Functions

In bond graph form:

For n=1 and k independent extensities:


only k–1 independent intensities
because for n=1 we find Gibbs' fundamental relation: E(q ) = e T ⋅q

by definition: dE = e T ⋅ dq

from the above equations follows the Gibbs-Duhem relation: ( de) T ⋅q = 0

Summerschool Bertinoro, 7-11 July 2003 (73) © Peter Breedveld

36
Legendre Transformations
Legendre transform of F(x) with respect to xi is by definition: L{F( x)} x = L x = F( x) − y i ⋅ x i
i i
∂F
with y i =
∂x i
k
and the total Legendre transform of F(x) is: L{F( x)} = L = F( x) − ∑ y i x i
Note that for n = 1: L=0 i =1

Now dL x = dF − d ( y i x i ) = dF − y i dx i − x i dy i = ∑ y jdx j − x i dy i
i j≠ i
or L x = L x ( x1 ,… , x i −1, y i , x i +1 ,… , x k )
i i

which means that xi is replaced by yi as independent variable or “coordinate”!


Hence
k
L = L(y); dL =− ∑ x i dy i =−(dy) T ⋅ x
i =1

Summerschool Bertinoro, 7-11 July 2003 (74) © Peter Breedveld

Co-Energy Functions
The co-energy E ∗q of E(q) with respect to qi is by definition:
i

E∗q =−L q = E∗q ( q1 ,… , q i − 1 , e i , q i +1 ,… , q k )


i i i

Hence E (q ) + E q (… , e i ,… ) = e i ⋅q i
i

The total co-energy E*(e) of E(q) is: E* = -L, hence E(q ) + E∗ (e) = e T ⋅q


For n=1: E (e) = 0
confirming earlier conclusion that there are only k-1 independent ei

1
For n=2: E(q ) = E∗ (e) = e T ⋅q
2

1 2
For n=3: E(q ) = e T ⋅q E ∗ ( e ) = e T ⋅q
3 3

Summerschool Bertinoro, 7-11 July 2003 (75) © Peter Breedveld

37
Relations for
Co-Energy Functions
dE ∗q = de i ⋅ q i − ∑ e j ⋅ dq j
i j≠1

k
dE∗ = ∑ de i ⋅q i = (de) T ⋅q = ( n −1)e T ⋅dq = ( n −1)dE
i =1

n −1 T  1
E∗ = ( n −1) E = ⋅e ⋅q =1− e T ⋅q
n  n

Summerschool Bertinoro, 7-11 July 2003 (76) © Peter Breedveld

Legendre transforms in simple


thermodynamics
In thermodynamic systems with
– internal energy U
– entropy S
– temperature T
– volume V
– pressure p
– total mole number N
– total material potential µtot
– mole number per species i Ni
– chemical potential µi: m −1 tot
Free energy F: F = L S = U − TS = − pV + ∑ µ i N i + µ ⋅ N
i =1 F(T, V, N , N ) (= N. f (T, v, c))
m −1
dF = −SdT − pdV + ∑ µ i dN i + µ tot ⋅ dN
i =1

Enthalpy H: H = L V = U − (−pV) = U + pV
H (S, p, N , N ) (= N. h(s, p, c))
m −1
dH = TdS + Vdp + ∑ µ i ⋅ dN i + µ tot ⋅ dN
i =1
Summerschool Bertinoro, 7-11 July 2003 (77) © Peter Breedveld

38
Legendre transforms in simple
thermodynamics
m −1
Gibbs free enthalpy G: G = LS, V = U − TS − (−pV) = µ tot ⋅ N + ∑ µi N i
i =1

m −1
dG = −SdT + Vdp + ∑ µ i dN i + µ tot ⋅ dN
i =1

G ( T, p, N , N ) (= N. g( T, p , c) )

For m=1: g = µ tot (T, p)

Summerschool Bertinoro, 7-11 July 2003 (78) © Peter Breedveld

Legendre Transforms and Causality

• If an effort is “forced” on a port of a C element (“derivative causality” or “flow


causality”), this means that the roles of e and q are interchanged in the set of
independent variables, which means that the energy has to be Legendre
transformed.
• This is particularly useful when the effort e is constant (e.g. an electrical
capacitor in an isothermal environment with T=Tconst):
u
q
. 
T
S
. s e
: T const

dF = udq − SdT = udq


u
q
. C
dF
P = u⋅q =
dt

Summerschool Bertinoro, 7-11 July 2003 (79) © Peter Breedveld


with

39
Constitutive Relations

• The function ei(q) is called constitutive relation, also called constitutive


equation, constitutive law, state equation, characteristic equation, etc.
• If ei(q) is linear, i.e. first order homogeneous, then E(q) is second order
homogenous, i.e. E(q) is quadratic. In this case, and only in this case:

E ( αq ) = α 2 E ( q )
1
E ( q ) = e T ⋅q
2
(de) T ⋅q = e T ⋅dq = dE

Summerschool Bertinoro, 7-11 July 2003 (80) © Peter Breedveld

Maxwell Reciprocity
From the principle of energy conservation (First law) can be derived that:

∂2 E ∂2 E
=
∂q i ∂q j ∂q j∂q i

∂e j ∂e i
=
∂q i ∂q j

i.e. the Jacobian matrix of the constitutive relations is symmetric.

This is called Maxwell reciprocity or Maxwell symmetry.

Summerschool Bertinoro, 7-11 July 2003 (81) © Peter Breedveld

40
Intrinsic Stability
Intrinsic stability requires that this Jacobian is positive-definite:

 ∂e 
det   > 0
 ∂q 

and that the diagonal elements of the Jacobian are positive:

∂e i
>0 ∀i
∂q i

Summerschool Bertinoro, 7-11 July 2003 (82) © Peter Breedveld

Legendre Transforms in Mechanics


In mechanical systems with
– kinetic energy T potential energy V
– displacements x momenta p
– velocities v forces F:

• Hamiltonian H: E ( q ) = H ( x , p) = T + V

• Lagrangian L: H∗p =−L p = v T ⋅ p − H = (T + T∗ ) − (T + V) = T∗ − V = L(x, v)

∂H
with v =
∂p

• co-Hamiltonian H
*
: H* = vT ⋅ p + FT ⋅ q − H =
x, p x, p

= (T + T* ) + (V + V* ) − (T + V) = T∗ + V∗ = H∗ (F, v)

∗ ∗ ∗ ∗ *
• co-Lagrangian or Hertzian H x : H x = FT ⋅q − H = ( V + V ) − (T + V) = V − T =−L ( F, p)

∂H
with F =
∂x
Summerschool Bertinoro, 7-11 July 2003 (83) © Peter Breedveld

41
Legendre Transforms in
Electrical Circuits

In electrical circuits with


– capacitor charges q and voltages u
– coil flux linkages Φ and currents i:

E(q , Φ ) = E c ( q ) + E L (Φ )

E∗ ( u , i ) = u T ⋅ q + i T ⋅ Φ − E = E

only in linear case!

Summerschool Bertinoro, 7-11 July 2003 (84) © Peter Breedveld

Example nonlinear 3-port C


• energy storage in GAS:
– Ideal −R s − s0
 v  cv
pv = RT T ( s, v ) = T0   e cv
 v0 
– Van-der-Waals
−R s − s0
 a
p+ 2  ( v − b ) = RT  v − b  cv
 v  T (v, s) = T0   e cv
 v0 − b 

Summerschool Bertinoro, 7-11 July 2003 (85) © Peter Breedveld

42
Regular causal form in
thermodynamics:
p = p(v,T ) T
differential
causality!
mechan-
thermal
ical
port C port

p(T,V,N)
S(T,V,N)
-dV = -V µ(T,V,N) -dS = -S
dt dt

material
µ = µ ( p,T ) -dN = -N
port dt
Summerschool Bertinoro, 7-11 July 2003 (86) © Peter Breedveld

Desired integral causal form:


p = p(v,T ) T
integral causality!
mechan-
thermal
ical
port C port

p(S,V,N)
T(S,V,N)
-dV = -V µ(S,V,N) -dS = -S
dt dt

material
µ = µ ( p,T ) - dN = - N
port dt
Summerschool Bertinoro, 7-11 July 2003 (87) © Peter Breedveld

43
Compute all conjugate variables:

Differential
v causality
p T
p(v,T)

s(v,T)
- v
µ (p,T) s
∫ d/dt

dV = −V V S dS = S

dt ∫ N dt

µ dN = N
dt
Summerschool Bertinoro, 7-11 July 2003 (88) © Peter Breedveld

Change to integral causality (T(v,s)?):

Integral
v causality
p T
p(v,T)
T(v,s)
- v
µ (p,T) s
∫ ∫
dV = −V V S dS = S

dt ∫ N dt

µ dN = N
dt
Summerschool Bertinoro, 7-11 July 2003 (89) © Peter Breedveld

44
Find S0 from T0 (?):
T0

v ????
p T
p(v,T)
T(v,s) S0
- v
µ (p,T) s
∫ ∫
dV = −V V S dS = S

dt ∫ N dt

µ dN = N
dt
Summerschool Bertinoro, 7-11 July 2003 (90) © Peter Breedveld

Form requires T0, not S0

s−s v   
1 ∂ p
T ( s, v ) = T0 exp  0
− ∫   dv 
 cv cv v  ∂ T dv =0  

 0 

Summerschool Bertinoro, 7-11 July 2003 (91) © Peter Breedveld

45
T0

v
p T
p(v,T)
T(v,s)
- v
µ (p,T) s
∫ ∫
dV = −V V S dS = S

dt ∫ N dt

µ dN = N
dt
Summerschool Bertinoro, 7-11 July 2003 (92) © Peter Breedveld

Concluding remarks

Port-based modeling:
– no a priori decision about input and output of a
bilateral relation
– two variables of bilateral relation (port):
• dynamically conjugated
• power conjugated in case of energy conservation:
– effort and flow
– behavior with respect to energy is domain
independent (C,I; R; TF,GY; Se,Sf;
Se,Sf; 0,1)
– causality assignment gives feedback on
modeling decisions

Summerschool Bertinoro, 7-11 July 2003 (93) © Peter Breedveld

46

You might also like