Chapter 10

Bond graph Modeling of Thermal Systems

10.1 Introduction

A student, after completing a few initial courses in mechanics develops a capability of

formulating and arriving at equations of motion for systems comprising of particles
and rigid bodies interacting with each other and acted upon by external forces. The
situation with thermodynamics is not so happy. If we closely watch (or recall our
own experience) the growth of capability of a student in thermodynamics, it will be
apparent that in general students dwell within a class of thermo-dynamical systems
varying marginally from a well layed-out set of situations. Fault is not with the
students but often with the pedagogies in this field.

The statements of the, so called, second law are often phenomenological in character
from which each time a set of suitable working rules is erected to approach a new
situation. The approaches become so arduous due to the tendency to delay the
introduction of notion of entropy and absence of a basic unit or a building block such
as a point in geometry and a particle in classical mechanics.

In what follows, thermodynamics would be handled the way it was used by Born [3]
and Prigogine [51]. The basic unit of thermodynamic modeling is a Homogenous
Block. A homogenous block is a body or a subsystem which is at uniform
temperature. The only way in which it’s entropy can be altered is through exchange
of heat.

Now, we state the second law of thermodynamics of a system comprising of

homogenous blocks.

(a) Each homogenous block of the system is associated with a single valued
function of time T, called temperature. The temperature of each
homogenous block is a positive number i.e., T > 0.

(b) If the total system is conceived as a set C of homogenous blocks, and say A
and B are two subsets (or subsystems) of C such that C =AB, then the
entropy of the system C i.e., Sc is related to the entropies of the subsystems
SA and SB and the entropy of the overlapping part SAB as follows.

S C  S A  S B  S AB
 BOND GRAPH MODELING OF THERMAL SYSTEMS 272
In other words entropy is an extensive quantity like mass, population,
volume, energy etc.

(c) The time rate of change of entropy of the system C has two components

S C  S e  S i .
S e is due to exchange from outside the system C and S i is the growth
within the system C.

(d) The S e may take place in two ways,

Nh 
Q
(i) S e  T
i 1
ei

i
.


where Q ei is rate at which the i th homogenous block imports heat from

agencies outside the system C, Ti is the temperature of i th homogenous

block and Nh is the number of homogenous blocks the system C is comprised
of.

(ii) The other way of growth of Se would be through addition of new

homogenous blocks. It may reduce if the homogenous blocks are removed
from C. Symbolically we may write this component of entropy growth as
S e  N
 S , where N
 is rate at which new homogenous blocks are added
to the system C and S is the entropy of blocks. In place of number of
homogenous block N any other suitable enumerator may be used. For
continuous systems mass or volume may as well play this role.

The total rate of growth of exchange entropy is sum of the changes in both
the above ways.
(e) S i is not defined but S i  0 . When S i is greater than zero the process is
called irreversible and when S  0 then the process is reversible.
i

All statements of second law traditionally presented may be derived from above
axioms. We will just prove one of them. Reader may try proving others.

Clausius statement: Unaided by work heat flows from a body at higher temperature
to a body at lower temperature.

Proof :- Let us consider a system C which contains two homogenous blocks A and B
as shown in Fig. 10.1.
273 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 10.1

The block A is at a temperature TA and block B is at a temperature TB. Say heat flows

Q
from A to B then, rate of change of entropy of A, S A   and that of the B,
TA
Q
S B   . The change of entropy of system C is
TB
T  T 
S C  S A  S B  Q A B .
T AT B
The change of entropy is entirely internal thus S i  0 which leads to
(T  TB )
Q A  0.
TATB

Now as TA > 0 and TB > 0 one may see that TA  TB. TA > TB means that if heat goes from
A to B then the temperature of A is higher than that of B. What about TA = TB, can heat
still go from A to B or from B to A, as both should be permissible? Yes, if a part of A is
worklessly removed and appended to B or vice versa. Such a transfer is reversible.
We now proceed to the actual business of modeling thermo-dynamical systems.

This chapter begins with modeling a volume of gas to which mechanical and thermal
energies are imparted and the amount of gas varies by exchange of material. The
thermal C-field for a volume of gas was initially considered by Beaman and
Breedveld [4]. Section 10.2 to 10.7 of this chapter are devoted to creation of the C-field
and its interfaces.

10.2 The thermal C - field for a volume of gas

Consider a collapsible chamber as shown in Fig. 10.2 with thermally conducting walls
and a hole through which gas may enter or leave. The gas within the volume is a
system comprising of homogenous block at the same temperature enumerated by the
mass of the gas.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 274

Fig. 10.2

The internal energy inside the volume V may be expressed as a function of volume of
the chamber, total entropy and total mass of the gas inside the chamber. It is assumed
that the gas under consideration is ideal.

U  U V , S , m  . (10.1)
where U = internal energy, V = chamber volume, S = total entropy and
m = total mass of the gas inside the chamber.

The rate of change of this internal energy with respect to time may be written as

U  U  U
U  V S m
 . (10.2)
V S m

U U U
Now  p , T and  .
V S m
where p is thermodynamic pressure, T is absolute temperature and  is chemical
potential.

A basic C - field now may be proposed as shown in Fig. 10.3.

275 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 10.3 Fig. 10.4

This C - field has three power ports. The flow and effort variables for mechanical port
are V and –p, that for thermal port are S and T and for material port are m  and
. This C - field is not complete for the process of generation of entropy S is
incomplete. As S is an extensive quantity we may define S = ms, where s is entropy
per unit mass.

Thus
S  m
 s  ms . (10.3)

The component m  s is due to mass transfer where as ms is the growth due to
heat transfer to the gas volume as we assume that there is no possibility of growth of
entropy in any other way inside the chamber.

Thus
S  m
 s  S ht (10.4)

where S ht denotes growth of entropy due to heat transfer.

Now an outer layer may be added to the basic C - field as shown in Fig. 10.4.

If the effort of the external material port is taken as hc, then

hc    Ts . (10.5)

If one interprets  as internal or Gibb's free energy per unit mass of gas entering the
chamber and as s is entropy per unit mass of the gas hc becomes enthalpy of unit mass
of the gas within the chamber. This may be shown as follows.

The Gibb's free energy function is defined as

G  U  pV  TS  H  TS ,
where
G = Gibb's free energy function for entire amount of gas,
H = enthalpy of entire amount of gas. Now

Now
 G   G   G 
dG    dp    dT    dm ,
 p  T ,m  T  p ,m  m T , p
and from
 BOND GRAPH MODELING OF THERMAL SYSTEMS 276
 U 
d U  pV  TS   Vdp  SdT   dm .
 m 

Thus chemical potential is

 G   U 
     .
 m  p ,T  m V ,S

Now if we take all quantities for unit mass of gas entering the chamber,

 G   H    TS 
as      ,
 m  p ,T  m  p,T m

 G  S
or    h T ,
 m  p ,T m

or   h  Ts ,

or hc    Ts .

This establishes the equation (10.5).

10.3 Thermodynamic relations for C - field for an ideal gas

For m kilograms of the gas in the chamber,

dU  TdS  pdV . (10.6)

mRT
or mCv dT  TdS  dV . (10.7)
V

where Cv is specific heat of the gas per kilogram and R is universal gas constant per
kilogram.

Thus

dT dV
dS  mCv  mR , (10.8)
T V

which leads to
277 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

 mCv mR 
 T  V  
S  S 0  ln    ,
T  V  (10.9)
 o 
  0 


or T  To V / Vo  R / C v e  S  S o  / mC v .
  (10.10)

The expression for internal energy may now be obtained

U  mCvTo V / Vo   R / C v e  S  S o  / mC v  U o . (10.11)

Now we take derivatives of U to obtain pressure and chemical potentials.

  S  So 
U mCvTo V 
p  
V

 e mC v
  R / CV  , (10.12)
V Vo  o 

 R / Cv  S  So 
and   U  C T  V  mC v   S  S0   ,
1 
e (10.13)
v o V  
V  o  mCv 

where   C p / Cv

For practical simulation on generalized software like SYMBOLS 2000 / Shakti, one
may use modulated sources of efforts to incorporate these functions as shown in Fig.
10.5 where F1 = (V/V0 )- and F2  e ( S  S 0 ) / mC v . In SYMBOLS 2000 / Shakti one may
even create an object to deal with such a C - field.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 278

Fig. 10.5

10.4 Modeling the thermal interface

Let us consider two chambers A and B at temperatures TA and TB respectively,

connected through a thermal resistor R as shown in Fig. 10.6.

Fig. 10.6

The rate of change of entropy at A is

T  TB
S A   A . (10.14)
TA R

and the rate of change at B is

T  TB
S B   A . (10.15)
TB R
279 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Assuming the conductor is ideal (neither does it generate entropy in any other way
nor does it store entropy), the net change of entropy in the system is given by

T  T  2
S A B  S A  S B  A B  0 ,
TATB R

which is always larger than (or equal to) zero, which ever way the heat may flow.
1
This is compatible with the second law of thermodynamics. Now if we denote
R
by Ct, the thermal conductivity, then the conductivity matrix may be written as

 Ct Ct 

S A   T A
 T A  T A 
  C Ct  TB 
, (10.16A)
S B   t 
 TB TB 


or S   Ct  T . (10.16B)

It is easy to see that matrix [Ct] is singular. If temperatures of the chambers are given,
entropies may be obtained but if rate of entropy change of the chambers are given,
the temperatures cannot be obtained. It is quite compatible with the indeterminism of
thermodynamics.

As most bond graph analysis software are created based on resistive fields and do not
attempt inverting these fields to arrive at the equations of state, two practical
proposals are given below to incorporate this interface in the model.

Proposal 1: Two unit gyrators may be incorporated to avoid an attempt to invert as

shown in Fig. 10.7.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 280

Fig. 10.7

Proposal 2: Modulated flow sources may be incorporated in the model to maintain

the rate of change of entropy as shown in Fig. 10.8.

Fig. 10.8

10.5 Modeling the material interface

Before we proceed to model the material interface, let us consider some important
relations for the steady flow of an inviscid compressible fluid.

(a) The energy equation without body forces and heat transfer is

D  v2 
 1 p
h   . (10.17)
Dt 
 2   t


where h is specific enthalpy, v is the velocity of the fluid, ρ is mass density

and p is the pressure [73].
281 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

For steady flow


 v2 

d h  0. (10.18)

 2 


(b) The variations of pressure and specific kinetic energy as governed by

velocity and vorticity given as follows.

 v2 

d
dp
 

 
  v    dr ,
2 
(10.19)

 

where  is vorticity vector and r is position vector within the fluid.

The variation of entropy is given as

 
Tds    v  dr . (10.20)

Along a stream line


 dp v2 

d


  0,
2 

(10.21)

and ds  0 . (10.22)

Thus entropy remains constant.

10.6 Interface through an orifice

Let us assume two chambers exchanging gas through an orifice. Mechanical work is
done on each volume of gas and heat is transferred through the walls of the chamber
as shown in Fig. 10.9.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 282

Fig. 10.9

Here the gas is essentially driven by difference of enthalpies per unit ass of the gas
volumes in the chambers. Thus we use equation (10.18) to determine the velocity of
the gas through the orifice

v2
 h1  h2 or v 2 h1  h2  . (10.23)
2

The rate at which mass of gas is exchanged

  Ao  2 h1  h2 
m (10.24)

Where Ao is the area of orifice.

So much of vorticity develops at the entry of chamber-2 that there is growth of

entropy (see equation (10.20)). One can not obtain this growth of entropy by simply
integrating equation (10.20) for that would need knowledge of the vorticity field. An
alternative way exists to do this. As the flow once again becomes stagnant in
v12
chamber-2, s 2  for unit mass of the gas exchanged. The change in the total
2T
entropy due to this would then be
Ao v 3
S 2  . (10.25)
2T

The bond graph model of this phenomenon shown in Fig. 10.10 where

H  x  1 for x0
0 for x0 .
283 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 10.10

10.7 Interface through finite resistor

Let us assume the chambers are interconnected through a resistive element as shown
in Fig. 10.11. The flow through such a resistor obeys linear law as follows.

 ,
p  RQ (10.26)

 is the volume flow rate through

where ∆p is pressure drop across the resistor and Q
it and R is the resistance.

Fig. 10.11
 BOND GRAPH MODELING OF THERMAL SYSTEMS 284
It is also assumed that the flow inside the chamber receiving the gas becomes
stagnated. Thus kinetic energy of the gas gets converted to heat there. This additional
entropy must be accounted for at the receiving chamber.

The bond graph model of this system is shown in Fig. 10.12 where

H  x  1 for x0
0 for x0

Fig. 10.12

The flow through the resistor does not change the volumes of the chambers, thus the
bonds connecting it to mechanical ports are flow activated. Where as flow of gas
through it affects the exchange of material between the chambers. The material mass
flow rate is determined by multiplying the volume flow rate by the density of the gas
(TF modulated by ρ). These ports are effort activated.
Often the pneumatic resister is defined as

  p ,
Rm (10.27)

where m mass flow rate then the bond graph of Fig. 10.12 would be modified as
shown in Fig. 10.13 where H (x) = 1 for x  0 and H (x) = 0 for x  0 .
285 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 10.13

10.8 Model of a toy hovercraft

A model of toy
Hovercraft is shown in
Fig. 10.14. The spring
with stiffness K is kept
compressed to start with.
On release it pushes the
piston P, which in turn
compresses the gas in
chamber -1. The gas from
chamber -1 goes to the
lower chamber -2
through a resistor R, then
from lower chamber it
diffuses into the
atmosphere.
Fig. 10.14
In fact it is the pressure in the lower chamber that keeps the craft afloat. The bond
graph of the device is shown in Fig. 10.15.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 286

Fig. 10.15

10.9 Pseudo bond graph for heat transfer and concept of derived bond
graph (DBG)

10.9.1 Pseudo bond graph for heat transfer

The modeling of the thermal interface was presented in the previous section. This
model was essentially that of heat transfer through a thermal resister. The base of the
thermal power and factor of power were taken as absolute temperature (as effort) and
rate of entropy change (as flow). This scheme gave manageable bond graph model for
volume of gas with exchange of thermal and mechanical energies and material
exchange. However, the model for heat transfer turned out to be complex. Reader
may look back the two proposals discussed there.

Often a simplified approach is adopted to model the phenomenon of heat transfer by

deviating from the basic idea of bond graph theory. This class of models are created
by forcing bond graphs to mimic the observed phenomenon or its mathematical
description without basing it on any extensive or set-additive, factorable entity like
total energy or entropy and their time rate changes. Such bond graphs are called
pseudo bond graphs. A pseudo bond graph model for the phenomenon of transfer of
heat and its accumulation in thermal capacitor may be created as follows.
287 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 10.16

To start with let us consider a thermal resister as shown in Fig. 10.16. The rate of heat
transfer Q  through an ideal linear thermal resister (with constant thermal
conductivity and no thermal capacity) may be written as

  hA T  T  ,
Q (10.28)
1 2

where A is the area of cross-section, h is coefficient of thermal conductivity, T1 and T2

are the temperatures at the two faces of the thermal conductor.

In terms of thermal conductivity k one may write the relation as

  kA  T  T  ,
Q (10.29)
1 2
L

where L is the length of the conductor. These relations are similar to that of flow of
current through an electrical resistance, if we give Q  the status of flow, T1 and T2
that of potentials or (T1 – T2) as the effort (potential difference) applied to the
conductor and 1/(hA) or L/(kA) the status of resistance.

That is

  1 T  T  .
Q (10.30)
1 2
R

Unlike electrical systems, where flow multiplied by effort gives power, in thermal
system Q  times (T1 – T2) relates to no meaningful quantity. Still, based on similarity
of constitutive relations one may draw a bond graph to represent the conduction of
heat through a conductor as shown in Fig. 10.17.

Next let us consider an ideal heat capacitor as shown in the Fig. 10.18.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 288

Fig. 10.17 Fig. 10.18

The ideal heat capacitor is, what is often termed as thermodynamic homogeneous
block. Which means the entire body is at the same temperature. Heat reaches all its
parts instantaneously. If a mass point is treated as a primitive building block of
classical mechanics, these homogeneous blocks are primitive building blocks of
thermodynamics. For a thermal capacitor,

1 t1
T  t1   T  t 0  
C  t0
  t  dt ,
Q (10.31)

where T(t1)–T(t0) is rise in temperature of the heat capacitor, C it's heat capacity and
Q the heat flowing to it. One may express C in terms of specific heat and total mass
of the substance as C = mc. This relation is similar to charging of an electrical
capacitor. The simple bond graph mimicking this phenomenon is shown in Fig. 10.19.

Fig. 10.19

Now we are prepared to create bond graph models of heat transfer in simple
situations. Let us consider a conducting rod placed between two chambers at
temperatures T1 and T2 as shown in Fig. 10.20. The rod is insulated such that heat
flows in axial direction. The rod is then subdivided in reticules (say six in this
example).

Fig. 10.20

The conceptual reticulation is done assuming that heat capacity of each reticule is
idealized in homogeneous blocks and the resistances representing the resistivity from
289 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

centre of one reticule to that of the next reticule are placed between these
homogenous blocks. The resistivity of half a reticules are placed at the ends
interfacing the temperature sources. This scheme is shown in Fig. 10.21.

Fig. 10.21

Pseudo bond graph model of the system as per the above reticulation is shown in
Fig. 10.22.

Fig. 10.22

Higher dimensional models for heat transfer may be likewise created. These models
may become nonlinear when other models of heat transfer are involved along with
variation of thermal capacity with temperature and total heat of the heat capacitor.
Let us consider a simple and idealized situation where a mass of ice is heated by
radiation to be melted to water and to superheated steam as shown in the Fig. 10.23.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 290

Fig. 10.23

The total heat and temperature relation of the pure substance may be as shown in
Fig. 10.24. The bond graph of the system is shown in Fig. 10.25.

Fig. 10.24

Fig. 10.25
291 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

The resistor Rr for radiative heat transfer between surface i and j may be expressed
ij

as follows.

1
 
 

Rri , j   A  Ti  T j T 2 i  T 2 j
1
1/  i  1/ 

 1 
 j

Aα = min(Ai, Aj) where Ai and Aj are the areas of ith and jth surfaces, σ is the Stefan-
Boltzmann constant Ti, Tj are the temperatures of the two surfaces and ei, ej are the
emissivity of the surfaces.

One will have to create a working bond graph from the conceptual bond graph by
appending suitable observers and modulated sources. For inclined surfaces one has
to incorporate view factors. Reader may see details in any standard text on heat
transfer.

10.9.2 Derived bond graphs - a path to pseudo bond graphs

(Note: reader may skip this section if he is not specially disturbed by pseudo bond
graphs of heat transfer).

Often system equations derived in one set of variables may be changed to some other
set of variables. The variables of the new set are function of those of the original set.
Such change of variable is done either by bringing the set of equations to easily
integrable form or to obtain other meaningful description of system behavior or both.
Such change of variables may even be done right at the system model in form of bond
graph. However, in doing so we may have to bring suitable changes in the
constitutive relation of the elements including the two-ports TF and GY and the
junction elements. In addition to this an icon like an enclosing box will have to be
introduced for representing the change of variables and scope of the change in the
system model. These new readjusted graphs may be called derived bond graphs. In the
following we would develop the idea of derived bond graph in steps.

10.9.3 Definitions

Scope-Box (SB): Scope-Box in short SB is an enclosure which encloses a part or whole

of a bond graph. For any portion of the graph which is within the scope-box, the
effort and flow variables are redefined as functions of original effort and flow
variables. On one or both sides of the box which do not intersect any bond, two
notches are created, one of these notches is triangular and other rectangular. Within
or just near the triangular notch the function mapping the original flow to new flow is
 BOND GRAPH MODELING OF THERMAL SYSTEMS 292
written and within or just near the rectangular notch the function mapping effort to
new effort is written.
Scope-Box Windows: The box has windows. Through these windows the transition
bonds enter the box. Each window divides a bond into two parts. One of which lies
inside the box the other remains out of it. The part out side has original flow and
effort variables and the part which falls inside has the mapped ones. Figure 10.26
shows a typical Scope-Box with windows, notches and functions. If a scope-box
encloses the entire bond graph it does not have any window.

Fig. 10.26 Fig. 10.27

Mapping Functions (MF): The mapping functions may be written as follows

f   Ge  f  and e  H f  e  .

where e and f are the effort and flow variables of original true bond graph model, e
and f  are that of derived bond graph. The function Ge( f ) is function of original
flow and e is treated as a parameter in the function likewise function Hf (e) is the
function of original effort and original flow, f is treated as a parameter of the function.

Thus G e 1  f    f and H f 1  e  e .

10.9.4 Derived constitutive laws

The constitutive relations for the junctions and the storage and dissipative bond
graph elements need readjustment and additional definition. Base, however, should
be the original true bond graph.

The Derived Junction Elements (DJE): Say in the original model there is an 1-
junction (Fig. 10.27).
293 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

then f1  f 2  f 3  f 4  f 5        f n ,
and e1  e2  e3  e4  e5         en  0 ,

For a derived part of the graph this junction would be marked as 1d (Fig. 10.28) and
junction laws would be modified as

Fig. 10.28

Ge 1  f1  Ge 1  f 2   Ge 1  f3         Ge 1  f n 

and H f 1  e1   H f 1  e2   H f 1  e3           H f 1  en   0

Likewise for a zero junction of the original model shown in Fig. 10.29.

Fig. 10.29

e1  e2  e3        en ,

and f1  f 2  f3  f 4        f n  0 .
 BOND GRAPH MODELING OF THERMAL SYSTEMS 294
Whereas for the corresponding Od -junction shown in Fig. 10.30 the junction laws
would get revised to

H f 1  e1   H f 1  e2   H f 1  e3           H f 1  en  ,

and Ge 1  f1  Ge 1  f 2   Ge 1  f3         Ge 1  f n   0 .

Fig. 10.30

The modeler has option of using underived junction 1 and 0 even within the scope-
box if it turns out meaningful.

Derived C Element: The constitutive relation for true bond graph may be written as

e  t  Q 
t
with Q 
 –
f    d .

Thus for derived C element the constitute relation should be

  t 
e   H f  t   t 
 

  –
Ge1    f     d 
 .


One may also introduce derived generalized displacement as

t
Q 
 –
f    d .

In some cases one may be able to write the derived constitutive relation

e   t  Q .

Derived I Element: The constitutive relation for true bond graph may be written as
295 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

f  t  p 
t
with p 
 e   d .
–

Thus the constitutive relation for derived I element should be

  t 
f   Ge  t  t 
 

  –
H f 1    e    d 



One may also introduce derived generalized momentum

t
p 
 e   d
–
.

It may be that some times the derived constitutive laws may be written as
e  t  p  .

Derived R Element: When R element is in its resistive form the constitutive relation
may be written as e  t  f  .

Thus it's derived form would be

e  H f  t    t Ge1t   f    .
 

Likewise if the element is in its conductive form, then it would be f  xt e .  

It's derived form may then be written as

 
f   Ge  t   xt  H f 1 t   e    .
  

Derived two-port Element: Let us now consider derived two-port elements.

(i) Derived Transformer: Say in a true bond graph model a transformer constitutive
relation for forward causality as shown in Fig. 10.31 is as follows.

f 2   f1 ; e1   e2 .

It's derived version would be

f 2  Ge   Ge1  f1  ; e1  H   H 1  e  


  f 
 f 2 

 BOND GRAPH MODELING OF THERMAL SYSTEMS 296

Fig. 10.31 Fig. 10.32

In reversed causality as shown in Fig. 10.32. The constitutive relation would be

1 1
e2  e1 ; f1  f2 .
 

The derived version would be

1  1 
e2  H f  H f 1  e1   ; f1  Ge  Ge 1  f 2   .
   

Thus the modulus operates on the original flow and effort variables. Otherwise there
is no change in the role of mapping functions. In a scope-box, the derived transformer
may be represented as shown in Fig. 10.33.

Fig. 10.33

(ii) The Derived Gyrator: The constitutive relations of a gyrator for two of its causal
orientation as shown in Fig. 10.34(a) and Fig. 10.34(b) would be as follows.

Fig. 10.34(a) Fig. 10.34(b)

For the gyrator of Fig. 10.34(a)

e2   f1 and e2   f1 .

The derived relation should then have been

297 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS


e2  H f  t   Ge t  1  f1  and 
e1  H f  t   Ge t  1  f 2   .

For the gyrator of Fig. 10.34(b) the true bond graph constitutive relation would be

1 1
f2  e1 and f1  e2 .
 

The corresponding derived relations should have been as follows.

1  1 
f 2  Ge t   H f  t  1  e1   and f1  Ge  t   H f  t  1  e2   .
   

But these relations are not valid. Let us consider a significant aspect of gyrator
element. The changes of flow to effort and effort to flow across a gyrator element
demands that the mapping functions should also be flipped. The flow mapping
function before a gyrator element should be flipped to an effort mapping function as
follow.

Ge  f   G f  e  .

Likewise the effort mapping function would be flipped to a flow mapping junction as
follows.

H f  e  H e  f  .

Thus one scope-box should terminate right at a gyrator and a new scope-box should
then begin with mapping functions flipped as shown in Fig. 10.35. This is in addition
to the constitutive relation.

Fig. 10.35
 BOND GRAPH MODELING OF THERMAL SYSTEMS 298
The right derived constitutive relation for gyrator element may now be created
considering all the changes which take place across this element.

For the causal orientation of Fig. 10.34(a) the derived constitutive relation would be


e2  G f  t   Ge 1  f1  and e1 
 H f  H e 1 f 2   .

Likewise the causal orientation of Fig. 10.34(b) the derived constitutive relation of the
gyrator would be

 1   1 
f 2  H e   H f 1  e1    and f1  Ge t    G f 1  e2    .
     
   

10.9.5 Back to the problem of heat transfer

Let us consider two homogeneous blocks at temperatures T1 and T2 interconnected

with an ideal linear thermal conductor of resistivity R or conductivity Ct = 1/R as
shown in Fig. 10.36. Say the thermal capacities of the homogeneous blocks be C1 and
C 2.

Fig. 10.36

True bond graph for this system may be created in two ways,

(a) Taking total heat Q as base measure and temperature as effort and rate of
change of entropy S as flow.

(b) Taking total entropy S as base measure and reciprocal of absolute

1
temperature,    as flow. The effort
as effort and rate of heat flow Q
T
 may be called 'Coldness'.
299 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

The Bond graph as per proposal (a): The phenomenon of heat transfer as per this
proposal is already discussed in the previous section of thermal interface. Thus we
straightaway come to the bond graph model of Fig. 10.37.

Fig. 10.37

The laws for thermal capacitors would be as follows.

CT  TS .

Thus T  T0 e ( S  S 0 ) / C , where T0 is the initial temperature and S0 is the initial

entropy. C is the heat capacity. Now we proceed for derived bond graph with
mapping functions as
Q e  
  G S  TS ,

and T  H f  T   T , identity mapping.

We would perform the mapping in two stages. First, we mark out the scope--boxes
and the mapping functions, covering the bond graph model (Fig. 10.38).

Fig. 10.38

The derived resistive field then would be

 BOND GRAPH MODELING OF THERMAL SYSTEMS 300


 T1  S1 
   Q 
  1 1  T1 
    1   Ct   .
 1  1 T2 
  S  Q 
 T1 2  2

The derived capacitor would be as

S  Q
 /T .

1 t Q   
 t0 T    d – 1 S0
c ,
T  T0 e e c

1 1 t Q   
– S0  t0 T    d
Q  t  1 ,
or T  T e c ec
T t C
0

 t / C ,
or T  Q

1 t
or T 
C  –
    d .
Q

The derive 0-junction, 0d is as follows

(Fig. 10.39).

T1  T2  T3  T4        Tn .
Q 1 Q 2 Q
   n  0 .
T1 T2 Tn
 Q
Thus Q   Q
  0 as all
1 2 n
T1  0 .
Fig. 10.39

So there is no difference between derived 0d and original 0-junction and one may see
the R - field has been reduced to a simple one port R element with resistivity Ct and
appended to a true 0-junction. Thus the derived bond graph is as shown in Fig. 10.40.
301 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 10.40

Now if the gyrator is removed by reinterpreting the resistance and junction on which
it is appended, the bond graph becomes as the one shown in Fig. 10.41.

Fig. 10.41

We now have the entire graph within one scope-box and this graph may be
immediately identified as the pseudo bond graph of heat transfer.

Thus the pseudo bond graph is infact a derived bond graph from a true bond graph
based on total energy Q with S as flow and T as effort.

The Bond graph as per proposal (b): The total entropies of a system is a set-additive
or a measure quantity and its time rate S may be factorized in two factors.

1
S   Q .
T

1
 as flow and
We now take Q  as effort.  may be called coldness.
T
The 1-junction (Fig. 10.42} with entropy as base has the following junction laws.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 302

Fig. 10.42 Fig. 10.43 Fig. 10.44

 Q
Q  Q
 Q  and               0 .
    Q
1 2 3 4 n 1 2 3 4 n

Likewise the 0-junction (Fig. 10.43) laws would be

1   2   3   4       n .

 Q
Q  Q
 Q
    Q
 0 .
1 2 3 4 n

The thermal resistor's constitutive law may be written as (Fig. 10.44).


 Q 1 
 Ct  1 1  1 
    .

Q2   1 2  1  1  2 

The conductivity field is definable in this approach. The thermal capacitor's

constitutive relation may be written as

  0
 t  
  0 t

.
Q    d  1
C 0

Now one may create the true bond graph shown in Fig. 10.45 on entropy base. The
unit gyrators are used such that the standard R - field is used in the model.
303 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 10.45

Now we proceed for creation of a derived bond graph with following mapping
functions.
1
T  H Q  t     ,
1 

 G
and Q    
  t  Q  Q , identity mapping.

The R - field becomes


 Q 1 
  1 1  T1 
    Ct    ,

Q2    1  1 T2 

and the capacitor's constitutive law transforms to

1 t
T  t   T  0 
C  Q   d .
0

Now we cover the graph in suitable scope-boxes (Fig. 10.46) and then bring the
changes, which render a simpler bond graph.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 304

Fig. 10.46

This R - field may be replaced by single port R element and GY may be removed
using R in it's resistive form.

Fig. 10.47

The bond graph (Fig. 10.47) enclosed in the scope-box is the so-called pseudo bond
graph. The junctions are normal 1 and 0 junctions.

Embedding derived bond graphs in a bond graph model: Let us consider a thermal
system with two chambers of gas exchanging heat and material, as shown in
Fig. 10.48. The mixed bond graph for this system is shown in Fig. 10.49. The scope-box
covering the derived portion of the graph, the window of the box and the mapping
functions are indicated in the diagram.
305 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 10.48

Fig. 10.49
10.9.6 Realization of windows

In the mixed bond graph model of Fig. 10.49 there are two windows indicated by W1
and W2. Through these windows two bonds are passing. Across these bonds the
change in variables takes place. For manual derivation of system equation, one needs
just to take cognizance of this fact. The software handling bond graphs do not have
 BOND GRAPH MODELING OF THERMAL SYSTEMS 306
facilities to incorporate the scope-boxes, window, mapping function etc. More such
extensions are essential if the junctions inside the scope-boxes turn out to be usual
junctions. The window then may be realized by activated two ports with suitable
modulations. The relevant portions of the original model of Fig. 10.49 and the
realized windows are shown in Fig. 10.50 and Fig. 10.51. Once the realization of
windows is accomplished the scope-boxes along with all the attributes may be
dropped. The model thus obtained may be handled by usual software and techniques
to analyze the models and simulations.

Fig. 10.50

Fig. 10.51

Problems

P1. The system shown in Fig. P10.1 consists of two chambers of same cross-
section separated by a light perfectly insulating piston. If the stops are
removed the piston can move freely. There is no leakage through piston.
307 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

There is a damper of
resistance Rd, one end of
which is connected to the
piston and other to the wall
of the chamber 1. The initial
volume of chamber 1 is V1.
The entire system is insulated
such that no heat can enter
the chambers from outside or
go out. The chamber 1 is Fig. P10.1
filled by ideal gas and
chamber 2 is evacuated. The heat generated by the damper is absorbed by
the gas. Now the stops are removed. The piston expands to a volume V2.
Draw a bond graph model to study this phenomenon. From the model,
prove that the change in entropy is given by
S2 – S1 = mR ln (V2 / V1),

Which happens to be the same as that for free adiabatic expansion of gas
from volume V1 to V2? As long as Rd > 0, show that this change of entropy is
independent of its value.

P2. Consider a perfectly insulated tube with ends closed partitioned in two
chambers by a light piston as
shown in Fig. P10.2. There is
a mass M rigidly attached to
the piston. The chamber 1 is
filled by m amount of ideal
gas and a damper Rd is
attached to the piston and
the rear wall of the chamber
1. The piston is also a perfect
insulator. The chamber 2 is
Fig. P10.2
very long and evacuated.

Initially the piston is held by the stops as shown. Obtain expressions for
temperature and entropy change of the gas with change in volume of the gas
after the stops are removed.
P3. A hydro-pneumatic
shock absorber is shown
in Fig. P10.3. Draw its
bond graph model.
Assume that there is no
exchange of heat with the

Fig. P10.3
 BOND GRAPH MODELING OF THERMAL SYSTEMS 308
environment or within the system. The hydraulic resistances at the oil ports
follow a law


p  Q Q
2C d2 A02

where  is the density of oil, Cd is the coefficient of discharge, A0 is port

 is volumetric flow and p is the pressure difference.
area, Q

P4. A simple entrapped gas

spring system is shown in
Fig. P10.4. The fraction
between piston and cylinder
is assumed to be zero. A load
ML is placed at the top of
piston as shown in Fig. P10.4.
The chamber is filled with
ideal gas at volume V0 and
sufficient pressure to balance
the load. The thermal
resistance of the wall is R
and the ambient temperature
is Ta. Draw bond graph
model of the system with
mechanical and thermal
ports.
Fig. P10.4
Linearise the
system model and derive system equations. Show that if R   then the
effective damping of free oscillation becomes zero and the same happens
when R  0 . For all intermediate values there will be positive damping.
P5. A thermo fluid system with a scheme of discretization of thermo fluid
aspects is shown in Fig. P10.5. The total inertia and hydraulic resistance of
the water in the tube may be considered as lumped in single I and R
elements. The out let valve is also represented by a resistance.
309 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. P10.5

Draw bond graph model of the system. The tube may assumed to be
uniform with area of cross section a and the reservoir of area A. The
temperatures in each section are much below the boiling temperature of
water, thus there is no phase change in the system.

Project type problems

Pr1. In compressible fluid with constant density flows through a tube of cross-
sectional area A. The tube is thermally insulated every where along its length
except a portion of length l. This portion of the tube is a heat conductor.
Around this portion a resistive coil of resistance Rc is wound. This coil is
covered from outside by an insulating tube with reflecting inner surface.
About 80% of the heat generated in the coil goes to the flowing water as
shown in Fig. Pr10.1

Fig. Pr10.1

If the thermal capacity of the fluid per unit volume is Cf, the thermal
conductivity is , the inlet temperature is Tin then create real and derived or
pseudo bond graph models of the system to predict the outlet temperature
Tout.
 BOND GRAPH MODELING OF THERMAL SYSTEMS 310

Pr2. A co axial counter flow heat exchanger is shown in Fig. Pr10.2.

Fig. Pr10.2

The thermal conductivity of the wall of inner tube is K and its thermal
capacity per unit volume is C . The heat capacity of unit volumes of
fluids is Cf, the thermal conductivity of fluids be Kf. Consider the system to
be lumped by dividing it by transverse planes such as aa' and bb' as shown
in the Fig. P10.7. For simplicity assume that temperature of each lump of
fluid is uniform. Create derived or pseudo bond graph model on the system.