You are on page 1of 26

Chapter 2

Causality
2.1 The notion of causality

Ever since man attempted to comprehend the connection between events, he has
been led again and again, to the notion of causation. Ancients from every part of
civilized world in their systems of philosophy or religion attempted to create picture
of genesis and causation. Aristotle may be identified as the first philosopher of the
west who initiated, what may be called the notion of empirical causality. Eastern
philosophers, even before Aristotle, attempted to question this notion. However,
these questions were asked in domains unrestrained and failed to take steps to foster
the empirical aspects of causality. Bertrand Russel in his essays (Cambridge Paper
1888-1899 and "Notion of Causality" in Human Knowledge-its Scope and Limits) points
out that Galileo Galilei (1564 - 1642) was the first natural philosopher to create an
empirical notion of causality on which the structure of classical system dynamics is
founded. Galileo's writings were, very justifiably, obscure. However, in refined
modern language one may state the ideas of Galilean causality as follows.

A signal P(t) should be identified as the cause of signal Q(t) if the history of P(t) prior
to the contemporary time determines the contemporary value of signal Q(t), i.e.,

Q  t   H t  P     ,     , t  (2.1)

Where Ht is a history functional which depends on the values of P from a distant


past to the present time t. The distant past is represented by -. However, if the
signal has finite values after a definite point of time in the past then the signal may be
considered to be zero before that point of time.

Let us consider a case when two signals are connected in such a way that both of
them play the role of cause of each other, i.e.,

Q t   H t  P    with    , t  ,
(2.2)

and P t   H t  Q   with    , t  ,


(2.3)

where Ht and Ht' are two history functionals. Then one may reduce
CAUSALITY 34


Q t   H t  H   Q      with     ,  ,    , t  ,
 
(2.4)
and Ht Ht' may then be treated as new history,


or Q  t   H t  Q    (2.5)

Now if an initial value of Q () is given (with an additional condition of Q     0


for    0 ), one may predict any later value of Q(t).

If a set of signals is such that their history relations may be so manipulated that every signal
of the set may be determined by its own history then the set is called causally closed . A
causally closed set may be treated as our first impression of what we generally call a
system. The Causal closer is a paramount concept of system modeling though not
complete in itself.

Let us consider a situation where the history functionals are integrations,

t
Q t     P   d , (2.6)


t
and P t     Q    d , (2.7)

From this one may deduce

t 
Q t      Q    d d . (2.8)
 

Incidentally, integration is a history functional of which an inverse exists, which is


derivative. Let this inverse be operated in succession on Eqn. (2.8).

dQ t 
t
   Q   d , (2.9)
dt 

d 2 Q t 
  Q t  . (2.10)
dt 2
35 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

This differential equation has the power to predict the future behavior of signal Q(t)
from given initial values (and rates).

2.1.1 Causation in physical system modeling

In so called physical system dynamics we confine to only one notion of history


functional that is time integration. If the system is energetically closed, the above
notion of causation would have been enough to obtain the predictive relations either
in the form of a set of differential or integral equations or their combination. Model of
a physical system is limited in such a way that a greater part of universe remains
outside its boundaries from which it imports power or rejects to it. In this chapter we
would enlarge the notion of causation to incorporate all aspects such as interaction
with exterior, storage and constraints.

2.1.2 Information exchange and its laws

Let us revert back to the bond graphs. It was discussed in Chapter - 1 that a bond in a
bond graph model represents exchange of power between elements or junctions
appended at its ends. This exchange has a physical appeal. Bond graphs, however,
are not merely portrait of physical reality but are more than that. They may be
subjected to systematic analysis to render mathematical model of the system
allowing prediction of its dynamical behavior. For this to happen, bonds have to be
associated with another kind of exchange i.e., exchange of information. The term
information needs explanation. It is assumed that at one end of a bond one of the
factors of power (either effort or flow) is determined by an element or a junction
appended at the other end, and vice-versa. In this sense there is a dialogue between
the interfaces at the ends through the bond or there is an exchange of information.
Paynter told (in some what loose terms) that ergs (energy) and bits (information)
flow across a bond.

In the general scheme of modeling classical systems by power exchange portraits


(bond graphs) there exists no element which can send both the information (effort as
well as flow) to any thing else to which it is interfaced. This is because if there could
be such an element then the system interfacing it will be over constrained by this
element without any possibility of general morphological and constitutive structure
and dynamical behavior. Say in electrical domain there an element exists which
determines both potential difference and current in a system interfacing it and is
power directed towards the interface. Say potential difference and current are in
phase. Then the only system which can be interfaced with it should be a resistance of
definite value or its equivalents.
CAUSALITY 36

This leads to laws of information exchange in bond graph theory as follows.

(a) One energy port (one end of the bond) can impart information of only one
of the factors of power to the interfacing element or junction.
(b) The two energy ports constituting two ends of a bond cannot impart
information of the same factor of power to their corresponding interfacing
element or junction.

These requirements make representation of information exchange easier. On each


bond one has to indicate the direction of flow of information of only one of the
factors of power. The information of the other factor flows in the opposite direction.
In bond graph theory the flow of information of effort is represented by putting a
small transverse stroke at the end of the bond (energy port) where this information is
imparted. Obviously the other energy port (unstroked end of the bond) imparts the
information of flow to the interfacing element or junction. Say we have two elements
or junctions A and B connected by a bond. Two alternative causal orientations could
be as shown in Fig. 2.1(a) and Fig. 2.1(b). In Fig. 2.1(a) A receives the effort
information and B the flow information. On the other hand in Fig. 2.1(b) B receives
the effort signal and A the flow signal.

Fig. 2.1(a) Fig. 2.1(b)

2.2 Causality of sources

We told that history functionals and causal closer are essential ideas for system
dynamics. But these ideas are not complete in themselves. There may be robust
exogenous sources which may drive the system without being loaded by the system.
There are two such source types which are possible. One which imparts the
information of flow and the other which imparts the information of effort.

To illustrate the basic concept of causality, let us assume that there is a prime mover
with a smart speed governor as shown in Fig. 2.2(a).

Fig. 2.2(a)
37 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 2.2(b)

The prime mover is driving the load i.e., power is going from the prime mover to the
load. Apart from sending power, the prime mover also decides that the load should
run at a particular speed depending on the setting of the governor. So, it may be said
that from the prime mover, the information of flow is generated which goes to the
load. The load generates the information of torque (effort) which the prime mover
receives and adjusts the inner mechanisms to compensate for it. The wavy lines in
Fig. 2.2(b) indicate the direction of flow of these information.

Let us consider another example. A cell and an electrical load are connected as
shown in Fig. 2.3(a).

Fig. 2.3(a) Fig. 2.3(b)

The cell here sends the information of voltage (effort) to the load and the load sends
the information of current to the cell as shown in Fig. 2.3(b). This information is used
by the cell to adjust the rate of chemical reaction to maintain the potential difference
across its terminals.

Each source element can generate only one of the two information. Thus if one of
them is shown in a diagram, then the other has a direction opposite to it. As already
mentioned the convention is to put a stroke at the bond end at which the effort
receiver is connected. The situation of Fig. 2.2(b) may be bond graphed as shown in
Fig. 2.4(a) where SF represents the prime mover with a prescribed speed. Similarly,
the situation in Fig. 2.3(b) may be represented as shown in Fig. 2.4(b).

Fig. 2.4(a) Fig. 2.4(b)

Once again we state that the receiver of effort information is indicated by a stroke
end and the receiver of flow information is indicated by an open end. Thus for a
CAUSALITY 38
source of flow, SF, the stroke end of the bond should be where element SF is
appended, whereas for a source of effort, SE, the stroke end of the bond should be
where it is appended to a junction.

2.3 Causality of I and C elements

So far what has been established is the causality of sources. Now we discuss the
causalities of I and C elements. The cause should precede the consequence. The
history of cause signal must decide the present value of consequent signal. If we
examine the mechanics of a mass point in this light by considering the impulse-
momentum equation, then we have

t 1 t
p  -
e dt or f 
m  -
e dt , (2.11)

Where p is momentum, e is force (effort), f is velocity (flow) and m is mass (inertia) of


the particle. In the above relation the integral is the history functional of Galilean
causality and hence we can say that force is the cause and velocity (hence
momentum) is the consequence.

Current in an inductor is given by

1 t
i
L  -
e dt , (2.12)

Where e is voltage, L is the inductance (inertia) and i is the current. If we term the
product Li as electrical momentum then

t
p  Li   -
e dt . (2.13)

From this integral we can say that voltage is the cause and current (hence electrical
momentum) is the consequence.

Thus for inertia or I elements the cause is effort (force, voltage etc.) and the
consequence is flow (velocity, current etc.).

For a spring of stiffness K, the deformation at any instant is given by

t t
Q  -
f dt and force e  K
-
f dt.
(2.14)
39 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Similarly for a capacitor of capacitance C, the charge accumulated on the plates is


given by
t 1 t
Q  -
i dt and voltage e
C  -
i dt . (2.15)

Therefore, for compliant or C elements the cause is flow (velocity, current etc.) and
the consequence is displacement (deformation, charge etc.) and hence effort (force,
voltage etc.).

In compliance with Galilean causality, I and C elements representing mass point,


inductor, spring, capacitor etc. should get the information of respective causes and
should give back the corresponding consequences to the junction to which these are
attached.

As per the above discussion, the proper causality for inertia and compliant elements
are shown in Fig. 2.5(a). The proper causality for an element is called integral causality.
Some times the causal stroke will have to be inverted i.e., the I element acquires
stroke at the junction end of the bond and C element at the element end of the bond.
Such a causal orientation is called improper or differential causality. Implication of such
inversion will be discussed when we encounter them at a later stage in this chapter.

Fig. 2.5(a)

For I and C - fields or multiport elements (discussed in Chapter - 3), the integral
causalities are as shown in Fig. 2.5(b).

Fig. 2.5(b)

These elements are called storage elements. What do they store? The idea based on
which the first impression of bond graph is created, is the idea of exchange of energy
as explained in Chapter - 1. Thus it is very natural to uphold I and C elements (and
fields) as storers of energy. However, in the light of above discussion it becomes
apparent that these elements (and fields) are primarily storers of the information
CAUSALITY 40
(cause) they receive. In Chapter - 3 where generation of system equations from a
bond graph model is discussed, the stored or integrated cause is treated as a state of
the system. A generalized view point is that these elements (and fields) are storers
of states of the system. In later chapters you will find that not all storers of states are
storers of energy. These elements (and fields) become storers of energy as well when
kinetic energy is associated with I elements (and fields) and potential energy is
associated with C elements (and fields). Their being storers of energy is not an
essential requirement in bond graph theory.

2.4 Causality of R element

The constitutive law of a resistive element is such that integration of any power
vriable does not occur. For a linear resistor the constitutive law is either

1
eR f or f  e.
R

Nonlinear generalization is

e    f ,t  or f    e, t  ,

where φ and  are single valued functions of their arguments. Hence there is no well
defined causal structure associated with a resistor. A resistor, unlike inertia or
compliant elements does not retain the system energetically closed. It either rejects or
imports energy from outside the system boundaries and unlike sources its behavior
depends on dynamical state of the system.

To arrive at the causal structure of the resistive element let us first negate the
resistive element itself. If a model of the whole universe which does not have
anything remaining out of it could be created, then that model would not have any
resistive element. If the model exhibits dynamical behavior abiding by the notion of
causation, then it should consist of only storage elements, junctions (0 and 1) and
two-port elements. The model of a finite system which we create is a part of that
universe. We incorporate exchange of power between this finite system and rest of
the universe, unmodelled, through the resistive elements. Their constitutive
relations, though brought to some mathematical forms satisfying requirements of
mechanics and thermodynamics, are essentially empirical in nature. The causal
structure of resistive elements is then not dictated by the system which we model but
by a doctrine that there is a perfectly causalled universe (with no improper
causalities) of which our system model is a part. Thus the causation of the resistive
elements should be such that they do not introduce any improper causality in the
model created. The causalities of these elements should be so assigned that, abiding
41 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

by the causal structure of junctions and two-port elements (to be discussed in the
sequel), instances of differential causality should be removed or atleast minimized.

In the light of the above discussion we argue that resistive elements have flexible
causal structure. They may appear in bond graphs in the junction causalled form as
shown in Fig. 2.6(a) or in the element causalled form as shown in Fig. 2.6(b).

Fig. 2.6(a) Fig. 2.6(b)

A junction causalled R element is said to be in resistive causality. When it is in


element causaled form it is said to be in conductive causality.

It appears from the above argument that our house is in order. It is so to a great
extent. Mechanical, electrical, hydraulic and magnetic domains do not pose any
problem. However, thermal domain does. A thermal resistor does not throw any
heat outside the system boundaries when it connects two bodies at different
temperatures. Through such a resistor heat goes from one body to the other. One
may argue that a thermal resistor in such a situation increases the entropy of the
system and thus it is an importer of entropy. A question may then be raised what is
the nature of the perfect universe which is entropically closed and of which our
model is a part? Such a universe is negated by the second law of thermodynamics
(the statement of Clausius). In bond graph models of thermal systems (discussed in
Chapter - 10) resistors admit only one causal posture. When thermal capacitors are
modeled as compliant elements, the thermal resistors appear only in the conductive
causal orientations. This loss of flexibility of causal orientation implies that thermal
resistors are not truly resistors. They are infact modulated sources of flows (entropy
rates). The reader will appreciate this view point in Chapter - 10 on "Bond graph
modeling of thermal systems".

2.5 Causality of junction elements

The junction elements equate the contemporary values of factors of power, either
flows or efforts, associated with all the bonds appended to them. The 1-junctions
equate the flows and the 0-junctions equate the efforts. There are no history functions
associated with these elements, thus their causal structure cannot be formulated by
Galilean notion of causation. The causal structures of these elements are formulated
from a different stand point.
CAUSALITY 42
The debate regarding the evolution of the picture of the world in man’s mind and the
role of mind in its creation and perpetuation came to its climax with the work of
Kant1 in the 18th century. The ideas which are most relevant to physical sciences as
outfall of this debate, not firmly attributing to any particular philosopher who
contributed to the debate, may be expressed very briefly as follows.

(a) The world picture in the mind of man is an outcome of the data he receives
from the world around him through his sensory organs.

(b) Deductive logic and thus also mathematics are constructs of mind which
seek consistency and order in this picture of the world, eliminating all
anthropomorphic elements. Thus both are children of sense data and mind.

(c) A model of any aspect of world created by man should be analyzable and
its behavior, as time flows, should be predictable by logic and mathematics.

It is the third component as expressed above which dictates the formulation of causal
structures of junction elements.

A 1-junction equates the flows


associated with the bonds appended to
it. Thus one and only one bond should
bring the information of flow which
then should be distributed to other
bonds. Thus the causal structure of a 1-
junction should be as shown in Fig.
2.7(a).

If more than one bond brings the


information of flow to 1-junction it will
Fig. 2.7(a).

1
Will Durant says that Kant is the last author whom one should read on Kant. One must
approach Immanuel Kant through works of others on him. We suggest "The story of
philosophy” by Will Durant and “History of Western Philosophy” by B. Russel, to start with.
43 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

lead to a conflicting situation and the


system model will not be analyzable.

We may give a simple argument as


follows.

Say there is a system model with a 1-


junction as shown in Fig. 2.7(b).

Now by taking some of the parameter


values in the rest of the system to
suitable limits the system model may
be brought to a form shown in Fig.
2.7(c). Fig. 2.7(b).

Fig. 2.7(c) Fig. 2.7(d)

Reader may well imagine how can it be done? Here is an answer. The 1-junction as
shown in Fig. 2.7(d) will get disjointed as in Fig. 2.7(e), if m .

Fig. 2.7(e) Fig. 2.7(f) Fig. 2.7(g)


CAUSALITY 44
Likewise a 0-junction shown in Fig. 2.7(f) may get disjointed as shown in Fig. 2.7(g) if,
K  0.

One may show that all other situations


may be also disjointed by taking suitable
parameters to some limiting values. We
leave it to the reader to explore. Now
when a model is brought to a form of Fig.
2.7(c), the disjointed pieces may send
independent values of flows to the bonds
appended to the 1-junction, as one
subsystem would not know what the
other is sending to the 1-junction and the
junction rule would be violated.
Alternatively if the mathematical
expressions for the flows on these bonds Fig. 2.8
are derived and equated and we proceed to analyse the system it will amount to
equating the flow value of one of the bonds to that of other bonds. This argument is
allegoric in nature far from being rigorous. More concrete arguments based on the
relation between bond graphs and linear graphs may be found in a paper by Martens
and Orts [46].

Now for a 0-junction, we may argue in similar lines that only one bond should bring
the information of effort and it should be distributed to other bonds by the junction
as shown in Fig. 2.8.

In the language of bond graphs we may say that,

(a) The causality of a 1-junction should be so assigned that one and only one
bond has flow causality towards the junction. All other bonds should be
effort causalled towards the junction,

(b) The causality of a 0-junction should be so assigned that one and only one
bond has effort causality towards the junction. All other bonds should be
flow causalled towards the junction.

A bond which is flow causalled towards a 1-junction is called the strong bond of that
1-junction. Likewise a bond which is effort causalled towards a 0-junction is called the
strong bond of that 0-junction. Each junction element in a bond graph model should
have one and only one strong bond.

2.6 Causality of two-port elements


45 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Let us consider a transformer first. There are two power ports or bonds connected to
a transformer. The flow in one of the bonds is proportional to the flow in the other
bond and likewise effort in one of the bonds is proportional to that in other bond.
Thus causal orientations of a transformer should be either as shown in Fig. 2.9(a) or
as in Fig. 2.9(b) where J1 and J2 stand for any junction type.

Fig. 2.9(a) Fig. 2.9(b)

The gyrator is also a two-port element. Unlike transformer, in a gyrator the effort in
one of its ports or bond is proportional to the flow in the other port or bond.
Likewise the flow in one of the ports of a gyrator is proportional to the effort in the
other. Thus the causal orientation of a gyrator should be either as shown in Fig.
2.10(a) or as in Fig. 2.10(b).

Fig.2.10(a) Fig.2.10(b)

2.7 Differential causality

In section 2.3 we argued that proper causal orientation for I and C elements (and
fields) should be as shown in Fig. 2.5(a) (or Fig. 2.5(b)). These causal orientations are
called integral causalities or integral causal orientations for these storage elements. It was
discussed earlier that with these causal orientations the elements return to the
junctions to which they are appended, signals which are proportional to time
integrations of the signals which their ports or bonds receive from the junctions. In
this sense storage elements with such causal orientations look towards the past i.e.,
record histories of the past values of the signals they receive from the system and
return back a signal which is proportional to these histories.

However, it may so happen that the causal orientation of a storage element is reverse
of the integral orientation. This may happen as the causal structure of the junction
elements does not allow more than one strong bond. Since R elements have flexible
causalities these elements may help in avoiding such situations. But in some models
such reversed causal orientations may become unavoidable without suitable
extension or reduction of the model.
CAUSALITY 46
Reverse of an integral causal orientation on storage elements is called differential
causality or differential causal orientation. A differentially causalled C element is shown
in Fig. 2.11. Similarly a differentially causalled I element is shown in Fig. 2.12.

Fig.2.11 Fig.2.12

Now the question is what is wrong with the differentially causalled elements? Firstly,
the signal returned by the differentially causalled element is time derivative of the
signal received by the element. For the C element of Fig. 2.11 the flow signal returned
(assuming the element is linear) is proportional to the derivative of the effort signal
i.e.,
1 de
f  .
K dt

What is wrong with that? Well, derivative is not a causal operator. If we are at time t,
then derivative of e may be defined (almost for every t barring only such time point
where this derivative is not continuous) in two ways,

de e t   e  t   t  e  t   t   e t 
 lim  lim .
dt  t  0 t  t  0 t

The second form tells us that if the derivative is known then the present value of e
may be obtained from the immediate future value of e. In a sense we are asking
future to decide the present which seriously violates the Galilean notion of causality.
This argument may, however, appear rather abstract. Let us consider a situation
which may indicate what it could imply in physical terms. We consider an ideal case
where a capacitor without any leak is being charged by an ideal cell i.e., a cell
without any internal resistance through wires which are perfect (super) conductors
as shown in Fig. 2.13. The bond graph model of the system is shown in Fig. 2.14.

Fig.2.13 Fig.2.14

In causalling the model we have honored the causal structure of 1- junction and
ended up with a differential causality on the element C2. Let us see if some thing has
gone wrong. The total energy sent out by the cell is
47 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS


Ecell  V  i dt  V Q,
0

where i is the current flowing in the circuit and Q is the total charge accumulated in
the capacitor. If the capacity of the capacitor is C then

Q2 Q
Ecell  as V  .
C C

The energy stored in the capacitor is

1 Q2
Ecap  .
2 C

Now the question is where has the half of the energy gone? The model does not
provide any clue to resolve this. In fact such a model could not be the model of a
physical system. No one can charge a capacitor this way to store all the energy which
comes out of a cell even if there is no resistivity in the wires of the circuit and internal
resistance in the cell. In reality there has to be some resistance in the circuit either in
the wires or inside the cell or both. Such case is shown in Fig. 2.15. The bond graph
model would be as shown in Fig. 2.16.

Fig. 2.15 Fig. 2.16

There is no differential causality. The conductive causality of R3 restores the causality


of the 1-junction. Still the capacitor may be charged to store half the energy sent by
the cell but now we know where the other half has gone.

One may say that a differentially causalled element may upset the energy balance of
the system. Their occurrences indicate incompleteness of the model.

We ask readers to resolve how the energy balances are upset in the following system
models with differential causalities as shown in Fig. 2.17(a), (b) and (c).
CAUSALITY 48

Fig. 2.17(a) Fig. 2.17(b)

Fig. 2.17(c) Fig. 2.18

Let us reconsider the system of Fig. 2.13 and its bond graph model. This time we do
not honor the causal structure of 1-junction but say that the C element ought to be
given integral causality. The bond graph model is shown in Fig. 2.18.

The 1-junction now, does not have any open bond. What does this situation imply?
The implication is that the source of voltage and the capacitor are always in
equilibrium in the sense that potentials across their terminals balance each other. On
the first sight it looks all right. But the model does not give any clue as to how did an
initially uncharged capacitor come up to maintain this equilibrium. The dynamical
process is irresolvable in Kantian sense and thus this model could not be a model of
any physical situation.

Two conclusions may be drawn from the above discussions.

(a) Each junction element having one and only one strong bond is an
uncompromisable requirement in bond graph theory.

(b) Occurrence of a differentially causalled storage element indicates


incompleteness of the model of the system.

2.8 Quintessence

So far we were presenting to the reader the philosophical stand points which lead to
the causal structures of various bond graph elements. Barring the sources, which are
impositions of exogenous signals with inflexibility of causal orientations, the causal
structures of other elements are formulated on the basis of following independent
attitudes.

(a) Cause preceding the consequence - Galilean notion of causality for I and C
elements and fields.
(b) Embedability of the system in a perfect universe - for R elements and fields.
(c) Analyzability of dynamical behavior - Kantian notion of causality for
junction elements.
49 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

The terms Galilean and Kantian are used only to indicate two independent attitudes.
We do not mean to strictly attribute them to these great philosophers.

Now we put the essence of this discussion in Table 2.1 and Table 2.2 for ready
reference of the modeler.

Table 2.1 Causality of bond graph elements


CAUSALITY 50

Table 2.2 Causality of bond graph elements


51 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS
CAUSALITY 52
2.9 Algorithm for assigning causality

A systematic way of assigning causality to a bond graph model was suggested by


Karnopp and Rosenberg [31]. Their algorithm is very effective for manual
assignment of causalities. However some computer programmes use some what
improved algorithms to reduce stages of iterations. The software SYMBOLS follows
basically the Karnopp-Rosenberg algorithm. In the following steps we give the
algorithm as was proposed by them in an extended form.

1. Select any source either SE or SF. The source types have fixed causalities
which cannot be altered. Thus in case of SE give a stroke at the junction end
of the bond and in case of SF give a stroke at the element (i.e., SF) end of the
bond.

2. If the bond newly causalled becomes strong bond of a junction then weak
causalities (i.e., a stroke for a bond on 1-junction and open for a bond on 0-
junction) may immediately be assigned to all other bonds of the junction
which are unassigned.

3. If any newly causalled bond happens to be one of the ports of TF or GY,


then causality may be assigned to the other port in accordance with the
causality structures of these elements.

4. If any of the bonds causalled through step -1 to 3 happens to be the strong


bond on a junction then causality may be immediately assigned to other
bonds of the junction

5. If through step -1 to 4 any bond becomes last but one bond to be causalled
on a junction then the last one should be causalled to be a strong bond on
that junction.

6. If a stage is reached when causality can not propagate further then select the
next source element which is still uncausalled and repeat the steps from 1
through 5.

7. If no other source element is left then select a storage element (I or C) which


is not causalled. Assign to it proper or integral causality. Then repeat steps
from 2 through 5.

8. In the event that the causality cannot be propagated further select the next
uncausalled storage (I or C) element and repeat the steps 2 through 5.
53 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

9. If no storage element is left uncausalled then choose any unassigned R


element and assign it any causality (either resistive or conductive) then
repeat the steps from 2 through 5.

10. In case this leads to change of causality of an already assigned storage


element and brings it to differential causality then reverse the causality of
the R element and again propagate this following the steps from 2 through
5.

11. If the step -10 leads to new occurrences of differential causalities which are
more in number then those after step -10 then retain the earlier one
(assigned in step -9 with propagation). If the present step leads to lesser
number or no differential causality then consolidate this causality. In case
occurrences of differential causalities cannot be avoided by either causalities
of the R element then either the model should be reformulated or the one
with minimum differential causalities should be retained.

12. If the causality cannot be propagated further after steps 9, 10 or 11 then


select the next unassigned R element and repeat the steps from 9 through 11.

13. After the step -12 only some of the internal bonds (bonds joining the
junctions either directly or through two--port elements) may remain
unassigned. Select one such bond and assign arbitrary causality to it. Repeat
the steps from 2 through 5.

14. If the step -13 demands reverting causality of a previously integrally


causalled storage element (I or C) then find out if this situation may be
avoided by reversing causality of a R element. If this can be done then the
causality of that R element should be reversed and propagated.

15. If no such R element in step -14 exists then the causality of the internal bond
should be reversed and step -13 and 14 should be repeated.

16. If causalling the internal bond both ways produces differential causalities
then either system model should be reformulated, or one producing
minimum occurrences of differential causalities should be retained.

17. Now the next uncausalled internal bond should be selected and steps from
13 through 16 should be repeated.

18. Even if every thing goes well after step -17 is completed and all the bonds
get assigned without any occurrences of differential causalities, one should
check for what is known as a causal loop. A causal loop is a set of junction
CAUSALITY 54
elements so connected that they form a close loop and each junction in such
a loop has a strong bond which is an internal bond in this loop. If causal
loop is detected then one should examine if reversal of any R element averts
such a loop without introducing differential causality. When this can be
done the system graph has been successfully causalled otherwise the model
should be reformulated.

2.10 What is wrong with a causal loop?

A causal loop is either irresolvable in the sense that equation of motion of a system
model containing them cannot be arrived at by a procedure outlined in Chapter - 3,
or they have differential causalities disguised as sober integral causalities.

Let us consider a system shown in Fig. 2.19 and its two alternative bond graph
models as shown in Fig. 2.20(a) and (b).

Fig. 2.19
55 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

Fig. 2.20(a) Fig. 2.20(b)

After going through Chapter - 3 readers may attempt to derive system equations for
the model of Fig. 2.20(a) and ascertain its unresolvability. The model of Fig. 2.20(b)
has two differential causalities which were disguised in the model of Fig. 2.20(a).

2.11 Examples on assignment of causality

Some examples of causalling bond graphs are shown in the following figures. A
spring-mass system and its causalled bond graph is shown in Fig. 2.21(a).

System Causalled bond graph

Fig. 2.21(a)

The bond graph corresponding to the system in Fig. 2.21(b) may be causalled in two
alternative ways.
CAUSALITY 56

System One way of causalling Another way of causalling

Fig. 2.21(b)

If the ground of system in Fig. 2.21(b) is excited by an effort instead of a flow source,
the system will have fixed causality as shown in Fig. 2.21(c).

System Causalled bond graph

Fig. 2.21(c)

A system with a transformer and its causalled bond graph is shown in Fig. 2.22(a).
57 MODELING AND SIMULATION OF ENGINEERING SYSTEMS THROUGH BOND GRAPHS

System Causalled bond graph

Fig. 2.22(a)

Figure 2.22(b) shows a gyrocompass and its bond graph.

Gyro Compass Causalled bond graph

Fig. 2.22(b)
CAUSALITY 58

Two different arguments are


used for drawing bond graph
for the system shown in Fig.
2.22(c). We leave it to the
readers to reconstruct the
arguments.

One way of modeling

System Another way of modeling

Fig. 2.22(c)

You might also like