103

Electrical Power Quality and Utilisation, Journal Vol. XII, No. 2, 2006
Extreme Possibilities of Circuital Models
of Electric Machines
Tadeusz J. SOBCZYK
Cracow University of Technology, Poland
Summary: The paper presents possibilities of so-called circuital mathematical models of electric
machines, in which the machine is treated as a set of magnetically coupled coils. It is a very
popular class of models used for analysis of operational conditions of electric machines. However,
the progress in numerical solutions of electromagnetic filed causes the interest of researchers
to shift into so-called field-models. This paper shows the authors opinion on extreme possibilities
of circuital models putting main attention to their creation, simplification and solving.
1. INTRODUCTION
Circuital models are the most popular research tools of
electric machines. Such models are commonly used for
investigation of electric machine properties itself, as well as
for solving operation problems of more complex systems, in
which electric machine is an important element. Usefulness
of such models is out of the question, but nowadays field
models basing on numerical solving of field equations,
become concurrent. The field models have a great advantage
for electric machine design when a shape and dimensions of
a machine electromagnetic circuit have to be found. However,
for prediction of machine properties the advantage of field
models is not so evident.
A very fast development of computer hardware and
software for electromagnetic, mechanic and thermal fields
causes the field models to be used for prediction of electric
machines technical data as well as for modeling operational
problems related to electric machines in drive or in power
system. Generally, the field models have to be solved
numerically, so there is no chance to predict qualitatively
machine properties by such models. Field models are time
consuming and need powerful computers and modern
software, so are rather expensive.
Circuital models treat an electric machine as a set of
magnetically coupled coils located on a stator and rotor and
an elementary mechanical system with constant parameters.
Such models ensure rather a simple dimensionless description
of electric, magnetic and mechanical phenomena in electric
machines. All of them have to be considered together to
model electromechanical energy conversion processes in
electrical machines.
It is commonly accepted that the Lagranges formalism,
which is based on the co-energy function of the whole system,
is the most effective way to model a process of energy
conversion. However, it requires an assumption that the co-
energy stored a machine magnetic circuit is a unique function
of currents and a mechanical angle. Interpreting this
assumption physically means that the currents paths have
to be well determined and characteristics of magnetic material
have to be unique, too. Making above assumptions has strong
physical consequencesconducting continuous areas and
ferromagnetic materials with a hysteresis loop is not accepted
theoretically in such models. Usually, conducting areas are
discretized enlarging a number of currents.
The co-energy function is determined by characteristics
of coils creating an electric machine. The coil characteristics
determine the coils linked fluxes as function of all currents
i
1
,..., i
N
and a mechanical angle j of a stator and rotor mutual
position:
1 N
( )
n n
i ,...,i , y y j for 1 N) n { ,..., (1.1)
The co-energy function is given by the formula:
N
2
1
1
0
1
( ) J ( 0 0 )d
2
n
i
om n n n
n
E , i ,...,i' , ,..., , i' j w y j

i
(1.2)
Leaving detailed considerations well known in literature,
the equations of circuital models of electric machines take
the general form:
d
( )
dt
, Y j - i R i u
(1.3a)
2
2
( ) d
J
dt
om
m
E ,
T
j j
j

i
(1.3b)
where Y(i, j), i(t), u(t) are vectors of linked fluxes, currents
and voltages respectively and R

is a square resistance matrix.
This paper presents advantages and limitations of circuital
approach to modelling of electric machines and focuses on
extreme possibilities of such models.
2. LINEAR CIRCUITAL MODELS
Circuital models obtained under the assumption on
linearity of machine magnetic circuit are especially attractive
due to their simplicity. Because of that they become a base
for modelling electromagnetic phenomena in electric
machines. Linearity of magnetic circuit leads to linear
relationships between linked fluxes and currents for all coils,
which is mathematically expressed by the formula:
Key words:
aa..
Tadeusz J. Sobczyk: Extreme Possibilities of Circuital Models of Electric Machines
104 Power Quality and Utilization, Journal Vol. XII, No 2, 2006
( ) ( ) , Y j j i L i (2.1)
and the co-energy function is given by the quadratic form of
currents:
1
2
( ) ( )
T
om
E ,j j i i L i (2.2)
Equations of machine models take the form:
( )
d
( )
dt
j - L i R i u (2.3a)
2
1
2 2
d ( )
J
dt
T
m
T
j j
j

L
i i (2.3b)
in which the most important is the inductance matrix

L(j).
Self and mutual inductances of all coils, depending on design
data of machine windings and geometry of machine magnetic
circuit, substitute it. The inductance matrix is symmetric
(L(j) = L
T
(j)), positively determined (det L(j) > 0) and
periodic with respect to the angle j (L(j) = L(j + 2p)). Each
of these mathematical properties has a physical background:
independence of stored energy from the way it has been
accumulated, no ideal magnetic coupling of coils representing
machine windings and periodicity of phenomena in the
machine with respect to the rotor position angle.
Determination of all inductances as functions of the angle
j is an essential problem. Usually two groups of coils
representing stator and rotor windings are distinguished and
the inductance matrix is written in the form:
( )
m m
ss ss rs
m m
rs rr rr
( ) ( )
( )
( ) ( )
T
j j
j
j j
(
-
(

(
-
(

L L L
L
L L L
(2.4)
where:

ss
L is a leakage inductances matrix;
m
ss
( ) j L is the main inductances matrix of stator coils;

rr
L is a leakage inductances matrix;
m
rr
( ) j L is the main inductances matrix of rotor coils;
m
rs
( ) j L is a mutual inductances matrix of stator and
rotor coils.
By means of this methodology it is possible to create a
very detailed model, in which each individual coil, both on a
stator and a rotor sides, is treated as an independent one. Let
such a set of coils be located in a cylindrical magnetic circuit
with a non-smooth air gap both on a stator and a rotor sides.
Such a model can be considered a base model and any
individual model for a given machine can be obtained from it.
It is relatively easy to find the inductance matrix for such a
basic model. Let M arbitrary coils, which have magnetic axes
on angular positions a
1
, a
2
, ..., a
M
with respect to a stator
reference axes, be located on the stator and N arbitrary coils,
which have magnetic axes on angular positions b
1
, b
2
, ..., b
N
with respect to a rotor reference axes, be located on the rotor
side. Let each coil produce the MMF with full Fourier
spectrum. In such a case all self and mutual inductances
depend on three variables and are periodic with respect to
each of them [11] and can be expanded onto triples Fourier
series. For self and mutual inductances of stator coils these
are: a location angle of magnetic axes of individual stator coil
a
m
, an angle between magnetic axes of two considered stator
coils a
m
a
k
and a rotor position angle j. For self and mutual
inductances of rotor coils as these three variables can be
chosen: a location angle of magnetic axes of individual rotor
coil with respect to stator reference frame j + b
n
, an angle
between magnetic axes of two considered rotor coils b
n
b
k
and the rotor position angle j. For mutual inductances
between stator and rotor coils the following three variables
can be used: a location angle of magnetic axes of individual
rotor coil with respect to a stator reference frame j + b
n
, an
angle between magnetic axes of two considered rotor coils
one on a stator and one on a rotor j + b
n
a
m
and a rotor
position angle j. The general forms of inductance matrices
L
m
ss
(j), L
m
rr
(j)

and L
m
rs
(j) are given in Appendix 1. In all
above mentioned matrices there appear triples sums in which
summing up index v depends on harmonics orders of coils
MMFs. Indexes r, s depend on harmonic orders of a
permeance function representing geometry of air-gap zone,
which, for a non-smooth air gap on stator and rotor sides,
can be written in the form of double Fourier series:
j j
( ) e e
rx s
r ,s
r s
x,
j
l j L

(2.5)
Leakage inductance matrices L
s
ss
and L
s
rr
can be also
very carefully analysed analyzed accounting for the
placement of individual coils in the slots and in the coil
outhangs.
This base model is very detailed and rather complicated
but it just shows the possibilities of linear circuital models of
electric machines. It requires a lot of parameters, mainly
inductances. Usually, the inductances for linear models are
calculated from a very simple analysis of magnetic field
distribution in the air gap. It is one-dimensional analysis, in
which a radial component of flux density in the air gap is
determined as a function of angular position along
circumference. There are many formulas in literature
determining the self and mutual inductances of an arbitrary
pair of coils located in the air gap zone [2, 9, 10, 11]. Coils
MMF functions and the permenace function of the air gap
are needed for that. For example, in [11] such a general formula
is given, but its presentation and interpretation would be too
long. The inductance for a linear base model could be
calculated by numerical solution of field equations in a
machine magnetic circuit, assuming linearity of magnetic
material. Nowadays computers and magnetic field solvers
allow doing it very fast and effectively. While computing
inductances it should be kept in mind that in any electric
machine a set of coils is specially arranged, which leads to a
specific structure of inductance matrices. It can happen that
a distortion of the matrix structure can lead to more serious
mistakes than a little bit less precise calculation of inductances
with the matrix structure.
In order to explain this statement, let us consider the
symmetrically located sets of identical coils on a stator and
analogously on a rotor. For such a case it is enough to know
105
three functions. The matrix L
m
ss
(j) is determined by one
function of three variables mentioned above

L
s
(a
m
a
k
, a
m
,
j), for the matrix L
m
rr
(j) it is enough to know the function
L
r
(b
n
b
k
, j + b
n
, j) and the matrix L
m
rs
(j) is determined by
the function L
rs
(j + b
n
a
m
, j + b
n
, j). As it has already
been written, all these functions are periodic with respect to
each of their variables. The structures of these matrices are
given in Appendix 2.
From a base model (operating with individual coils in a
cylindrical magnetic circuit with a non-smooth air gap on
stator and rotor sides) models of any individual machine
having some specific features of magnetic circuit and coils
arrangement can be obtained. For a magnetic circuit with a
non-smooth air gap on a stator or a rotor side inductances
become functions of two variables and for a smooth air-gap
inductances they depend on one variable only an angle
between magnetic axes of coils. The numbers of summings
up in the matrices shown in Appendix 1 are reduced
respectively. In an electric machine the elementary coils are
connected in multi-turn coils located in one slot, which are
then grouped, and the groups constitute phases. The way of
coils connections is mathematically expressed by a Kron
constrain matrix. By means of Kron constrain matrices every
internal asymmetry in a machine, both designed as well arisen
a result of a fault can be mathematically expressed. Denoting
Kron constrain matrices of stator and rotor coils by C
s
and
C
r
respectively, the inductance matrix of a machine takes the
form:
( )
m m
s s ss ss rs
m m
r
r
rs rr rr
( ) ( )
( )
( ) ( )
T
T
T
j j
j
j j
(
(
- (
(
(
(
(
(
-
(

C
C 0 C 0 L L L
L
0 C
0 C
L L L
(2.6)
Introducing Kron matrices a machine is described by
independent currents, so the mathematical description of a
machine is unique. However, a unified approach starting from
a base model cannot be the fastest way to a machine model
for each case. Because linear models can be created very
easily, an individual model should use, depending on
requirements, directly group of coils or a whole phase, instead
of elementary coils.
Any model has to be effectively solved, which is a separate
problem for electric machine models. Commercial software
packages like MatLab, MatCad, Mathematica and others are
very well prepared for solving high order differential equations,
even with complex coefficients, which in fact winds up the
problem. Thus, equations created above can be numerically
solved together with equations of co-operating systems:
mechanical, power electronics, etc. However, they are rather
complicated and any of their simplification is profitable and
leads to reduction of computing time, does not require a very
advanced hardware and reduces the time necessary for
preparation and performance of research. Simplification of
electrical model equations is important from the cognitive point
of view, facilitating physical interpretation of phenomena inside
a machine and properties of model solutions.
An electric machine at circuital approach is treated as a
set of coils, so for simplification of its model equations, a
system analysis methods can be applied. The most popular
and a very effective way is a transformation of variables,
which leads to decomposition of machine equations into
independent sub-systems, decreasing order of equations,
which have to be solved together. Application of
symmetrical components of currents and voltages both on
stator as well as rotor sides, which are defined by linear
transformation with a matrix of the general form, is especially
effective:
2 (K 1)
2 4 2(K 1)
(K 1) 2(K 1) (K 1)(K 1)
1 1 1 1
1 c c c
1
1 c c c
K
1 c c c

(
(
(
(

(
(
(
(

S
L
L
L
M M M O M
L
,
2
j
K
c e
(2.7)
The inductance matrix transformed into symmetrical
components is Hermitian and takes the form:
( )
s s
r
r
m m
ss ss rs
m m
rs rr rr
1
s s
1
r
r
( )
( ) ( )
( ) ( )
T
T
T
j
j j
j j

(
(
(
(
(

(
-
(

(
-
(

(
(
(
(
(

S
S 0 C 0
L
0 S
0 C
L L L
L L L
C 0 S 0
0 C
0 S
(2.8)
Description by symmetrical components puts in order
machine equations and every symmetry of magnetic circuit
and of coils arrangement reflects on their structure so
description by symmetrical components is specially effective
for fully symmetrical machines.
An important simplification of machine equation can be
obtained using the other transformation, which eliminates
dependence of the inductance matrix L
s
(j) on the rotor
position angle. Matrices of such transformations depend on
that angle and they turn the previously defined symmetrical
components on the complex plain. Unfortunately, not for every
case these matrixes can be found. Leaving the details, general
conditions for the structure of the inductance matrix L
s
(j)
allowing to eliminate its angle dependence are given below.
Firstly, the L
s
(j) matrix should be approximated by the Fourier
series with a limited number of terms:
K
j
K
( ) e
k
k
k
j
j

S
L L
(2.9)
choosing harmonic orders so as to have at most one term
L
m
e
jmj
of this series in any matrix element. These
transformation matrices have the general form:
Tadeusz J. Sobczyk: Extreme Possibilities of Circuital Models of Electric Machines
106 Power Quality and Utilization, Journal Vol. XII, No 2, 2006
s,1 s,M
jk jk
s
( ) diag e e
j j
L j
(

L
(2.10a)
r,1 r,N
jk jk
r
( ) diag e e
j j
L j
(

L
(2.10b)
where M and N are respective dimensions of stator and rotor
quantities. Secondly, the equations (2.11) should be checked.
When an element of the matrix L
s
(j) does not contain any
term L
m
e
jmj
a respective equation should be removed from
the equations (2.11). If these equations can be fulfilled, a set
of numbers {k
s,1
,...,k
s,M
} and {k
r,1
,...,k
r,n
} is not unique, if
not the Fourier series (2.9) the number of considered harmonic
should be limited or the orders of harmonic taken into account
in the inductance matrix should be changed.
1,2
1,3
1,M
1,M 1
1,M 2
1,M N
2,3
2,4
2,M
2,M 1
2,M 2
2,M N
M N-1,N M
m
m
m
m
1 1 0 0
m
m
m
m
m
m
m
m
m
-
-
-
-
-
-
- -
( (
( (
(
(
( (
( (
( (
( (

( (
( (
(
(
( (
( (

(
( (
( (
( (
( (
( (
( (
( (
( (
( (
( (
( (
( (

(
(
(
(
( (
M
L
M
M
M
M
[ ]
s,1
s,2
s,M
r,1
r,2
r,N
0
k 1 0 1 0 0
k
1 0 0 0 1
k
0 1 1 0 0 0
0 1 0 1 0 0 k
k
0 1 0 0 0 1
k
0 0 0 0 1 1
( (
(
(
( (
( (
( (
( (
( (
(
(
( (
(
(
( (
(
( ( (
(
(
(
(

(
(
(

(
(
(

(
(
( (
(
(

(
(
(

(

L
L L
M M M M M
M L L
L
L
M M M M M M
L M
M
L
(
(
(
(
(
(
As a result of transformations of matrices depending on the
angle (2.10) a new inductance matrix, Hermitian also, but with
constant elements, is obtained:
( )
s s s
r r
r
m m
ss ss rs
m m
rs rr rr
1
s s s
1
r
s r
( )
( )
( ) ( )
( ) ( )
( ( ) )
( ( ) )
T
T
T
L j
L j
j j
j j
L j
L j

(
(
(
(

(

(
-
(

(
-
(

(
(
(
(
(

N
S 0 C 0
L
0 S
0 C
L L L
L L L
S 0 C 0
0 C
0 S
(2.12)
and model equations take the form:
d
j ( )
dt
-
N
N N N N N N
i
L K L i R i u (2.13a)
( )
2
2
d
J j ( )
dt
*
T
m
T
j
-
N N N N
i L K KL i (2.13b)
in which by K a diagonal matrix of the form is denoted:
s,1 s,M r,1 r,N
diag k k k k (

K L L (2.14)
The equations in the form (2.13) are effortless for numerical
integration and essentially simplify an analysis of steady
state at a constant angular velocity.
3. MODELS ACCOUNTING FOR MAGNETIC
NON-LINEARITY
Non-linearity of magnetic circuit highly complicates the
electric machine equations. The equations (1.3) seem to be
simple, but in fact they are rather complicated because of
non-linear coil characteristics, i.e. relations between linked
fluxes and currents. A set of functions describing those
characteristics:
1 N
( ) for n {1 N}
n
i ,...,i , , ,..., y j (3.1)
has to insure uniqueness of the co-energy function, so the
functions (3.1) have to fulfill the conditions:
n k
k n
i i
y y

for {1 N} n,k ,..., (3.2)
For sets having numerous coils a determination and
adjustment of characteristic description are very difficult
problems, practically unsolved till today. It is very difficult to
find relations between characteristics like those for
inductances in the matrices (2.6). Nowadays it is possible to
describe analytically the characteristics for a set of two or
three coils. Undoubtedly, an analytical description of
characteristics for a set of non-linear coils is an open problem,
which has to be solved for a further development of electric
machines circuital models.
An alternative is determination of the co-energy as a
function of all currents and a rotor position angle E
om
(i
1
,...,
i
N
, j), from which the coil characteristics can be found:
1 N
1 N
( )
( )
om
n
n
E i ,...,i ,
i ,...,i ,
i
j
y j

(3.4)
Remembering that the values of co-energy can be
calculated numerically only, by solving magnetic field
equations, the co-energy function can be only approximated.
When a number of independent variables (mainly currents)
is large the co-energy function can be approximated by a
(2.11)
107
multivariable power series, which requires a lot of coefficients.
This method, quite evident, has not been practically used till
now in circuital models yet.
However, this approach can be simplified by looking for
qualitative properties of the co-energy function, keeping in
mind that the state of electric machine magnetic circuit
depends on all coils. A heuristic determination of the co-
energy as a function of the total MMFs of all machine
windings is very effective. It is possible when leakage fluxes
are considered separately, which is a strong assumption, but
gives a chance to introduce magnetic non-linearity of the
main magnetic circuit into a circuital model by not a very
complicated way. Thus, in the co-energy function two terms
are indicated: one representing the co-energy in leakage
zones and the other representing the co-energy in the main
magnetic circuit:
1 N 1 N 1 N
( ) ( ) ( )
om o og
E i ,...,i , E i ,...,i E i ,...,i ,
s
j j -
(3.5)
Magnetic circuits of leakage fluxes are very often locally
separated zones and their magnetic state is determined by
one or two coil currents. To reduce a list of variables of the
second term in (3.5) harmonic equivalent magnetizing
currents [11], which in fact represent a sum of MMFs of all
machine windings for individual space harmonic, can be
introduced. A very spectacular effect can be obtained
considering the main space harmonic only. In this case co-
energy depends on the parameters of the total MMF of this
harmonic, which are its magnitude and angular position
along circumference. The MMF magnitude is represented
by an equivalent magnetic current, which depends on
positions and currents of all machine windings and also on
a rotor position. Considering higher space harmonic, each
of them add a new pair of variables: a respective equivalent
current and an angle. Remarks mentioned above allow to
change a list of the other term of the co-energy function
E
og
(i
1
,..., i
N
, j):
1 N ,k k ,l l
( ) ( k l )
og og
E i ,...,i , E ,i , ,i , , j h h L L
(3.6)
where {...,k, l,...} denote a set of harmonic orders taken into
account. In practice, only MMF space harmonic with highest
amplitudes decide on saturation of the main magnetic circuit,
so a number of variables of the co-energy function is very
limited.
The total MMF of all machine windings, expressed by
equivalent harmonic magnetizing currents i
m,r
and respective
angles rh
r
has the form:
( )

1
( )
2
( ) ( )

,
i t
x,t cos x t
r
r
r
r
r

(3.7)
A square of the equivalent magnetizing current of r
harmonic is given by a quadratic form:
( )
N N
2

1 1
(w k ) (w k ) ( )
, n n, n k k , k n k
n k
i i i cos
r r r
r


(3.8)
where:
w
n
is a coil turn number,
k
n,r
is a winding factor,
a
1
,..., a
N
are the position angles of magnetic axes of all
windings.
The angle rh
r
is determined by the formula:
( )
N
1
N
1
(w k ) ( )

(w k ) ( )
n n, n n
n

n n, n n
n
i sin
tan
i cos
r
r
r
r

(3.9)
Introducing new variables into the co-energy function, the
Lagrange equations of electrical machines with a non-linear
magnetic circuit take the form:
og og

( )
d d
R
d d ( )
,
o
n n n
n , n n
E i E
E
u i
t i t i i i
r r
r r r
rh
rh

- -
(
(
(

,
,

for {1 N} n ,...,
(3.10a)
2
og og
m
2

( )
d
J
( )
d
,
,
E i E
T
i
t
r r
r r r
rh
j
j rh j

- -
(
(

,
(3.10b)
Appendix 2 presents their detailed structure. It should be
noticed that the matrices in these equations follow from
localization of windings along circumference of the air gap,
so the application of symmetrical components can simplify
the matrix structure, analogously to linear models.
In simpler cases qualitative properties of the co-energy
function E
og
follow from engineering intuition. For an electric
machine with mono-harmonic windings of p order and a
smooth air gap the co-energy function can be predicted as
one variable odd function of the equivalent magnetizing
current for p harmonic i
m,p
:
og ,p 0 ,p
( p ) ( ) E i , E i h
(3.11)
For mono-harmonic salient pole machines with poles on a
stator or a rotor side, the co-energy function should be
predicted as two variables function: an odd function of the
equivalent magnetizing current for p harmonic i
m,p
and a
periodic function of an angle

ph, calculated with respect to
the axis of symmetry of a stator or a rotor side, respectively:
og ,p 0 ,p 2 ,p
( p ) ( ) ( ) 2p E i , E i E i cos h h - -L
(3.12)
For machines having salient poles on both sides and
mono-harmonic windings, the co-energy function depends
on three variables: as before on the equivalent magnetizing
current for p harmonic i
m,p
and on two angles ph and
j ph, when the angle ph is measured with respect to the
symmetry axis of a stator side or on angles ph and

j + ph,
when the angle ph is measured with respect to symmetry
axis of a rotor side. It can be predicted in the forms:
Tadeusz J. Sobczyk: Extreme Possibilities of Circuital Models of Electric Machines
108 Power Quality and Utilization, Journal Vol. XII, No 2, 2006
og ,p 0 ,p 2 2 ,p
( p ) ( ) ( )
2 p 2 ( p )
r , s
r s
E i , E i E i
cos r cos s
h
h j h
-

(3.13a)
or:
og ,p 0 ,p 2 2 ,p
( ) ( ) ( )
2 p 2 ( p )
r , s
r s
E i , p E i E i
cos r cos s
h
h j h
-
-

(3.13b)
The co-energy function can be qualitatively predicted also
for some cases when higher space harmonic should be taken
into account. For instance, the co-energy function for a
machine with a smooth air gap considering MMFs of orders
p and 3p can be predicted in the form of Fourier series:
og 3p ,p ,3p p 3p
0
( ) 3p( )
k
k
E E i ,i cos k h h

(3.14)
Accounting for space harmonic p and 5p the co-energy
function is of the form:
og 5p ,p ,5p p 5p
0
( ) 5p( )
k
k
E E i ,i cos k h h

(3.15)
However, a heuristic qualitative prediction of co-energy
function properties becomes very difficult or impossible at
all when more than two space harmonic should be considered
or geometry of a magnetic circuit loses the symmetry.
A quantitative determination of the co-energy functions,
heuristically predicted, can be done in two steps. First, the
co-energy values should be found from a numerical
computation of magnetic field equations in a machine
magnetic circuit for properly chosen set of winding currents
and rotor positions. In the second step, the equivalent
magnetizing currents and respective angles should be
calculated for the same set of winding currents and rotor
positions.
The approach basing on properties of the co-energy
function is not commonly used yet, but it is very profitable
because it allows to introduce magnetic non-linearity into
the classical circuital model of electric machines. It is
relatively simple and has a physical background by using an
equivalent magnetizing currents. It should be noticed that
this methodology could be used to create linear models of
electric machines. Assuming magnetic linearity, the co-energy
function can be predicted as a quadratic form of equivalent
magnetizing currents and obtained models do not use
inductances at all.
Consideration of magnetic non-linearity in circuital models
of electric machines has its own history and a lot of papers
have been devoted to this problem. Most often the parameters
of linear models are subjected to respective currents, which
in many cases leads to sufficient accuracy of models solutions
with measurements carried out on real machines. However,
that approach is not general enough and parameters have to
be chosen for an individual problem. An approach
introducing into d-q models so called cross saturation effect
is also rather popular. It gives satisfactory results from the
technical point of view, but very often the conditions (3.2)
which are fundamental for proper modelling of energy
conversion, are not fulfilled.
4. SOLVING OF ELECTRIC MACHINE MODELS
Generally, circuital models of electric machines are
described by ordinary non-linear differential equations. This
non-linearity has two sources. The first one is a magnetic
non-linearity of the machine magnetic circuit, which up to
now has been made from ferromagnetic materials. The other
one follows from the phenomenon of electromechanical
energy conversion. To convert electrical energy into
mechanical, the energy stored in the magnetic field of an
electric machine has to depend on some mechanical variable.
But this energy is a quadratic function of a current which
creates the structural non-linearity of electric machine
equations. For more than one hundred years history of electric
machine development, solution of full models, including a
motion equation has been a serious problem. Engineers
simplified the circuital models in order to apply methods of
classical circuit theory. First of all, electrical and mechanical
equations were separated, solving electrical equations under
the assumption that mechanical motion is known, and for
solving mechanical equation, it was assumed that all
currents are known. It means that electrical and mechanical
phenomena were treated separately. Each part of equations
was linearized, mainly assuming linearity of magnetic
materials, and even solved analytically. In last decades the
problem of solving has practically disappeared thanks to
the development of computers and specialized software.
Nowadays, there are a lot of commercial software packages
which can solve very large sets of non-linear differential
equations at a time acceptable by engineers. This statement
could be used to finish further considerations on top
extreme possibilities of solving methods for electric
machine models.
However, circuital models give a chance for quantitative
analysis of solutions, i.e. they allow to predict same
properties of electrical machines without calculations in
numbers. Considering the equations in a general form shown
in Appendix 3, it should be stated that mathematics does
not give any possibilities to find their solutions, and,
therefore, the only problem to be raised is the problem of
qualitative analysis of solutions of the electrical equations
in the form:
( )
n
d
( )
dt
, j - L i i R i u (4.1)
For periodic voltages:
e
j
e
( ) ( T ) e
k t
k
l
t t

u u u , where: e
e
2
T

(4.2)
109
assuming also that a rotor angular velocity is periodic also
with apriori known period:
m
j
m
( ) ( T ) e
l t
l
l
t t w w w

, where:

m
m
2
T

(4.3)
the steady state solution of (4.1) can be foreseen in the form
of double Fourier series:
e m
j( )
e
k l t
k ,l
k l

-



i I

(4.4)
The Fouriers coefficients have to fulfill an infinite algebraic
equation set, balancing the series on both sides of the
equation (4.1):
U = (R + j

9L
n
(I)) * I (4.5)
These equations are non-linear because the elements of
the L
n
(j, i)

matrix depend on a solution which is being looked
for. Analyzing carefully, the structure of this set the current
can be determined qualitatively. Basing on this equation, the
current can be also determined quantitatively. Therefore,
equations (4.1) should be solved iteratively. The simplest
procedure consists in a direct iterative algorithm solving in
successive iterations a set of linear equations, but limited to
finite dimensions:
U = (R + j

9L
n
(I
i
)) * I
i+1
(4.6)
For arrangement of Fourier coefficients I
k,l
of the current
vector, according to the notation given below:
1 0 1
T

(

I I I L L I

1 0 1
T
k k , k , k ,
(

I I I L L I
(4.7)
the matrices in the equation (4.6) take the forms:
[ ]
[ ]
1 0 1
= diag
= diag
= diag W W W

(

R R R
R R R
L L
L L
L L
R
R
9
(4.8)
e m e e m
= diag
k
k k k - (

L L 9 (4.9)
0 1 2
n 1 0 1
2 1 0
( ) =


(
(
(
(
(
(
(

L L L
L L L
L L L
M M M
L L
L L
L L
M M M
L I
(4.10)
0 1 2
1 0 1
2 1 0
k , k , k,
k k , k , k ,
k , k, k ,

(
(
(
(
(
(
(

L L L
L L L
L L L
M M M
L L
L L
L L
M M M
L
(4.10)
where L
k,l
are the Fourier coefficients of the matrix L
n
(j, i):
e m
j( )
n
( ) e
k l t
k ,l
k l
, j

-



L i L
(4.11)
substituting in it the time function of the angle and time
functions of currents obtained in a previous iteration. A vector
of voltages U in the detailed notation takes the from:
1 0 1
=
T
T
k k

(

(

U U U
0 u 0
L L
L L
U
U
(4.12)
For solving non-linear equation set (4.5) Newton-Raphson
algorithm can be also used. It requires to solve in the
successive iteration the equation set of the form
U + j9(L
d
(I
i
) L
n
(I
i
)) * I
i
= (R + j

9L
d
(I
i
)) * I
i+1
(4.13)
In this equation there appears the matrix L
d
(I), which is
created analogously to the matrix L
n
(I), but on the basis of
the Fourier series of a dynamic inductance matrix:
e m
j( ) d
d
( )
( ) e

k l t
k ,l
k l
,
,
Y j
j

-

i
L i L
i
(4.14)
where

n
( ) ( ) , , Y j j i L i i . The Newton-Raphson algorithm
is much faster converged.
To solve both the equations (4.6) as well as the equation
(4.13), the Fourier coefficients of the Fourier series (4.11) of
matrix L
n
(j, i) and (4.14) of the matrix L
d
(j, i) have to be
calculated in each iteration. Elements of these matrices are
non-linear functions of currents and the rotation angle, so
calculation of those Fourier coefficients have to be done in
three steps. First, the values of all currents and the angle are
calculated from the Fourier series obtained in previous
iteration for a set of chosen time instant. Next, the time
functions of all elements of the matrices L
n
(j, i) and

L
d
(j, i),
are calculated in the same set of time instant, and finally, new
values of the Fourier coefficients of the series (4.11) and
(4.14), required in the equations (4.5) or (4.13) can be found.
The electromagnetic torque of an electric machine can be
obtained from the general formula presented in Appendix 3.
It is described by a quadratic form:
em
( )
T
T , j i W i i
(4.15)
The matrix of this form depends on the currents and the
rotation angle too, so it can be written in the form of double
series, like the matrices L
n
(j, i) and

L
d
(j, i):
Tadeusz J. Sobczyk: Extreme Possibilities of Circuital Models of Electric Machines
110 Power Quality and Utilization, Journal Vol. XII, No 2, 2006
e m
j( )
( ) e
k l t
k ,l
k l
, j

-



W i W
(4.16)
The formula for electromagnetic torque can be presented
in the double Fourier series:
e m
j( )
em
e
k l t
k ,l
k l
T

-



T
(4.17)
and its coefficients can be written as the quadratic forms of
the Fourier coefficients of currents [9, 11].
The method of steady state analysis presented above,
has really extreme possibilities, but in a general case is rather
complicated. For simpler cases, i.e. for models of machines
having same internal symmetry, this method is very effective
and allows to predict qualitative properties of both currents
and the electromagnetic torque, and it also allows calculating
their values. Such an analysis is more profound than a
determination of machine properties from numerical solutions
of differential equations. However, such a tendency is
nowadays observed.
5. CONCLUSIONS
The ideas presented in the paper reflect the authors
opinions on extreme possibilities of circuital models of electric
machines. For more than one hundred year history of
development of electric machines these models have been
dominant and the knowledge about them is enormous.
Therefore, the opinions presented in the paper have to be
selective. They have become limited to the critical problems
of mathematical modeling of electric machines: generating of
equations of linear models with the highest degree of
complexity, simplification of these equations, generating of
equations which take into account magnetic non-linearity,
and the most general method of predicting steady state
solutions.
The increasing role of field models makes circuital models
be confronted with new tasks, which involves their close
relations to numerical analysis of electromagnetic field in a
machine. Circuital models, however, should not be
considered to be in competition with field models, but rather
as their essential supplement. There is no doubt that circuital
models of electric machines will continue for a long time to be
the basic tool to investigate phenomena in machines,
especially, to research their performance with more and more
complex co-operating systems.
REFERENCES
1. Adkins B., Harley R.G.: The General Theory of
Alternating Current Machines. London, Chapman Hall, 1975.
2. Concordia C.H.: Synchronous Machines, Theory and
Performance. J. Wiley & Sons Inc., New York, 1951.
3. Kovacs K.P., Racz I.: Transiente Vorgange in
Wechselstrommaschinen. Verlag der Ungarische Akademie der
Wissenschaften, 1959.
4. Kl u s z c z y n s k i K. , Mi k s i e wi c z R . : Parasitic torque
of cage induction motors (in Polish). Pub. by PTETiS, Warsaw-
Gliwice, 1993.
5. P a s z e k W.: Transients of Alternating Current Machines (in
Polish). WNT, Warsaw, 1986.
6. P a s z e k W.: Alternating Current Machine Dynamics (in
Polish). Helion, Gliwice, 1998.
7. P u c h a a A . : Dynamics of Electric Machines and
Electromechanical Systems (in Polish). Warsaw, WNT, 1977.
8. Va s P. : Electric Machines and Drives: A Space-Vector Theory
Approach. Oxford University Press, 1992.
9. R u s e k J . : Computer Analysis of Induction Machines by
Harmonics Balance. Pub. by Univ. of Mining & Metallurgy,
Cracow, 2000.
10. S k wa r c z y s k i J . : Electric Machines (in Polish), Parts. I i
II. Pub. by Univ. of Mining & Metallurgy, Cracow,1997.
11. S o b c z y k T. J . : Methodology of Mathematical Modeling of
Induction Machines (in Polish). WNT, Warsaw, 2004.
12. Wh i t e D . C . , Wo o d s o n H . H .: Electromechanical
Energy Conversion. J. Wiley & Sons Inc., New York, 1959.
Tadeusz J. Sobczyk
was born (1944) and educated in Poland. He received
M.Sc., Ph.D. and Dr. Habil. (D.Sc.) degrees in electrical
engineering from the Faculty of Electrical
Engineering, Automatics and Electronics at the
University of Mining and Metallurgy, Cracow, in 1967,
1974, and 1977 respectively and the scientific title
of a Professor from the President of Poland in 1991.
Since 1991 he is a member of the Committee of
Electrical Engineering of the Polish Academy of Science. In 2000 he
was awarded the honorary title Doctor Honourees Causa of the Russian
Academy of Sciences. He worked at the University of Mining and
Metallurgy in Cracow (19671989) and next at the Cracow University
of Technology at the Faculty of Electrical & Computer Engineering.
In the years 1993-1999, he was the Dean of this Faculty. Presently he
is the Director of the Institute on Electromechanical Energy
Conversion and the Head of Department of Electrical Machines in
this Institute. Since November 2000, he has also been employed at the
Electrotechnical Institute, presently at the Department of Electric
Machines. His main research fields are: electrical machines and drives,
electromechanical systems, electrical energy conversion and
transformation by power electronic systems. He has published a book
Methodological aspects of induction machines modelling (in Polish)
and more then 170 scientific papers in Polish and international journals
and proceedings of conferences.
Address:
Cracow University of Technology,
Institute of Electromechanical Energy Conversion,
ul. Warszawska 24, 31-155 Krakw,
e-mail: [email protected]
111
Appendix 1 - Inductance matrices of a base model
1 2 1 M 1
2 1 2 M 2
M ! M 2 M
j ( ) j ( ) 1,1 1,2 1,M j
j ( ) j ( ) 2,1 2,2 2,M j
j ( ) j ( ) M,1 M,2 M,M j
L L e L e e
L e L L e e
( ) diag
L e L e L e
r
,r,s ,r,s ,r,s
r
m ,r,s ,r,s ,r,s
ss
r
,r,s ,r,s ,r,s
j



( (
( (
(

(
(
(

L
L
L
M M O M M
L
j
e
s
s r
j
n

(
(
(
(
(
(
(
(
(
(
,

1 N 1 2 1
2 N 2 1 2
N 1 N
j ( ) j ( ) M 1,M 1 M 1,M 2 M 1,M N j
j ( ) j ( ) M 2,M 1 M 2,M 2 M 2,M N j
m
rr
j ( ) M N,M 1 M N,M j
L L e L e e
L e L L e e
( ) diag
L e L e
r
,r,s ,r,s ,r,s
r
,r,s ,r,s ,r,s
r
,r,s ,r,s
j
- - - - - -
- - - - - -
- - -
(
(
(

(
(
(

L
L
L
M M O M M
N 2
j
j ( ) 2 M N,M N
e
e L
( r s )
s r

,r,s
j
n
-
- - -
(
(
(
(
(
(
(
(
(
(
(
(
(
,

L
1 1 1 2 1 M 1
2 1 2 2 2 M 2
N 1 N
j ( ) j ( ) j ( ) M 1,1 M 1,2 M 1,N j
j ( ) j ( ) j ( ) M 2,1 M 2,2 M 2,N j
m
rs
j ( ) M N,1 j
L e L e L e e
L e L e L e e
( ) diag
L e L e
r
,r,s ,r,s ,r,s
r
,r,s ,r,s ,r,s
r
,r,s ,r,
j
- - -
- - -
-
(
(
(

(
(
(

L
L
L
M M O M M
N 2 N M
j
j ( ) j ( ) M N,2 M N,N
e
e L e
( r s )
s r

s ,r,s
n j
n
- -
- -
(
(
(
(
(
(
(
(
(
(
(
(
,

L
Appendix 2 - Inductance matrices of a base model for symmetrically located coils
s s s
s s s m
ss
s s s
(0 0 ) ( 0 ) ( 0 )
( ) (0 ) (2 )
( ) ( 2 ) (0 )
L , , L , , L , ,
L , , L , , L , ,
L , , L , , L , ,
j j j
j j j
j j j
(
(
(

(
(

(

L
L
L
M M O M
L
r r r
r r r m
rr
r r r
(0 ) ( ) ( )
( ) (0 ) (2 )
( ) ( 2 ) (0 )
L , , L , , L , ,
L , , L , , L , ,
L , , L , , L , ,
j j j j j j
j j j j j j
j j j j j j
(
(
- - -
(

(
(

(

L
L
L
M M O M
L
rs rs rs
rs rs rs m
rs
rs rs rs
( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( )
L , , L , , L , ,
L , , L , , L , ,
L , , L , , L , ,
j j j j j j j j j
j j b j j j j j j j
j j j j j j j j j
- (
(
- - - - - - -
(

(
(
-
(

L
L
L
M M O M
L
Tadeusz J. Sobczyk: Extreme Possibilities of Circuital Models of Electric Machines
112 Electrical Power Quality and Utilization, Journal Vol. XII, No 2, 2006
Appendix 3 - General equations of circiutal models accounting for non-linearity of a main magnetic circuit
2
1 1 2 1 2 1 N 1 N 1 1
2
2 2
2 1 2 1 2 2 N 2 N
, ,
2
N N
N 1 N 1 N 2 N 2 N
( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
d d 1
d d
( ) ( ) ( ) ( ) ( )
,
og ,
,
cos cos
i
E i
cos cos
t t i i
i
cos cos
r r r
r r y
y
r r
y
r r
(

(
(
(

(
(
-
(
(
(
(
(
(

L
L
M M
M M O M
L

(

(

(
-
' `
(

(

(

' '
1 2 1 2 1 N 1 N 1
2 1 2 1 2 N 2 N 2
2
,
N 1 N 1 N 2 N 2 N
0 ( ) ( ) ( ) ( )
( ) ( ) 0 ( ) ( )
d 1
d ( )
( )
( ) ( ) ( ) ( ) 0
og
sin sin i
E sin sin i
t
i
sin sin i
r
r
r r
r r
rh
r r
( (
( (


( (
-
' `
( (

( (


( (
' '

L
L
M M O M M
L
1 1 1
2 2 2
N N N
R
R
R
u i
u i
u i
( ( (
( ( (
( ( (

( ( (
( ( (
( ( (

M O M
2
1 1 2 1 2 1 N 1 N
1
2 2
2
2 1 2 1 2 2 N 2 N 1
1 2 N
2 2
, ,
2
N
N 1 N 1 N 2 N 2 N
( ) ( ) ( ) ( ) ( )
( )cos ( ) ( ) ( ) ( ) d
1
J
d
( ) ( ) ( ) ( ) ( )
og
cos cos
i
E i
cos
i i i
i i
t
i
cos cos
r r r
r r
r r j
j
r r
(

(
(

(
' `
(


(

(

' '

L
L
L
M
M M O M
L
(
(
(
-
(
(
(

1
2
1 2 N
2
,
N
2
1 1 2 1 2 1 N 1 N
2
2 1 2 1 2 2 N 2 N
2
N 1 N 1 N 2 N 2 N

1
( )
( )

( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
og
E
i i i
i
cos cos
cos cos
cos cos
r
r
j
j
rh
j
r r
r r
r r
(
(

(
(

(
- (

(
(
(

(
(

L
O
L
L
M M O M
L
1
2
m
N
i
i
T
i
(
(
(
(
(
(
-
(
(
(
(
(
(