I

Calculation of the impedance of rotor cage end

rings

S.Williamson
M.A. Mueller

Indexing termq: Modelling, Induction motors

of images. Kovacs model ignores the circular nature of

Abstract: The authors show that the resistance the rings, which in itself produces a nonuniform current
and leakage reactance of the end rings of a rotor distribution, even under DC conditions. In addition, it is
cage can be evaluated from the impedance of an not suitable for determining the reactance of the end
identical ring which carries a circulating current of rings.
the appropriate frequency. Two mathematical The calculation of end-ring reactance is diflicult. The
methods for calculating end-ring impedance using field in the end region of a cage motor is exceedingly
this physical model are described, the first based complicated, and an accurate model for its calculation
on the use of modal currents together with the would need to allow for relative motion, intricate bound-
method of images, and a second employing finite ary shapes, a mixture of magnetic and nonmagnetic
elements. The influence of skin effect is invest- materials, and magnetic saturation and anisotropy. Such
igated, as is the proximity of the rings to the shaft a model is still beyond the capabilities of current analysis
and to the core-pack end. The method is also techniques. As a consequence, the expressions used for
applied to the consideration of the impedance of the calculation of rotor end-ring leakage reactance owe
the end rings in a double-cage machine. much to engineering pragmatism.
Cochran [SI states that rotor end-ring leakage react-
ance is generally (i) ignored; (ii) lumped into the stator
1 Introduction end-turn leakage equation by the use of an empirical
factor or adjustment; or (iii) evaluated using a separate
The end rings of the rotor cage in a cage induction motor reactance equation. The authors are aware of the expres-
may contribute a significant proportion of the rotor sion for end-ring reactance quoted by Liwschitz-Garik
impedance, and can exert an appreciable influence over [SI, but have not been able to discover its origin. Our
the performance of the machine. Efforts continue to be experience of current industrial practice is that option (iii)
made to improve the reliability with which performance above is adopted, but that the expressions used differ
can be predicted at the design stage, and it is therefore widely, as do the empirical correction factors they
appropriate that consideration be given to the calcu- contain.
lation of this important component of impedance. The purpose of this paper is to propose a single physi-
End-ring resistance is generally calculated by multi- cal model for the calculation of both end-ring resistance
plying resistivity by the mean circumference of the rings, and leakage reactance. Two alternative mathematical
divided by their cross-sectional area. This approach approaches to the calculation of end-ring impedance used
ignores both skin effect and the nonuniformity of the on the proposed physical model, are discussed. One of
current flow in the rings. Trickey [I] and Williamson and these techniques, the finite-element method, is now SUE-
Begg [2] have studied the current distribution in rings ciently well known to require only the briefest of descrip-
which have a radial depth that is large when compared to tions, but the second technique, based on the use of
the depth of the bars that feed current into them. Their modal currents and the method of images, is less well
results show that, for rotors with wide end rings, the known and so more details are given. The model that is
simple calculation method can lead to poor accuracy, proposed is inherently capable of taking skin effect into
particularly when the pole number is high. Both studies account.
ignored the influence of skin effect, concentrating instead
on the current flow path from bar to bar under what
were effectively DC conditions. Skin effect in end rings 2 M o d e l and assumption
has been studied by Kovacs [3], employing a circuit tech-
nique based on modal currents, also used by Graneau [4] The effect of end-ring resistance is customarily accounted
to study current distribution in power cables. Kovacs for by adding an appropriate amount to the bar resist-
used a linearised model for an end ring, replacing it by a ance.
long, straight bus bar of the same cross-sectional area.
The rotor core end and shaft were then modelled as infi- (1)
nitely permeable planes, and replaced using the method
where
Paper 9067B (Pl), first received 24th March and In revised form 1st July
1992
R , = bar resistance
The authors are with the University of Cambridge, Department of
N , = number of rotor bars
Engineering, Trumpington Street, Cambridge CB2 lPZ, United p = pole-pair number
Kingdom R , = resistance of one complete end ring
IEE PROCEEDINGS-B, Vol. 140, No. I , JANUARY 1993 51
I

Eqn. 1 is derived from the consideration of rotor joule replaced, using the method of images. It is assumed that
loss. It is a relatively straightforward matter [2] to show the motor frame and rotor shaft are sufficiently distant
that the loss that occurs in a cage of bar resistance Rh from the end ring to have a negligible effect, and these
and zero end-ring resistance, which carries a 2 p pole surfaces are ignored. The resulting model then consists of
current distribution, is the same as that in the same cage the end ring and its image, as shown in Fig. 2.
with the same current distribution if the bar resistance is
R, and the end-ring resistance R e . The end-ring resist-
ance used in eqn. 1 is calculated simply from the resis-
tivity and dimensions of the end rings, although a
correction may be needed if the radial width of the end
rings is comparable with the pole pitch, as discussed in
the preceding Section.
An entirely analagous approach may now be adopted
with respect to the rotor end-ring leakage reactance. The
effective rotor leakage reactance per bar may be calcu- ring image centrelot shaft end ring
lated using the expression

x;= x,+
$1
xe

2 N , sin2[

where X , = rotor leakage reactance per bar due to all

non end-ring causes, and X, = the reactance of one end I
ring. In fact, eqn. 2 may be derived in a similar fashion to boundary of
eqn. 1, but considering reactive power balance instead of core pack
power balance. Fig. 2 The rotor end ring and ifs image in the core pack
This approach gives a simple physical model that
allows end-ring reactance to be calculated. X , is the reac- The next step is notionally to subdivide the end ring
tance presented by one end ring to current that flows in and its image into a number of elementary subrings, as
an uninterrupted path around it. In a real end ring, the shown in Fig. 2, and to calculate the mutual inductance
current distribution varies around the periphery because between all possible pairs of subrings in the real ring,
of the bar currents which are feeding into and out of it. taking the presence of the image into account. Silvester
This effect is taken into account approximately by the [7] gives a suitable expression for the inductance between
denominator in eqn. 2. Having outlined this physical two coaxial filamentary rings. With reference to Fig. 2,
model, we will now propose two alternative mathemati- the mutual impedance between the ith and j t h subrings is
cal models that will enable end-ring impedance to be cal- given by:
culated for any specific end-ring geometry.

2.1 Circuit model

where
The circuit model proposed by the authors is based on
the use of modal currents and the method of images. Fig.
1 shows a sectional representation of the end region of a F(z) = {[: - k]K(k) ~ 2 i
E(k) (4)

in which
stator core
I I / I

K ( k ) and E(k) are complete elliptic integrals of the first

and second kinds. They can be evaluated rapidly and
efficiently using fourth-order polynomials, as shown in
Reference 8.
Eqns. 3-5 may be used to evaluate z,,
, provided i # j .
When i = j , which means that we are attempting to cal-

I p--y j j
culate the self-inductance of the ith ring, eqn. 4 becomes
infinite. This anomaly arises from the well known singu-
larity in the field at the surface of a filamentary conduc-
tor. The problem may be overcome using Grover's 191
formula for the self-inductance of a circular loop, radius
stator
werhong a, of a conductor of square cross-section of side length c.
stator care
L ~~ I I

Fig. 1 A side elevation oJthe rotor end rinq

cage motor from which the rotor bar extensions to which

the end ring is attached have been omitted, for the sake
of clarity. Fig. 2 shows how the same end ring is model-
- 0.84834 + 0.2401
(3l
-

The self-impedance of the ith subring is therefore

(6)

led. In essence, the ends of the stator and rotor cores are
regarded as an infinitely permeable surface and then Z,, = R , + jw[L, + p,, ri F(2z,)] (7)
52 / L E P R O C E E D I N G S - B , Vol. 140. N o . I , J A N U A R Y l Y Y 3
R , is the D C resistance of the ith loop, and the second These procedures are relatively straightforward and can
term in the brackets accounts for the influence of its be programmed to run on a basic PC.
image.
If we imagine that a voltage q IS needed to drive 2.2 Finite-element model
r,
current round the ith subring, we may write a voltage An alternative approach to the calculation of end-ring
balance equation for that subring in the form impedance is to make use of an advanced field calcu-
N
lation technique. The main advantage of this type of
q = ,=1 Z', r, (8) approach is that the presence of the iron surfaces that
bound the end region can be represented more realistic-

1
'-
1

in which a total of N subrings is assumed. Similar equa- ally. Fig. 3 shows a typical solution domain AB and C D
tions may be written for all N subrings. In matrix form l

frame

end
coreofpack
stator cover

The impedance matrix of eqn. 9 may be inverted to

obtain an admittance matrix, which allows the subring
currents to be expressed as a function of the subring volt-
ages, thus
core Duck
shaft surface
... (10)
.. . . . .. . . . E- F
. . .
...
.................................................................................
If the modal currents in the subrings are to represent centre of Shaft
current flow in the massive end ring, the subring voltages Fig. 3 T h e problem area .surriiunding the end ring t o be solued using
must all be equal, i.e. finite-element analysis
- -
v-I -- v-2-- v -3 -- . . . = v - v
N- (1 1) represent the stator and rotor surfaces facing each other
Eqn. 10 may therefore be written in the form across the air gap; BC is a flux line; AJ and BD are the
stator and rotor core-pack ends, respectively; HIJ is the
Y,,+ Y,, + Y,, + _ ' ' + VI, machine case; HGF represents the surface of the bearing
housing, and E F is the surface of the shaft.
(12) The problem is axisymmetric about the shaft centre,
which is marked on Fig. 3. The field problem may be
Y,, + Y,, + y,, + . + r v ,' '
couched in terms of the circumferentially directed mag-
netic vector potential as
The total current flowing in the ring is obtained by
summing the currents flowing in all of the subrings
i = i, + i, + i, + ' . . +I;, (1 3)
where
so that eqn. 12 may be written
I = pv
where
and
N
,'=II F J
, = I ,=1
N

The impedance of the end ring is obtained from the

C = axial length of ring x log, - I;;[ (19)

reciprocal of this admittance, i.e. A, is the average value of A over the conductor cross-
section, i.e.
1
=7
Re, +.ix,, (16)

The procedure for calculating the end-ring resistance and

leakage reactance is therefore where S is the cross-section through the conductor, and
(i) Notionally divide the end ring into a number of J , is the average current flowing in the ring ( = I / S ) .
subrings; The boundary surfaces corresponding to the stator
(ii) Calculate the elements of the impedance matrix and rotor core packs (i.e. BAJ and CDE) are regarded as
using eqns. 3-7; consisting of infinitely permeable iron (i.e. aA/dn = 0),
(iii) Invert the impedance matrix to obtain the admit- and the flux line boundary BC becomes A = constant.
tance matrix; The remaining boundaries to the solution correspond to
(iv) Add together all the elements of the admittance unlaminated iron surfaces. The study of the eddy currents
matrix and find the reciprocal of the sum. induced in these surfaces (i.e. case, bearing housing and
I E E PROCEEDINGS-B, Vol. 140, N o . I , J A N U A R Y 1993 53
I

shaft) is beyond the scope of this paper, but we may gain 3.1 Effect of frequency
some insight as to their possible influence by considering Figs. 5u and 56 show the calculated end-ring resistance
the two extremes. These correspond to an ideal conduct- and reactance for the 4 kW motor, as a function of fre-
ing surface (in which case EFGHIJ becomes a flux line) quency in the range 0-50 Hz. Both figures include results
or an ideal magnetic surface (in which case the flux is obtained using the circuit method and the finite-element
normal to EFGHIJ). method using the two alternative boundary conditions.
Eqn. 17 may be solved by any convenient numerical Fig. 5a shows that the resistance of this end ring is little
technique. We used an axisymmetric formulation of the affected by frequency variations a result that is hardly
~

finite-element method [lo]. surprising in view of the fact that its radial depth is less
The end-ring impedance is determined from the field than one skin depth at 50 Hz. Fig. 5b shows that the end-
solution by equating the power and reactive volt-amps ring inductance is also substantially independent of fre-
calculated from the field solution to that given by the quency, and again this is a result that might he expected
usual lumped circuit expressions because the majority of the leakage flux does not pass
through the conductor itself, but links through the hole
(21) in the middle of the ring. Skin effect can therefore be
expected to play only a small role.
Figs. 5c and 5d show the calculated end-ring resistance
and reactance for the 100 kW motor, over the same fre-
quency range. At 40mm, the depth of the end rings is
now more than four times the skin depth at 50 Hz, and
3 Calculation of end-ring impedance
so skin effect might be expected to play a more important
role. This proves to be the case, as is shown in Fig. 5c,
The circuit method and the finite-element method have which indicates an increase in end-ring resistance of
both been implemented and used to calculate the imped- approximately 50%)at 50 Hz, when compared to the DC
ance of two motors of different size and pole number. value. Fig. 5d confirms that end-ring leakage inductance
The first is a four-pole machine rated at 4 kW, and the is sensibly frequency-independent.
second is a two-pole machine rates at 100 kW. The rele- Figs. 5a-5d show that the frequency trends predicted
vant dimensions of these two machines are given in Table by both the finite-element and the circuit methods are
1, which refers to the diagram shown in Fig. 4. similar, but that the numerical values obtained can differ
appreciably. It is interesting to note, however, that the
Table 1 : Dimensions for end-ring model (see Fig. 4). All circuit method, in which the rotor shaft and case are
values in mm
completely ignored, gives values which lie in between
4 kW motor 100 kW motoi those obtained using the finite-element method for the
I 80 220 two limiting boundary conditions. The only exception to
2, 1 35 this is the resistance of the 4 kW motor, for which the
22 6 20 difference between the maximum and the minimum
24 69 165 resistance at 50 Hz is less than 1%. In fact, for all four
r 100 301.5
r1 64 161 figures the answers given by the circuit method differ by
r3 13 40 less than 4% from the mean of the two answers given by
'4 42 110 the finite-element method.
15 21 5 41.85
re 36 100
9, 10 20
9 0.4 0.8 3.2 Effect of proximity to the shaft surface
Cage-type Cast Fabricated The effect that proximity to the shaft surface has on end-
material aluminium copper ring impedance can be investigated by varying the inner
ring radius (r4 in Fig. 4) while keeping all other dimen-
sions constant. Figs. 6 a and 6b show the results of this
exercise for the 4 kW motor, and 6c and 6d for the
100 kW motor. For all of these figures a constant fre-
quency of 50 Hz has been assumed.
The resistance of the end rings of the 4 kW motor
(Fig. 6a) is seen to be insensitive to the proximity of the
shaft, as witnessed by the closeness of the results calcu-
lated using all three models. The main effect, therefore, is
a linear variation in resistance due to the corresponding
linear variation in mean diameter as r, is varied. Fig. 6c
shows that the proximity of the shaft surface is a signifi-
cant factor for the 100 kW motor only when r4 is reduced
from its design value of 110 mm to approximately half of
that value. This corresponds to a clearance between the
shaft and the end ring of about 15mm, a value that
would not normally be adopted in any practical design.
Two results emerge from Figs 66 and 6d. The first is
that end-ring reactance varies linearly with end-ring
radius, and the second is that the circuit model (which
ignores the presence of the shaft) again gives results that
1 centre of shaft are closely equal to the average of the two finite-element
Fig. 4 The dimensions used to define the fe model calculations using the alternative boundary conditions.
54 I b E PROCEEDINGS-B, Vol. 140, N o . I , J A N U A R Y IY93
I

LO 0 -
39 8 - 160-
53 9 6 - c,
"U

$ 392-
svi
ILO-
vi
P

386-
38 4 4 I I
0 10 20 30 40 50 0 IO 20 30 40 50

020

frequency,Hr frequency,Hz
b d
Fig. 5 The variation ofend-ring resistance and inductance with frequenryfor the 4 kW motor and IO0 kW motor
a Reststance, 4 kW motor h Inductance, 4 k W motor c Resistance, 100 kW motor d Inductance, 100 kW motor
..... Circuit model ~~ fe model (iron boundary) ~ le model (flux line boundary)

This last result is seen to hold even when the end ring-to- moved far enough from the core, their inductance can
shaft clearance is reduced to unrealistically small values. start to increase again as the influence of the case and
bearing housing begins to be felt. It is thought, however,
3.3 Effect of proximity to the core-pack ends that this will only happen at unrealistically large dis-
The effect that proximity to the core ends has on end- tances from the core pack, so that this phenomenon has
ring impedance can be investigated by varying the clear- no practical importance. These figures also show once
ance (zl in Fig. 4) while keeping all other dimensions again that the circuit method gives values for end-ring
constant. Figs. 7 a and 7 b show the results of this exercise inductance that are closely equal to the average of the
for the 4 k W motor, and 7c and 7d for the 100kW values obtained using the more detailed finite-element
motor. For all of these figures a constant frequency of method with the two alternative models for the shaft,
50 Hz has been assumed. bearing housing and case.
The insensitivity of the resistance of the end rings of
the 4 kW motor to the proximity of the core-pack ends,
which is displayed in Fig. 7a, is to be expected in view of 3.4 Effect of stator end windings
their relatively small dimensions compared to the skin The reactance calculated using the model and methods
depth. By contrast, the variation of resistance for the discussed in Section 2 is the total reactance of the end
100 kW motor is dramatic (Fig. 7c). As the end rings are ring. Our work on stator end-winding leakage reactance
moved closer to the core pack, the skin effect produces a [ l l ] has shown that the stator windings and rotor cage
rapid change in resistance. The curves indicate that if the are coupled through their end-region fluxes, so that the
designer had placed the end rings flush against the core above end-ring reactance must be regarded as consisting
pack (as in the cast rotor of the 4 kW motor) the end-ring of the sum of 'mutual' and 'leakage' components. This
resistance would have been as much as 25% higher at mutual reactance should properly form part of the mag-
starting on a 50 Hz supply. In fact, the designer placed netising reactance in the equivalent circuit. For any prac-
the ring at a distance equivalent to just over three skin tical machine, it will be small compared to the reactance
depths from the core pack, at which its presence is seen due to the main airgap flux. The residue is the end-ring
to have little effect on resistance. leakage reactance.
Figs. 76 and 7d show the effect on the end-ring induct- In order to calculate the true end-ring leakage react-
ance of the two motors of varying zI. As would be ance, this end-winding mutual should be subtracted. The
expected, the inductance increases as the rings are moved procedure we recommend is as follows:
closer to the core for all practical values of end-ring (i) Calculate X, using the methods described here and
clearance. The results also show that, if the rings are determine the effective reactance per bar via eqn. 2
I E E PROCEEDINGS-B, Y d . 140, N o . / , J A N U A R Y l Y Y 3 55

I
I

I
201 1 61 I I
20 30 LO 50 40 60 80 I00 120
rA , m m rL mm
C
4

Fig. 6 The oariafion ofend-ring resisfunce and inductance with radius from the shalt centrefor the 4 kW motor and 100 kW muto?
UResistance, 4 k W motor h Inductance. 4 k W motar c Resistance, 100 kW motor d Inductance, 100 k W motor
Circuit model -~ le model (iron boundary) ~ - fe model (flux line boundary)

(ii) Refer Xi to the stator, in the normal manner, to cages, whereas separate end rings are employed with fab-
obtain the referred rotor leakage reactance X2 ricated cages. The model developed in Section 2 may be
(iii) Calculate the stator end-winding leakage reactance used without modification for calculating the impedance
per phase, both with and without the presence of rotor of a shared end ring, but further consideration must be
currents taken into account (see [l 11). The difference given to separate end rings.
between these two values is the mutual inductance per
phase due to end-winding coupling, X,, 4.1 Circuit model
(iv) Subtract X,, from X 2 to obtain a better approx- Fig. 8 shows the arrangement of the rings and their
imation for the rotor leakage reactance. image in the core-pack end used in the circuit model. The
general procedure closely follows that described in
4 Double-cage machines Section 2.1, except that now the outer ring may be dis-
1 cretised into N subrings, and the inner ring into M sub-
The two cages in a double-cage machine may share a rings. The resulting impedances, which a r e calculated in
common end ring or have two separate end rings. accordance with eqns. 3-7 may be used to set up the
Common end rings are usually found in rotors with cast following matrix

(23)

56 I E E PROCEEDINGS-B, V o l 140, N o . I , J A N U A R Y 1993

I

!
5 LO-c,
.I_---

S !
$
: 3 5 -1i
E !
!
30-! actual position
1 of the end ring
of theendring !
25 i I I I I lL I ! , I I I
0 20 LO 60 80 I00
z1,mm
C
0 4 1 0

1 @ 3[

RC

001 I I I I 021 I I 1 I I I
0 10 20 30 LO 0 20 LO 60 80 100
z , rnm z,.mrn

b d

and vice versa.

As before, the impedance matrix may be inverted to
io,,= TI + i2+ T3 + . . + iN (27)
obtain an admittance matrix, which may be written as - -
shown in eqn. 24, to retain the partitioned structure I,, = 1 N + 1 + rN+2 f rN+, + " ' + iN+M (28)

Following the procedure in Section 2.1, we now recognise ,mposing eqns, 25-28 on matrix eqn, 23 produces
that the voltages acting around the subrings in the outer
ring must be equal (to, say E.,). Similarly, the voltages in
(29)
the inner ring must also be equal to each other, i.e.
IEE PROCEEDINGS-B, Vol. 140, N o . I , J A N U A R Y 1993 57
I

where 4.2 Finite -element model

N N
The solution domain shown in Fig. 2 can readily be
60 = j = l k = l ?k adopted to suit a double-cage rotor by the simple addi-
tion of an extra end ring. The principal difference is that
N+M N+M the current density term ( J in eqn. 17) will now be
r,; = j -1
N+I k=N+1
c 7k nonzero over two regions, rather than one, so that C, A,,
and J , will have to be separately defined over each ring
N N+M
cross-section. Computationally this presents no difficulty,
but further consideration must be given to the interpreta-
tion of the results.
Eqns. 30-32 show that the admittances contained in eqn. LJsing the nomenclature of the preceding Section, the
29 are obtained by summing the elements of the corres- total power loss in the two end rings is given by
ponding admittance submatrices, as identified in eqn. 23.
P,", = R,, I:", + R,i I t . + 2R,, Re [TO",k l (37)
and the average stored energy by
W,,,= fL,, I:u, + t L , , + L,, Re [To",T;] (38)
From a single finite-element solution we may calculate
\

/
i
image of
f
image of
inner ring
i
inner
end ring
outer
end ring
only P,,, and W,,,, from

P,", = L:"cL"tc'ri"T
- dvo'
outer ring ~

centre; of shaft

+ g1
m j inner ring
JJ* d V d ] (39)

W,,,= 2 Re [I JA* d V d ] (40)

To identify the resistances and inductances contained in

eqns. 37 and 38, we carried out three separate field solu-
tions using the finite-element model, each using a differ-
boundary of ent pair of end-ring currents To,, and l i n .The values of
core pock
P,,, and Yo,obtained from these solutions enabled two
Fig. 8 Separate end rrngs /or double-cage rotor and their image ~n the sets of three simultaneous equations to be set up, and the
core pack
resistances and inductances were obtained by solving
these simultaneous equations. Numerical experiments
The 2 x 2 admittance matrix contained in eqn. 29 is showed that, as expected, the inductance and resistance
inverted to obtain, finally values so obtained were independent of the pairs of end-
ring currents chosen, provided, of course, that the three
(33) current pairs used were linearly independent.

Z,, is the self-impedance of the cuter rjng, and Z,, is the 4.3 Numerical results
self-impedance of the inner ring. Z,, = Zi, is the coupling The resistances and inductances of the end rings of two
impedance between the two end rings. These impedances industrially manufactured double-cage motors have been
may be written in the form computed using both circuit and finite-element tech-
niques. Scaled end-region outlines for the two machines
z,,= R , , + jwL,, (34) are shown in Figs. 9a and 9b. The end region shown in
Fig. 9u is part of a 3.3 kV, four-pole motor, rated at
zi,= Rii + j u L i i (35) 373 kW, and that shown in Fig. 9h is part of a 1050 V.
Z,, = Z,, = Rio + jwLi, (36) 4/8 change-pole motor rated at 112/56 kW.
Table 2 gives the values computed for the 373 kW
The fact that the 'self' impedances of the outer and inner motor, and Table 3 gives those for the 112/56 kW motor.
end rings have resistance as well as reactance will come Both sets of results relate to a frequency of 50 Hz, that is,
as no surprise. The 'mutual resistance' term Rio is less to locked rotor with a 50 Hz supply. The results obtained
common. It is a manifestation of the proximity effect, by using the finite-element method are given for the same
which the current flow in one end ring influences the two extremes of boundary condition that were used in
current distribution and loss in the other. The procedure Section 3.
for calculating these impedances is therefore
(i) Notationally divide the end rings into a number of Table 2 : End-ring resistances and inductances for 373 kW
subrings ; double-cage motor
(ii) Calculate the elements of the impedance matrix;
(iii) Invert the impedance matrix to obtain the admit-
tance matrix;
~

Circuitmethod 496 348 148 0732 0274 0360

(iv) Sum the elements of the admittance matrix in Fe (Boundaries 498 386 151 0722 0 2 7 3 0356
groups corresponding to the location of the subrings to infinitely
form the end-ring admittance matrix (eqns. 29-32); permeable)
(v) Invert this 2 x 2 matrix to obtain the end-ring Fe(Fluxline 498 382 151 0 7 1 9 0272 0353
boundaries)
impedance matrix (eqn. 33).
58 I E E P R O C E E D I N G S - B , Vol 140, N o . I , J A N U A R Y 1993
Table 3: End-ring resistances and inductances for puter, and it is therefore worth discussing their relative
112156 kW double-cage motor merits.
The circuit method is conceptually the simpler of the
two techniques, requiring no special numerical skills. The
Circuit method 525 406 157 0798 0338 0425 main computational cost arises from the need to invert
Fe (Boundaries 527 437 160 0790 0337 0421 the complex impedance matrix, which is a dense matrix
infinitely of dimension equal to the number of subrings into which
permeable) the end ring, or rings, is divided.
Fe(Fluxline 526 432 160 0786 0335 0419
boundaries) The finite-element method is now widely used
throughout the motor manufacturing industry, but
largely as a development tool rather than for day-to-day
03
use in the design office. For everyday use by designers,
this method would most probably need a purpose-
written interface that would take input in the form of the
dimensions shown in Fig. 4, and would run the finite-
element model transparently (i.e. without needing any
further action from the user).
Sample timings for the two methods run on a single-
cage test problem, using 256 nodes in the end ring for the
finite-element model, and 256 subrings for the circuit
model, gave solution times of 63.1 s and 96.6 s, respec-
tively, on a modern workstation. For a typical double-
01
i cage machine the timings for the circuit method and the
finite-element method were 136.2 and 655.1 s, respec-
tively. The comparatively large increase for the finite-
0 05 element method arises from the need to carry out three
finite-element solutions in order to segregate the various
components of resistance and inductance. These times
will fall rapidly as computer speeds continue to increase.
00 005 01 015 02 025 03
0

6 Conclusions
03
A physical model for the determination of end-ring
impedance has been described, and two techniques for
performing the relevant calculations outlined. The first of
these, which employs the concept of modal currents and
the method of images, is relatively simple to implement.
The second makes use of the finite-element method and is
inherently capable of modelling the complex boundary
shapes that are found in the end regions of machines. The
0 difficulty arises when consideration is given to the bound-
ary conditions, and we have therefore explored the influ-
01 ence of the proximity of the various physical boundaries
using two alternative extremes of boundary conditions to
represent them for a single-cage motor.
0 05 The results obtained suggest that, for practical sizes of
rings, neglecting the presence of the boundaries alto-
gether will give a value of impedance that is closely equal
oooj , , , , , , , , , , , , to the average obtained using these two extremes. We are
00 005 01 015 02 025 03 therefore led to conclude that, in the absence of other
b data, the pragmatic approach is to ignore the presence of
Fig. 9 Scaled diagram of motor end-region. For key t o boundaries see the shaft, bearing housing, end cover and case when cal-
Fig. 3. Dimensions are in mefres culating end-ring impedance for a single-cage rotor.
a 373 kW double-cage motor The results obtained for the two double-cage rotors
h 112156 kW change-pole double-cage rnolor
with separate end rings indicate that the boundaries to
Tables 2 and 3 indicate a close correlation between the the end ring have little effect on resistance and induct-
resistance and inductance values computed using the ance. The circuit method appears to have a clear advant-
alternative models, suggesting that each is equally valid. age in this instance, because the various end-ring
The greatest percentage discrepancy arises in the calcu- resistances and inductances are obtained from a single
lation of ROi but, as this resistance is a relatively small solution.
component, the error involved in, say, total power loss
calculation will be small.
7 References
5 Computational aspects 1 TRICKEY, P.H.: Induction motor resistance ring width, Trans.
Am. Inst. Elect Eng., 1936, 55, pp. 144-150
Either of the two methods discussed in Sections 2 and 4 2 WILLIAMSON, S., and BEGG, M.C.. Calculation of the resistance
may be implemented using a relatively inexpensive com- of induction motor end rings, Proc. IEE. 1986, 133, B, (2), pp. 54-60

IEE PROCEEDINGS-B, Vol. 140, N o . I , J A N U A R Y 1993 59

3 KOVACS, G.: Skin effect in the end ring of asynchronous motors, 8 ABRAMOWITZ, M., and STEGUN, I.A.: Handbook of mathe-
Pror. I n t . Con/: Elect. Machines. Budapest, 1982, pp. 31-34 matical functions (Dover Publications, New York, 1965)
4 GRANEAU, P.. Underground power transmission, the science, 9 GROVER, F.W.: Inductance calculations (Dover Publications,
technology and economics of high-voltage cables (John Wiley & New York, 1946)
Sons, New York, 1979) I O SILVESTER, P., and FERRARI, R.L.: Finite elements for electrical
5 COCHRAN, P.L.: Polyphase induction motors, analysis, deslgn engineers, 2nd edn (Cambridge University Press, 1990)
and application (Marcel Dekker, New York, 1989) I 1 WILLIAMSON, S., and MUELLER, M.A.: Induction motor end-
6 LIWSCHITZ-GARIK, M.: Electric machinery, Vol. 1, Fundamen- winding leakage reactance calculation using the Biot-Savart
tal and DC machines (D. Van Nostrand Company, Toronto. 1946) method, taking rotor currents into account, Proc. Inc. Con/: Elect.
7 SILVESTER, P.: Modern electromagnetic fields (Prentice-Hall, Machines, Cambridge, Mass., 1990, pp. 480-484
Englewood Cliffs, New Jersey, 1968)

60 IEE PROCEEDINGS-B, Vol. 140, N o . I , J A N U A R Y 1993