24 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 33, NO.

1, JANUARYFEBRUARY 1997

CO f tic a
i rs
David G. Dorrell, Member, IEEE, William T. Thomson, Senior Member, IEEE, and Steven Roach

Abstract- This paper provides new information for the on- identified as being suitable for monitoring to assess the degree
line diagnosis of airgap eccentricity in 3-phase induction motors. and type of rotor eccentricity present.
A new theoretical analysis of the interaction between harmonic Rotor eccentricity in induction motors takes two forms-
field components due to static and dynamic rotor eccentricity
which previous research has not considered is put forward. The static eccentricity (where the rotor is displaced from the
resultant non-supply-frequency current components produced stator bore center but is still turning upon its own axis) and
in the supply current which are highlighted by the analysis dynamic eccentricity (where the rotor is still turning upon the
are found to exist experimentally and indeed shown to be a
stator bore center but not on its own center). The causes of
function of the combined effect of both dynamic and static
eccentricity. Further vibration analysis is put forward to identify either type of rotor eccentricity are many, such as incorrect
which particular form of rotor eccentricity is dominant; hence bearing positioning during assembly, worn bearings, a bent
illustrating how faults can be identified in a motor using condition rotor shaft, operation at a critical speed creating rotor “whirl,”
monitoring techniques. etc. Eccentricity causes a force on the rotor that tries to
pull the rotor even further from the stator bore center. ln
I. INTRODUCTION the case of static eccentricity this is a steady pull in one
ONDITION monitoring strategies for large cage induc- direction. This makes the unbalanced magnetic pull (UMP)
tion motors such as those proposed by Thomson et al. difficult to detect unless specialist experimental equipment
[l], [2] usually detect the presence of faults in the machine such as that used in either [4] or [5] is utilized, which is not
by monitoriiig high-frequency vibration and terminal current possible for motors in service. Dynamic eccentricity produces
components over a period of time. These strategies can be used a UMP which acts on the rotor and rotates at rotor rotational
to detect rotor eccentricity. The algorithms used to identify velocity. This makes the UMP easier to detect by vibration
particular frequencies that are suitable for monitoring are monitoring at the rotational velocity frequency, as illustrated
derived usually using airgap permeance modulation of the in [6]. Both types of eccentricity cause excessive stressing of
stator MMF waves. The airgap permeance series consists of the machine and greatly increase bearing wear. In addition,
several Fourier series accounting for the rotor eccentricity and the radial magnetic force waves produced by eccentricity can
both rotor and stator slotting effects. Detailed derivations of also act on the stator core and subject the stator windings
these algorithms are given by Yang [3]. However the high to unnecessary and potentially harmful vibration. It is also
frequency current and vibration components usually include feasible for a rotor-to-stator rub to occur with consequential
slotting permeances, this can cause problems if the motor damage to the core and windings and the rotor cage. This may
cannot be dismantled to find the number of the stator slots lead to insulation failure of the stator windings or breaking
and rotor bars and this information is not freely available. of rotor cage bars or end-rings and hence a costly repair in
This paper will study the effects of combined dynamic and a large high-voltage induction motor. Therefore, the on-line
static rotor eccentricity on the airgap fields without recourse to diagnosis of rotor eccentricity i s highly desirable to prevent
slotting effects. From the theoretical model put forward, lower serious operational problems.
frequency components in the current and vibration spectra are Previous studies concentrated on either type of eccentricity
and rarely considered both types being present simultaneously.
Paper IPCSD 96-43, approved by the Electric Machines Committee of the
IEEE Industry Applications Society for presentation at the 1995 IEEE Industry To some extent, this is an artificial situation. It is well
Applications Society Annual Meeting, Lake Buena Vista, FL, October 8-12. known that there will be an inherent level of static airgap
Manuscript released for publication July 24, 1996. eccentricity in 3-phase induction motors due to manufacturing
D. G. Dorrell is with the Department of Engineering, The University of
Reading, Reading RG6 6AY, U.K. and assembly methods. In practice, this means that when
W. T. Thoinson is with the School of Electronic and Electrical Engineering, dynamic eccentricity occurs, then both types of eccentricity
The Robert Gordon University, Aberdeen AB10 lFR, U.K:
S. Roach is with Scottish Nuclear, East Kilbride G74 SPR, U.K. exist together. If there is always some degree of both types
Publisher Item Identifier S 0093-9994(97)00990-0. of eccentricity, the theoretical analysis of the airgap harmonic
0093-9994/97$10.00 0 1997 IEEE
DORRELL et al.: ANALYSIS OF AIRGAP FLUX,CURRENT, AND VIBRATION SIGNALS 25

fields and the associated rotor and stator surface MMF waves by permeance modulation of the fundamental MMF wave by
can be expanded from that usually put forward in studies of the rotor eccentricity. The magnitude of the sideband currents
rotor eccentricity. is very much dependent on the degree of both dynamic and
It is interesting to note that the UMP force characteristics for static eccentricity and also the damping of the modulated field
a cage induction motor when operating in the steady-state have components by the rotor cage.
been recently shown to be quite different. The experimental
work carried out by Bradford [4] and Dorrell[5] into UMP due
A. Application of Ampere’s Circuital Law
to static eccentricity showed the UMP to increase rapidly as
the motor is loaded from no-load. However, Dorrell, in another The motor can be modeled in terms of rotating airgap field
study into the effects of dynamic eccentricity [6], found that harmonics and stator and rotor surface MMF harmonic waves.
the UMP due to dynamic eccentricity decreased as the motor Assuming sinusoidal variation of the stator MMF wave, the
is loaded from no-load. The effects of combined static and current density on the stator inner surface can be expressed as
dynamic eccentricity on UMP are, therefore, difficult to assess
without more experimental work. However, this paper will
js(y, t ) = J s sin(wt -pky) (1)
concentrate on the detection of dynamic and static eccentricity where Js is the peak of the current density, w is the supply
by monitoring of the low-frequency sideband currents of the frequency (in rads), IC is the inverse of the average airgap
line current, although some stator vibrational components that radius and y is the linear distance round the airgap circumfer-
are associated with nodal stator vibration and UMP will be ence from some base point (such that liy = angle round the
studied. airgap from the base point).
An initial investigation into the combined effects of both Reference [5] derives the airgap field components from
types of eccentricity was reported on by Thomson [7]. It Ampere’s circuital law. If it is assumed that the flux crosses
was found that the magnitudes of these sideband currents are the airgap normally, even with an eccentric rotor, then the
dependent on both the dynamic and static eccentricity. Stavrou airgap field can be derived as in [5] from
and Penman 181 observed sideband currents in a study of
static eccentricity, they assume that they are due to equalising
circulating currents in stator windings which are connected in
t ) = A(!/> t ) / P 03’(Y > t >.dY (2)

parallel. Reference [5] shows these circulating currents are where A(y, t ) is the airgap permeance.
of supply frequency if the rotor is statically eccentric and
that it is dynamic eccentricity that would induce damping
B. Airgap Permeance
currents with frequencies similar to those observed. Salon
et al. [9] also show results that confirm this is indeed the The airgap is usually constant with length g. but if both
case. Also, these circulating damping currents would not be static and dynamic eccentricity are present then the airgap
observed in the line current spectrum since they are circulating length can be expressed as
currents within the stator windings. It can be hypothesized that
g(y, t ) = g[1 - 6 s cos(ky) - S d c o s ( u T t - Icy)] (3)
they may have been looking at sideband currents due to the
interaction of a rotor with a high degree of static eccentricity where 6, is the degree of static eccentricity, 6 d is the degree
and some inherent dynamic eccentricity. This will be shown of dynamic eccentricity and wT is the rotor rotational speed
later. such that
(1 -
w, = - (4)
P
11. THEORETICAL
ANALYSIS
6d < 1. Inversion
with slip s. To avoid a rotor-stator rub, 6,i-
Freize and Jordan [lo], Fruchenicht et al. [ll], and Kron of the expression for the airgap length leads to the airgap
[ 121 have shown that dynamic eccentricity produces airgap permeance being expressed as a Fourier series. However, to
field components varying with respect to time at fs f f, aid clarity we shall assume that both the dynamic and static
Hz at any point and with corresponding p & 1 spatial pole- eccentricity are low so that we can write
pairs in addition to the usual p pole-pair field (where fs
is the supply frequency, fr is the rotational frequency of
the rotor and p is the fundamental pole-pair number of
the motor). If the stator winding is assumed to be series This expression can then be used to obtain the stator airgap
connected, these additional fields should not induce EMF’S field components using Ampere’s circuital law. The coeffi-
into the stator winding due to the mismatch of field pole cients of the Fourier series produced by inversion of the airgap
numbers and the stator Fourier winding harmonics. However, length when only static or dynamic eccentricity is present (i.e.,
current components have been observed in the stator current either 6, or 6d in (3) is zero) are shown in [6] where it can
spectrum with frequencies corresponding to f s 4~ f, Hz [7], be seen that, at 40% eccentricity, the second harmonic is still
[8]. This is due to some inherent static eccentricity being only 20% of the first harmonic and also that the first harmonic
present with a dynamically eccentric rotor. This can be shown coefficient has a magnitude of about 0.45 on a normalized
26 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 33, NO. 1, JANUARYIFEBRUARY 1997

characteristic. Therefore, it can be hypothesized that ( 5 ) should The rotor currents are produced by the EMF’s induced into
be a reasonable approximation at 40% eccentricity-as used the rotor by the stator electric field as detailed in [6].The stator
in the experimental verification. electric field is derived from the stator magnetic field in (7) and
is given in the Appendix. The magnitudes of the electric field
C. Stator Airgap Field Components components and the rotor harmonic impedances will govern
the rotor harmonic current densities given in (9). Over a typical
Combining (1)-(3) yields the expression operating range where s << 1 then static eccentricity induces
1-10JS
p 3Z 1 MMF waves in the rotor with current frequency of
b,(y, t ) = -cos(wt - pky) approximately w / p . However, dynamic eccentricity leads to
kP9
+ +
x (1 S, COS(^^) Sd C O S ( U , ~- k y ) ) . (6) slip frequency current components for the p 3Z 1 MMF waves.
It can be hypothesized that the MMF p k 1 waves on the
Expanding and rearranging, the stator airgap field is rotor surface will increase rapidly in magnitude when the
motor is loaded for waves associated with dynamic eccentricity
b s ( y , t ) = BP,cos(wt - p k y ) compared to those associated with static. This is borne out in
+ cos(wt - ( p - 1)ky) [5] and [6].In [5] the UMP due to static eccentricity increases
rapidly as the motor is loaded from no-load, showing that the
+ +
Bg; cos(wt - ( p 1)ky) auxiliary p & 1 stator fields are not increasingly being damped
+ I?;:; cos((w - w r ) t - ( p - 1)ky) as the line current increases sharply. This is in contrast to the
results in [6] where the UMP drops as the rotor is loaded
+ ~“s:d; +
COS((W +
w r ) t - ( p 1)ky) (7) from no-load, only beginning to increase close to full-load
where speed. It should be pointed out that these experiments were
carried out at reduced voltage and this contrast may be less
clear at full voltage due to saturation effects. However, all
rotor MMF waves should increase to some degree with load
over the steady-state operating range due to the increase in
These fields, in tum, will induce EMF’s into the rotor cage
and hence produce rotor surface MMF waves. By inspection, terminal current.
it can be seen that both static and dynamic eccentricity produce
their own p & 1field components, the only divergence between E. Non-Supply-Frequency Current Components
them being the rotational velocity. It should also be noted that
The rotor MMF waves in Section 1I.D. will generate their
this analysis is for a motor with p > 1; i.e., two pole motors
own field components. Modulation by the airgap eccentricity
represent a special case where the p - 1 field is a homopolar
will lead to p pole-pair airgap field components that induce
flux which may or may not flow.
non-supply-frequency EMF’s in the stator windings. Concen-
trating only on these field components, they are obtained
D.Rotor MMF Waves by application of Ampere’s circuital law again. Converting
the relevant rotor MMF waves to stator coordinates and
To obtain an expression for the rotor MMF wave com-
multiplying by the 1st airgap permeance harmonic gives
ponents it is necessary to convert the expression for the
stator field into the rotor reference frame using the coordinate
relationship br (V) t> j
(#U)

[
(#P*l)

ky = w,t t ky’ (8) = Ji.: cos(wt - (1,- 1)kY - P42)

q P - 1)g
where y‘ is the distance around the rotor circumference from a

1
Jp+l
rotor reference point. The stator field components will produce
rotor currents such that the rotor surface MMF wave can be
+ k ( PR+( s1) ) g cos(& - (p + 1)ky - p & )
expressed as x cos(w,t - ky)

+ J;::) sin(swt - ( p - 1)ky’ - ~ $ 4 ) It can be seen that rotor MMF waves due to static eccentricity
are modulated by the dynamic eccentricity airgap permeance
+ $7;) sin(swt - ( p + 1)ky’ - ~ 4 5 ) . (9) and vice versa, in order to obtain the supply frequency
 ~

DORRELL et al.: ANALYSIS OF AIRGAP FLUX, CURRENT, AND VIBRATION SIGNALS 21

sideband currents. Working through and extracting only the then the airgap field can be written as
relevant p pole-pair field components gives
b(9,t ) = BPcos(& - pky - 8,)
(ZW) + qsjlcos(wt - ( p - 1)ky - 0,)
P + B:' cos(wt - ( p -t- 1)ky - 83)
7P-1
+ B:;' cos((w - w,)t - ( p - 1)ky - 0,)
+ B-$' cos((w + w,)t - ( p + 1 ) k y - 85). (13)
Squaring [ 13) will lead to the rotating force wave components
7P-1
acting on the stator surface. By combining (12) and (13), the
rotating radial force per unit area at any point approximates to
(14), shown at the bottom of the page.This expression assumes
the main field to be much larger than the field combnents due
to eccentricity and therefore some of the components have
been neglected in the squaring process.
Hence p pole-pair fields with frequencies of w f w, are Static eccentricity will produce additional 100-Hz vibration
present. These fields will induce nonmains-frequency EMF's components only (for a 50-Hz supply). Since the main field
in the stator winding which are measurable by monitoring of already creates a 100-Hz vibration due to the associated
sideband currents about the line current. rotating force wave, and that the magnitude of this field is
dominant, then it may be expected that the increase in 100-Hz
vibration due to static eccentricity may be noticeable but not
F. Vibration Monitoring great. This can be further explained by reference to (7) a d
As previously stated, dynamic UMP causes a rotating radial (9). The p f 1 stator airgap fields are scaled by the degree
force on the rotor, this is due to the interaction of space- of eccentricity from the fundamental p pole-pair field but (9)
harmonic field components with pole-pair numbers differing shows that, at synchronous speed, while the p pole-pair is
by one [ 5 ] , [ 6 ] . In addition, there may be other rotating undamped by the rotor, the p f 1 flelds do induce EMF's
force waves acting on the stator bore surface which are a into the rotor and are, therefore, damped. Hence, in (13), the
function of the rotor eccentricity. These can be monitored magnitudes of the second and third terms which are due to
using accelerometers fixed to the outer casing of the motor. static eccentricity should be much smaller than the magnitude
This will help identify if eccentricity is present. It should be of the fundamental field in the first term. A recent study by
remembered that mechanical imbalance of the rotor will also Dorrell [I31 showed that the UMP due to static eccentricity
cause rotating velocity vibration. Therefore, it may not be is steady from no-load to full-load if only the fields in (13)
possible to identify dynamic eccentricity solely by monitoring are considered. Therefore, if the fundamental field decreases
the rotational velocity vibration. slightly as the load is increased, then the p f 1 fields should
The attractive force between the rotor and stator at any show an increase in magnitude as the motor is loaded. This
particular point is usually approximated to should effectively keep the 100-Hz vibration steady as the
motor is loaded.
B2 The 100-Hz vibration component will be unpredictable due
cr=-
2Po to the phase angles 0 and also the varying spatial harmonic
numbers of the different force waves. Hence at some points the
where B is the airgap field at any particular point. If the major force waves will add while at other points they will subtract.
space-harmonic components are considered to be those in (7) Therefore, the vibration monitoring of the 100-Hz component

1
a(y,t) x -
2Po
(Bp)2
[- 2
+
cos(2wt - 2pky - 281) Bp x [BTSj1cos(2wt - (2p - 1)ky - 01 82) -
+ + +
BTZ1cos(2wt - (2p 1)ky - d l - 0 3 ) B&l(cos((2w - u,)t - (2p - 1)ky - 81 - d 4 )
+ cos(w,t - ky - 81 - 6 ' 4 ) )
~ - (2p + 1) x ky - el - e5) + cos(wrt - ky + o1 + e,))]
+ B : $ ' ( C O S (+( ~w,>t
(4 os(2w,t - 2 k y
+ gP-lBP+l
(4 + 84 - 8,)
28 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 33, NO. 1, JANUARYIFEBRUARY 1997

to assess the degree of static eccentricity may not be consistent 111. EXPERIMENTAL
VERIFICATION
and clear. However, this is investigated later. Experimental verification was carried out on a standard
Dynamic eccentricity produces other non-twice-supply fre- 415 V, 20.5 A, 11 kW, 4-pole cage induction motor connected
quency vibration in addition to the rotational velocity vibra- to a 50 Hz supply. The stator consisted of 36 slots containing
tion. In (14), the force waves are derived from the sums a double-layer series-connected 3-phase winding with short
and differences of the cosine terms in (13). and the non- pitching of one slot. The rotor was a cast aluminum type with
twice-supply frequency terms due to dynamic eccentricity are 28 bars and a skew of one rotor slot pitch. The bell-ends of
shown to be f r (which cause UMP), 2fs & f r , 2fT, and the motor were removed and the rotor mounted on pedestals.
2 ( f s ;t f r ) . There is also a small 2 f 3 vibration but this may Movement of the stator on its mountings with respect to the
be difficult to assess due to reasons previously mentioned. rotor allowed the variation of static eccentricity, this was
Equation (9) shows that the fourth and fifth terms of (13) measured accurately using feeler gauges though a tolerance
are wholly stator fields at synchronous speed, i.e., they are of about 8% of the mean airgap length was incurred with
not damped by the rotor. This is illustrated in 1131 where the measuring equipment. Dynamic eccentricity could also be
the UMP is much higher for dynamic eccentricity than for varied. This was obtained by turning down the shaft ends
the equivalent UMP due to static eccentricity at no-load. where the bearings are located and fitting sleeves. The axes
The UMP then decreases as the motor is loaded because of the inner and outer surfaces of these sleeves were not
the fourth and fifth terms in (7) are increasingly damped by common, therefore creating dynamic rotor eccentricity when
the rotor. Therefore the non-twice-supply frequency vibration inserted between the rotor shaft and the bearings. This time
teims due to dynamic eccentricity should be clearly defined the measurement tolerance level was much less at about 4% of
as a function of eccentricity since they should be small with a mean airgap length since more accurate clocking techniques
nominal concentric rotor and increase rapidly with eccentricity, could be used. The degree of eccentricity down the axial length
especially at no-load where the rotor damps the p i 1 pole- of the motor was kept uniform.
pair fields lightly. As the motor is loaded it is expected Tests were carried out at full rated voltage with varying
that the magnitude of these vibrations should decrease in the degrees of both static and dynamic eccentricity. Three load
same manner as the UMP. Again, experimental verification is points were used and the average taken of the line current
necessary for these suppositions. (no variation in line current due to eccentricity was observed).
The UMP in the direction where y = 0 (direction of static With the motor uncoupled the line current was 10 A, at a
eccentricity) can be calculated from slip of 0.01 the current was 13.6 A and at full load ( s =
0.02) the current was 20.2 A. These give lower sideband
current frequencies of 25, 25.25, and 25.5 Hz and upper
sideband frequencies of 75, 14.75, and 74.5 Hz, respectively.
The sideband currents themselves were measured using a
which leads to the expression showing that the UMP is due spectrum analyzer. This had averaging facilities and recording
to the interaction of fields with pole-pair numbers differing of the current magnitudes was in decibels, converting to
by one. The force o ( y , t ) is made up of all the stator and milliamperes by reference to the main current magnitude.
rotor airgap components as shown in (14) plus higher space The current sampling was vi;* current transformers and a low
harmonics due to other winding harmonics [6]. The UMP inductance resistor. Vibration1 monitoring was obtained using
due to the product of two field components with pole-pair an accelerometer with the spectrum analyzer used to process
numbers differing by one and when the rotor exhibits dynamic the signal.
eccentricity the UMP will be of the form:
A. Variation of Sideband Currents With the Presence
(16) of Both Static and Dynamic Eccentricity
Figs. 1 and 2 show the variation of the sideband current
if the phase angles in (14) are neglected. Static eccentricity magnitudes with both types of eccentricity present. It is
leads to quite clear that these sideband currents are essentially due
to the combined effects of static and dynamic eccentricity.
It can be seen that there is a decrease in lower sideband
(17) current from no-load to full-load, when there is either a
nominal zero dynamic or static eccentricity, but there is
where p , is the fundamental pole-pair number or a higher little change when loading thc motor if there are substantial
winding harmonic (i.e., n = 1,5, 7,11, etc.). The total UMP degrees of both types of eccentricity. However, while the upper
is then the sum of all the different combinations of stator sideband characteristic vanes with dynamic eccentricity such
and rotor field products meeting the requirements of (16) or that the no-load magnitude is much higher than the full-load,
(17). These show that the dynamic UMP manifests itself as static eccentricity causes the sideband current to be almost
a rotational speed vibration and that static UMP is a steady constant with motor load. This is also shown in Fig. 4. If
pull. both types of eccentricity are present then the no-load upper
 DORRELL et al.: ANALYSlS OF AIRGAP FLUX, CURRENT, AND VIBRATION SIGNALS
29

-*-
-+- 20% 75Nz
40% 25Hz
-+- 40% 75Hz

Fig. 1. Variation of sideband currents with combined static and dynamic eccentricity at no-load.

0.14
fl
Dynamic ecc. & freq. I
-*- 0 46 25.5 Hz
04674.5Hz
20 % 25.5%
-+- 20% 74.5Nz
+
-+- 40% 25.5Hz
40% 74.5Hz

t:
e

1------

I I I I I
50 %
!
0% 8.3 I 45 of static
16.7Degree 25 %eccentricity
33.3 % 41.7 X

Fig. 2. Variation of sideband currents with combined static and dynamic eccentricity at full-load.

L u u u a w b WILU me lower sideband characteristic.

.c.llo Levels of Static or Dynamic Eccentricity
There is clear evidence for the potential of condition moni-
Sideband currents were observed even with inherent (nom-
toring to identify increasing rotor eccentricity over a period of
inally zero) levels of static or dynamic eccentricity. Figs. 3
time by observing the change in magnitudes of these sideband
and 4 show an increase in the sideband currents with one type
currents. Whether it would be possible to identify which is
the major type of eccentricity present should be the basis of eccentricity when the other was nominally set to zero. It
can be seen that the magnitudes are nenerallv hipher
- - - ~ - - -in
~ 2
for f w h r research. In addition, there may be other causes .( -I* I AB. J,

nf ":,loL..-,l -------
where there was nominally zero static eccentricity, because of
I

=u=urllJu wrenis not associated with rotor eccentricity.

--- U . . ~ l l , l , G l l l
I
"I

While there is a contrast in sideband characteristics it should a small degree of static eccentricitv oresent
.---
I .v
~ I I Ptn % K - - - - - ~
hp romamL,--~ L
L ~ l l L I J ~ L ine motor 1s small with only seven rotor
l
L-A
~ l l l u ~ ~ ~ ~ tolerance ("inherent eccentricity"). This is to be expected since
bars/pole and a large degree of skew. Condition monitoring the rotor had been specially machined to remove dynamic
is usually applied to large motors with much higher bars/pole eccentricity whereas static eccentricity is removed by stator-
numbers and no rotor skew. rotor alignment using feeler gauges-clearly leading to more
30 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 33, NO. 1, JANUARY/FEBKUARY 1997

Fig. 3 . Dynamic eccentricity with inherent static

I I I I I
0% 8.3 % 16.7 % 25 % 33.3 % 41.7 % 50 %
Degree of Static Eccentricity
Fig. 4. Static eccentricity with inherent dynamic.

inaccuracy. It is interesting that the upper sideband in Fig. 3 As previously mentioned, there is a possibility of us-
and lower sideband in Fig. 4 follow the same characteristic ing condition monitoring to identify the type of eccentricity
with a large decrease in magnitude with initial loading then present. For this motor, if bearing wear causes increasing static
levelling out with further loading. Equation (1 1) shows the eccentricity this would be identified as an increasing lower
sideband currents to be due to different MMF waves, i.e., the sideband current with increasing eccentricity-this sideband
lower sideband current is due io a six-pole MMF wave created also decreases as the motor is loaded. However, the upper
by static eccentricity and a two-pole MMF wave created by sideband increase with eccentricity is less pronounced and
dynamic eccentricity; whereas, the upper sideband current is remains almost constant with load. If a shaft fault causes
due to a two-pole MMF wave created by static eccentricity bending and hence dynamic eccentricity, the upper sideband
and a six-pole field created by dynamic eccentricity. This still increases with eccentricity but loading causes a large drop
may be the reason for the similar characteristics of different in magnitude. Monitoring of the casing vibration may also be
sidebands. an aid to condition monitoring as discussed in the next section.
 ~

DORRELL et al.: ANALYSIS OF AIRGAP FLUX, CURRENT, AND VIBRATION SIGNALS 31

25 ,
0 % Dynamic eccentricity
+
-. 20 % Dynamic eccentricity
n

ecn2 0 -
-+-- 40 % Dynamic eccentricity
\

a
B
~ 1 5 -
3 ,
8
v
si
.g 10 -
B
B>

0 0.01 0.02
Slip
Fig. 5. UMP vibration due to dynamic eccentricity

C. Vibration Monitoring of UMP due to Dynamic Eccentricity shows a general small increase in rotational velocity UMP as
Dynamic eccentricity causes a rotating force and monitoring the static eccentricity is increased but it is very small.
of the rotational speed vibration may aid further identification
of rotor eccentricity. Figs. 1-4 showed that, whatever the in- D. Casing Vibration
crease in eccentricity, there is an increase in sideband current. Static eccentricity can be detected by monitoring the twice-
If there is an increase in the rotational velocity vibration ‘then supply frequency vibration, in this case 100 Hz. Fig. 7 shows
this should be associated with dynamic eccentricity. If there is the variation of the 100-Hz vibration with both static and
little change in rotational speed vibration even with increasing dynamic eccentricity at no-load and full-load. It can be seen
sideband current magnitudes then this may indicate increasing that there is an increase in vibration with static eccentricity
static eccentricity. Figs. 5 and 6 illustrate this. The vibration though the no-load results seem variable. However, the full-
was monitored at a point on top of the motor casing, at the load results seem to verify the discussion in Section II-F. where
center of the core length. The UMP vibration due at zero slip it was proposed that dynamic eccentricity has little effect on
is 25 and 24.75 Hz at 0.01 slip and 24.5 Hz at 0.02 slip. the twice-supply-frequency vibration.
The vibration is on a linear scale rather than in decibels to The variation of the 2f, - f r vibration is shown in Fig. 8.
try to assess the change in UMP. This assumes that the UMP These where obtained at no-load (where the slip is virtually
is approximately proportional to the vibration level since the zero) and full-load. It can be seen that there is a very marked
change in frequency of the components identified as suitable increase in vibration with dynamic eccentricity and some
for monitoring is less than 1 Hz over a steady-state range of increase with static eccentricity.
no-load to full-load. It should be remembered that it is not +
The 2 f s f r vibration components are shown in Fig. 9.
strictly correct to correlate the vibration magnitudes directly Again, the vibration is found to increase markedly with dy-
to the force-wave magnitudes between different vibration namic eccentricity and also, to degree, with static eccentricity.
components since the stator frame will react differently to The increase with static eccentricity was found to be much
force-waves with varying pole-numbers and rotational speeds. greater than that found with the 2 f s - fr vibration as can be
However, the results in [13] suggest that UMP is proportional seen by comparison of Figs. 8 and 9.
to the rotational velocity vibration since the change from no- Figs. 8 and 9 also show a decrease in vibration when there
load to full-load is 25 Hz-24.5 Hz. Resonant frequencies for is some degree of dynamic eccentricity present as the motor
this particular stator frame have been found to be present only is loaded (as expected). Also, Fig. 7 shows that load has
at much higher frequencies [14]. little influence on the 2f, vibration as discussed earlier. It
Fig. 5 shows a great increase in rotational velocity vibration was found however, that static eccentricity seems to have
with dynamic eccentricity. It also shows a drop in vibration a large influence on the 2fs + f r vibration components but
as the motor is loaded, i.e., the UMP is decreasing as the load little influence on the rotational velocity vibration. Whereas
is increasing. This was also found in [6] with the same motor the latter was expected, the former was difficult to assess;
but at reduced voltage. It is interesting to note that the drop it should be noted that the rotational velocity is caused by
in UMP is not as pronounced at full voltage as that found at a distinct radial decentring force acting between the stator
reduced voltage. This is probably due to core saturation. Fig. 6 and rotor, but the other vibrational components are rotating
32 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 33, NO 1, JANUARYFESRUARY 1997

20 % Dynamic 0 Slip
-- 20 % Dynamic 0.01 Slip
*
T
20 % Dynanic 0.02 Slip
*25
4 q b
* -+&- 40 % Dynamic 0 Slip

_ _ _ _ - ---0
--
-..*... 40 % Dynamic 0.02 Slip
+-------

..................................
.........................

0
0% 8.33 % 16.7 % 25 % 33.3 %
Static eccentricity
Fig. 6 . Effect of static eccentricity on UMP vibration.

80

n
70
VI
\

260

50
;
;EI
W
40
F1
.4
a 30
44 -It- O%DynamicOSlip
$0
+ 2020 %X Dynanic 0 SEp
10 -+-
-+-
Dynamic 0.2 Slip
40 Dynamic 0 Slip
%
-+-40 % Dynamic 0.02 Slip 1
" I I I I I I I
0% 8.33 % 16.7 % 25 % 33.3 % 41.67 % 50 %
Static eccentricity
Fig. 7. Vibration of twice-supply-frequency vibration

radial force waves with several different nodal patterns. This (13) (which are increasing with static eccentricity) with terms
probably accounts for the clarity of the results from the 4 and 5; therefore, the 2 f s f f T vibrations increase with
other vibrational components being poorer. However, they static eccentricity (these products were not included in (14)
would help identify if the rotational velocity vibration is when (13) was squared since they were assumed small). This
due to UMP caused by rotor eccentricity or by mechanical hypothesis relies on the assumption that the f r vibration is
imbalance. dominated by UMP and that the interactions between terms 2
Noting that there was an increase in 100-Hz vibration with and 3 and 4 and 5, which produce f r vibration but not UMP,
static eccentricity, it could be argued that, if the main p can be neglected. Hence accounting for the differences in the
pole-pair field is constant with increasing static eccentricity, characteristics in Figs. 6, 8, and 9.
then the f r vibration will be constant with increasing static
eccentricity (since it is caused by UMP due to the interactions
of terms 1 with 4 and 5 of (13); these are functions of dynamic IV. CONCLUSION
eccentricity) as shown in Fig. 6. However, the 2f, I f r This paper identified the low pole-number airgap field com-
vibrations also include the interactions of terms 2 and 3 of ponents which are produced in a cage induction motor when
DORRELL et al.: ANALYSIS OF AIRGAP FLUX, CURRENT, AND VIBRATION SIGNALS 33

- - 0 X Dynamic 0.02 Slip, 75.5 Wz

n 4 20 % Dynamic 0 Slip,75 Hz
a20- -+- 20 96 Dynamic 0.02 Slip,75.5 HZ
r -+- 40 46 Dynamic 0 Slip, 75 Hz

Fig. 8. Variation of 2f, - fr vibration.

25
, *

-&-0% Dyn Os, 125Hz

/a--
0% Dy 0.02~124.5Wl
v
’ +-
-4--
20% Dyn Os 125Hz
/’
20% Dyn 0.02s 124.5Hz
------*‘ --+- Dyn Os 12%~
E
rf! -+--40%
40% Dyn 0.02s 124-

01U
0%
I
8.33 %
I
16.7 56
I
25 % 33.3 56
r I
41.67 %
1
50 96
Static eccentricity
Fig. 9. Variation of 2fs + fr vibration.

the rotor exhibits both dynamic and static eccentricity. From APPENDIX
this it was shown that non-supply-frequency currents will Equation (7) can be expressed in terms of the rotor reference
appear in the current spectrum and additional casing vibration frame using (8):
will be produced. The theoretical results and experimental
work put forward in this paper show that it is possible to use
condition monitoring to identify rotor eccentricity in a cage
induction motor without using slot-passing frequencies. The
analysis showed that dynamic and static eccentricity should be
considered simultaneously since it is highly likely that there
would be some degree of both types of eccentricity present.
Monitoring of sideband currents and certain low frequency
+ B g ; cos (-1 + (Pp + 1)s “t - ( P + W Y ’ - P 4 3
1
casing vibration components should lead to clear identification + Bps;dl)cos(swt - ( p - 1)Cy’ - p$4)
of increasing dynamic eccentricity and a good indication of
increasing static eccentricity. + B”srd:cos(swt - ( p + 1)ky’ - p$s) (AI)
34 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 33, NO. 1, JANUARYlFEBRUARY 1997

where the magnitudes of the fields all increase with slip. [7] W. T. Thomson, “On-line current monitoring to diagnose shaft misalign-
ment in three-phase induction motor drive systems,” in Proc. ICEM’94,
EMF’s will be induced into the rotor due to the airgap electric Paris, France, 1994, pp. 238-243.
field associated with the airgap flux. An expression for the [SI A. Stavrou and J. Penman, “The on-line quantification of air-gap
airgap field can be obtained using eccentricity in induction machines,” in Proc. ICEM’94, Paris, France,
1994, pp. 261-266.
[9] S. Salon, M. DeBortoli, D. Burrow, and C. Slavik, “Calculation of
circulating current between parallel winding in induction motors with ec-
centric rotors by finite element method,” in Proc. ICEM’92, Manchester,
U.K., 1992, pp. 371-375.
where x is the axial direction. Therefore the airgap electric [lo] W. Freize and H. Jordan, “Unilateral magnetic pulls in three-phase
field is machines,” EZ-A, Translation C. E. Trans. 7836, vol. 83, no. 9, 1962.
[11] S. Fruchericht, E. Pittuis, and H. Seinsch, “A diagnostic system for
three-phase asynchronous machines,” in Proc. 4th EMDA Conc, IEE
no. 310, London, U.K., 1989.
[12] G. Kron, Equivalent Circuits ofElectric Machines. New York: Dover,
1967.
[I 31 D. G. Dorrell, “The sources and characteristics of unbalanced magnetic
pull in cage induction motors with either static or dynamic rotor
eccentricity,” in Stockholm Power Tech. Con$, Stockholm, Sweden, June
18-22, 1995, pp. 229-234.
[14] M. Lucas, “Vibratory response of stator cores of large induction motors
operating in an offshore installation,” M. Phil. thesis, CNAA, Robert
+ E!&,, cos(swt - ( p - 1)ky‘ - p&) Gordon’s Institute of Technology, Sept. 1986.

+ E;&\ cos(swf - ( p + I)ky’ - p4s) (A31

where the coefficients, in terms of the magnetic field and slip, David G. Dorrell (M’96) was born in St. Helens,
are U.K., in 1965. He received the B.Eng. (Hons.)
degree from the University of Leeds, Leeds, U.K.,
in 1988, the MSc. degree from the University of
Bradford, Bradford, U.K , in 1989, and the Ph.D. de-
gree from the University of Cambridge, Cambridge,
U.K., in 1993.
From 1994 to 1996, he was a Lecturer with The
Robert Gordon University, Aberdeen, U.K. He is
-sw B$A currently with the University of Reading, Reading,
n U.K., in the position of Lecturer. His main interests
are electrical machines and drives.

BY inspection, it can be seen that the electric field com-

ponents for the main field and the dynamic-eccentricityfields
William T. Thomson (SM’94) was born in
will increase with both slip and magnetic field but the static- Perthshire, Scotland, in 1946. He received the B.Sc.
eccentricity fields will approximately vary with magnetic field (Hons.) and MSc. degrees from the University of
only if the slip is much less than one, i.e., under steady-state Strathclyde, U.K. The M.Sc. degree was awarded
for research on vibration and noise in small power
conditions. The EMF’s induced into the rotor by the stator will electrical motors while employed by Hoover Ltd. as
be proportional to these electric field components. a graduate Research Engineer from 1973 to 1975.
He is currently Professor of Electrical Engi-
neering, The Robert Gordon University, Aberdeen,
REFERENCES U.K., His principal research interest is in condition
monitoring of electrical drive systems.
[I] J. R. Cameron, W. T. Thomson, and A. B. Dow, “Vibration and cunent Prof. Thomson is a fe:Ilow of the Institution of Electrical Engineers, London,
monitoring for detecting airgap eccentricity in large induction motors,” U.K., and a Chartered Engineer, U.K.
Proc. Inst. Elect. Eng.,, vol. 133, pt. B, no. 3, pp. 155-163, May 1986.
r21 W. T. Thomson, J. R. Cameron, and A. B. Dow, “On-line diagnostics of
large induction motors,” in Ebbrations and Audible Noise in kternating
Machines, R. Belmans, K. J Binns, W. Geysen, and A Vandenput,
Eds. Norwell, MA. Kluwer, 1988. Steven Roach received the B.Eng. (Hons ) in elec-
[ 3 ] S. J. Yang, Low-Noise Electrical Motors. Oxford, U K.: Oxford Univ. tronic and electrical engineering from The Robert
Press, 1981 Gordon University, Aberdeen, U.K. in 1995.
[4] M Bradford, “Unbalanced magnetic pull In a 6-pole induction motor,” He is presently with the Engineering Department,
Proc. Inst Elect. Eng., vol. 115, no. 11, pp. 1619-1627, 1968. Scottish Nuclear, East Kilbride, U.K., as a Graduate
[ 5 ] D. G. Dorrell, “Calculation of unbalanced magnetic pull in cage in- Trainee Engineer.
ductlon machines,” Ph.D. Dissertation, University of Cambridge, U.K ,
1993.
[6] -, “The effects of dynamic rotor eccentricity in cage induction mo-
tors,” in Proc. UPEC’94, University College Galway, Galway, Ireland,
1994, pp. 402405.