ETEP

Methods of ConditionMonitoring
and Fault Diagnosis for Induction Motors
0.V. Thorsen, M. Dalva

Abstract
The study gives a synopsis over condition monitoring methods both as a diagnostic tool and as a techniquef o r
failure identification in high-voltage induction motors. Methods of fault detection are reviewed and analyzed:
mechanical (vibrations, shockpulses, acoirstic, speedfluctuations), electro-mechanical (currents, surges, par-
tial discharges, leakagejluxes) and temperature-, oil particle-, gas analysis, and performance- and visual me-
thods. The study indicates how different c p e s of faults generate different frequency components in vibration
spectra, in motor supply current, in torque and in magnetic field, and relationships between fault types and
spectra have been developed. Supply current waveform analysis has been applied to demonstrate the detec-
tion of broken rotor bars and bearing faults.

1 Introduction Bearing faults can be caused by improper or not suf-

ficient lubrication, heavy radial- and axial stress due to
Within industry high-voltage motors have a broad shaft deflection. poor mounting, alignment or founda-
application area, for instance for the drive of gas com- tion. This stress may lead to vibrations and thereby to in-
pressors and pumps. In these important processes much creased wearing.
effort is laid on reducing the number of failures to an ab- Bearing faults can best be detected by vibration-,
solute minimum. Growing fault developments should be shock pulse- and acoustic emission monitoring.
warned about to better plan the maintenance. There are two types of air-gap eccentricities,static
Within off-shoreactivities the motors are often start- and dynamic. With static eccentricity the minimum air-
ed directly on-line. This leads to large start currents and gap position is fixed in space, while for the dynamic the
torque pulsations, and these conditions may contribute center of the rotor and the rotational center do not coin-
to end winding-, rotor- and bearing faults. cide, so that the minimum air gap rotates. Static eccen-
Voltage unbalance in the mains can also cause fault tricity leads to detectable frequency components in the
in the windings, and a relatively small voltage unbalance stator current.
can cause a many times higher current unbalance. Dynamic eccentricity causes side band components
In addition some motors are subject to the risk of in- around slot harmonics in stator line current frequency
trusion of water. spectra.
Several methods are available for fault detection. Static eccentricity by non-uniform air gap causes
The main groups are mechanical, like vibrations, shock unbalanced magnetic torque and vibration with frequen-
pulses, acoustic, speed fluctuations, and electro-me- cy of twice the supply frequency, and mechanical unbal-
chanicaf, like currents, surges, partial discharges and ance produces vibrations of rotational frequency, or
leakage fluxes. In addition come temperature-, oil parti- twice, three or four times that frequency. Misalignment
cle-, gas analysis, and performance- and visual methods. between motor and load causes vibrations of double
However, a high-voltage motor represents a com- rotor speed frequency [31.
plex electromagnetic system, and hence the faults and Stator faults may be due to inter-turn winding faults
fault reasons are so involved that a reliable fault detec- caused by insulation breakdown. This breakdown leads to
tion must be based on a combination of methods. unbalance in the stator accompanied by changes in har-
monic air-gap flux and current time harmonics. Because
the relative change in the supply current is small, it may
2 Faults and Reasons for Faults be difficult to detect a fault in this way [4]. The alternative
could be to detect the axially transmitted fluxes giving a
The various faults on motors are grouped as follows reliabledetecting of small unbalancesin magnetic- as well
[I]: as in electric circuits. In addition one should detect the ec-
centricity and phase-unbalance. So the stator faults are
- Bearings. most reliably detected by axial leakage flux monitoring.
- Stator windings. Rotorfaults are mainly breakage of joints between
- Rotor bars and -end rings. bars and end rings as a result of pulsating load or direct
on-line starting. The result is that the current increases
- Shaft or couplings. in the remaining bars with an increasing risk of more
- External devices. fractures. Rotor failures lead to torque pulsations, speed

ETEP Vol. 8. No. 5 , SeptembedOctober 1998 383

E TEP
fluctuations, vibrations, changes of the frequency com- stator slots and a 28 rotor slot die-cast rotor. The motor
ponents in the supply current and axial fields, combined developed its nominal full-load torque at 1435 min-'.
with acoustic noise, overheating and arcing in the rotor
and damaged rotor laminations. 3.1 Performance Monitoring

Performance monitoring is on-line measuring of the

3 Motor Condition Monitoring Methods motor parameters like supply voltage and current, input-
and output power, mechanical stress in couplings and so
Condition monitoring means continuously or peri- on. In weak supplies. as on board platforms and ships,
odically condition investigation of the motor. Sudden or unsymmetry in the supply voltage can occur, as a worst
unexpected changes in the parameters under monitoring case one can have single-phasing (loss of one of the
indicate significant changes in the condition of the phases). This will cause unsymmetrical motor currents,
motor. Condition monitoring is applied as a diagnostic and overheating of the windings, with insulating faults
tool for fault detection and as one of the bases for main- as a result.
tenance planning.
The alternatives for maintenance planning are 3.2 Vibration Monitoring
fixed maintenance cycles or maintenance after the fail-
ure has occurred. Condition monitoring offers the pos- Vibration monitoring is based on vibration trans-
sibility of a better planned maintenance, so that shut ducers. virtual measuring accelerometers of piezo-re-
downs and repair can be performed at the most suitable sistive types with linear frequency spectrum (micro-
time. phones are also in use). Vibrations are caused by mag-
Common methods for condition monitoring are netic-, mechanical- and/or aerodynamic forces. The
[3,5,61: measuring parameters can be displacement, velocity
and acceleration. Vibrations are measured directional-
Performance monitoring. ly, e. g. radially or axially, and the transducers are often
Vibration monitoring. placed on the bearings for detecting mechanical faults.
Shock-pulse monitoring (SPM), mainly for rolling However, by placing probes on the stator as well. i t is
element bearings. also possible to detect uneven air gap, stator-winding-
or rotor faults, asymmetrical power supply and unbal-
Acoustic emission monitoring. ances in the driven load [6]. This is based on the fact that
Speed fluctuations monitoring. any change in the normal tlux distribution in the motor
Current monitoring. will cause a change in the vibration spectrum. The
measured or registered results, either the whole spec-
Air-gap torque monitoring. trum or certain frequency components, are compared
Instantaneous stator power monitoring. with measurements registered when the machine was
Surge test winding diagnosis, direct current high po- new (or in a known condition). The sensitivity of the
tential testing and high-frequency surge testing. method is fairly high.
The signal processing is realized in a spectrum ana-
Partial discharges monitoring. lyzer, and in its simplest form it can be measuring of rms
Leakage flux (end winding leakage flux and axial values. However, rms do not give the possibility forfault
flux) monitoring. localization. The alternative is a more advanced fre-
Temperature monitoring. quency analyzer with that possibility.
It is possible to express the relationship between the
Oil particle monitoring. particularfault and the ciominnntfrequencies in the vi-
Gas analysis. bration spectrum for an electric motor.
Visual monitoring. As a fault in rolling element bearings begins to de-
velop there will be an increase in vibrations. When a roll-
Experimental Work ing element encounters a discontinuity,a pulse of vibra-
tion results with a frequency determined by the bearing
A test rig has been developed consisting of a test geometry. The rolling element bearing frequencies will
bench provided with a fourquadrant loading machine. be when assuming pure rolling motion [7].
This machine is connected to the mains via converter, - For an outer race defect:
and equipped with tacho-impulse generator and an elec-
tronic torque meter. J;,(in Hz) =(b/2)f,,[l-(db/dp)cosPI.
The test motor has transducers attached to sense - For an inner race defect:
voltage, current, power, flux and vibration.
Other instrumentation on the test rig are a Universal f; (inHz)=(b/2)fm[I +(db/dp)COSP].
Analyzer and a Dynamic Signal Analyzer, a two-chan-
nels FFT spectrudnetwork analyzer with frequency
- For a ball defect:
0 kHz to I 0 0 kHz. Control and processing of data are ac- fb (in HZ) = (dp/db)f,m[I -((db/dp)cosP)'].
complished by a PC.
The test motor is a standard three-phase, 50 Hz,
- For a cage defect:
4 kW, 380 V, 9.2 A, four-poles induction motor with 36 j i (in Hz)= ( 1/2)Jm[ 1 -(d~/cf,)cos~l.

384 ETEP Vol. 8,No. 5 , Septernber/October 1998

 CTCD
L I L I

Here are: fect parts. a crack on the innedouter race or on the roll-
h number of balls or rollers. ing elements in the bearing. A crack on the stationary
race gives pulses of the same strength, a crack on the ro-
fi contact angle of the balls on the races. tating race gives impulses that will vary with changes in
jnl, relativespeed(in s-I) between innerandouterraces,
ball diameter, the rolling-element load [ 7 ] .
bearing pitch diameter.
3.4 Acoustic Emission Monitoring
There will alsobevibrationsof very high frequencies.
The test machine has the following bearing data: Acoustic emission monitoring works with ultrason-
Shaft end bearing: ic as well as audible frequencies. Contact between roll-
b = 9, dh = 9.52 mm, d, = 53. I mm. ing elements with and without cracks generates waves
Fan end bearing: that propagate through the machine with the speed of
b = 9, d,, = 7.94 mm, dp= 45.25 mm. sound. The waves have little energy, but their high fre-
quencies can be detected by piezoelectric or capacitive
Nominal rotor speed: I 435/60 = 23.92 s-'. transducers. The method is particularly promising for
Calculating the frequencies for the bearings for test the detecting of bearing faults.
motor:
- Shaft end bearing: j ,= 88.3 Hz, J = 126.9 Hz. 3.5 Speed Fluctuations Monitoring
fh = 129. I H Z , fc.= 9.8 HZ.
Speed fluctuations monitoring can detect defects by
- Fan end bearing: ji, = 88.8 Hz, f;= 126.5 Hz. measuring fluctuations in the rotational period of the
f h = 132.3 Hz, = 9.9 Hz. motor. The method is particularly useful at rotor faults.
Unhtrltrncetl rotor can be detected by the increas- vibrations, air-pap eccentricity, rotor asymmetry, dam-
ing of vibrations with rotational frequency (n/60).or aged bearings/couplings and misalignment.
twice. three or four times that frequency. Misalignment The current in a normal rotor bar will fluctuate si-
can be detected by the increase of vibrations of twice nusoidally with slip frequency, and give a contribution
the rotational frequency of the shaft 171. Since magnet- to thedeveloped torque that will vary sinusoidally with
ic forces are proportional to the square of the flux den- twice the slip frequency. A defective bar will not con-
sity, vibrations will come into existence, also at normal tribute to the shaft torque. The shaft will consequently
run, with the frequency 25, and slot harmonic compo- slow twice per slip cycle. Summed up for the whole
nents as shown in the following ( , f ,is the supply fre- rotor. one can say that the resulting torque consists of
quency). two components. one constant and one that varies with
In the core vibmtion specrriim the rotor vibration twice slip frequency. Provided all rotor bars are in
slot harmonics will be:f; [ ( Z ? / p () 1 -s) f v ' ] . v ' = 0,2,4, order, the mean ofthe variable part ofthe torque is zero.
6, ...,and sidebands+2sfi. Broken rotorburs will mod- However. in most cases the motorhas a loadof variable
ulate stator core vibrations with f2sfi. torque, and the measuring instruments must be capable
Statorasymmetr_v(winding fault or unbalanced sup- of distinguishing between load tluctuations and fluctu-
ply/single phasing) produces vibration signals contain- ations of twice slip frequency indicating rotor faults.
ing 25, for asymmetric rotor also twice slip frequency The selective measurement therefore requires a refer-
side band ( f 2 s f i ) on the shaft frequency [ 7 ] . ence signal at slip frequency. [8] describes principles
Static eccentricity gives frequency components of of instruments capable of detecting faults on a single
twice supply frequency (2J ) in addition toslot harmon- rotor bar of a squirrel cage motor by speed fluctuation
ic components:fi [ ( v Z l ) (I - s ) / p f v ' ] , v' = 0. 2,4, ..., measurements.
v = 1,2,3, ...; Z? is the number of rotor slots andp is the Determination of slip and speed can be performed
number of pole pairs. indirectly by isolating the fundamental slot harmonic
Dynamic eccentricity also produces twice slip fre- frequency component from the supply current waveform
quency side bands on the shaft frequency in addition to [9]. With siniisoidal mmf the frequency of the slot har-
the slot harmonic components: fi [ ( v & -t nd)(1 - s)/p monic component for the stator current can be ex-
f v ' ] , n d = I , 2, ... [4,7]. pressed:
An alternative signal processing method for vibra-
tional analysis of bearing monitoring is the statistically Z2(1 -S,fi/Pffi.
treatment of the measuring data from accelerometers. A To take stator mmf time h a h o n k s into considera-
bearing that runs satisfactorily gives a Gaussian distri- tion, f i must be multiplied by h; the slot harmonic fre-
bution of the frequency components. Defects in the bear- quency for the stator current will then be:
ing disturb this distribution. This method gives an abso-
lute measure of the bearing condition, one does not need Zz(1 - s ) f i / p * h f i .
to know the original condition. h being the harmonic of the stator mmf time harmonics.
h = l , 3 , 5 , 7 .....
3.3 Shock-Pulse Monitoring The motor speed is given by:
n = 60fi ( I - s)/p.
Shock-pulse monitoring (SPM) applies piezo-
electric transducers fixed to the bearings to detect shock where the slip s is found from the measurement of the
waves caused by impact between moving parts and de- frequence components.

ETEP Vol. 8, No. 5. September/October 1998 385

ETEP
Frequency components caused by eccentricity will a) n 50Hz (1-2s) Diff. ( I +2s) Diff.
be: mid dB .50Hz dB dB .50Hz dB dB
I435 -0.1 45.7 -56.9 56.8 54.3 -58 57.9
1454 -1.8 46.9 -63.5 61.7 53.1 -58 56.2
v = 1,2,3, ..., h = 1,3,5,7, ...; 1471 -3.3 48.1 -66.2 62.9 51.9 -63.1 59.8

ndis an integer (eccentricity order number 0 for static ec- b) n 50Hz (1-2s) dB Diff. (1+2s) Diff.
centricity and 1,2,3,4, ... for dynamic eccentricity) [7]. mid dB .50Hz dB .50Hz dB dB
1435 -0.2 45.7 -48.3 48.1 54.3 -55 54.8
3.6 Current Monitoring 1454 -1.5 46.9 -51.1 49.6 53.1 -56.1 54.6
1471 -3 48.1 -55.4 52.4 51.9 -57.1 54.1
Current monitoring deals with the analysis of the C) n 50Hz (1-2s) dB Diff. (1+2s) Diff.
motor supply current because this is influenced to agreat m i d dB .50Hz dB .50Hz dB dB
extent by the motor condition. Current analysis may re-
I435 -0.4 45.7 -38.2 37.8 54.3 -42.7 42.3
veal damaged rotor bars as well as mechanical problems I454 -1.8 46.9 -43.1 41.3 53.1 -44.8 43
associated with rotating machines (MCSA - Motor Cur- 1471 -3.1 48.1 -49.8 46.7 51.9 -49.5 46.4
rent Signature Analysis) [ 101.
Faults in rotor bars will introduce different frequen- d) n 50Hz (1-2s) dB Diff. (1+2s) Diff.
cy componentsin the supply current to the stator. Because min-' dB .50Hz dB .50Hz dB dB
of asymmetry in the rotor circuit as a result of damaged I435 -0.1 45.7 -34 33.9 54.3 -38.6 38.5
rotor bars the rotor current will produce positive- and 1454 - 1.6 46.9 -39.3 37.7 53.1 -39.5 37.9
negative-sequencerotormmf at frequencies sol. Here a 1 1471 -3.1 48.1 -45.2 42.1 51.9 -43.6 40.5
is theangular frequency of the supply fundamental.
One assumes as a starting point that the mmf is sinu- Tab. 1. Overview of the results by measurements on the test
motor (Diff.: Difference)
soidally distributed. Positive rotor mmf experiences an a) Sound motor
angularspeedsw, + wl(1 -s)= a1relative tostator, while b) One broken rotor bar
angular speed of negative sequence rotor mmf relative ro c) Two broken rotor bars
stator is soI- a,( 1 - s) = (2s - I)al,which gives an in- d) Three broken rotor bars
duced stator current of the angular frequency (2s - I )al.
If we considered the v-th space harmonic we will
have:sol +.vwI(1 -s), v = 1,3,5,7, I1,13, ... [ I I ] being In asimilar way one can describe the next components.
the order of space harmonics. This gives the following One can conclude by sayingthat damaged rotor bars
frequencies for different values of v: WI, (2s - I)wI, and generally rotor asymmetry will produce the follow-
(3- 2s)Wl, (4s - 3)a1,( 5 - 4s)ol, (6s - 5)a1, (7 - ~ S ) O I , ing frequency components in the srator current:
( 8 ~ - 7 ) ~...
l , [7, 121.
( 2 s - l ) 0 1 , ( 4 ~ - 1 ) ~ 1 , ( 2 sl)wl,(6s-
+ I)o1,(4s+I)w1.
Between the angular frequencies of the positive- and
negative-sequencestator currents components there will These are sidebands f2 s f , , 2 4sfi, f6sfi, ... around
hence be an angular frequency difference of 2 ~ 0 1that , the supply frequency fi [lo]. For the test motor this
again leads to rorque pulsations of the angular frequen- means: 45.7 Hz, 54.3 Hz,41.3 Hz,58.7 Hz,37 Hz,
cy 2 s a l and their accompanying speed fluctuations 63 Hz, ... for 1 435 min-I.
2 ~ 0 The
1 . speed fluctuations relative to stator is: 2 s w 1 + The magnitudes of the above described frequency
0l(1 -s) = (S + I)WI and 2~01- W I ( 1 -s) = (3s- I)Ol. components are small compared with the main supply
The stator flux with its electrical angular frequency wI frequency component and are decreasing with the har-
therefore produces additional rotor currents with angu- monic order, and therefore difficult to detect. Particular-
lar frequencies swI and 3so1. 3sal produces positive- ly at low load or no load, these frequency components
and negative rotor mmf that relative to stator has an an- are close to the mains frequency, and that also leads to
gular speed: 3swl - al( -
1 - s) = (4s I)wI, and 3s01 measuring complications.
+ OI(1 - s) = (2s + l ) o l which induce stator currents of Tab. 1presents an overview of the results by meas-
the angular frequencies (4s- l ) w l and (2s + 1)ol. If we urements on the mentioned test motor. First measure-
considered the v-th space harmonic we will have: ments (a) are performed on a sound motor, thereafter on
a motor with one (b), two (c) and three broken bars (d),
3SOl 2 V O l ( 1 -S).
respectively. For each case the measurements are carried
New torque pulsations get the angular frequency out with a load of 100 % (1 435 min-'), 75 % (1 454 min-I)
4sw1,and associated angular speed fluctuations referred and 50 % ( 147 1 mid), respectively,Fig. 1to Fig. 4 show
to the stator: 4 s a 1+ al(1- s) = (3s + I)wI, and 4 s a 1 the current spectra.
- 01 ( 1 - s) = (5s - l)ol.The stator flux produces new [ 131 applies a method for statistical signal process-
rotor currents of the angular frequency 5 ~ 0 1 . 5 ~pro- 0 1 ing for detecting rotor faults on-line.
duces positive- and negative-sequence rotor mmf that Statistical analysis can be used for signal process-
again will cause stator currents with angular frequencies ing, often as a supplement to the direct method.
(6s- l)wl and(4s+ l)al. If weconsidered the v-thspace A current component with a frequency lower than
harmonic we will have: the main supply frequency by twice the slip frequency
will cause an analog ammeter in the supply lead to
fluctuate slowly. A spectrum analyzer will provide

ETEP Vol. 8. No. 5, September/October 1998

 ETEP
iI -
I I / I 1
I ! l n i i i T

I
i

gT
01

5 I5 25 35 45 55 65 75 85 Hz 105 5 15 25 35 45 55 65 75 85 Hz 105
f--’ f-
Fig. 1. Current spectrum for motor with one broken rotor bar Fig. 3. Current spectrum for motor with three broken rotor bars
(black)together with that of a sound motor (grey), 100 % load (black) together with that of a sound motor (grey), 100 % load

mT
01
21

5 15 25 35 45 55 65 75 85 Hz 105
f -*
Fig. 2. Current spectrum for motor with two broken rotor bars
(black)together with that of a sound motor (grey), 100 % load
5 15 25 35 45
f -.
55 65 75 85 Hz 105

Fig. 4. Current spectrum for motor with three broken rotor bars
(black) together with that of a sound motor (grey), 50 % load

more information, the signal processing is similar to

that for vibrational analysis. The ratio of current com- The principal slot harmonic frequency for the test
ponent amplitude to that of the supply frequency com- *
motor will be: 28 ( 1 - 0.0433)50/2 50 = 6 19.7 Hz and
ponent gives an indication of the degree of fault or 7 19.7 Hz.
damage [3]. To take stator mmf time harmonics into considera-
All machines have a certain asymmetry, so the fre- tion,fi must be multiplied by h, h being the harmonic of
quency spectrum will contain frequency components the stator mmf time harmonics, h = 1,3,5,7, ....The fre-
with distance 2 s 0 , away from the mains frequency. In quencies will then be: v&( 1 - s)fi / p f:hfi.
practice often only the side band frequency ( 1- 2s)ol is Sidebands at twice the slip (2sfi)also appear around
applied for the fault diagnosis. Measurements on the these frequencies.
original condition is not necessary. Also static and dynamic rotor eccentricity, as well
Even in a case where a motor is connected to a sym- as mechanicalfaults on bearings and driven equipment,
metrical three-phase supply the air-gap flux will contain will affect the frequency spectrum of the stator current.
space harmonics. The reason for that is the non-sinu- The air-gap eccentricity can occur due to shaft deflec-
soidal spatial distribution of the stator windings, stator- tion, bad alignment, bearing wear etc.
and rotor slot harmonics. Because of dynamic eccentricity also side-band
With sinusoidal mmf the fundamental slot harmon- components around the slot harmonics in the stator cur-
icfrequency for the stator current can be expressed: rent will occur in the frequency spectrum of the stator
current 141.
-s)filp-cfl. The frequency components in the stator current be-
Taking into account all slot harmonic frequencies cause of air-gap eccentricity and slot harmonics are:
the expression is:
fr [( Y Z 2~h )( 1 - S)/P -t h 1, v = 1,2,3, ...,h = 1,3,5,7, ...,
VZ2(1 -s)fllp*:fi, nd is an integer (eccentricity order number0 for static ec-
where v = 1,2, 3, ... is any integer. centricity and 1.2.3.4, ... for dynamic eccentricity) [7].

ETEP Vol. 8, No. 5, September/October 1998 387

ETEP ___
The stator spectral components because of dynamic
eccentricity alone, gives the following harmonics in the
stator current [ 12, 141:
f i l l + . v ( l - s ) l p l = I f i ~ v f , , I , v = 1,2,3(spatialharmon-
ic order).J,,, is the shaft rotational frequency (in Hz).
For the test motor this will be: i

50 [ 1 2 1( 1 - 0.0433)/2]=73.9 Hz and 26.1 Hz m

(s = 0.0433and v = 1);
50 [ 1 2 2( 1 - 0.0433)/2] = 97.8Hz and 2.2 Hz
(s = 0.0433 and v = 2).
5 25 45 65 85 105 125 145 165 Hz 205
If the load torque varies with multiples of rotclrion-
al speed, the load will generate spectral harmonics in the f-
motor current that will overlap the harmonics caused by Ti
the fault conditions [lo].To avoid influence of the load
on the fault registrations one has to monitor frequency A
components not obscured by the oscillations from the
load. I

An oscillation in load torque at a multiple of rota-

tional speed will produce stator currents with frequen-
cies:
9
2-

f i [ I ~ m ( l - s ) / p ] , m=1,2,3,... [lo].
This can be the same frequencies as given by certain 205 225 245 265 285 305 325 345 365 Hz 405
unsymmetries and faults in rotor bars. f b
For a two-poles motor the following frequencies are Fig. 6.Current spectrum for motor with an outer race fault on
given: sfi, (2- s)fl, (2s - I )fi, (3 - 2s)fi,(3s - 2)fi, both bearings and cage fault on one of them (black), together
(4- 3s)fi,(4s- 3)fi,(5- 4s)fi, form = 1,2,3. with current spectrum for sound motor (grey), 100 % load
For a four-poles motor the following frequencies are a) Frequency range 5 Hz ..,205 Hz
given: ( I -s)fi/2, (3-s)f,/2,sf!,(2-s)fi,(1 - 3s)J/2, b) Frequency range 205 Hz ..,405Hz
(5 - 3s)fi/2,for rn = I , 2,3.
A defect in rolling element bearings causes the air
gap of the motor to vary, and generates stator currents at fan end 138.8 Hz,38.8Hz,227.6 Hz,127.6Hz etc. For
frequencies: an inner race defect the frequencies will be 176.9 Hz,
76.9 Hz, 303.8 Hz,203.8 Hz, ..., 176.5 Hz, 76.5 Hz,
ffikmJ.1, m = 1,2,3,... 1141. 303 Hz,203 Hz etc. For a ball defect the frequencies will
f v is one of the vibration frequencies for the bearings as be 179.1 Hz,79.1 Hz,308.2Hz,208.2 Hz, ..., 182.3Hz,
given before, calculated from the bearings dimensions. 82.3 Hz,314.6Hz. 214.6Hz,..., and for a cage defect
To the test motor several bearing faultswere imposed. 59.8Hz,40.2Hz,69.6Hz,30.4Hz ,...,59.9Hz,40.1Hz,
An outer race defect on the shaft end of the motor gener- 69.8Hz,30.2 Hz etc.
ates currents of frequencies: (50 2 88.3)Hz = 138.3 HZ Fig. 6 presents current spectra with fault on outer
and 38.3 Hz,226.6Hz and 126.6 Hz, ..., Fig. 5, on the race for both bearings, and cage defect for one of them,
together with current spectra for sound motor, 100%
load.
Air-gap dynamic eccentricity, caused by bearing
faults, produces anomalies in the air-gap flux density
and thereby harmonic current components as shown.

tI
3.7 Air-Gap Torque Monitoring

Air-gap torque monitoring can detect faults in the

motor during run up as well as in service. The method
$T
01 can detect cracked rotor bars and shorted stator coils
[9,17, 181.The shape of the air-gap torque distinguish-
es between unbalances caused by cracked rotor bars and
unbalanced stator windings. As the method is based on
5 15 25 35 45 55 65 75 85 Hz 105 the detection of more parameters than the current, more
f -* information is available. Any unbalance in the motor
Fig. 5. Current spectrum for motor with an outer race defect can be indicated. The instrumentation consists just of
on the shaft end (black) together with that of a sound motor transducers for currents and voltages, the signal pro-
(grey), 100 % load cessing is carried out by a PC.

388 ETEP Vol. 8. No. 5, SeptembdOctober 1998

 E TEP
According to the general theory for induction ma- most of the other insulation test methods just measure
chines, an expression for air-gap torque may be [7]: the insulation condition against earth or between the
phases. A surge with steep front penetrating a winding
T = .13 P ( w I -N ~ L -Z v L 2 - N iL I >. gives voltage differences between windings and coils so
The flux linkages are calculated as integral of the that insulation faults are revealed.
difference between stator voltage and the product of sta-
tor resistance and current: 3.10 Measuring of Partial Discharges
YLI-N=I(uLI-N-iLIR)dt, Measuring of partial discharges may provide a
~LZ-=
N ~(ULZ-N - iLzR)dt. measure of the insulation condition within a motor. It
has, however, been some uncertainty about the connec-
The phase voltages are expressed by the line voltag- tion between the degree of partial discharges and the re-
es: maining life time.
-
~ L I - N= ( I / ~ ) ( u L I - L 1,
~
ULZ-LI - D e j n ition
ULZ-N = (I/~)(uL?-LI - UW-LZ), According to [20] partial discharges (PD) are local-
ized electrical discharges in insulating media, restricted
uL3-N = ( ~ / ~ ) ( U L-~u-LL
1 -Z
W).
to only a part of a dielectric and only partially bridging
The torque expressed by parameters obtained direct- the insulation between conductors. PD may occur in cav-
ly from measurements is consequently: ities (voids or occlusions) in solid insulation, in gas bub-
bles in liquid insulation or between layers of insulation
T = (pl.13) [ ~ L Z ~ L I - -ULZ-LI
U -3i~1R)dt with different dielectric characteristics. They may also
- iLIJ(uL2-LI - U L ~ - L ~3i~zR)dtl.
-
occur at sharp edges or points of metallic surfaces.

After a simple rearranging one arrives at [ 121: - The abc Model

The most common model to describe the phenonie-
T = ( p / a ) ( ( i L l - ~ L Z ) ~ [ - U L I - W - R ( ~ L ~ - ~ L I ) I ~ ~ non of PD in a solid insulation material, is known as the
“abc model”, and can be traced back as far as to 1932
- ( ~ L ~ - ~ L I ) ~ [ - u L z - L I-R(iLI - i ~ d l d r j .
[21]. a, b and c stand for three capacitances Ca,c b and
Asymmetry in the stator results in positive- and neg- C,. C, represents the capacitance of the occlusion (for in-
*
ative-sequence stator mmf at angular speed 01. Posi- stance a gas-filled void in a solid dielectric), c h is the ca-
tive mmf produces a constant torque, whereas negative pacitance of the dielectric in series with the void, and C,
mmf produces an harmonic torque with angularfrequen- the rest of the capacitance of the dielectric (Fig. 7). In the
cy:-Ol-wl=-2O,. circuit model c b cc C,c< Ca.
Asymmetry in the rotor results in positive- and neg-
ative-sequence rotor mmf at frequencies *SO. Positive - Qualitative Description of the P D Process
rotor mmf produces a constant torque, whereas negative Now consider a dielectric with the relative permittiv-
mmf together with positive mrnf of the stator produces a ity &r with a little sphere shaped void in it, filled with a gas
torque with angular frequency 2 01. of relative permittivity 1.0. If the electric field strength in
the dielectric is E, then the field strength in the void is E,
3.8 Instantaneous Stator Power Monitoring = E(3&,)/(1 + 2 ~ , [22].
) Similarly,if the occlusion is a nar-
row slot between two dielectric layers, one has E, = EE,.
Instantaneous stator power monitoring gives infor- However, the permittivity of the gas inside the void is not
mation about the motor condition because it is affected always 1.O, and both the geometry of the void and the per-
by current magnitude, voltage and phase shift between mittivity inside may be difficult to find. Nevertheless, a
current and voltage waveforms. In [ 181 instantaneous void will most likely experience values of the field
stator power is used for signature analysis of induction strength that are considerably higher than for the rest of
motors. [ 181 shows that the spike generated in the fre- the insulation, and a local breakdown,or a PD, may occur.
quency spectrum of the power appears directly at the fre- The PD in the void causes the field strength and the
quency of the disturbance, independently of the number voltage across the void to collapse to nearly zero, and im-
of poles of the motor. mediately thereafter the voltage and field strength will
start to rebuild until the breakdown values again are
3.9 Surge Test reached, and the process will repeat itself until the voltage
across the void cannot reach its breakdown starting value.
The surge test is a well-known method for diagnos- However, when the test voltage has reached a sufficient
ing of winding faults. In the surge comparison test, two low value, so that the voltage across the void experiences
identical high-voltage, high-frequency pulses are simul- a sufficient negative value, the process will repeat itself
taneously imposed on two phases of the motor winding, with opposite polarity, so that for each half period of the
while the third phase is grounded. The reflected pulses impressed voltage there will be two series of PD.
are compared on an oscilloscope [ 191.The imposed and Each little PD in these series implies a movement of
the reflected pulses will cover each other provided all the a little electric charge q,nrfrom the C, to C, (or vice versa).
windings are healthy. Surge tests reveal insulation faults This little charge is related to the jump in voltage VA
between windings, coils and groups of coils, whereas across the void from the starting voltage to the rest volt-

ETEP Vol. 8, No.5 , SeptembedOctober 1998 389

ETEP
- More Definitions [ZO]
Test voltage The PD pulse reperition rate n p is the average num-

I T
$1 ber of PD pulses (above a specified magnitude) per sec-
ond measured over a selected time. Integrated quantities
are current I, qiiadrntic rate D and discharge power P . I
is the sum of the absolute values of the apparent charg-
es during a certain time intervai divided by this time
interval. D is the sum of the squares of the apparent
charges, during a certain time interval, divided by this
+ & time interval and Pis the average power fed into the ter-
Fig. 7. The abc model for partial discharges in a dielectric minals of the test object due to PD. Alternating voltages
with a void related to PD are given by their peak values divided by
a) Test object b) Equivalent circuit 42. Inception voltage U,is the lowest voltage at which
partial discharges are observed in the test arrangement,
when the voltage applied to the test object is gradually
age. The little internal movement of charge is nearly a increased from a lower value at which no (or specified
Dirac pulse. that is a time-current graph of nearly zero low values of) such discharges are observed. PD extinc-
time but a defined area. The voltage drop UAacross the tion voltage U,is the lowest voltage at which PD are ob-
void caused by the Dirac pulse, causes a voltage drop in served in the test arrangement when the voltage applied
the outer circuit that can be expressed as a voltage divi- to the object is gradually decreased from a higher value
sion between c b and C,, since one can suppose that dur- at which such discharge are observed.
ing the extreme short time of this process, there will be
- Alternative t o the abc Model
no current in the outer circuit. However, the voltage drop
over C,is detectable across the terminals oftheelectrodes The abc model has been successfully used for many
as UA = u ,cb/(c b + ca). years despite many shortcomings. Strictly speaking, the
Immediately after the Dirac pulse is over, a current concept of void capacitance is not correct, and the other
will start to flow in the outer circuit. This current is also capacitances involved cannot be calculated [23,24].By
of short duration, but lasts much longer than the Dirac using classical electrostatic theory one can find expres-
pulse. This current represents a charge 9 which can be sions for apparent charge and current due to PD for a die-
expressed as: q = uA(C, + CbCc)/fCh+ C,) or nearly: lectric with given defect geometries. It will lead too far
q = u A C ~ The
. charge q is defined apparent charge. AC- to describe the method more precisely, but one of the
cording to [20] the apparent charge of a partial discharge conclusions is that a defect (a void or another occlusion)
is that charge which, if injected instantaneously between which causes a PD with a certain charge in one system
the terminals of the test object, would momentarily will produce a PD magnitude of half that value in a
change the voltage between its terminals by the same system of double size but with the same defect.
amount as the partial discharge itself. The apparent
charge is expressed in pC, and is not equivalent to the - Reasons and Locationsfor PD
charge qlnrat the PD site, that is, in the above description, The insulation of a modern high-voltage rotating AC
q is different from q,",.Let the voltage across the test ob- machine stator winding is mainly epoxy/mica (former-
ject be: u = Usin(wt). Then the voltage across the ly asphalt/mica) groundwall insulation covered with a
voidpriortothePDis: uI = [ c b / ( c b + C c ) ] f l Usin(wt). semi-conductive layer. and stress grading coating. The
These two voltages are drawn in Fig. 8. The voltage at bars are normally pressed against the bottom of the slot
which the PD starts is U,(inception voltage, see below), by wedges at the top of it. By casting or extruding insu-
and the rest voltage just after the PD is Urns,. lation materials it is in practice nearly impossible to
Referring to Fig. 8, when uI at the instant tl has completely avoid small occlusions of air or gas (voids,
reached U,, the breakdown starts, and the voltage across cavities) in the solid. Insulation systems build up of
the void drops to Umt,the differencebeing: UA= Ui- URyl. layers of different or equal materials will often have
Immediately thereafter the voltage across the void will
build up according to the equation u2 = U I - UA(that is,
according to the sloping part of the nearly saw-tooth part
of the graph (parallel to ul). Normally u2 will reach U,,
and another PD will occur (t2), and the voltage across the '1 2
void will increase according to u 3= u I - 2 UA. Now it may u3
well happen that 1.43 will reach U,, and hence another sin-
gle PD starts. However, in the example of Fig. 8 it does +
not do so until it has become negative, and the single PD t
processes starts over again with negative polarity ( t 3 ) .
The whole PD process is characterized by a series of sin-
gle PD in the first part of the positive sine wave, and a
similar series in the first part of the negative part of the
sine wave. In accordance with the particular cases, there Fig. 8. Test voltage and voltage across a void during PD
may be more or less single PD in each series. in a dielectric

390 ETEP Vol. 8. No. 5 . Srptember/October 1998

 CTCD

small slips between the layers, or there may be small par- Impedance or filter
0 -
ties where the insulation does not adhere perfectly to the T I
metallic electrodes, or a pre-insulated bar of a stator Coupling -L
winding may not be side-packed properly in the slot, or +Test object capacitanceT
the wedge may have become loose. As a result of poor 1 Connecting lead I
side-packing and/or loose wedges the electromagnetic
forces under load will cause small movements that may
abrade the field-equalizingpaint in the slot section of the
coil. Further one can often observe PD at the end wind-
ings, mainly because of dirt or oil contaminations. PD
may also occur around a conductor insulated with a gas Fig. 9. Example of a test circuit; the dotted Z, is an alterna-
(corona). tive position [20]
We may sum up here that PD may take place:
in voids in the dielectric itself in the slot part,
pressor for noise from the mains. The so-called Phase
in voids between the copper conductors and the Resolved Partial Discharge Analysis (PRDA) applies
groundwall insulation, very fast responding computer techniques so that it has
been made possible to analyse the PD as to the number
in voids between the surface of the stator insulation
of discharges, the charge in each of them, and, if supplied
and the core iron (slot discharge), and
with other sensors in addition to the RFCT, the location
in voids in the dielectric itself in the end winding, or of the PD in the motor and the type of fault that have led
surface discharges there due to contamination or pure to PD. In addition to the principle that noise pulses from
design as to clearances or stress grading (altogether the mains will be suppressed by cables or condensers,
called end winding discharges). one applies the principle that PD pulses and noise puls-
es have given different distances to travel to the sensors.
Consequences of PD
PD will deteriorate the insulation in three ways [ 2 2 ] : - Calibruting PD Measuring Instruments
Calibration is a major concern with PD. To date it has
Bombardment of the insulation around the discharge not been possible to establish an accepted and reliable
spots from ions and electrons accelerated in the dis- method
of absolute measurement so that one for instance
charge paths. could design a standard transportable calibrating instru-
Chemical reactions in materials adjacent to the PD ment. That means among other implications,that it is dif-
spots caused by the PD, particularly organic materi- ficult to establish a sort of standard recommended limits
als are subject to these reactions. for the “severity of the PD’. The calibration that has been
possible to be established, is only valid for a particular
Radiation from PD (PD are extremely fast phenom-
machine with particular instrumentation. However.
ena, pulse rise time and duration are in the order of
much research is currently going on in this field, and even
ns); ultraviolet light may have enough energy to de- at the present level the PD monitoring has proved to be a
teriorate organic compounds.
very useful. not too expensive, and possible to be run by
PD will erode insulating materials such as polyester non-experts after a short training period. PD quantities
and epoxy resins. Mica on ;he other hand is very resist- that may directly be recorded and displayed by a PD in-
ant against attack from PD. The compound of epoxy strument arepuise-repetition rate np(pulse count rate) in
/mica is therefore relatively long living as far as PD is the unit “pulses persecond”,npparentchargeq in the unit
concerned, so that it will be necessary to watch the trend pC (or pulse magnitude in the unit mV) andpulse-phase
of the deterioration of the insulation in order to predict angle in the unit degree. The three quantities may be pre-
the condition in the future. sented in the form of a three-dimensional graph that dis-
tinguishes between positive and negative pulses. A less
- Methods of Detecting PD complicate variant presents pulse count rate and magni-
Until some few years ago detecting and measuring tude with polarity in a two-dimensional graph.
of PD was performed off-line, and Fig. 9 shows a basic For many years the insulation of high-voltage ma-
test circuit [20,25]. The impedance (or filter) prevents chines was based on mica flakes and bitumen- or shellac
discharge pulses from being bypassed through the high resins. As the voltage and power ratings increased, one
test-voltage supply, it also will attenuate disturbance experienced localized premature failures. Traditional
(noise) from the supply. The recording instrument is a methods for monitoring the bulk insulation condition, e.
very fast responding oscilloscope or a recorder. Much g. the “Loss Tangent Variation”, showed to be insuffi-
effort is placed on designing such test circuits for dif- cient to detect localized premature failures in a reliable
ferent test object and different noise condition and other way. And when the more modem thermoset insulation,
demands. based on mica chipping and glass fibres, was introduced,
For measurements of PD in high-voltage machines the need for detecting localized faults was still there. The
on-line, high (radio) frequency current transformers localized faults (attack on one coil or conductor) on this
(RFCT) are clamped on the machines supply cables. For modem insulation can appear unexpectedly and devel-
machines supplied by a bus duct system, the sensoring op relatively rapidly, so one has to follow the situation
is more complicate. The cable itself serves as noise sup- with not more than months interval.

ETEP Vol. 8, No. 5, September/October 1998 39 I

E TEP
An insulation condition monitoring today may have - Stator Slotting
three stages: ASa result of stator slotting the rotor current in ad-
- At least twice a year on-line monitoring tests are car- dition has the frequency components [ 121:
ried out based on Capacitor Couplers, RFCT or Ro- fi [ vZ1(1- s ) l p k h s ] , Z ,beingthenumberofstatorslots,
gowski coils. These tests will indicate a problem on time harmonics h = 1,3,5, ... and v = 1,2,3, ... .
an early stage, and help to make the decision wheth-
er the machine is to be taken out of service, and if so, For the test motor with Z, = 36, s = 0.0433 and
when. The demand for calibration and quantifying is p = 2 , h = 1 andv= 1 onehas:
hence not strong at this stage. 50/2*1.36(1 - 0.0433) * 1 -0.0433.50 HZ= 863.2 HZ
- If the on-line monitoring tests, and supplying obser- or 858.9 Hz.
vations, have led to the decision of taking the ma- Similarly for h = 1 and w = 2:
chine off-line, of-line tests are carried out. At this
50/2.2.36( 1 - 0.0433) 2 1-0.0433.50 HZ= 1724 HZ
stage the machine will remain fully assembled, and
or 1719.9 Hz,
the tests should be able to confirm the on-line indi-
cations, and in addition give information on the se- and forh = 3 and w = 1:
verity of the attacks, that is, the measuring instru-
ments must have more accurate calibration than in
50/2.1*36(I - 0.0433) k 3.0.0433.50 HZ = 867.5 HZ
or 854.5 Hz.
the first stage.
Also the new harmonic frequencies in the rotor cur-
- Finally dismantled locating tests, with removed rotor, rent due to saturation in the mmf frequency region and
are carried out to localize the problem. slot harmonic sidebands are identified [30]:
3.11 Magnetic Field [ 1 k k ( 1 - s)]fi, k = 3,5,7, 1 1, ... (mmf and saturation),
and
The magnetic field in the air gap during normal
conditions varies sinusoidally in space and time. Some [vZ1(1- s ) l p - m ( l *s)lf~,
stator- and rotor faults cause deviations from the sinu-
soidal variations. Rotor faults can be detected by a ([VZI(l -s)lp-(2+m)(l 2s,llfi,
search coif fixed to the stator. Changes of the air-gap m = 5,7, 1 1, ... (sidebands due to saturation),
flux density caused by rotor or stator will in addition v = 1 , 2 , 3.....
produce an axialflux which can be detected by a meas-
uring coil around the shaft or other sensing devices - Rotor Slotting
such as Hall probes. Faults are revealed by analysis of
the frequency spectrum of the measuring signal [7,291. Rotor slotting gives the following frequencycompo-
By monitoring the axial leakage fluxes it is possible to nents in stator signal spectrum:
identify various asymmetries and fault conditions as fi[VZ2(1 -s>lp+- 1 I. [ l a .
broken rotor bars, stator winding intertum short circuit,
loss of a phase, eccentricity and so on. One component - Static and Dynamic Eccentricity
of the axial leakage flux is caused by the stator current,
and one is caused by the rotor current, so that the har- Static eccentricity will not give rise to rotor cur-
monic is direct related to the harmonic in stator- and rents, but dynamic eccentricity will cause rotor currents
rotor current. and could be confused with a broken bar [ 161.
Dynamic eccentricity spectral components in sta-
- Space-Harmonics tor current are:
For the reason of space-harmonics the air-gap flux fi[-+v(l-s)/p+ 1],[12].
density distributionis not purely sinusoidal. If the angu- Static- and dynamic rotor eccentricity will have ef-
lar frequency of the stator current is wI, the slip-depen- fect on the stator current and on the magneticflux. AS
dent frequenciesin the mtorcurrentare:s o t . (6-5s)01, shown already,the frequency componentsin the magnet-
(7s-6)~1,(12-Ils)wl,(13s- 12)wl,etc.[7,29].Gen- icflux because of air-gap eccentricityand slotting are:
erally the rn-th spatial harmonicinduces in the rotor cur-
rents with frequencies: * *
fi [(v2 2 n d ) ( 1 - S ) / p h 1,
nd = 0, 1,2,3, ... (0 for static eccentricity),
[I+m(l-s)]fi, rn=5,11,17,...,
h = 1,3,5,7, ..., v = 1,2,3, ... .
[l - m ( l - ~ ) ] f i , m = 1,7, 13, 19, ... [30]. - Fundamental Sidebands of the Supply Frequency
For the test motor at a slip of s = 0.0433 one has: For the fundamental sidebands of the supply fre-
(6-5*0.0433)50Hz=289.2Hz (rn=5), quency one has a flux in rotor at frequencies:
(7.0.0433 - 6)50 Hz = 284.8 Hz (m = 7), f i [ I * h ( l - ~ ) / p ] , h = l , 2 , 3 , ...[121.
In [30] also the new harmonic frequenciesdue to sat-
(12 - 1 1 .O.O433)5O HZ= 576.2 HZ (m= 11). uration in the mmf frequency region and slot harmonic
(13 -0.0433- 12)50 Hz = 571.9 Hz (m = 13). sidebands are identified.

392 ETEP Vol. 8. No. 5. September/October 1998

 ETEP
- Stator Voltage Asymmetry 3.13 Oil-Particle Analysis
Provided sinusoidal mmf, a stator voltage asymme-
Oil-particle analysis is not much in use for motors.
try [7,29],will result in positive- and negative-sequence
but in particular cases with circulatory lubricant system
rotor currents with angular frequencies WI - Wmr = SWI
it may be a valuable supplement. Some systems include
mdW,,+Wl=(2-~)0l.
a unit of motor and a pump immersed in the pumped me-
However, one must take space harmonics into ac-
dium. The supply to the unit is through a so-called um-
count. Third space harmonics will result in positive-
bilical. It provides wires for electric power and control,
and negative-sequence rotor currents with angular fre-
and oil that provides insulation, cooling and lubrication,
quencies: [ ( w 1 / 3- o,,)/(wl/3)]w~ = (3s - 2)01,and
[2-(3S-2)]Wl =(4-3S)Wl.
and also maintains an over-pressure in the motor com-
The fifth harmonic gives in a similar way (5s- 4)Wl pared to the surrounding medium. In such systems oil
and (6 - 5s)ol, the seventh gives (7s - 6)01 and particle analysis is a method in point.
(8 -7s)wl, etc. The analysis aims to examine the oil for particles of
The rotor currents will consequently produce leak- metal, fibres or dust. The causes of the particles may be
age axialflux components of angular frequencies: SWI, friction of streaming oil, bearing wear and tear or partial
(2 - s ) w ~(3s
, -2)01, (4- 3S)Ol, (5s - 4 ) 0 1 ,(6 - 5 S ) 0 1 ,
discharges or corona, or they can originate from chemi-
cal impurities like water, or oxides. Detection of parti-
(7s - ~ ) w I( 8, - 7 s ) etc.,
~ ~generally:
cles and other impurities is realized in different ways:
, 1 , 3 , 5 , 7,....
w ~ ~ v w ~ ( I - s ~) =
- A magnetic plug will attract ferromagnetic particles
- Broken Bars in lubricating and hydraulic oils.
Considering only the fundamental space harmonic, - Mobile field equipment for the collecting of particles
asymmetries on the rotor, e. g. broken bars, generate in from filters for later analysis.
a similar way positive- and negative-sequence rotor cur-
rents with angular frequencies solwhen only the funda- - Spectrometry for analysing what elements the parti-
mental space harmonic is considered. Because of torque cles contain.
pulsations one has, as already shown, rotor currents with - Particle counting and analysing to reveal particle size
angular frequencies swI.3swl,5swl, ... . and concentration.
The negative-sequence slip frequency S W I will in- - Ferrography, deposit of ferro-magnetic particles by
duce stator current with the frequencies: SWI - OI( 1 -s)
magnetism.
=(2s- 1)Wl.
Generally, the v-th space harmonics for instance for - Ferroscopy, microscope and optical detection.
SOIgive angular frequencies in the stator current: - Partigraphy, change in resistance in nickel film ex-
SWI -+ VWI(1 -s), v = I , 3,5,7, ..., posed to wearing from oil particles.
that is (2s - I)@,, (3 - 2s)ol, (4s - ~ ) W I(5, - 4s)wl, Water detection can be realized by
(6s-5)01, (7-6s)wl. (8s-7)01,etc. - distillation,
3swl give angular frequencies in the stator current:
- the Karl-Fischer method,
3SOl f V a l ( 1 - S),
- crackle test,
that is (2s + I), (4s - I), ...,and 5swl give
- chemical reactions, and
5 S 0 1 2 V W I( 1 -s), that is (4s + I), (6s - I), ... .
- capacitance measurements.
One can conclude that the followingfrequency compo-
nents due to rotor asymmetrieswill be in the stator current:
3.14 Gas Analysis
(2s - 1)0i, (4s- ~ ) w I(2s , + I)oI, (6s - I)wI,
(4s+ l)Ol, ... . Gas analysis has, according to [3], shown to be ause-
Axial leakage flux have the same frequency com- ful method for large motors. The cooling air is analyzed
ponents. for possible carbon monoxide gas content from degra-
Supply faults occur when one phase is lost. This will dation of the electrical insulation, and in this way one
generate spectral components even multiples of$. may detect overheating of the windings.

3.12 Temperature Measurements 3.15 Visual Monitoring

Temperature measurements with thermocouples Visual monitoring means either inspection with the
and thermistors have a long tradition. Bearing tempera- bare eye or closed circuit television cameras, or a by pho-
tures give useful information on possible friction prob- tography or video tape to record trends.
lems, and are surveyed on a routine basis. Coolant bulk
outlet temperature is often monitored, particularly when
the machine is stressed beyond its rated data. 4 Application of Monitoring Methods
Winding temperatures on the hottest spots are difi-
cult to detect because of insulation problems, especially Use of monitoring methods are continual increasing
in high-voltage motors. May be a fibre-optic system is for medium to large induction motors. Both off-and on-
an answer to that problem. line systems are in use.

ETEP Vol. 8, No. 5, Septernber/October 1998 393

ETEP
Originally systems where basedon the analysis of vi- electric tield strength in a void of a die-
brations with spectrum analysers, which required an ex- lectric related to PD
pert to use and interpret the results. Then the systems uti- frequency of supply fundamental
lized software packages to perform the analysis required. vibration frequency due to outer- and
Systems for current analysis are connected to per- inner race-, ball- and cage defects
manent mounted current transformers in motor cables. relative rotating frequency
The inputs will be switched through to a signal-process- vibration frequency
ing unit individually for analysis, and the multiplexing order of time harmonics
can be canied out at the computer or remotely. instantaneous line currents
factors
speed of rotor (or shaft)
5 Maintenance and Faults eccentricity number
PD pulse repetition rate
Maintenance may generally follow three strategies: number of pole pairs
Curarive or run-to-breakdown is the simplest and least internal charge of PD
expensive, and is suitable for small and less important resistance per stator phase
motors which are easy to replace. Time-based mainte- slip
nance, either according to a time schedule or running air-gap torque
hours may prevent many failures, but may as well cause time
many unnecessary shut-downs. Condition-based main- instantaneous test voltage related to PD
tenance gives theoretically an optimal service, but pre- result of U, on the terminals, related to PD
supposes a well planned system with sensors as well as voltage drop in a void related to PD
data analysis equipment. ~ L I - L ~~L. Z - L I .
If an interruption in the normal running is expensive, L3-L2 instantaneous line-to-line voltages
or a potential fault may cause great damages, or the motor L~LI-N, ~ L ? . N ,
has a particular position so it is complicate to replace, UL~-N instantaneous line-to-neutral voltages
then an investment in a condition monitoring system can V factor (order of space harmonic)
be cost saving. V' factor
The maintenance philosophy must be adapted to the ZI number of stator slots
whole plant, all three strategies are often applied within Z? number of rotor slots
a factory. P contact angle
~ L I - N v, L 2 - N flux linkages in phase L1 and L2
WI angular frequency of supply fundamental
6 Conclusions Wrot angular rotor (shaft) speed
Er relative permittivity of a dielectric
The study presents an overview over condition mon-
itoring methods as a diagnostic tool and as a technique - Abbreviations
for failure identification. Methods of fault detection are BC Bus Coupler (small epoxy-mica capacitors
reviewed and analyzed. mounted on the high-voltage terminals; ap-
Different types of faults generatesdifferent frequency plied on smaller machines (TGA with BC))
components in vibration spectra, in motor supply current, FFT Fast Fourier Transformation
in torque and in magnetic field, and relationshipsbetween MICA trademark for an expert system diagnosis
fault types and spectra have been developed. Practical ap- computer program (the program compares
plications and experimental results are presented, particu- present input data with originally reference
larly for rotor- and bearing faults at different loads. data and gives out a diagnosis)
Condition-based maintenance offers an optimal ser- mmf magnetomotive force
vice, but presupposes a well-planned system with sen- PD Partial Discharge
sors as well as data analysis equipment. PDA Partial Discharge Analysis
If an interruption in the normal running is expensive, PRPDA Phase Resolved Partial Discharge Analysis
or a potential fault may cause great damages, or the (a sophisticated computer-based PDA system)
motor has a particular location so it is complicated to be RFCT Radio Frequency Current Transformers (to
replaced, then an investment in a condition monitoring be clamped on the supply cable of the
system can be cost saving. machine; RFCT may also be used in connec-
tion with TGA (TGA with RFCT)when
surge voltage capacitors are installed on the
7 List of Symbols and Abbreviations phase terminals; the RFCT are mounted in
the earth conductor of the surge capacitors)
- Symbols ssc Stator Slot Couplers (ruler-shaped antennas
b number of balls (or rollers) mounted under some of the stator wedges,
Cd,Cb, C, equivalent capacitancesfor PD calculations may be applied with the TGA for large turbo
&.d p diameter of bearing ball (or roller) and of generators (TGA with SSC))
bearing pitch TGA Turbo Generator Analyzer (for detecting on
D quadratic rate of apparent charges turbo generators and motors)

394 ETEP Vol. 8,No.5. September/October 1998

 ETEP
References [El Johcrnsen. 1.: Slethak. J.: Hogspenninpsteknikk. Trond-
- . -
Thorsen, 0. V ;Dalvu. M.: A Survey of Faults on Induc- heim/Norwny: Tapir, I978
tion Motors in Offshore Oil Industry. Petrochemical In- [23] fetlersm, A.: Partial discharges in voids in solid dielec-
dustry,Gas Terminals and Oil Refineries. IEEETrans. on trics. An alternative approach. Annu. Rep. of the Conf.
Ind. Appl. IA-3 I ( 1995) no. 5, pp. 1 186- 1 197 on Electr. Insul. & Dielectr. Phenomena. Gaithersburg,
IEEE Comm. Rep.: Report of Large Motor Reliability MDlUSA 1987,pp. 58-64
Survey of Industrial and Commercial Installations. Pt. I [24] Sedding. H. G.: Theory and practice of partial discharge
and Pt. 11. IEEE Trans. on Ind. Appl. [A-2 I ( 1985)no. 4, testing of rotating machines. Int. Conf. on Generator a.
pp. 853-872 Motor Partial Discharge Testing. Totonto/Canada 1994
Tnvner, P. J.; Gaydon. B. C.; Ward, D. M.: Monitoring [25] Stone. G. C.; Sedding, H . G.: In-Service Evaluation of
generators and large motors. IEE Proc. B- 133 ( 1986) Motor and Generator Stator Windings Using Partial Dis-
no. 3, pp. 169- 180 charge Tests. IEEE Trans. on Ind. Appl. IA-3 1 ( 1995)
Thomson, W T ;Leonarel. R. A.: Milne, A. J.; Penman, J.: no. 2, pp. 299-303
Failure Identification of Offshore Induction [26] h h i c l a . H.: Umemoto, K.: Insulation Diagnosis for Ro-
Motor Systems using On-Condition Monitoring. 4th tating Machine Insulation. IEEE Trans. on Electr. Insul.
Nat. Reliability Conf., BirminghadUK 1983, Proc. EI-2 1 ( 1986) no. 6, pp. I 02 I - I 025
pp. 2C/3/1-2C/3/1 I 1271 Krecke, M.; Goffaux, R.: Attempt at Estimating the Re-
Ashen. R. A.; Steele, M. E.; Knight. L. G.: Motor Condi- sidual Life of the HV Insulation of AC Rotating Ma-
tion Monitoring Methods. Conf. on Drives, Motors and chines. ParisFrance: CIGRE-Rep. 1 1-12, 1988,pp. 1-8
Controls, Harrogate, North YorkshireNK 1983, Proc. [28] Bonnett, A. H.: Analysis of Winding Failures in Three-
pp. 163-171 Phase Squirrel Cage Induction Motors. IEEE Trans. on
Leonard. R. A.; Thomsm, W T : Vibration and Stray Flux Ind. Appl. IA-14 ( 1978) no. 3. pp. 223-226
Monitoring for Unbalanced Supply and Inter-turn Wind- [291 Penmein. J.: Dry. hi.N.; Elit. A. J.; B y m ,M! E.: Condi-
ing Fault Diagnosis in Induction Motors. British J. of tion monitoring of electrical drives. IEE Proc. B-133
NDT (1986) pp. 21 1-215 (1986) no. 3. pp. 142- 148
&IS, 19: Parameter Estimation, Condition Monitoring, [30] Nutcirajmi. R.; Kohler, J. L.; Sottile. J.: Condition Mon-
and Diagnosis of Electrical Machines. OxfordGreat itoring of Slipring Induction Motors. Electr. Power Syst.
Britain: Oxford Univ. Press, 1993 R ~ s .I5 ( 1988)pp. 189- I95
Gcryfon,B. G.: An Instrument to Detect Induction Motor [3 I ] Uselmi. T.: Teshimci,H.; Ohtci. S.;Koizumi, T.: Ohno. E.:
Rotor Circuit Defects by Speed Fluctuation Measure- A Dynamic Monitoring System for Bearings Using a Mi-
ments. IEE Conf. Publ. 174 ( 1979) pp. 5 -8 croprocessor. 6th Ind. Electron. a. Control Instrument.
Watson, J. E ; Cow. K. S.: Thornson, W T.: Non-invasive (IECI) SOC.Annu. Conf., Philadelphia. Pa./USA 1980.
instrumentation for on-line management and monitoring Proc. pp. 24- I28
of electrical plant. Life management of power plants, I2-
14des.,Conf.Publ.no.401, 1994 Manuscript received on June 21, 1996
Schoen. R. S.; Htibetlrc 7: G.: Effects of Time-Varying
Loads on Rotor Fault Detection in Induction Machines. The Authors
1EEETrans.on Ind. Appl. IA-3 I ( 1995)no. 4, pp. 900-906
Deleroi. W: Der Stabbruch im Kafigliiufer eines Asyn- Olav Vaag Thorsen (1935) received his
chronmotors. Arch. f. Elektrotech. 67 ( 1984) pp. 9 I -99 degree in Electrical Engineering (M.Sc.)
Franceschini, G.; Tassoni, C.: Time Harmonic Investi- from the Norwegian University of Sci-
gation of Flux Density in the Stator and Rotor Frame ence and Technology in 1963. From
Oriented to Diagnostics of Cage Induction Machines. 1963 to 1965 he worked iw a resexch-
IEE Conf. Publ. ( 1993) no. 376, pp. 103- 108 and development engineer with the A/S
Ho, S. L; Chan, W L.: Leung. H. W : Application of Sta- National Industry (ABB) in the field of
tistical Signal Processing for Condition Monitoring of electric machines. Since 1965 he has
Rotor Faults in Induction Motor. IEE Conf. Publ. 376 been with the Bergen College, Ber-
(1993) pp. 97- 102 gedNonvay, from 1985 associate pro-
Schoen, R. R.; Habetier: T G.; Kamran, E ; Bartheid, R. fessor, and from 1987 professor in
G.: Motor Bearing Damage Detection Using Stator Cur- electricalengineering.His fieldsofinterestmElectricMachines,
rent Monitoring. IEEETrans. on Ind. Appl. IA-3 l ( 1995) Industrial Drives and Power Electronics. (Bergen College,
no. 6, pp. I 274- I279 Nygidsgaten I 12. 5008 BetgenOJorway, Phone:+475558
Hargis, C.; Gaydon, B. G.; Kamash, K.: The Detection
of Rotor Defects in Induction Motors. IEE Conf. Publ. 7500, Fax: +475558 7790, E-mail: [email protected])
213(1982)pp.216-220 Magnus Dalva ( 1935) received his de-
Kliman. G. B.; Stein, J.: Methods of Motor Current Sig- gree in Electrical Engineering (MSc.)
nature Analysis. Electr. Mach. a. Power Syst. 20 (1992) from the Norwegian University of Sci-
pp. 463-474 ence and Technology in 1963. From
Hsu, J. S.: Monitoring of Defects in Induction Motors I963 to 1965 he worked as a research-
Through Air-Gap Torque Observation. IEEE Trans. on and development engineer with the Nor-
1nd.Appl. IA-31(1995)no.5,pp. 1016-I021 wegian Electric and Brown Boveri A / S
Legowski. S. E; Tcvnadlowski, A. M.: Instantaneous (ABB) in the field of electric machines.
Stator Power as a Medium for the Signature Analysis of
Induction Motors. IEEE Ind. Appl. Soc.Annu. Meeting From 1965he hasbeen with the Stavang-
Conf. 1995, Conf. rec. pp. 619-624 er College, except for the two years from
Schlichting, S.: StoDspannungspriifung von Elektromo- 1970 to 1972 when he was an assistant
toren zur Friiherkennung von Isolationsfehlern. Elek- professor at the University of NairobiKenya. Since 1984he is an
trit.-wirtsch. 85 (1986) no. 10, pp. 406-409 associate professor in electrical power engineering. His fields of
IEC-hbl. 60270( 198I): Partial discharge measurements. interest are Electric Machines, Industrial Drives and Power Elec-
Offenbach . BerlidGermany: VDE-VERLAG GMBH tronics. (Stavangeer College, Kristine Bonnevie’s vei. 404 Sta-
Gemant, A.; v.PhilippoS; W :Die Funkenstreckemit Vor- vanger/Norway. Phone: +47 5 I 83 2043. Fax: +475 I 83 I7 50, E-
kondensator. Z. f. tech. Phys. (1932) no. 9, pp. 425-430 mail: [email protected])

ETEP Vol. 8. No. 5. September/October I998 395