See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/268326342

Estimation of the Number of Rotor Slots and Rotor Speed in Induction Motors
Using Current, Flux, or Vibration Signature Analysis

Article  in  Australian Journal of Electrical and Electronics Engineering · January 2008

DOI: 10.1080/1448837X.2008.11464192

CITATIONS READS

6 1,631

6 authors, including:

Nesimi Ertugrul Douglas Andrew Gray

University of Adelaide University of Adelaide
151 PUBLICATIONS   2,520 CITATIONS    197 PUBLICATIONS   2,107 CITATIONS   

SEE PROFILE SEE PROFILE

Colin H Hansen
University of Adelaide
295 PUBLICATIONS   5,196 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Wind Farm Noise View project

Sound refraction from hot exhausts View project

All content following this page was uploaded by Douglas Andrew Gray on 31 January 2015.

The user has requested enhancement of the downloaded file.

 Estimation of the Number of Rotor Slots and Rotor Speed
in Induction Motors Using Current, Flux, or Vibration
Signature Analysis
Randy Supangat, Nesimi Ertugrul, Wen L. Soong, Douglas A. Gray, *Colin Hansen
and Jason Grieger
School of Electrical and Electronic Engineering, *School of Mechanical Engineering
University of Adelaide
Adelaide, SA, Australia
Email: rsupang@eleceng.adelaide.edu.au
ABSTRACT
Effective condition monitoring of induction motors
requires knowledge of the number of rotor slots and the
rotor speed. These were primarily obtained in the
literature by multiple measurements and using design
knowledge of the motor under test. Furthermore, the
earlier studies have mainly concentrated on stator
current analysis. This paper investigates the use of the
stator current, axial flux, and vibration signals in order
to provide accurate rotor speed and rotor slot number
estimation by utilising the eccentricity harmonics, slip
frequency, rotor frequency, and rotor slot passing
frequency. The proposed simple methods are verified
using an extensive series of measurements on a 2.2 kW
squirrel-cage and a 5 kW wound-rotor induction motor
over a wide range of loading conditions.
1. INTRODUCTION
Induction motors are one of the most widely used
electrical machines in industrialised nations as they are
reliable and robust. However, a combination of
environmental, duty-cycle, installation, and
manufacturing related factors will deteriorate the
condition of such motors. This deterioration can cause
unexpected down-time and loss of production. In order
to prevent this, induction motors need to be maintained.
One of the most cost-effective maintenance approaches
is based on an online condition monitoring system. This
monitors the condition of a motor through non-invasive
sensors (e.g. current sensors, flux sensors, vibration
sensors, etc.) while it is operational.
Most of the well-known condition monitoring techniques
are based on the Fourier Transform (FT) [1-4]. These
techniques usually compare and measure the magnitude
of the fault frequency components, where the magnitude
of these components in faulty motors is usually higher
than in healthy motors.
However, identification of the exact frequency of the
fault frequency components often requires knowledge
about the operating speed. For example, the broken rotor
bar (BRB) sidebands in (1) depend on the slip, which
can be calculated if the rotor speed and the number of
poles are known. The number of poles is given on the
machine nameplate but other parameters such as the
number of rotor slots is usually not provided.
Knowledge of the number of rotor slots is important to
determine the eccentricity fault frequency components
(2), which is used to detect static and dynamic
eccentricity [4-6]. The number of rotor slots and
eccentricity fault frequency harmonics have also been
used for speed estimation in the past [7-9].
f brb = f (1± 2s ) (1)
⎡
f ecc1 = f ⎢(kR ± nd )
(1 − s ) ± v ⎤ (2)
⎥
⎣ p ⎦
where fbrb is the broken rotor bar sideband frequencies, f
is the supply frequency, s is the slip, fecc1 is the
eccentricity fault frequencies, k is any integer, R is the
number of rotor slots, nd = 0 (static eccentricity) and nd =
1, 2, 3… (dynamic eccentricity), p is the number of pole
pairs, and ν is the order of the stator time harmonics.
It was reported in [8] that it is possible to estimate the
steady-state speed of an inverter-driven motor (28 kW,
415 V, 50 A) by tracking the speed-dependent slot
harmonic component, which is related to the number of
rotor slots, within a certain frequency range. In this
reference, the extracted harmonic was then compared to
the 19th harmonic of the fundamental frequency in order
to determine the rotor rotational speed. It was also
reported that the number of rotor slots could be
determined by examining the stator current spectrum as
the load was decreased. The speed dependent harmonic
would reduce in amplitude and change in frequency as
the load decreased. The technique, however, required
data taken under various loading conditions.
A similar speed estimation technique on an inverter-
driven induction motor (4 kW, 4 poles, 28 rotor slots)
was reported in [7]. It was shown in this work that the
transient speed can be estimated by utilising the Short-
Time FT (STFT). In addition, the research also
incorporated zero padding (i.e. frequency interpolation)
into the FT in order to improve the detection of speed
dependent rotor slot harmonics. Another similar
technique was also utilised in [9] to estimate the rotor
speed of an inverter-driven motor (10 hp, 4 poles), which
used decimation to alias the rotor slot harmonics to a
single frequency. This process improved the ability to
detect the desired slot harmonic. It was also reported that
the number of rotor slots could be found by finding the
most significant slot harmonics in (2) when R is varied
between 30-54 (even numbers only), nd is varied for 0
and ±1, and ν is varied for ±5, ±3, ±1. However, the
technique was not tested for a wide range of motors (i.e.
motors with rotor slots other than 30-54) and the process
was found to be computationally intensive.
Three different speed estimation techniques not
employing rotor slot harmonics were also reported in the
literature. In [10], the FT and Hilbert transform
demodulation approach were utilised, and in [11] an
extended Kalman filter (EKF) and a real-coded genetic
algorithm (GA) optimisation were incorporated to
estimate the rotor speed. The work in [12] involved
injecting an extra rotor current with controllable
frequency and matching that frequency to the rotating
frequency of an induction motor.
In general, it can be concluded that most of the speed
estimation techniques based on the rotor slot harmonics
have focussed only on inverter-driven induction motors,
under limited loading conditions, and have used only the
stator current. Furthermore, the rotor slot number
estimation techniques mentioned either requires data at
various loads or is computationally intensive. Therefore,
this paper aims to improve on the existing methods and
to provide simple and accurate estimation of the speed
and rotor slot number by utilising the current, flux, or
vibration sensor signals. Several methods for speed and
rotor slot estimation based on different signal
frequencies and sensor types are investigated and their
performance is compared over a wide range of loads.
2. EXPERIMENTAL SETUP
The experimental results given in the following sections
of the paper were obtained using a purpose built test rig
based on a high-speed data acquisition system (2 x
DAQPAD-6052E boards from National Instruments,
capable of simultaneous sampling at 5 MHz) with a
custom written software (LabVIEW based). The test
motors include two 2.2 kW, 415 V, 4 pole, 32 rotor slots,
star-connected squirrel cage induction motors and two 5
kW, 415 V, 4 pole, wound rotor, star-connected wound
rotor induction motors. The 2.2 kW motors were
installed using a precision laser alignment tool (Optalign
Plus from Prűftechnik) and a torque wrench (Norbar, set
at 35 Nm) in order to ensure the accuracy and the
repeatability of the data obtained. The motors were
loaded by a separately-excited 5 kW dc generator.
During the tests, two current sensors, one external axial
flux search coil, and a vibration sensor (drive-end
horizontal) were attached to the motor under test and the
sensor outputs were sampled simultaneously via an anti-
aliasing filter, which was an 8th order Butterworth filter.
Two sampling rates were utilised:
• 400 Hz sampling frequency and 100 seconds
sampling time (low-frequency measurement), which
allows a very high-resolution frequency analysis (0.01
Hz resolution),
• 8000 Hz sampling frequency with a sampling time of
5 seconds (high-frequency measurement).
The FT analysis was performed using zero padding
(interpolation in the frequency domain) which is 16
longer than the number of samples, and spectral
averaging, which divides the data into 5 segments for the
400 Hz data and no averaging for the 8 kHz data (due to
the limited resolution), in order to improve the accuracy
of the peak detection and to reduce noise interference.
Experimental data from the motor signals was collected
using healthy motors (2.2 kW and 5 kW) and a faulty
motor (2.2 kW) with 4 broken rotor bars (BRB) over a
range of loads from no-load to full-load. The broken
rotor bar fault was introduced by cutting a narrow slot
through the end-ring next to the lamination stack using a
small diameter milling cutter to break the electrical
connection between the rotor bar and the end-ring. This
method was used to minimise the disturbance to rotor
end-ring currents which occur if more of the end-ring
was removed.
3. ROTOR SPEED ESTIMATION TECHNIQUES
The rotor speed is an important piece of information in
an on-line condition monitoring system. In this section
we will examine speed estimation techniques which
utilise stator current, vibration, and axial flux signals.
3.1. STATOR CURRENT
The eccentricity fault frequency components shown in
(2) are easily detected in the stator current [4-6] and
axial flux signals. These fault frequencies are a function
of the supply frequency, the number of rotor slots, the
number of poles and the slip. It is possible to reduce this
eccentricity fault frequency equation into a function
(fecc2) of only the supply frequency and the rotor
frequency, fr, as shown in (3) by setting the variables k =
0, nd = ±1, ν = 1 [9]. This relation allows the rotor speed
to be found using (4).
⎡
f ecc2 = f ⎢(± 1)
(1 − s ) ± 1⎤ = f ± f , (3)
⎥ r
⎣ p ⎦
f r = f − f ecc 2 (4)
0
X(f)*X(f) in dBs
-50
-100
20 30 40 50 60 70 80
Frequency (Hz)
Figure 1: A sample frequency spectrum of stator
current signal at 23% load illustrating the
eccentricity harmonics (3).
The frequency spectrum given in Fig. 1 illustrates that
the eccentricity fault frequency components (3) are not
particularly strong in the current signal but they are still
detectable and can be used to calculate the rotor speed.
Furthermore, the experimental results suggest that the
lower sideband is stronger than the upper sideband.
Hence, this paper will use the lower sideband to
calculate the speed.
3.2. VIBRATION
Vibration sensors are widely used in on-line condition
monitoring systems. The vibration signal can be used to
detect eccentricity, stator and supply unbalance faults
and is also useful for speed estimation. In particular the
vibration signal contains a large amplitude component at
the rotor frequency fr (see Fig. 2). Based on the
experimental results, this rotor frequency component has
been proven to be strong under all loading conditions.
As a result, rotor speed can be readily estimated using
the vibration signal.
-70
X(f)*X(f) in dBs
eccentricity harmonics (3), and (bottom) the slip
frequency.
Figure 3 (top) shows that the eccentricity harmonics,
which are used for detecting the speed, are quite strong
in the flux signal. The figure also suggests that the lower
sideband is stronger than the higher sideband. Therefore,
the lower sideband will be used in this paper to calculate
the motor speed.
Figure 3 (bottom) shows that the slip frequency
component in the flux signal is detectable and it can be
used to calculate the speed. However, this slip frequency
component is found to be weaker than the speed related
component of the first method, which is based on the

-80
eccentricity harmonics. Therefore, the first speed
-90 estimation method is considered significant in the axial
-100 flux signal.
-110
3.4. SPEED ESTIMATION COMPARISON
-120
10 20 30 40
Frequency (Hz) Table 1 presents a comparison of the performance of the
speed estimation methods discussed in the previous
Figure 2: Frequency spectrum of a vibration signal sections. It shows measured results from a 2.2 kW
at 23% load showing the rotor frequency.
motor under loading conditions from no-load to full-
load. The table shows the load (in %, calculated from the
3.3. AXIAL LEAKAGE FLUX
rotor speed), the measured rotor speed (nmeas, using an
An axial leakage flux sensor is another non-invasive optical tachometer), the estimated rotor speed and the
sensor, which can be easily installed without any speed error (∆n, in %).
modification to the motor under test. This sensor is a Table 1: Comparison of Speed Estim ation Methods
simple search coil located at the non-drive end of the test
motor. There are two speed estimation methods that can CURRENT: VIBRATION:
Eccentricity Har monics Rotor Frequency
be applied to the axial leakage flux signal. The first nmeas Load Speed ∆n Load Speed ∆n
method is the same method as in the stator current (rpm) (%) (rpm) (%) (%) (rpm) (%)
signal, which is based on the eccentricity fault frequency 1498.0 2.0 1498.7 0.05 2.2 1498.5 0.03
harmonics in (4). The second method uses the fact that 1489.0 13.3 1488.8 0.01 13.6 1488.6 0.03
1472.0 33.1 1472.2 0.01 32.9 1472.4 0.03
the axial leakage flux from the rotor end-windings 1436.0 75.0 1436.3 0.02 75.0 1436.3 0.02
contains a strong component at the slip frequency, sf. 1424.0 88.8 1424.4 0.03 89.0 1424.3 0.02
This slip frequency can be used to determine the speed 1416.0 97.5 1417.1 0.08 97.5 1417.1 0.08
through the relationship described in (5). FLUX: FLUX:

fr
(n − s * ns ) ,
= s (5) nmeas
Eccentricity Har monics
Load Speed ∆n Load
Slip Frequency
Speed ∆n
60 (rpm) (%) (rpm) (%) (%) (rpm) (%)
1498.0 2.0 1498.7 0.05 - - -
where ns is the synchronous speed of the motor in rpm. 1489.0 13.3 1488.8 0.01 13.6 1488.6 0.03
1472.0 33.1 1472.2 0.01 33.1 1472.3 0.02
0 1436.0 75.0 1436.3 0.02 75.1 1436.2 0.01
1424.0 89.0 1424.2 0.01 88.9 1424.3 0.02
X(f)*X(f) in dBs

1416.0 97.5 1417.1 0.08 97.6 1417.0 0.07

-50
The results given in Table 1 illustrate that the speed
estimation methods, which are based on eccentricity
-100
20 30 40 50 60 70 80
harmonics and rotor frequency, are very accurate under
Frequency (Hz) all loading conditions. It can be noted that these methods
-50 can detect the speed within 0.1% accuracy when
-60
compared against an optical tachometer. This accuracy is
X(f)*X(f) in dBs

well within the tolerance of the optical sensor.

-70
-80
On the other hand, the speed estimation based on the slip
frequency was found not to be as reliable as the other
-90
methods. This slip frequency method is found to be good
-100
0 5 10 15 for high loads but unreliable at light loads (i.e. less than
Frequency (Hz) 10%). At light loads, the slip frequency component is
non-existent because it is very close to the DC
Figure 3: A sample frequency spectrum of an axial
leakage flux signal at 23% load showing: (top) the component and its magnitude may be small. In this
situation, the slip frequency component will blend with
the DC component and the peak detection algorithm will
mistakenly detect the DC component as the slip
frequency. Therefore, the slip based speed estimation
method cannot detect the rotor speed when the load is
less than 10%, as shown in Table 1.
Further clarification about how each of the speed
estimation techniques performs at different loading
conditions is shown in Fig. 4. It shows that the signal to
noise ratio, SNR, which is described in (6), of the
eccentricity component based speed estimation
techniques are better at light load, and the SNR reduces
as the load increases. However the techniques still
maintain respectable SNR under all loading conditions.
The SNR of the rotor frequency and slip frequency based
speed estimation techniques shows an increasing pattern
as the load increases. The rotor frequency based
technique shows good SNR under all loading conditions
while the slip frequency based technique fails at very
light load (less than 10%).
SNR (dB) = Signal (dB) − Noise Level(dB)
70
60
50 Flux: f-fr
SNR (dB)
40 Vibration: fr
Flux: sf
30
Current: f-fr
20
10
0 -40
0 20 40 60 80 100 120
Load (% )
Figure 4: SNR of rotor speed estimation techniques as a
function of load.
In summary, simple and reliable speed estimation can be
achieved from any one of the stator current, vibration,
and axial flux signals. The most reliable speed
estimation techniques are those which are based on the
rotor frequency in the vibration and on the eccentricity
component in the flux because of their high SNR at all
loading conditions. Speed estimation based on the slip
frequency is found to be the worst because of the low
SNR at very light load.
4. ROTOR SLOT NUMBER ESTIMATION
After the rotor speed has been estimated, the next step is
to estimate the number of rotor slots. This paper
investigates two rotor slot number estimation techniques
using the stator current, axial flux, and vibration signals.
4.1. STATOR CURRENT AND AXIAL FLUX
The first rotor slot number estimation technique is based
on the eccentricity fault frequency components in (2),
which are present in the stator current and axial flux.
This eccentricity fault frequency equation can be
reduced to a combination of rotor frequency and supply
frequency by setting nd = 0 (for static eccentricity) and k
= 1. Furthermore, experimental results suggest that the
first static eccentricity harmonics (i.e. ν = 1) are one of
the stronger and more reliable components. As such, the
rotor slot number estimation technique will focus on this
eccentricity fault frequency component (fecc3).
⎡ (1 − s ) ⎤
f ecc 3 = f ⎢ R + 1⎥ = Rf r + f (7)
⎣ p ⎦
The relationship described in (7) allows the estimation of
the number of rotor bars (or slots), R, by determining the
value of R which corresponds to the largest amplitude
component in the current or flux spectrum, that is not a
harmonic of the fundamental.
Figure 5 shows typical power spectra for the stator
current and axial flux signals under full-load for the 2.2
kW test motor. As highlighted in the figure, the
eccentricity component (7) is the most dominant
component, other than the harmonics of the fundamental.
Furthermore, this component is also found to be
consistently dominant across all tested loading
conditions.
0
X(f)*X(f) in dBs
(6)
-50
-100
0 1000 2000 3000 4000
Frequency (Hz)
0
-20
X(f)*X(f) in dBs
140
-60
-80
0 1000 2000 3000 4000
Frequency (Hz)
Figure 5: Typical frequency spectrum of a stator
current (top) and an axial flux signal (bottom).
Typical results of the proposed rotor slot number
estimation technique on a 2.2 kW motor at full load are
depicted in Fig. 6. The figure shows the magnitudes of
the frequency components given by (7) in the current
and flux spectra, for a range of possible number of rotor
slots R. The highest magnitude component in both the
current and flux spectra is achieved when the number of
rotor slots is 32, which corresponds to the actual number
of rotor slots. It was found that the highest magnitude
component was more easily identified in the current
spectrum compared to the flux spectrum.
Table 2 summarises the results of the rotor slot number
estimation technique on four motors. This includes two
2.2 kW motors, one healthy and one with four broken
rotor bars (BRB), and two 5 kW motors, both healthy,
under multiple loading conditions. The results indicate
that the proposed rotor slot number estimation technique
can successfully determine the number of rotor slots for
the four tested motors under all loading conditions. It
was also proven that the rotor slot number estimation
technique proposed here is equally effective in the
current and flux signals, although the current signal is
preferred due to the reason explained previously.
 -20 slot number estimation using the current signal is also
calculated from the current.
X(f)*X(f) in dBs
-40

-60 4.2. VIBRATION

-80 The rotor slot number estimation technique that is based
-100
on the eccentricity fault frequency component (7),
10 20 30 40 50 60 cannot be used on the vibration signal, as this component
Rotor Slots is not present.
-20
The electromechanical interaction between the stator and
the rotor slots is known to produce a strong component
X(f)*X(f) in dBs

-40
-60
-80
-100
10 20 30 40 50
Rotor Slots
Figure 6: Magnitudes of the possible eccentricity
component (6) against the possible number of
rotor slots in the stator current (top) and the axial
flux (bottom) spectra.
However, there are few precautions that should be noted
when utilising this rotor slot number estimation method.
An issue may arise when the motor load is very low (less
than 10%) and the number of rotor slots is few (less than
15 rotor slots). In this condition, the rotor frequency will
be very close to the synchronous frequency and the
frequency of the eccentricity component in (7) may be
very close to a strong harmonic of the fundamental. As a
consequence, the two components can be difficult to
distinguish. This problem can be reduced by making sure
the sampling period is long enough to ensure a high
frequency resolution.
Table 2: Summary of the Rotor Slot Number
Estimates Based on Eccentricity Harmonics
2.2 kW, MOTOR 1 (Healthy) 2.2 kW, MOTOR 2 (4BRB)
CURRENT FLUX CURRENT
at the rotor slot passing frequency (frsp) in the vibration
sensor. The rotor slot passing frequency depends on the
number of rotor slots and the rotor frequency as in (8).
This relationship allows the estimation of the number of
60
rotor slots, R, by finding the largest amplitude
component at the frequency given by (8) that is not a
harmonic of the fundamental when R is varied over the
possible values. Note that the rotor slot passing
frequency is a special case of the eccentricity fault
frequency components given in (2) when nd is set to 0
(for static eccentricity), k to 1, and ν to 0.
f rsp = Rf r (8)
Figure 7 shows a typical power spectrum of a vibration
signal, with the rotor slot passing frequency highlighted,
when a 2.2 kW motor is run at full load. The figure
shows that the rotor slot passing component is the largest
component in the spectrum. However, the magnitude of
this component is found to be sensitive to the level of
load, the type of load, the type of the motor, and the
installation of the motor. As a consequence, the rotor slot
passing component may not be the largest amplitude
component of all possible combinations of R. Hence it is
possible that this rotor slot number estimation technique
may fail in certain conditions.
FLUX -60
X(f)*X(f) in dBs

Load # of Load # of Load # of Load # of

(%) Slots (%) Slots (%) Slots (%) Slots -80
3.4 32 3.4 32 1.1 32 1.2 32
13.7 32 13.7 32 4.6 32 4.6 32 -100
24.0 32 24.0 32 16.2 32 16.2 32
33.7 32 33.7 32 21.2 32 21.2 32 -120
43.7 32 43.7 32 28.8 32 28.8 32 500 1000 1500 2000 2500 3000 3500 4000
53.0 32 53.3 32 41.9 32 41.9 32 Frequency (Hz)
63.8 32 63.8 32 61.0 32 61.1 32
74.6 32 74.6 32 83.3 32 83.3 32 Figure 7: Typical power spectrum of a vibration
85.2 32 85.2 32 104.9 32 104.9 32 signal at full-load.
96.6 32 96.6 32 119.1 32 119.1 32
5 kW, MOTOR 1 (Healthy) 5 kW, MOTOR 2 (Healthy) Table 3 presents the results of the rotor slot number
CURRENT FLUX CURRENT FLUX estimation technique based on the rotor slot passing
Load # of Load # of Load # of Load # of frequency, on the 2.2 kW and 5 kW motors. The table
(%) Slots (%) Slots (%) Slots (%) Slots shows that the technique works 90% of the time for the
6.3 12 6.3 12 6.6 12 6.6 12
28.4 12 28.4 12 27.1 12 27.1 12 2.2 kW motors (squirrel-cage type) but it fails for the 5
49.2 12 49.2 12 49.2 12 49.2 12 kW (wound-rotor) motors. It should be noted here that
71.8 12 71.8 12 72.0 12 72.0 12 the 2.2 kW motors were installed with precision tools,
85.3 12 85.3 12 87.8 12 87.8 12 while the 5 kW motors are typical motors with an
unknown installation procedure.
In Table 2 there are slight differences between the
calculated loads from the current signal and the flux In the 2.2 kW motors, the rotor slot number estimation
signal. These differences come from the fact that the fails at light loads for the healthy motor and fails at
loads are calculated based on the rotor speed from the medium loads for the faulty motor. The reasons for this
corresponding signal. For example, the loads in the rotor failure are that the rotor slot passing frequency
component is not significantly stronger than the other
components in the vibration signal, and the vibration
 signal from the faulty motor also contains further strong
signal components.
Table 3: Rotor Slot Number Estimation Based on
Rotor Slot Component in the Vibration Signal
financial support provided by National Instruments
Corporation, USA, and the technical support provided by
the workshop staff of the School of Electrical and
Electronic Engineering.

2.2 kW, MOTOR 1 (Healthy) 2.2 kW, MOTOR 2 (4BRB) 7. REFERENCES

Load # of Rotor Load # of Rotor
(%) Slots (%) Slots [1] I. Ahmed, R. Supangat, J. Grieger, N. Ertugrul, and W. L.
3.6 36 1.7 32
Soong, "A Baseline Study for On-Line Condition
13.8 32 4.7 32
23.2 32 16.1 32 Monitoring of Induction Machines," Australasian
33.3 32 21.0 35 Universities Power Engineering Conference (AUPEC),
43.4 32 29.2 32 Brisbane, Australia, 2004.
53.0 32 41.4 32 [2] M. E. H. Benbouzid, "A Review of Induction Motor
63.4 32 61.3 32
74.0 32 83.9 32
Signature Analysis as a Medium for Faults Detection,"
85.3 32 105.3 32 IEEE Transactions on Industrial Electronics, vol. 47, no.
96.9 32 119.0 32 5, pp. 984-993, October 2000.
5 kW, MOTOR 1 (Healthy) 5 kW, MOTOR 2 (Healthy) [3] G. B. Kliman, W. J. Premerlani, B. Yazici, R. A. Koegl,
Load # of Rotor Load # of Rotor and J. Mazereeuw, "Sensorless, Online Motor
(%) Slots (%) Slots Diagnostics," IEEE Computer Applications in Power, pp.
6.5 19 6.1 20 39-43, April 1997.
28.1 19 27.0 18
49.9 18 49.2 20
[4] S. Nandi and H. A. Toliyat, "Condition Monitoring and
72.4 19 71.6 19 Fault Diagnosis of Electrical Machines - A Review,"
85.3 12 87.8 20 IEEE Industrial Applications Society Annual Meeting,
vol. 1, pp. 197-204, Oct. 1999.
In the 5 kW motors, the rotor slot number estimation [5] R. R. Schoen and T. G. Habetler, "Effects of Time-
fails at almost all of the loading conditions. This is Varying Loads on Rotor Fault Detection in Induction
because again the vibration signal of these motors is Machines," IEEE Transactions on Industry Applications,
vol. 31, no. 4, pp. 900-906, July-Aug. 1995.
found to contain further strong components, which may
be caused by installation issues, the type of load, and the [6] W. T. Thomson, D. Rankin, and D. G. Dorrell, "On-Line
type of motor. Current Monitoring To Diagnose Airgap Eccentricity In
Large Three-Phase Induction Motors - Industrial Case
5. CONCLUSION Histories Verify The Predictions," IEEE Transactions on
Energy Conversion, vol. 14, no. 4, pp. 1372-1378, Dec.
1999.
This paper investigated the use of stator current, axial
leakage flux and vibration sensor measurements to detect [7] R. Blasco, M. Sumner, and G. M. Asher, "Speed
the rotor speed and the number of rotor slots of a three- Measurement of Inverter Fed Induction Motors Using the
phase induction motor based on measurements at a FFT and the Rotor Slot Harmonics," Power Electronics
and Variable-Speed Drives, London, Oct 1994.
single operating point. Operating points varying from
no-load to full-load were tested. Rotor speed estimation [8] A. Ferrah, K. G. Bradley, and G. M. Asher, "Sensorless
techniques based on the eccentricity harmonics, slip Speed Detection of Inverter Fed Induction Motors using
Rotor Slot Harmonics and Fast Fourier Transform,"
frequency, and the rotor frequency were investigated. Power Electronics Specialists Conference, Toledo, 1992.
The rotor slot number estimation techniques proposed
are based on the eccentricity harmonics and the rotor slot [9] K. D. Hurst and T. G. Habetler, "Sensorless Speed
passing frequency. A summary of the key results are as Measurement Using Current Harmonic Spectral
Estimation in Induction Machine Drives," IEEE
follows: Transactions on Power Electronics, vol. 11, no. 1, pp. 66-
• the rotor speed can be accurately determined from any 73, Jan. 1996.
of the current, flux and vibration sensor [10] D. Shi, P. J. Unsworth, and R. X. Gao, "Sensorless Speed
measurements, at any value of load; Measurement of Induction Motor Using Hilbert
• speed estimation techniques which utilise eccentricity Transform and Interpolated Fast Fourier Transform,"
harmonics and rotor frequency, are able to operate IEEE Transactions on Instrumentation and Measurement,
vol. 55, no. 1, pp. 290-299, Feb. 2006.
down to lower loads compared to utilising the slip
frequency; [11] K. L. Shi, T. F. Chan, Y. K. Wong, and S. L. Ho, "Speed
Estimation of an Induction Motor Drive Using an
• the number of rotor slots can be detected accurately
Optimized Extended Kalman Filter," IEEE Transactions
from the stator current and axial flux signals by on Industrial Electronics, vol. 49, no. 1, pp. 124-133, Feb.
utilising the rotor speed and the eccentricity 2002.
harmonics;
[12] B. Kaku, I. Miyashita, and S. Yasukawa, "Novel Speed
• rotor slot number estimation using the vibration signal Estimation Method for Induction Motor Based on Extra
is dependant on the installation, the type of load, and Rotor Synchronous Current Control," IEEE Power
the type of motor. Electronics Specialists Conference, Fukuoka, May 1998.

6. ACKNOWLEDGEMENT
This project was supported by an ARC Linkage Grant,
LP0453951. The authors would like to acknowledge the

View publication stats