Académique Documents
Professionnel Documents
Culture Documents
DECLARATION
I certify that
a. The work contained in this thesis is my own and has been done by me under the
guidance of my supervisor.
b. The work has not been submitted to any other institute for any degree or diploma.
c. I have followed the guidelines provided by the institute in preparing the thesis.
d. Whenever I have used material (data, theoretical analysis, figures and text) from other
sources, I have given due credits by citing in the text of the thesis with details in the
references.
Date:
Neelam Mehala
(2K07-NITK-Ph.D.-1160-E)
iii
Certificate
Certified that the thesis entitled, CONDITION MONITORING AND FAULT
DIAGNOSIS OF INDUCTION MOTOR USING MOTOR CURRENT SIGNATURE
ANALYSIS, submitted by Ms. NEELAM MEHALA is in fulfillment of the requirements
for the award of the degree of DOCTOR OF PHILOSOPHY from NATIONAL
INSTITUTE OF TECHNOLOGY, KURUKSHETRA. The candidate has worked under
my supervision. The work presented in this thesis has not been submitted for the award of
any other degree/diploma.
Date:
Dr. Ratna Dahiya
Department of Electrical Engineering
National Institute of Technology
Kurukshetra (Haryana)
iv
Acknowledgements
During my Ph.D. study at National Institute of Technology Kurukshetra, I have been
fortunate to receive valuable suggestions, guidance and support from my mentors, colleagues,
family and friends.
First of all, I would like to express my most sincere gratitude to my supervisor
Dr. Ratna Dahiya. She has been a wise and trusted guide throughout the entire process. Her
guidance helped me to solve engineering problems and improve my communication. If it has
not been for her vision, encouragement, and her confidence in my ability, much of this work
would not have been completed.
I express my sincere gratitude and indebtedness to Dr. K.S. Sandhu,
Chairman, Department of Electrical Engineering, National Institute of Technology
Kurukshetra for his moral support and continuous encouragement.
I must thank to Sh. Satpal and Sh. Suresh Kumar, Sr. Instructors, YMCA University
of Science and Technology, Faridabad who were always available and willing to help with
laboratory experimental set up.
I would like to thank my husband Dr. Vikas Kumar for his moral support and
continuous encouragement. Finally, I extend my sincere gratitude to all those people who
helped me in all their capacity to complete this work.
ABSTRACT
Condition monitoring of induction motor have been a challenging task for the engineers and
researchers mainly in industries. There are many condition monitoring methods, including
vibration monitoring, thermal monitoring, chemical monitoring, acoustic emission
monitoring but all these monitoring methods require expensive sensors or specialized tools
whereas current monitoring out of all does not require additional sensors. Current monitoring
techniques are usually applied to detect the various types of induction motor faults such as
rotor fault, short winding fault, air gap eccentricity fault, bearing fault, load fault etc. In
current monitoring, no additional sensors are necessary. This is because the basic electrical
quantities associated with electromechanical plants such as current and voltage are readily
measured by tapping into the existing voltage and current transformers that are always
installed as part of the protection system. As a result, current monitoring is non-intrusive and
may even be implemented in the motor control center remotely from the motors being
monitored. Motor current signature analysis (MCSA) and Park's vector approach fall under
current monitoring. The MCSA uses the current spectrum of the machine for locating
characteristic fault frequencies. When a fault is present, the frequency spectrum of the line
current becomes different from healthy motor. Such fault modulates the air-gap and produces
rotating frequency harmonics in the self and mutual inductances of the machine. It depends
upon locating specific harmonic component in the line current.
Extensive literature survey has been done for understanding the various faults and
signal processing techniques available. It was observed that fault frequencies occur in the
motor current spectra are unique for different motor faults. These fault frequencies can be
easily detected with help of Motor Current Signature analysis (MCSA). Therefore, MCSA
based techniques are used in present work for detection of common faults of induction
motors. In addition, Park's vector approach is also applied for fault detection of induction
motor. The proposed methods in this research allows continuous real time tracking of various
types of faults in induction motors operating under continuous stationary and non stationary
conditions. These methods recognize the fault signatures produced in induction motor and
estimate the severity of the faults under different load conditions. The effects of these faults
on motor current spectra of an induction motor are investigated through experiments. In order
to perform the analysis on induction motors, an experimental set up is designed that can
vi
accurately repeat the measurements of current signals. In the present research work,
LabVIEW software is used to diagnose the faults of induction motor with direct online
monitoring. The experiments were conducted in four phases.
The first phase experimentally describes the effect of rotor faults on stator current of
motor. Three algorithms are proposed to track and detect the rotor faults in induction motors
operating under different load conditions: Fast Fourier Transform algorithm (FFT), Short
Time Fourier transform algorithm and Wavelet Transform based multiresolution analysis
algorithm. FFT Method is easy to implement. However, this method does not show the time
information. This is a serious drawback of FFT. More interesting signals contain numerous
transitory characteristics such as drift, trends, and abrupt changes. These characteristics are
often the most important part of the signal, and the Fourier analysis is not suitable for their
detection. Therefore, other methods for signal analysis such as STFT, Wavelet transform are
subsequently used to detect the rotor faults experimentally.
The second phase investigates short winding faults of induction motor. A short turn
fault in induction motor can result in complete failure and shut down of the machine unless
the fault is detected early, and evasive action is taken. In the research, this fault has been
detected successfully using four types of algorithms: FFT, Gabor Transform, Wavelet
Transform, Park's Vector Approach.
The air gap eccentricity faults are studied in third phase of the research. Same
experimental set up is used for this purpose. Special methods were applied to implement
static eccentricity and dynamic eccentricity in induction motor. Experimental results show
that it is possible to detect the presence of air-gap eccentricity in operating three phase
induction motor, by computer aided monitoring of stator current. Qualitative information
about severity of fault can be obtained by using power spectrum.
The forth phase of research work investigates the application of advanced signal
processing techniques for detection of mechanical faults such as bearing faults and gear box
faults. It is experimentally demonstrated that faults in bearings may be detected by
monitoring the voltage/current of the stator. This may offer an inexpensive alternative to
vibration diagnostics that require sensors which are expensive. It is observed that the
characteristic frequencies are not visible in the power spectrum for a smaller size outer race
fault and inner race fault. As severity of fault increases, the characteristic frequencies become
vii
visible. Wigner-Ville Distribition (WD) is also implemented for detection of bearing faults.
In addition, Parks Vector approach is also applied for detecting the bearing faults. It is
verified from experiments that the Parks vector current spectrum of healthy motor is
different from the current spectrum of the motor having faulty bearing. To detect the gear
box fault, an experiment has also been conducted. The results obtained from this experiment
show that any fault in either the pinion or the driven wheel generates a harmonic component
in the motor current spectrum which can be detected in power spectrum of induction motor.
The conclusions, contributions, and recommendations are summarized at the end.
viii
Table of Contents
Acknowledgements
Abstract
List of Tables
List of Figures
CHAPTER 1 Introduction
1.1 Overview..1
1.2 Objectives of research work.3
1.3 Orientation............................................................................................................... 6
ix
xi
5.3 Diagnosis of stator winding faults using FFT based power spectrum...91
5.3.1. Data acquisition parameters and LabVIEW programming.92
5.3.2. Observations and discussion....93
5.4 Stator winding fault diagnosis using Gabor Transform.99
5.4.1 Data acquisition parameters and LabVIEW programming...99
5.4.2 Observations and discussion ....101
5.5 Stator winding fault analysis using Wavelet Transform....102
5.5.1 Data acquisition parameters and LabVIEW programming..102
5.5.2 Observations and discussion.106
5.6 Park's Vector approach for diagnosis of short winding fault ....106
5.6.1 Data acquisition parameters and LabVIEW programming..106
5.6.2 Observations and discussion.109
5.7 Chapter summary.......109
xii
References 164
List of publications from the research work..175
xiii
List of Tables
Table 2.1 Statistics on motor faults/failure modes......12
Table 4.1: Expected fault frequencies at various load condition.61
Table 4.2: Parameters of experimental induction motor..62
Table 4.3: Specifications of data acquisition card (NI-PCI 6251)...63
Table 4.4: Data acquisition parameters67
Table 4.5: Power spectrum analysis of one broken bar at various loading conditions69
Table 4.6: Power spectrum analysis of five broken bars at various loading conditions..69
Table 4.7: Power spectrum analysis of twelve broken bars at various loading conditions.70
Table 4.8: Data acquisition parameters...77
Table 4.9: Decomposition details81
Table 5.1: Expected fault frequencies at various load conditions...91
Table 5.2: Experimental conditions for short winding fault detection93
Table 5.3: Power spectrum analysis for short circuited winding fault95
Table 5.4: Data acquisition parameters.101
Table 6.1: Expected fault frequencies at various load conditions.115
Table 6.2: Power spectrum analysis for 25% static eccentricity...118
Table 6.3: Power spectrum analysis for 50% air gap eccentricity119
Table 6.4: Power spectrum analysis for mixed eccentricity..119
Table 7.1: Expected fault frequencies for inner race fault at various load conditions..131
Table 7.2: Expected fault frequencies for outer race fault at various load conditions...131
Table 7.3: Experimental conditions for bearing fault detection134
Table 7.4: Power spectrum analysis for inner race fault of motor with 2mm hole134
Table 7.5: Power spectrum analysis for induction motor with 4mm inner race fault135
Table 7.6: Power spectrum analysis for induction motor with 2mm outer race fault136
Table 7.7: Power spectrum analysis for induction motor with 4mm outer race fault136
Table 8.1: Comparison of techniques applied for diagnosis of motor fault...162
xiv
List of Figures
Figure 1.1: Research plan..5
Figure 2.1: The process for fault diagnosis..13
Figure 3.1: Various types of short winding faults....39
Figure 3.2: Power spectrum of a healthy motor...45
Figure 3.3: STFT of healthy motor..47
Figure 3.4: Gabor spectrogram of a healthy motor..48
Figure 3.5: WVD representation of a faulty motor..49
Figure 3.6: Two channel perfect reconstruct filter..52
Figure 3.7: Discrete Wavelet Transform.52
Figure 3.8: Frequency range cover for details and final approximation..54
Figure 3.9: Current Parks vector for ideal condition..56
Figure 4.1: Idealized current spectrum60
Figure 4.2: Experimental set up...64
Figure 4.3: Data acquisition card (PCI-6251). 64
Figure 4.4: Data acquisition board (ELVIS)65
Figure 4.5: Motor fault detection and diagnosis system..66
Figure 4.6: Block diagram for obtaining power spectrum using LabVIEW programming.67
Figure 4.7: Power spectrum of healthy motor at no load.71
Figure 4.8: Power spectrum of faulty motor with 1 broken bar under no load condition...71
Figure 4.9: Power spectrum of faulty motor with 5 broken bars under no load condition..72
Figure 4.10: Power spectrum of faulty motor with 12 broken bars under no load condition..72
Figure 4.11: Power spectrum of healthy motor under half load..73
Figure 4.12: Power spectrum of faulty motor with 1 broken bar under half load...73
xv
Figure 4.13: Power spectrum of faulty motor with 5 broken bars under half load.....74
Figure 4.14: Power spectrum of faulty motor with 12 broken bars under half load...74
Figure 4.15: Power spectrum of healthy motor under full load .75
Figure 4.16: Power spectrum of faulty motor with 1 broken bar under full load...75
Figure 4.17: Power spectrum of faulty motor with 5 broken bars under full load..76
Figure 4.18: Power spectrum of faulty motor with 12 broken bars under full load76
Figure4.19:Block
diagram
for
obtaining
STFT
spectrogram
using
LabVIEW
programming ......78
Figure 4.20: STFT spectrogram for healthy motor..79
Figure 4.21: STFT spectrogram for faulty induction motor with broken bars79
Figure 4.22: Block diagram for Multiresolution analysis using LabVIEW programming..82
Figure 4.23: Multiresolution analysis for healthy motor.83
Figure 4.24: Multiresolution analysis for faulty motor with broken bars84
Figure 4.25: Slotted disc used in experiment...85
Figure 4.26: Experimental set up.86
Figure 4.27: Power spectrum of motor (Bolts placed on inner position of slotted
disc).86
Figure 4.28: Power spectrum of motor (Bolts placed in outer position of slotted
disc).87
Figure 5.1: Experimental set up...92
Figure 5.2: Power spectrum of healthy motor under no load condition..95
Figure 5.3: Power spectrum of faulty motor with 5% shortened under no load condition..96
Figure 5.4: Power spectrum of faulty motor with 15% shortened under no load condition....96
Figure 5.5: Power spectrum of faulty motor with 30% shortened under no load condition....97
Figure 5.6: Power spectrum of healthy motor under full load.97
xvi
Figure 5.7: Power spectrum of faulty motor (5% shortened) under full load..98
Figure 5.8: Power spectrum of faulty motor (15% shortened) under full load98
Figure 5.9: Power spectrum of faulty motor (30% shortened) under full load99
Figure5.10:Block
diagram
for
obtaining
Gabor
spectrogram
using
LabVIEW
programming....100
Figure 5.11: Gabor spectrogram for healthy induction motor...100
Figure 5.12: Gabor spectrogram for short circuited induction motor101
Figure 5.13: Block diagram for Multiresolution analysis using LabVIEW programming103
Figure 5.14: Multiresolution analysis for healthy motor...104
Figure 5.15: Multi resolution analysis for 30%short circuited induction motor105
Figure 5.16: Block diagram for experimental detection system....107
Figure 5.17: Block diagram for obtaining Current Park's vector pattern using LabVIEW
programming.................................................................................................107
Figure 5.18: Current Parks vector pattern for healthy motor...108
Figure 5.19: Current Parks vector pattern for short circuited motor....108
Figure 6.1: Healthy electric motor.112
Figure 6.2: Difference between static and dynamic eccentricity...113
Figure 6.3: Implementation of static eccentricity in induction motor..115
Figure 6.4: Parts of motor machined for implementing air gap eccentricity.116
Figure6.5:Block
diagram
for
obtaining
power
spectrum
using
LabVIEW
programming117
Figure 6.6: Power spectrum of healthy motor under no load condition120
Figure 6.7: Power spectrum of faulty motor with 25% static eccentricity under no load
condition...120
Figure 6.8: Power spectrum of faulty motor with 50% static eccentricity under no Load
condition..121
xvii
Figure 6.9: Power spectrum of healthy motor under full load condition...121
Figure 6.10: Power spectrum of faulty motor with 25% static eccentricity under full load..122
Figure 6.11: power spectrum of faulty motor with 50% eccentricity under full load..122
Figure 6.12: Power spectrum of healthy motor under no load condition..123
Figure 6.13: Power spectrum of faulty motor with mixed eccentricity under no load
condition...124
Figure 6.14: Power spectrum of healthy motor under full load125
Figure 6.15: Power spectrum of healthy motor with mixed eccentricity under full Load.126
Figure 7.1: Ball bearing dimensions..130
Figure 7.2: Inner race fault 132
Figure 7.3: Outer race fault....132
Figure 7.4: Power spectrum of healthy motor under no load condition...137
Figure 7.5: Power spectrum of faulty motor with 2mm hole in inner race of bearing under no
load condition (m=1)137
Figure 7.6: Power spectrum of faulty motor with 2mm hole in inner race of bearing under no
load condition (m=2)....138
Figure 7.7: Power spectrum of faulty motor with 4mm hole in inner race of bearing under no
load condition (m=1)138
Figure 7.8: Power spectrum of faulty motor with 4mm hole in inner race of bearing under no
load condition (m=2)139
Figure 7.9: Power spectrum of healthy motor under full load condition.139
Figure 7.10: Power spectrum of faulty motor with 2mm hole in inner race of bearing under
full load condition.140
Figure 7.11: Power spectrum of faulty motor with 4mm hole in inner race of bearing under
full load condition.140
Figure 7.12: Power spectrum of healthy motor under no load condition.141
Figure 7.13: Power spectrum of faulty motor with 2mm hole in outer race of bearing under
no load condition..141
xviii
Figure 7.14: Power spectrum of faulty motor with 4mm hole in outer race of bearing under
no load condition..142
Figure 7.15: Power spectrum of healthy motor under full load condition142
Figure 7.16: Power spectrum of faulty motor with 2mm hole in outer race of bearing under
full load condition.143
Figure 7.17: Power spectrum of faulty motor with 4mm hole in outer race of bearing under
full load condition.143
Figure 7.18: Block diagram for obtaining Wigner-Ville Distribution (WVD) representation
using LabVIEW programming.................................................................144
Figure 7.19: Wigner-Ville Distribution (WVD) representation for motor with healthy
bearing..145
Figure 7.20: Wigner-Ville Distribution (WVD) representation for motor with faulty bearing
(4mm hole in outer race)..145
Figure 7.21: Block diagram for obtaining Current Park's vector pattern using LabVIEW
programming147
Figure 7.22: Current Parks Vector pattern for healthy motor......147
Figure 7.23: Current Parks vector pattern for faulty bearing with 4 mm diameter hole in
inner race..148
Figure 7.24: Current Park vectors pattern for faulty bearing with 4 mm diameter hole in
outer race...148
Figure 7.25: Worm and worm gear151
Figure 7.26: Parts of gear box151
Figure 7.27: Worm wheel with damage tooth...152
Figure 7.28: Experimental set up...152
Figure 7.29: Motor coupled with load...153
Figure 7.30: Power spectrum for healthy gear box....154
Figure 7.31: Power spectrum of motor with faulty gear box.154
xix
CHAPTER 1
Introduction
1.1 Overview
The studies of induction motor behavior during abnormal conditions due to presence
of faults and the possibility to diagnose these abnormal conditions have been a challenging
topic for many electrical machine researchers. There are many condition monitoring methods
including vibration monitoring, thermal monitoring, chemical monitoring, acoustic emission
monitoring but all these monitoring methods require expensive sensors or specialized tools
where as current monitoring out of all does not require additional sensors. This is because the
basic electrical quantities associated with electromechanical plants such as current and
voltage are readily measured by tapping into the existing voltage and current transformers
that are always installed as part of the protection system. As a result, current monitoring is
non-intrusive and may even be implemented in the motor control center remotely from the
motors being monitored. [1-2].
It is observed that the technique called Motor Current Signature Analysis (MCSA)
is based on current monitoring of induction motor; therefore it is not very expensive. The
MCSA uses the current spectrum of the machine for locating characteristic fault frequencies.
When a fault is present, the frequency spectrum of the line current becomes different from
healthy motor. Such a fault modulates the air-gap and produces rotating frequency harmonics
in the self and mutual inductances of the machine. It depends upon locating specific
harmonic component in the line current [3-4]. Therefore, it offers significant implementation
and economic benefits. In the research work, Motor Current Signature Analysis (MCSA)
based methods are used to diagnose the common faults of induction motor such as broken bar
fault, short winding fault, bearing fault, air gap eccentricity fault, and load faults. The
proposed methods in the research allows continuous real time tracking of various types of
faults in induction motors operating under continuous and variable loaded conditions. The
effects of various faults on current spectrum of an induction motor are investigated through
experiments.
The various advanced signal processing techniques such as Fast Fourier Transform,
Short Time Fourier Transform, Gabor Transform, and Wavelet Transform are used to
diagnose the faults of induction motor. A suitability of the signal for different type of faults is
also discussed in detail. FFT is easy to implement but the drawback of this technique is that it
is not suitable for analyzing transient signals. Although Short-Time Fourier Transform
(STFT) can be used for analyzing transient signals using a time-frequency representation, but
it can only analyze the signal with a fixed sized window for all frequencies, which leads to
poor frequency resolution [5-6]. However, Wavelet Transform can overcome this problem by
using a variable sized window.
In order to perform accurate and reliable analysis on induction motors, the installation
of the motors and measurement of signal need to be accurate. Therefore, an experimental
procedure and an experimental set up has been designed that can accurately repeat the
measurements of signals and can introduce a particular fault to the motor in isolation of other
faults. Stator current contains unique fault frequency components that can be used for
detection of various faults of motor. Therefore, this research work investigates how the
presence of common faults, such as rotor bar fault, short winding fault, air gap eccentricity,
bearing fault, load fault, affects on different fault frequencies under different load conditions.
In the research work, signal processing techniques are used for condition monitoring
and fault detection of induction motors. The signal processing techniques have advantages
that these are not computationally expensive, and these are simple to implement. Therefore,
fault detection based on the signal processing techniques is suitable for an automated on-line
condition monitoring system [7]. Signal processing techniques usually analyze and compare
the magnitude of the fault frequency components, where the magnitude tends to increase as
the severity of the fault increase. Therefore, the various signal processing techniques are used
in present work for detection of common faults of induction motor. Signal processing
techniques have their limitations. For example, the reliability of detecting the rotor fault
using Fast Fourier Transform (FFT) depends on loading conditions and severity of fault. If
the loading condition is too low or the fault is not too severe, Fast Fourier Transform may fail
to identify the fault. Therefore, different techniques such as Wavelet Transform (WT) are
investigated in the research work to find better features for detecting common faults under
different loading conditions.
In present research work, twelve experiments are performed to diagnose the common
faults of induction motors using six different currents monitoring techniques. The results and
observations obtained are discussed and then final conclusions are made.
by using LabVIEW also, but they have not diagnosed all common types of faults of induction
motor.
In order to perform accurate and reliable analysis on induction motors, the installation
of the motors and measurement of their signal need to be reliable. Therefore, the first aim of
this thesis is to design an experimental procedure and an experimental set up that can
accurately repeat the measurements of signals and can introduce a particular fault to the
motor in isolation of other faults.
Stator current contains unique fault frequency components that can be used for
detection of various faults of motor. The methods proposed in this research work allow
continuous real time tracking of faults in induction motors operating under continuous
stationary and non stationary conditions. Therefore, second aim of this research work is to
investigate how the presence of common faults, such as rotor bar fault, short winding fault,
air gap eccentricity, bearing fault, load fault, affect on different fault frequencies under
different load conditions .
In this research work, condition monitoring and fault detection of induction motors is
based on the signal processing techniques. The signal processing techniques have advantages
that these are not computationally expensive and these are simple to implement. Therefore,
fault detection based on the signal processing techniques is suitable for an automated on-line
condition monitoring system. Signal processing techniques usually analyze and compare the
magnitude of the fault frequency components, where the magnitude tends to increase as the
severity of the fault increase. Therefore, the third aim of this thesis is to utilize the various
signal processing techniques for detection of common faults of induction motor.
Signal processing techniques have their limitations. For example, some faults could
be not diagnosed using Fast Fourier Transform, if the loading condition is too low or the fault
is not too severe. Therefore, the final aim of this thesis is to investigate new features using
different techniques such as Wavelet Transform (WT), to find better features for detecting
common faults under different loading conditions.
Broken
rotor bar
fault
Exp 1
Short
winding
fault
Exp 2
Exp 3
Air gap
Ecce.
fault
Exp 4
Gear
box
fault
Bear
ing
failure
Exp 5
FFT
Exp 6
STFT
Exp 7
Wigner
distribution
Exp 8
Exp 9
Gabor
transform
Wavelet
transform
Exp 10
Conclusions
Figure 1.1: Research Plan
Exp 11
Parks
Vector
Exp 12
1.3 Orientation
The research work is presented in eight chapters of this thesis. Chapter 1 presents
overview on condition monitoring of induction motors and objectives of research work along
with the organization of the thesis.
Chapter 2 deals with the detailed literature survey and review of previous work on
induction motor condition monitoring. It also provides the motivation to work on common
faults of induction machine and their diagnostics techniques.
In chapter 3, common faults of induction motor such as rotor fault, short winding
fault, air gap eccentricity fault, load fault and bearing fault has been introduced. Various
signal processing techniques such as Fast Fourier Transform, Short Time Frequency
Transform, Gabor Transform, Wigner-Ville Distribution and Wavelet Transform along with
mathematical equation is given.
Experimental investigation of the rotor faults of induction motors operating under
different load conditions is considered in chapter 4. The fault algorithm monitors the
amplitude of fault frequencies and tracks changes in their amplitudes over time. Experiments
are performed with using current based fault detection techniques such as Fast Fourier
Transform, Short Time Fourier Transform, and Discrete Wavelet Transform. To diagnose the
fault with these techniques, a laboratory test bench was set up. It consists of a three-phase
squirrel cage induction machine coupled with rope brake dynamometer. The rated data of the
tested three-phase squirrel cage induction machine were: 0.5 hp, 415V, 1.05 A and 1380(FL)
r/min. The speed of the motor was measured by digital tachometer. The Virtual Instrument
(VIs) was built up with programming in LabVIEW 8.2. This VIs was used both for
controlling the test measurements and data acquisition, and for data processing. The data
acquisition card (PCI-6251) and acquisition board (ELVIS) were used to acquire the current
samples from the motor under different load conditions. In order to test the system in
practical cases, several measurements were made, where the stator current of a machine with
known number of broken rotor bars was read. Current measurements were performed for a
healthy rotor and also for the same type of motor having different number of broken rotor
bars. Tests were carried out for different loads with the healthy motor, and with similar
motors having broken rotor bars. The rotor faults were provoked interrupting the rotor bars
by drilling into the rotor. The measured current signals were processed using the Fast Fourier
Transformation (FFT). Another experiment is performed to diagnose the broken rotor bar
fault using STFT. Multiresolution analysis has also been applied to diagnose the broken rotor
bar fault under varying load conditions. In addition, the effect of unbalance rotor is also
studied in the research work. To unbalance the rotor, a slotted disc with attached weights is
mounted on the shaft of motor. Then power spectrum is obtained using Virtual Instrument
(VIs). This power spectrum is compared with power spectrum of healthy motor to search out
the characteristic frequencies for studying the effect of unbalance rotor.
Chapter 5 presents the experimental work for diagnosis of stator winding faults in
induction motors operating under different load conditions. To diagnose the short winding
fault, MCSA based fault detection techniques such as FFT, Gabor Transform, Wavelet
Transform (WT) and Parks vector approach are implemented. Several experiments were
performed on motor under no load condition and with load coupled to shaft of motor. Short
winding fault was diagnosed with FFT for 5%, 15% and 30% short circuit of winding. The
results were compared to make the conclusions. After this, Gabor Transform and Wavelet
Transform was applied to diagnose the same fault with 30% short circuit of winding. The
Parks vector approach was also introduced for detecting the short winding faults. An
undamaged machine shows a perfect circle in Parks vector representation whereas an
unbalance due to winding faults results in an elliptic representation of the Parks vector. The
results obtained from the experiments present a great degree of reliability, which enables
these techniques to be used as monitoring tool for short circuit fault of motor.
The air-gap eccentricity fault in three phase induction motor is discussed in chapter 6.
The rated data of the tested three-phase squirrel cage induction machine were: 0.5 hp, 415V,
1.05 A and 1380(FL) r/min. To detect the eccentricity fault, Fast Fourier Transform (FFT) is
implemented. It was very difficult to create air gap eccentricity fault in motor because air gap
was very smaller in amount. Therefore, the special methods were used to replicate the air gap
eccentricity fault in laboratory. Experimental results demonstrate the effectiveness of the
proposed technique for detecting presence of air gap eccentricity in operating three phase
induction machine. Qualitative information about severity of air gap eccentricity fault can be
easily obtained by using FFT.
Chapter 7 proposes the experiments to investigate the load and bearing faults of
induction motor and their effect on the motor current spectrums. Gear defects and bearing
defects are replicated in the laboratory. The bearings were made failed by drilling the hole in
inner race and outer race of the bearing with help of Electric Discharge Machine (EDM).
Defective rolling element bearings generate eccentricity in the air gap with mechanical
vibrations. The air gap eccentricities cause vibrations in the air gap flux density that produces
visible changes in the stator current spectrum. The techniques such as FFT, Wigner-Ville
Distribution, Parks vector approach are applied to detect the bearing faults of motor. In the
research work, an experiment has also been conducted to defect the load fault. The load fault
8
is created by deforming gears tooth of gear box. The defective gear box (worm and worm
gear) is coupled to motor with help of coupling and experiment was conducted. Whenever
deformed tooth reaches the worm, the motor experience a Bump in its load which gives rise
to two frequency components symmetrically around the main frequency. This experiment
verifies the fault in gear box coupled to motor by monitoring the current in induction motor.
Chapter 8 presents the conclusion, contribution and scope for future work. The
research investigates the applications of advanced signal processing techniques to detect
various types of faults of motor such as rotor bar fault, stator winding fault, air gap
eccentricity fault, bearing failure, and load fault. The research work helps in understanding
the applications and limitations of fault detecting techniques. It is observed that LabVIEW is
user friendly software and may be helpful in detecting the faults on and off line. It also helps
in saving computational time of diagnosis. The new detecting methods proposed in this work
are able to diagnose motors faults more sensitively and more reliably.
CHAPTER 2
Literature Review
2.1 Introduction
In this chapter, the literature on condition monitoring of electric machine is reviewed.
This review covers some important topics such as condition monitoring, fault diagnosis,
thermal monitoring, vibration monitoring, electric monitoring, noise monitoring, motor
current signature analysis, Current parks vector approach, Fast Fourier Transform, STFT,
Wavelet transform, signal processing techniques, etc. In addition, this review also covers the
major developments in this field from early research to most recent.
circuits, rotating with respect to each other and in which power is transferred from one circuit
to the other by electromagnetic induction. It is an electromechanical energy conversion
device in which the energy converts from electric to mechanical form [8]. The energy
conversion depends upon the existence in nature of phenomena interrelating magnetic and
electric fields on the one hand, and mechanical force and motion on the other. The rotor
winding in induction motors can be squirrel-cage type or wound-rotor type. Thus, the
induction motors are classified into two groups [9]:
Squirrel-cage and
The squirrel cage induction motor consist of conducting bars embedded in slots in the rotor
iron and short circuited at each end by conducting end rings. The rotor bars are usually made
of copper, aluminum, magnesium or alloy placed in slots. Standard squirrel cage rotors have
no insulation since bars carry large currents at low voltages. Another type of rotor, called a
form-wound rotor, carries a poly phase winding similar to three phase stator winding. The
terminals of the rotor winding are connected to three insulated slip rings mounted on the
rotor shaft. In a form-wound rotor, slip rings are connected to an external variable resistance
which can limit starting current and associated rotor heating. During start-up, inserting
external resistance in the wound-rotor circuit produces a higher starting torque with less
starting current than squirrel-cage rotors [9]. This is desirable for motors which must be
started often.
The squirrel-cage induction motor is simpler, more economical, and more rugged
than the wound-rotor induction motor. A squirrel-cage induction motor is a constant speed
motor when connected to a constant voltage and constant frequency power supply. If the load
torque increases, the speed drops by a very small amount. It is therefore suitable for use in
constant-speed drive systems [8,9]. On the other hand, many industrial applications require
several speeds or a continuously adjustable range of speeds. DC motors are traditionally used
in adjustable drive systems. However, since DC motors are expensive, and require frequent
maintenance of commutators and brushes. Squirrel-cage induction motors are preferred
because they are cheap, rugged, have no commutators, and are suitable for high-speed
applications. In addition, the availability of solid state controllers has also made possible to
use squirrel-cage induction motors in variable speed drive systems. The squirrel-cage
11
induction motor is widely used in both low performance and high performance drive
applications because of its roughness and versatility.
Electric machines are frequently exposed to non-ideal or even detrimental operating
environments. These circumstances include overload, insufficient lubrication, frequent motor
starts/stops, inadequate cooling, etc. Under these conditions, electric motors are subjected to
undesirable stresses, which put the motors under risk of faults or failures [10]. There is need
to improve the reliability of motors due to their significant positions in applications.
According to IEEE Standard 493-1997 [11], the most common faults and their statistical
occurrences are listed in Table 1. This table is based on a survey on various motors in
industrial applications. According to the table, most faults happen to bearings and windings.
A 1985 statistical study by the Electric Power Research Institute (EPRI) provides similar
results, i.e., bearing (41%), stator (37%), rotor (10%) and other (12%) [12]. Several
contributions deal with these faults.
Types of faults
Bearing
Winding
Rotors
Shaft
Brushes or slip rings
External device
Others
152
75
8
19
-40
10
2
16
1
6
7
9
10
6
4
-8
1
--
2
-2
2
All motors
166
97
13
19
16
18
51
12
also possible to schedule future preventive maintenance and repair work. This can result in
minimum down time and optimum maintenance schedules [2]. Condition monitoring and
fault diagnosis scheme allows the machine operator to have the necessary spare parts before
the machine is stripped down, thereby reducing outage times. Therefore, effective condition
monitoring of electric machines is critical in improving the reliability, safety, and
productivity.
Data acquisition
Feature extraction
Decision making
Figure 2.1: The process for fault diagnosis
13
The condition monitoring of electrical and mechanical devices has been in practice for quite
some time now. Several methods have evolved over time but the most prominent techniques
are thermal monitoring, vibration monitoring, and electrical monitoring, noise monitoring,
torque monitoring and flux monitoring.
detecting bearing faults, the increased bearing wear increases the friction and the temperature
in that region of the machine. This increase in temperature of motor can be a detected by
thermal monitoring.
15
aerodynamic origin [20]. The largest sources of vibration and noise in electric machines are
the radial forces due to the air gap field. Since the air gap flux density distribution is product
of the resultant m.m.f. wave and total permeance wave. The resultant m.m.f. also contains the
effect of possible rotor or stator asymmetries, and permeanance wave depends on the
variation of the air gap as well , the resulting magnetic forces and vibrations are also depends
on these asymmetries. Thus by analyzing the vibration signal of an electric machine, it is
possible to detect various types of faults and asymmetries [22]. Bearing faults, rotor
eccentricities, gear faults and unbalanced rotors are the best candidates for vibration based
diagnostics. The vibration monitoring of electric machines is accomplished through the use
of broad-band, narrow-band, or spectral (signature) analysis of the measured vibration energy
of the machine. Vibration-based diagnostics is the best method for fault diagnosis, but needs
expensive accelerometers and associated wiring. This limits its use in several applications,
especially in small machines where cost plays a major factor in deciding the condition
monitoring method.
Li et al. [23] carried out vibration monitoring for rolling bearing fault diagnoses. The
final diagnoses are made with an artificial NN. The research was conducted with simulated
vibration and real measurements. In both cases, the results indicate that a neural network can
be an effective tool in the diagnosis of various motor bearing faults through the measurement
and interpretation of bearing vibration signatures. In this study, the vibration features are
obtained from the frequency domain using the FFT technique. Five vibration signatures are
constructed. They are created from the power spectrum of the vibration signal and consist of
the corresponding basic frequencies, with varying amplitudes based on the defect present.
Time domain information, such as the maximum and mean value of the amplitude vibration
waveform and the Kurtosis factor of the vibration waveform, are also considered. Thus, the
complete neural network has six input measurements. Researchers showed how the neural
network can be used effectively in the diagnosis of various motor bearing faults through
appropriate measurement and interpretation of motor bearing vibration signals. In Jack &
Nandi [24], there is an approach that brings better results. In this, the artificial neural network
is helped by a genetic algorithm. In this study, statistical estimates of the vibration signal are
considered as input features. The study examines the use of a genetic algorithm to select the
most significant input features in the machine condition monitoring contexts. By doing this, a
16
subset of six input features from a large set of possible features is selected, giving a very high
classification accuracy of 99.8 %. Li et al. [23] and Jack & Nandi [24] are devoted to
detecting mechanical faults; a similar approach could be extended to analyse the vibration
pattern when an electrical machine is working with an electrical fault.
The major disadvantage of vibration monitoring is cost. For example, a regular
vibration sensor costs several hundred dollars. A high product cost can be incurred just by
employing the necessary vibration sensors for a large number of electric machines. Another
disadvantage of vibration monitoring is that it requires access to the machine. For accurate
measurements, sensors should be mounted tightly on the electric machines, and expertise is
required in the mounting [25-27] In addition, sensors themselves may fail.
In most
applications, the stator current of an induction motor is readily available since it is used to
protect machines from destructive over-currents, ground current, etc. Therefore, current
monitoring is a sensor-less detection method that can be implemented without any extra
hardware [28].
2.4.5.1. Current signature Analysis
Numerous applications of using MCSA in equipment health monitoring have been
published among the nuclear-generation, industrial, defense industries. In most applications,
stator current is monitored for diagnosis of different faults of induction motor. Randy R.
Schoen et. al. [29] addressed the application of motor current signature analysis for the
detection of rolling-element bearing damage in induction machines. This study investigates
the efficacy of current monitoring for bearing fault detection by correlating the relationship
between vibration and current frequencies caused by incipient bearing failures. In this study,
the bearing failure modes are reviewed and the characteristic bearing frequencies associated
with the physical construction of the bearings are defined. The effects on the stator current
spectrum are described and the related frequencies determined. Experimental results which
show the vibration and current spectra of an induction machine with different bearing faults
17
are used to verify the relationship between the vibrational and current frequencies. The test
results clearly illustrate that the stator current signature can be used to identify the presence
of a bearing fault.
Randy R. Schoen [30] presented a method for on-line detection of incipient induction
motor failures which requires no user interpretation of the motor current signature, even in
the presence of unknown load and line conditions. A selective frequency filter learns the
characteristic frequencies of the induction machine while operating under all normal load
conditions. The generated frequency table is reduced to a manageable number through the
use of a set of expert system rules based upon the known physical construction of the
machine. This list of frequencies forms the neural network clustering algorithm inputs which
are compared to the operational characteristics learned from the initial motor performance.
This only requires that the machine be in good operating condition while training the
system. Since a defect continues to degrade the current signature as it progresses over time,
the system looks for those changes in the original learned spectra that are indicative of a fault
condition and alarms when they have deviated by a sufficient amount. The combination of a
rulebased (expert system) frequency filter and a neural network maximizes the systems
ability to detect the small spectral changes produced by incipient fault conditions. Compete
failure detection algorithm was implemented and tested. An impending motor failure was
simulated by introducing a rotating mechanical eccentricity to the test machine. After
training the neural network, the system was able to readily detect the current spectral changes
produced by the fault condition.
Schoen and Habetler [31-32] investigated the effects of a position-varying load torque
on the detection of air gap eccentricity. The torque oscillations were found to cause the same
harmonics as eccentricity. These harmonics are always much larger than eccentricity-related
fault harmonics. Therefore, it was concluded that it is impossible to separate torque
oscillations and eccentricity unless the angular position of the eccentricity fault with respect
to the load torque characteristic is known.
Randy R. Schoen and Thomas G. Habetler [33] presented an analysis of the effects of
position-varying loads on the current harmonic spectrum. The load torque-induced harmonics
were shown to be coincidental with rotor fault-induced harmonics when the load varies
synchronously with the rotor position. Furthermore, since the effect of the load and fault on a
18
single stator current harmonic component is spatially dependent, the fault induced portion
cannot be separated from the load portion. Therefore, any on-line detection scheme which
measures the spectrum of a single phase of the stator current must rely on monitoring those
spectral components which are not affected by the load torque oscillations.
John S. Hsu [18] suggested a method to monitor defects such as air gap eccentricity,
cracked rotor bars and the shorted stator coils in induction motors. Air-gap torque can be
calculated while the motor is running. No special down time for measurement is required.
Data of the air-gap torque for a motor kept periodically for comparison purposes. Since more
data than just a line current are taken, this method offers other potential possibilities that
cannot be handled by examining only a Line current. Experiments conducted on a 5-hp motor
showed the validity and potential of this approach.
Hamid A. Toliyat et. al. [34] developed a new induction machine model for studying
static rotor eccentricity. It is based directly on the geometry of the induction machine and the
physical layout of all windings. The model can simulate the performance of induction
machines during transients as well as at steady state, including the effects of static rotor
eccentricity. Since the dynamic model of the motor includes the mechanical equation, any
arbitrary time function of load torque can be specified from which the resulting stator current
is calculated. To illustrate the utility of this method, a conventional three phase induction
motor with 50% rotor eccentricity was simulated. Digital computer simulations have been
shown to yield satisfactory results which are in close agreement with experimental results of
previous studies.
Stanislaw F. Legowski et. al. [35] has been demonstrated that the instantaneous
electric power, proposed as a medium for signature analysis of induction motors, has definite
advantages over the traditionally used current. The characteristic spectral component of the
power appears directly at the frequency of disturbance, independently of the synchronous
speed of the motor. This is important in automated diagnostic systems, in which the
irrelevant frequency components, i.e. those at multiples of the supply frequency, are screened
out.
Randy R. Schoen and Thomas G. Habetler [36] presented a method for removing the
load torque effects from the current spectrum of an induction machine. They found that
previously presented schemes for current-based condition monitoring ignore the load effect
19
or assume that it is known. Therefore, a scheme for determining machine health in the
presence of a varying load torque requires some method for separating the two effects. This
is accomplished by comparing the actual stator current to a model reference value which
includes the load effects. The difference between these two signals provides a filtered
quantity, independent of variations of the load that allows continuous on-line condition
monitoring conducted without concern for the load condition. Simulation results showed the
effectiveness of this model reference estimation scheme at removing the load torque effects
from the monitored spectra. Experimental results illustrated the feasibility of the proposed
system. They demonstrated that the characteristic spectral components are present in the
difference current and that the load effects can effectively be removed from the monitored
spectrum to improve their detectability.
M.E.H. Benbouzid and H. Nejjari et. al. [37] stated that preventive maintenance of
electric drive systems with induction motors involves monitoring of their operation for
detection of abnormal electrical and mechanical conditions that indicate, or may lead to, a
failure of the system. Intensive research effort has been for sometime focused on the motor
current signature analysis. This technique utilizes the results of spectral analysis of the stator
current. Reliable interpretation of the spectra is difficult, since distortions of the current
waveform caused by the abnormalities in the drive system are usually minute. Their
investigations show that the frequency signature of some asymmetrical motor faults can be
well identified using the Fast Fourier Transform (FFT), leading to a better interpretation of
the motor current spectra. Laboratory experiments indicate that the FFT based motor current
signature analysis is a reliable tool for induction motor asymmetrical faults detection.
W. T. Thomson et. al. [38] presented an appraisal of on-line monitoring techniques to
detect airgap eccentricity in three-phase induction motors. On-line current monitoring is
proposed as the most applicable method in the industrial environment. The analyses of the
current spectra for different motors are presented in the study. The results verify that the
interpretation of the current spectrum proposed in this study was successful in diagnosing
airgap eccentricity problems.
Birsen Yazc and Gerald B. Kliman [39] discussed an adaptive timefrequency
method to detect broken bar and bearing defects. It was shown that the timefrequency
spectrum reveals the properties relevant to fault detection better than the Fourier spectrum in
20
the transform domain. The method is based on a training approach in which all the distinct
normal operating modes of the motor are learned before the actual testing starts. This study
suggests that segmenting the data into homogenous normal operating modes is necessary,
because different operating modes exhibit different statistical properties due to non stationary
nature of the motor current. Overlooking this fact will deteriorate the performance of the
detection. The result of this study showed that signals from faulty motors are several hundred
standard deviations away from the normal operating modes, which indicates the power of the
proposed statistical approach. Finally, it was suggested that the proposed method is a
mathematically general and powerful one which can be utilized to detect any fault that could
show up in the motor current.
Jafar Milimonfared et. al. [40] presented a new method for detecting broken-rotor-bar
faults by analyzing the stator-induced voltage after removing the mains. The method is
attractive because source non-idealities like unbalance time harmonics will not influence the
detection. Also it is clear from the nature of the test that it can be performed even with an
unloaded machine. Harmonic components predicted by theoretical analysis are clearly
matched by simulation results. However, due to inherent asymmetries of the machine, some
of these components may already exist, even in a healthy machine. It is also apparent from
the simulations and experiments that, although the number of broken bars does not have
much effect on the magnitude of the harmonic components, one can distinguish between a
faulty and a healthy machine. Interbar currents, dependence of the spectral amplitude on the
instance of disconnection, and short length of data also adversely affect on the detection
technique.
Benbouzid et. al. [7, 37] investigated the efficacy of current spectral analysis on
induction motor fault detection. The frequency signatures of some asymmetrical motor faults,
including air gap eccentricity, broken bars, shaft speed oscillation, rotor asymmetry, and
bearing failure, were identified. This work verified the feasibility of current spectral analysis.
Current spectral analysis was applied to other types of electrical machines too. For example,
Thomson [38,41] verified that the use of the current spectrum was successful in diagnosing
air gap eccentricity problems in large, high-voltage, three-phase induction motors. Le Roux
[42] monitored the current harmonic component at the rotating frequency (0.5 harmonic) to
detect the rotor faults of a permanent magnet synchronous machine.
21
Alberto Bellini et. al. [43] presented the impact of control on faulted induction
machine behavior. The diagnostic indexes usually used for open-loop operation are no longer
effective. Simulation and experimental results show that the spectrum of the field current
component in a field-oriented controlled machine has suitable features that can lead to an
effective diagnostic procedure. Specifically, in the case of stator and rotor faults, the
spectrum components at frequencies 2f and 2sf respectively, are quite independent of control
parameters and dependent on the fault extent.
Benbouzid [5] made a review of MCSA as a medium for fault detection. This study
introduces in a concise manner the motor signature analysis for the detection and localization
of abnormal electrical and mechanical conditions that indicate, or may lead to a failure of
induction motors. The MCSA utilizes the results of spectral analysis of the stator current for
the detection of airgap eccentricity, broken rotor bars and bearing damage. It is based on the
behavior of the current at the side band associated with the fault. For that, intimate
knowledge of the machine construction is required. It is explained that when the load torque
varies with rotor position, the current will contain spectral components, which coincide with
those caused by the fault condition. The torque oscillation results in stator current harmonics
that can obscure, and often overwhelm, those produced by the fault condition. Researcher
concluded that Fourier analysis is very useful for many applications where the signals are
stationary. However, it is not appropriate for analyzing a signal that has a transitory
characteristic such as drifts, abrupt changes and frequency trends. To overcome this problem,
Fourier analysis has been adapted to analyze small sections of the signal in time; this
technique is known as the short time fast Fourier transform (STFFT). STFT represents a sort
of compromise between time- and frequency-based views of a signal and provides
information about both.
Joksimovic & Penman [44] studied the interaction between faulty stator winding and
a healthy rotor cage. The faulty asymmetric stator winding may produce spatial harmonics
into the air-gap field. However, all these harmonics vary at a single frequency, i.e. the supply
frequency of the sinusoidal voltage source. The stator harmonics induce currents in the rotor
cage and reflect back from the rotor as new air-gap field harmonics. The air-gap harmonics
caused by the induced rotor currents vary at specific frequencies. The air-gap field harmonics
induce electromotive forces in the stator winding and generate harmonic stator currents at
22
these same frequencies. These are the same frequencies at which a healthy machine produces
harmonic stator currents. According to this analysis, a stator fault may generate only
harmonic stator currents, which vary at the fundamental and rotor-slot harmonic frequencies.
A fault in a stator winding may change the amplitudes of the stator-current harmonics, but it
will not produce any new frequencies in the stator-current spectrum. This significant result
implies that it may be difficult to detect a stator fault from a current spectrum using current
signature analysis.
Masoud Haji, and Hamid A. Toliyat [45] developed a pattern recognition technique
based on Bayes minimum error classifier to detect broken rotor bar faults in induction motors
at the steady state. The proposed algorithm uses only stator currents as input without the need
for any other variables. First rotor speed is estimated from the stator currents, then
appropriate features are extracted. Once normalized mean and variance plus mean and
covariance of each class are determined for an ac induction motor, the technique can be used
in online condition monitoring of the motor. Theoretical approach plus experimental results
from a 3 hp induction motor show the strength of the proposed method. Without loss of
generality, the algorithm can be revised to include other faults such as eccentricity and phase
unbalance.
Arkan et al. [46] presented a non-invasive online method for the detection of stator
winding faults in three-phase induction motors from the observation of the negative sequence
supply current. A power decomposition technique (PDT) was used to derive positive and
negative sequence components of measured voltages and currents. This study carried out
experimental studies, which showed that the negative sequence impedance could vary
between 10 % and 50 % during an inter-turn short circuit.
Tallam, Habetler, and Harley [47] monitored the negative-sequence voltage to detect
a turn-to-turn short circuit in a closed-loop drive-connected induction motor. A neural
network was used to learn and to estimate the negative-sequence voltage of a healthy motor,
which is used as the threshold. This helped to reduce the effects of machine non-ideality and
unbalanced supply voltage. According to [47], most of the turn-to-turn short circuit-related
fault signatures exist in the stator voltage because of the regulation of the drive controllers.
However, the influence of mechanical load was neglected. In practice, the distribution of
fault information between the stator voltage and current depends on drive controllers, as well
23
as mechanical load and operating conditions. Monitoring either stator current or voltage
alone cannot ensure an accurate prediction of motor conditions.
Miletic and Cettolo [48] acknowledged that Motor Current Signature Analysis
(MCSA) is one of the widely used diagnostic methods. This method is based on
measurement of sidebands in the stator current spectrum. These sidebands are usually located
close to the main supply frequency. Frequency converter causes supply frequency to slightly
vary in time and, as a result, some additional harmonics in the current spectrum are induced
and sidebands are reduced. These harmonics can be easily misinterpreted as the sidebands
caused by the rotor faults. In this study, the experimental results of fault diagnosis carried out
using standard supply and using frequency converter were compared and presented. All tests
were performed on 22 kW induction motor.
In current spectral analysis, the actual harmonics measured from a running machine
are always compared with known values (thresholds) obtained from a healthy motor. In
practical applications, the thresholds change with motor operating conditions. Therefore,
Obaid [49] proposed tracking the normal values of a healthy motor at different load
conditions. For each load condition, a corresponding threshold was determined and compared
with the on-line measurement to determine the motor condition. Besides the FFT technique
in spectral analysis, other techniques in advanced digital signal processing and pattern
recognition were applied to motor condition monitoring as well.
Mohamed El Hachemi Benbouzid, and Gerald B. Kliman [50] briefly presented
signal (mainly motor current) processing techniques for induction motor rotor fault detection
(mainly broken bars and bearing deterioration). The main advantages and drawbacks of the
presented techniques are also briefly discussed. In many cases, the conventional steady state
techniques may suffice. From the discussions, it appears that, for the most difficult cases,
time-frequency and time-scale transformations, such as wavelets, provide a more optimal
tool for the detection and the diagnosis of faulty induction motor rotors. On the one hand,
they remedy the main drawbacks of motor current signal processing techniques for fault
detection (i.e., nonstationarity). These techniques exhibit some interesting application
advantages, such as for coal crushers, where speed varies rapidly and for deteriorated
bearings where speed and signatures may vary in an unpredictable manner.
24
Szab Lornd et.al. [51] presented some results on detecting broken rotor bars in
induction motors. Five different motor conditions were studied (the healthy machine and
having up to 4 broken bars), each at 9 different loads. The results of this study show that if
there is any broken bar in the rotor it will directly affect the induced voltages in the stator
windings and the waveform of the stator currents. Therefore the spectrum analysis of the line
current (motor current signature analysis) is one of the best non-intrusive methods.
Szab Lornd et. al. [52] utilized the result of spectral analysis of stator current to
diagnose rotor faults. The diagnosis procedure was performed by using virtual
instrumentation (VIs). Several virtual instruments (VIs) were built up in Labview. These VIs
were used both for controlling the test measurements and data acquisition and for the data
processing. The tests were carried out for seven different loads with healthy motor, and with
similar motors having up to 5 broken rotor bars. The rotor bars were provoked interrupting
the rotor bars by drilling into the rotor. The measured current signals were processed using
the Fast Fourier Transformation (FFT). The power density of the measured phase current was
plotted. The results obtained for the healthy motor and those having rotor faults were
compared, especially looking for the sidebands components having the special frequencies.
The significance presence of some well defined sidebands frequencies in the harmonic
spectrum of the measured line current clearly indicated the rotor faults of the induction
machine.
Jason R. Stack et. al. [53] introduced the notion of categorizing bearing faults as
either single-point defects or generalized roughness. This is important because it divides
these faults according to the type of fault signatures they produce rather than the physical
location of the fault. The benefit of this categorization is twofold. First, it ensures that the
faults categorized as generalized roughness are not overlooked. The majority of bearing
condition monitoring schemes in the literature focus on detection of single-point defects.
While this is an important class of faults, a comprehensive and robust scheme must be able to
detect both generalized roughness and single-point defect bearing faults. Second, grouping
faults according to the type of fault signature they produce provides a clearer understanding
of how these faults should be detected. This provides improved insight into how bearing
condition monitoring schemes should be designed and applied. Experimental results obtained
25
from this research suggest generalized roughness faults produce unpredictable (and often
broadband) changes in the machine vibration and stator current.
Jason R. Stack et. al. [54] proposed a method for detecting developing bearing faults
via stator current. Current-based condition monitoring offers significant economic savings
and implementation advantages over vibration-based techniques. This method begins by
filtering the stator current to remove most of the significant frequency content unrelated to
bearing faults. Afterwards, the filtered stator current is used to train an autoregressive signal
model. This model is first trained while the bearings are healthy, and a baseline spectrum is
computed. As bearing health degrades, the modeled spectrum deviates from its baseline value;
the mean spectral deviation is then used as the fault index. This fault index is able to track
changes in machine vibration due to developing bearing faults. Due to the initial filtering
process, this method is robust to many influences including variations in supply voltage,
cyclical load torque variations, and other (nonbearing) fault sources. Experimental results
from ten different bearings are used to verify the proficiency of this method.
Srgio M. A. Cruz and A. J. Marques Cardoso [55] proposed two different methods
for the diagnosis of stator faults in DTC induction motor drives. Through a qualitative
analysis of the phenomena involving the behavior of this type of drive after the occurrence of
a stator fault in the motor, it was demonstrated that the flux and torque hysteresis controllers
tend to introduce a significant third harmonic component in the motor supply currents. The
presence of a strong third harmonic component in the motor supply currents is thus an
indicator about the presence of a stator fault. The results obtained with this diagnostic
technique demonstrated its effectiveness for the detection and quantification of the extension
of the fault in DTC induction motor drives.
Humberto Henao et. al. [56] presented the experimental and the analytical validation
of the equivalent internal circuit approach applied to the three-phase squirrel-cage induction
machine. The proposed model is the only one which allows simulation of the induction
machine state variables under normal or faulted operation in both stator and rotor sides. In
this study, space harmonic components predicted by analytical calculation are matched with
simulation results. The little differences in the frequency computation was caused by the
resolution of the Fast Fourier Transform. The proposed model is very good to predict fault
influence in the induction machine behavior.
26
Lyubomir et. al. [57] conducted an experiment to diagnose the broken rotor bar fault.
Motor Current Signature Analysis (MCSA) was used to diagnose the fault of motor. For this,
experiment was conducted on 0.5 kw induction motor. The rotor bar was damaged by drilling
the rotor. The spectra of health and faulty motor were compared. Stator current spectrum of
faulty motor shows the side bands at particular frequencies due to presence of broken rotor
bars with great reliability. Finally, researchers concluded that Motor Current Signature
Analysis (MCSA) is a reliable technique for diagnosis of broken rotor bar faults.
Jung et. al. [58] proposed an online induction motor diagnosis system using MCSA
with advanced signal and data processing algorithms. The diagnosis system was composed of
the DSP board for high-speed signal processing and advanced signal-and-data-processing
algorithm including the PCuser interface. The advanced algorithms were made up of the
optimalslip- estimation algorithm, the proper sample selection algorithm, and the frequency
auto search algorithm for achieving MCSA efficiently. The optimal slip estimation algorithm
suggested the optimal-slip estimator based on the Bayesian method of estimation. In addition,
the proper-sample-selection algorithm determined the standard of suitable samples for the
MCSA process from the characteristics of a measurement noise and spread spectrum. Finally,
the frequency auto search algorithm detected the abnormal harmonic frequency under
unspecified harmonic numbers with the tendency of the candidate spectrum magnitudes. To
verify the generality of the suggested algorithms, laboratory experiments were performed
with 3.7-kW and 30-kW squirrel-cage induction motors. The proposed system was able to
ascertain four kinds of motor faults and diagnose the fault status of an induction motor.
Experimental results successfully verified the operations of the proposed diagnosis system
and algorithms.
Szab Lornd et.al. [59] compared different fault diagnosis methods by means of data
processing in LabVIEW. The results obtained by experiments verified that the three-phase
current vector, the instantaneous torque, and the outer magnetic filed can be used for
diagnosing the rotor faults. At last, authors stated that due to its simplicity of motor current
signature analysis (MCSA), this method is the mostly used in industrial environment.
Chidong Qiu et. al. [60] developed a multitaper-based detection method for incipient
motor faults in order to detect weak fault eigen frequency submerged in noises environment.
The tradeoff problem between frequency resolution and variance was studied, and the
27
optimal tradeoff value was chosen to be applied on detecting motor faults. By selecting high
energy tapers, the root leakage of eigen frequency was eliminated, and the shape of eigen
frequency was changed to be distinguishable. Simulation studies were conducted and results
show that multi-taper method has a more steady and antinoise performance compared with
other methods. Finally, an experiment was arranged in laboratory, and the bearing faults were
put into the motor. By using the proposed method, it is validated that multi-taper method is
effective for detecting the motor incipient faults.
Frosini, and L. Bassi [61] proposed a new approach to use stator current and
efficiency of induction motors as indicators of rolling-bearing faults. This study illustrates
the experimental results on four different types of bearing defects: crack in the outer race,
hole in the outer race, deformation of the seal, and corrosion. Another novelty introduced by
this study is the analysis of the decrease in efficiency of the motor with a double purpose: as
alarm of incipient faults and as evaluation of the extent of energy waste resulting from the
lasting of the fault condition before the breakdown of the machine.
Load variation along with static and dynamic eccentricities degrees is one of the
major factors affecting the dynamic behaviors of eccentricity signatures which is utilized for
precise mixed eccentricity fault diagnosis. Without taking the effect of load variation into
account properly and just by considering the change in the static and dynamic eccentricity
degrees, inaccurate fault detection is acquired. This is of noticeable effects that load
variations have on side-band components that are used as fault detection indices. These
indices are extracted from the current spectrum of healthy and faulty motor. Therefore, Faiz
et. al. [62] developed an approach to recognize mixed eccentricity and determine the static
and dynamic eccentricities degrees individually at different load levels. In order to evaluate
the impact of load-dependent indices on eccentricity detection and fault-severity estimation, a
systematic relation between each other and eccentricity degree is proposed in this study.
Correlation coefficient and mutual information are applied to assess abilities of the obtained
indices for eccentricity detection in terms of their relation to static and dynamic eccentricities,
their degrees and dependency on the load of motor. The classification results indicate that the
elicited indices estimate the eccentricity type and degree exactly.
28
29
condition, and it allows a correct diagnosis of a healthy machine in some particular cases
where Fourier analysis leads to an incorrect fault diagnosis.
Cusido et. al. [67] proposes a different signal processing method, by combination of
Wavelet and Power Spectral Density techniques. It presents good theoretical and
experimental results. This study concluded that MCSA is a good method for analyze motor
faults over constant load torque, but in case of not constant load torque, an improvement is
needed. Wavelets Decomposition is the right technique for non stationary signals and Power
Spectral Density would be the right solution for introduce it on Industrial applications.
30
Douglas et. al. [72] proposed a new technique Extended parks Vector Approach
(EPVA), which was successfully applied in the steady diagnosis of rotor faults, inter-turn
stator faults and unbalanced supply voltage, and mechanical load misalignment. This
technique was based on the parks vector approach; however, it provides greater insight into
the severity of the faults.
Izzety Onel et. al. [73] investigated the application of induction motor stator current
signature analysis (MCSA) using Parks transform for the detection of rolling element
bearing damages in three-phase induction motor. This study presents bearing faults and
Parks transform and then gives a brief overview of the radial basis function (RBF) neural
networks algorithm. Data acquisition and Parks transform algorithm were achieved by using
LabVIEW. The neural network algorithm is achieved by using MATLAB programming
language. The diagnosis process was tested on a 0.75kW, squirrel-caged induction motor.
Experimental results showed that it is possible to detect bearing damage in induction motors
using an ANN algorithm. ANN was trained, giving 100% correct prediction for training data.
When ANN was presented a set of Parks vector pattern, the diagnosis system was found to
provide very good performance.
The research carried out by Szab Lornd et. al. [74] shows that how the Park's vector
approach based method can be used for detecting the rotor faults of the squirrel cage
induction machine. The squirrel cage induction machine was tested with two rotors, a healthy
one, and one having broken rotor bars. The line currents of the motor were visualized on an
oscilloscope using a special electronic circuit which was able to synthesize the two
orthogonal components of the current, voltage and flux phasors. Beside this the line currents
were acquired by a DAQ board from a PC using advanced virtual instruments (VIs) built up
in LabVIEW environment. Several characteristics of the motor under study were plotted. Due
to the broken rotor bars, there was significant fluctuation in the torque of the machine, and
the amplitude of the line current at the end of the starting period was quite high. The shape of
the current's phasor of faulty motor was not of perfect circular shape, which was the clear
indication a fault in the squirrel cage induction machine.
Izzet onel and Benbouzid [75] diagnosed the problem of bearing failure in induction
motors by using park vector approach. They also compared two fault detection and diagnosis
techniques, namely the Park transform approach and the Concordia transform. Experimental
31
tests were carried out on a 0.75 kW two-pole induction motor with artificial bearing damage.
The results indicate that the Park transform approach has better diagnosis capabilities than
the Concordia transform.
32
magnitude of the corresponding stator current harmonic components also increases. This is
because the mechanical vibration modulates the air gap at that particular frequency. These
frequency components then show up in the stator inductance, and finally in the stator current
[3]. For this reason, the MCSA can be used to detect rotor and bearing faults. As the flux
density in the air gap is defined as the product of the winding magneto-motive force (MMF)
and the air-gap permeance, variations in either of these will cause anomalies in the flux
distribution. The changes in the winding MMF mainly depend on the winding distribution.
On the other hand, the air-gap permeance depends on numerous effects including stator slots,
out-of-round rotors, air-gap eccentricities caused by mechanical unbalance and misalignment,
and mechanical shaft vibrations caused by bearing or load faults [4]. MCSA detects changes
in a machines permeance by examining the current signals. It uses the current spectrum of
the machine for locating characteristic fault frequencies. The spectrum may be obtained
using a Fast Fourier Transformation (FFT) that is performed on the signal under analysis.
The fault frequencies that occur in the motor current spectra are unique for different motor
faults. This method is the most commonly used method in the detection of common faults of
induction motors. Some of the benefits of MCSA include [3, 4, 5, 7, 50, 58]:
a) Non-intrusive detection technique:
With the technological advances in current-measuring devices, inexpensive and easyto-use clamp-on probes are more affordable and convenient to use for sampling current
without having to disconnect the electrical circuit or to disassemble the equipment.
b) Remote sensing capability:
Current sensors can be placed anywhere on the electrical supply line without
jeopardizing the signal strength and performance.
c) Safe to operate:
Since there is no physical contact between the current sensor and the motor-driven
equipment, this type of monitoring technique is particularly attractive to applications where
safety is of major concern.
Wavelet Transform can be used for fault diagnosis of induction motor. It works on
principle that all signals can be reconstructed from sets of local signals of varying scale and
amplitude, but constant shape. It is an easy and fast to implement data processing technique.
33
It analyses the signal at different frequency bands with different resolution by decomposing
the signal into coarse approximation and detail information.
Current Parks vector is most frequent used method in literature review applied to
diagnose the common faults of induction motor. The analysis of the three-phase induction
motor can be simplified using the Park transformation. The method is based on the
visualization of the motor current Parks vector representation. If this is a perfect circle the
machine can be considered as healthy. If an elliptical pattern is observed for this
representation, the machine is faulty. From the characteristics of the ellipse the fault's type
can be established. The ellipticity increases with the severity of the fault [67-70]. From the
literature cited, the following observations can be made:
(i) Condition monitoring has great significance in the business environment because there is
need to improve reliability of machine and to reduce the cost of maintenance.
(ii) The major disadvantage of vibration monitoring is cost. A regular vibration sensor costs
several hundred dollars. A high product cost can be incurred just by employing the
necessary vibration sensors for a large number of electric machines. Another
disadvantage of vibration monitoring is that it requires access to the machine. For
accurate measurements, sensors should be mounted tightly on the electric machines, and
expertise is required in the mounting. On other hand, there is no physical contact between
the current sensor and motor-driven equipment in electric monitoring therefore electric
monitoring is particularly attractive to applications where safety is of major concern.
(iii)In current based fault detection, various types of faults may cause broadband changes in
power spectra of stator current. Therefore, researchers choose the signal processing as the
tool for stator current based fault detection.
(iv) Investigations reveal that the fault frequencies occur in motor current spectra are unique
for different motor faults.
(v) It has been a broadly accepted requirement that a diagnostic scheme should be noninvasive and capable of detecting faults accurately at low cost. Therefore, Motor Current
Signature Analysis {MCSA) has become a widely used method because its monitoring
parameter is a motor terminal quantity that is easily accessible.
34
(vi) Numerous applications of using electric monitoring in motor health monitoring have been
published among the nuclear-generation, industrial, defense industries. In published work,
researchers used the variety of motors of different rating to diagnose the faults. But very
little work has been done to diagnose the all possible common fault of induction motor by
using the motor of same rating and same signal processing technique. So, there is need to
use the same type of motor and same signal processing technique to diagnose common
faults of induction motor so that effectiveness of signal processing techniques can be
studied.
(vii)
It is observed that very few experimental studies have been published which may
diagnose the single fault of induction motors with variety of signal processing techniques.
Therefore, an experimental study must be conducted to diagnose the single fault with
different signal processing techniques so that limitation of each signal processing
technique can be studied.
(viii) The effectiveness of signal processing techniques for non-stationary signals has not
been addressed appropriately in the literature. Therefore, more experiments need to be
carried out with different signal processing techniques so that it may be examined which
technique is best suited for non-stationary signals.
35
CHAPTER 3
3.1 Introduction
The detection of common faults of induction motor with help of signal processing
techniques is main focus of this research. A variety of faults can occur within three phase
induction motor during the course of normal operation. These faults can lead to a potentially
catastrophic failure if undetected. Consequently, a variety of condition monitoring techniques
have been developed for the analysis of abnormal condition. Signal processing techniques are
also very effective for fault detection. Due to continuous advancement of signal processing
techniques and related instruments, online monitoring with signal processing techniques has
become very efficient and reliable for electrical machines. The objective of this chapter is to
36
present the classification of three phase induction motor faults and various advanced signal
processing techniques for fault diagnosis of electric machines.
Electrical faults
Mechanical faults
Electrical faults include faults caused by winding insulation problems, and some of the rotor
faults. Mechanical faults include bearing faults, air gap eccentricity, load faults and
misalignment of shaft.
During the brazing process in manufacture, non uniform metallurgical stresses may
be built into cage assembly and these can also lead to failure during operation.
A rotor bar may be unable to move longitudinally in the slot it occupies, when
thermal stresses are imposed upon it during starting of machine.
Heavy end ring can result in large centrifugal forces, which can cause dangerous
stresses on the bars.
37
Because of the above reasons, rotor bar may be damaged and simultaneously unbalance rotor
situation may occur. Rotor cage asymmetry results in the asymmetrical distribution of the
rotor currents. Due to this, damage of the one rotor bar can cause the damage of surrounding
bar and thus damage can spread, leading to multiple bar fractures. In case of a crack, which
occurs in a bar, the cracked bar will overheat, and this can cause the bar to break. Thus, the
surrounding bar will carry higher currents and therefore they are subjected to even larger
thermal and mechanical stresses which may also start to crack [2]. Most of the current which
would have flowed in the broken bar now will flow in the two bars adjacent to it. Thus, the
large thermal stresses may also damage the rotor laminations. The temperature distribution
across the rotor lamination is also changed due to the rotor asymmetry. The cracking of the
bar can be presented at various locations, including the slot portion of the bars under
consideration and end rings of bar joints. The possibility of cracking in the region of the end
rings of bar joints is the greatest when the start up time of the machine is long and when
frequent starts are required [81].
38
of insulation is reduced by 50 % for every 100 C increase above the stator winding
temperature limit [82]. It is thus necessary to monitor the stator winding temperature so that
an electrical machine will not operate beyond its thermal capacity. For this purpose, many
techniques have been reported [83]-[86]. However, the inherent limitation of these
techniques is their inability to detect a localized hot spot at its initial stage.
A few mechanical problems that accelerate insulation degradation include movement
of a coil, vibration resulting from rotor unbalance, loose or worn bearings, airgap
eccentricity, and broken rotor bars [81]. The current in the stator winding produces a force on
the coils that is proportional to the square of the current. This force is at its maximum under
transient overloads, causing the coils to vibrate at twice the synchronous frequency with
movement in both the radial and the tangential direction. This movement weakens the
integrity of the insulation system [81]. Mechanical faults, such as broken rotor bar, worn
bearings, and air-gap eccentricity, may be a reason why the rotor strikes the stator windings.
Therefore, such mechanical failures should be detected before they fail the stator winding
insulation [87, 88]. Contaminations due to foreign materials can lead to adverse effects on the
stator winding insulation. The presence of foreign material can lead to a reduction in heat
dissipation [89]. It is thus very important to keep the motors clean and dry, especially when
the motors operate in a hostile environment.
39
Regardless of the causes, stator winding-related failures can be divided into the five
groups: turn-to-turn, coil-to-coil, line-to-line, line-to-ground, and open-circuit faults as
presented in Figure 3.1. Among the five failure modes, turn-to-turn faults (stator turn fault)
have been considered the most challenging one since the other types of failures are usually
the consequences of turn faults. Furthermore, turn faults are very difficult to detect at their
initial stages. To solve the difficulty in detecting turn faults, many methods have been
developed [90]-[96].
40
41
are contaminated by dirt and other foreign matter that are commonly often present in the
environment of industries. Bearing corrosion is produced by the presence of water, acids,
deteriorated lubrication and even perspiration from careless handling during installations [99,
100]. Once the chemical reaction has advanced sufficiently, particles are worn-off resulting
in the same abrasive action produced by bearing contamination. Under and over-lubrication
are also some other causes of bearing failure. In either case, the rolling elements are not
allowed to rotate on the designed oil film causing increased levels of heating. The excessive
heating causes the grease to break down, which reduces its ability to lubricate the bearing
elements and accelerates the failure process. In addition, Installation problems are often
caused by improperly forcing the bearing onto the shaft or in the housing. This produces
physical damage in form of brinelling or false brinelling of the raceways which leads to
premature failure. Brinelling is the formation of indentations in the raceways as a result of
deformation caused by static overloading. While this form of damage is rare, a form of false
brinelling occurs more often. In this case, the bearing is exposed to vibrations while even
though lightly loaded bearings are less susceptible, false brinelling still happens and has even
occurred during the transportation of uninstalled bearings [99]. Misalignment of the bearing
is also a common result of defective bearing installation.
42
43
computationally efficient way to calculate the DFT. By making use of periodicities in the
sines that are multipled to do the transforms, the FFT greatly reduce the amount of
calculation required. Functionally, the FFT decomposed the set of date to be transformed into
a series of smaller data sets to be transformed. Then, it composes those smaller sets into even
smaller sets. At each stage of processing, the results of the previous stage are combined in
special way. Finally, it calculates the DFT of each small data set. FFT algorithm can be used
to detect the various types of motor fault.
The Power spectrum is computed from the basic FFT function. The power spectrum
shows power as the mean squared amplitude at each frequency line. The FFT in LabVIEW
and LabWindows returns a two-sided spectrum in complex form (real and imaginary parts),
which must scale and convert to polar form to obtain magnitude and phase. The frequency
axis is identical to that of the two-sided power spectrum. The amplitude of the FFT is related
to the number of points in the time-domain signal. The following equation can be used to
compute the amplitude and phase versus frequency from the FFT [105].
Amplitude spectrum in quantity peak
=
Magnitude [FFT(A)]
=
N
real[FFT(A)]2 + imag[FFT(A)]2
N
.(3.1)
imag[FFT(A)]
..(3.2)
real[FFT(A)]
where the arctangent function here returns values of phase between - and + , a full
range of 2 radians.
Using the rectangular to polar conversion function to convert the complex array
FFT(A)
to
N
To view the amplitude spectrum in volts (or another quantity) rms, divide the non-DC
components by the square root of 2 after converting the spectrum to the single-sided form.
Because the non-DC components were multiplied by two to convert from two-sided to
single-sided form, The rms amplitude spectrum can be calculated directly from the two-sided
amplitude spectrum by multiplying the non-DC components by the square root of two and
44
discarding the second half of the array. The following equations show the entire computation
from a two-sided FFT to a single sided amplitude spectrum.
Amplitude spectrum in rms = 2.
=
180
.Phase.FFT ( A)
.(3.3)
The amplitude spectrum is closely related to the power spectrum. Single-sided power
spectrum can be computed by squaring the single-sided rms amplitude spectrum. Conversely,
the amplitude spectrum can be computed by taking the square root of the power spectrum. In
LabVIEW and LabWindows, the two-sided power spectrum is actually computed from the
FFT as follows [105].
The Power spectrum SAA ( f ) =
FFT ( A).FFT *( A)
N
..(3.4)
Where FFT*(A) denotes the complex conjugate of FFT (A). To form the complex conjugate,
the imaginary part of FFT(A) is negated.
45
Here, the speed of the power spectrum and the FFT computation depend on the number of
points acquired. If N is a power of 2, LabVIEW uses the efficient FFT algorithm. Otherwise,
LabVIEW actually uses the discrete Fourier transform (DFT), which takes considerably
longer. LabWindows requires that N be a factor of two and thus always uses the FFT. Typical
bench-top instruments use FFTs of 1,024 and 2,048 points. The Power spectrum of healthy
motor is shown in Figure 3.2.
st ( ) = s( )h( t )......(3.5)
The modified signal is a function of two times, the fixed time t, and the running time, . The
window function is chosen to leave the signal more or less unaltered around the time t but to
suppress the signal for times distant from the time of interest. That is,
st ( )
Since the modified signal emphasizes the signal around the time t, the Fourier transform will
reflect the distribution of frequency around that time,
st ( ) =
1
2
e j .st ( ) d .......(3.7)
1
e j .st ( ) h ( t ) d .......(3.8)
2
PSP ( t , ) = st ( ) =
1
e j .st ( ) h ( t ) d (3.9)
2
Thus, the magnitude of squared of the STFT yields the spectrogram of function, which is
usually represented like color plots.
46
To analyze the signal around time t, window function has chosen that is peaked around t.
Hence the modified signal is short and its Fourier transform (equ. 3.8) is called short-time
Fourier transform [6, 102]. STFT spectrogram can be used for fault detection of motor. The
STFT of a healthy motor is shown in Figure 3.3.
s (t ) =
m 1 n 1
cm ,n hm,n (t )
m=0 n =0
47
.(3.10)
where hm ,n (t ) is the time frequency elementary function, cm, n is the weight of hm ,n (t ) and
cm,n is the Gabor coefficients. The Gabor Transform computes the coefficients cm,n for the
signal s(t).
The following equation defines the time shifted and frequency modulated version, hm ,n (t ) ,
cm,n =
(3.12)
where y(t) is the analysis window, y(t) and h(t) are a pair of dual functions.
Gabor spectrogram can be used for fault diagnosis of induction motors. Gabor spectrograph
for a healthy motor is shown in Figure 3.4
48
W (t , ) =
1
1
1
s * t s t + e j d .(3.13)
2
2
2
1
1
1
S * s + e jt d
2
2
2
(3.14)
The equivalence of the two expressions is easily checked by writing the signal in terms of
spectrum. WVD can be used for fault detection of induction motor. The Wigner-Ville
Distribution of a healthy induction motor is shown in Figure 3.5.
49
50
Two-channel perfect reconstruction (PR) filter banks are a common and efficient way to
implement the DWT [105, 110]. Figure 3.6 shows a typical two-channel PR filter bank
system. The signal X[z] first is filtered by a filter bank consisting of G0(z) and G1(z). The
outputs of G0(z) and G1(z) then are down sampled by a factor of 2. After some processing,
the modified signals are upsampled by a factor of 2 and filtered by another filter bank
consisting of H0(z) and H1(z).
If no processing takes place between the two filter banks, the sum of outputs of H0 (z)
and H1(z) is identical to the original signal X(z), except for the time delay. This system is a
two-channel PR filter bank, where G0 (z) and G1(z) form an analysis filter bank, and H0(z)
and H1(z) form a synthesis filter bank. Traditionally, G0(z) and H0(z) are low pass filters, and
G1(z) and H1(z) are highpass filters. The subscripts 0 and 1 represent low pass and high pass
filters, respectively. The operation 2 denotes a decimation of the signal by a factor of two.
Applying decimation factors to the signal ensures that the number of output samples of the
two low pass filters equal the number of original input samples X(z). Therefore, no redundant
information is added during the decomposition. Two-channel PR filter bank system can be
used and consecutively decompose the outputs of low pass filters, as shown in Figure 3.6.
Low pass filters remove high-frequency fluctuations from the signal and preserve slow
trends. The outputs of low pass filters provide an approximation of the signal. High pass
filters remove the slow trends from the signal and preserve high-frequency fluctuations. The
outputs of high pass filters provide detail information about the signal. The outputs of low
pass filters and high pass filters define the approximation coefficients and detail coefficients,
respectively. Symbols A and D in Figure 3.7 represent the approximation and detail
information, respectively.
Detail coefficients can be called wavelet coefficients because detail coefficients
approximate the inner products of the signal and wavelets. This manual alternately uses the
terms wavelet coefficients and detail coefficients, depending on the context. The Wavelet
Analysis Tools use the subscripts 0 and 1 to describe the decomposition path, where 0
indicates low pass filtering and 1 indicates high pass filtering. For example, D2 in Figure 3.7
denotes the output of two cascaded filtering operationslow pass filtering followed by high
pass filtering. Therefore, this decomposition path can be described with the sequence 01.
Similarly, DL denotes the output of the filtering operations 000...1 in which the total number
51
of 0 is L1. The impulse response of 000...1 converges asymptotically to the mother wavelet
and the impulse response of 000...0 converges to the scaling function in the wavelet
transform [103, 105, 112].
G1 ( Z )
H1 ( Z )
2
Processing
Signal
G0 ( Z )
Re constructed
signal
H 0 (Z )
G1 ( z )
D1
G1 ( z )
Signal
D2
A1
G0 ( z )
G0 ( z )
G1(z)
G0 ( z )
A2 AL1
52
DL
AL
Gs ( w, t ) = s(u ) g (t u )eiwu du
. (3.15)
Shifting and scaling a different window function, called in this case mother wavelet, it is
obtained the so called Wavelet Transform.
Gs ( w, t ) = s
1
t u
du
a
a
(3.16)
where a is the scale factor, u is the shift, (t ) is the mother wavelet and Gs ( w, t ) is the
wavelet transform of function s(t).
The discrete version of Wavelet Transform, DWT, consists in sampling not the signal
or not the transform but sampling the scaling and shifted parameters. This result in high
53
frequency resolution at low frequencies and high time resolution at high frequencies,
removing the redundant information.
A discrete signal s[n] could be decomposed:
s [ n] =
a jo,k jo,k [ n ] +
k
j 1
j = jo k
d j ,k j ,k [ n ]
(3.17)
where
[ n ] = scaling function
jo,k [ n ] = 2
j0
( n ) : mother wavelet
j
Approx.
Level 3
Detail
level 3
Detail
level 2
Detail
level 2
f
fs
16
fs
8
fs
4
Figure 3.8: Frequency range cover for details and final approximation
54
fs
2
ia , ib , ic
id =
iq =
2
1
1
ia
ib
ic
3
6
6
1
1
ib
ic
2
2
..(3.18)
.(3.19)
Under ideal conditions, three phase currents lead to a Parks vector with the following
components:
id =
6
I sin t
2
.(3.20)
iq =
I sin t
2
2
..(3.21)
where
I
=supply frequency
=time variable
Its representation is a circular pattern centered at the origin of the coordinators as illustrated
by Figure 3.9. This is very simple reference figure that allows the detection of abnormal
conditions by monitoring the deviations of acquired patterns.
55
56
Wavelet signal processing can represent signals sparsely, capture the transient features of
signals, and enable signal analysis at multiple resolutions.
Current Parks vector is another method which is presented in this chapter. The
analysis of the three-phase induction motor can be simplified using the Park transformation.
This method is based on the visualization of the motor current Parks vector representation.
The techniques discussed in this chapter may be used to diagnose the common faults of
induction motor. The experimental results obtained with help of these techniques are
presented in subsequent chapters.
57
CHAPTER 4
Experimental Study Of
Rotor Faults Of
Induction Motor
4.1 Introduction
The need for detection of rotor faults at an earlier stage, so that maintenance can be
planned ahead, has pushed the development of monitoring methods with increasing
sensitivity and noise immunity. Broken rotor bars can be a serious problem with certain
induction motors due to arduous duty cycles. The objective of this chapter is to
experimentally demonstrate the effects of induction motor rotor fault on the motor terminal
quantities (Current, Voltage) using three different signal processing techniques. The effect of
unbalance rotor on current spectrum is also studied experimentally.
58
I(dB)
f [ Hz ]
f1 (1 2s )
f1
f1 (1 + 2s )
120 f1
p
(4.1)
where f1 is the supply frequency and p the poles.
We know that
Slip ( s ) =
n1 n
n1
.(4.2)
..(4.3)
n2
n1
Thus, n2 = s.n1
n1 n = s.n1
n = n1 s.n1
60
n = n1 (1 s )
..(4.4)
The backward rotating magnetic field speed produced by the rotor due to broken bars
and with respect to the rotor is:
nb = n n2
nb = n1 (1 s ) s.n1
nb = n1 n1.s s.n1 = n1 2n1.s
nb = n1.(1 2 s )
.(4.5)
(4.6)
Therefore, twice slip frequency sidebands occur at 2s f1 both side of the supply frequency
[52]:
fb =(1 2s)f1
(4.7)
The lower sideband and upper side bands are specifically due to broken bar and consequent
speed oscillation. In fact, researchers show that broken bars actually give rise to a sequence
of such sidebands given by [51, 52, 59]:
fb =(1 2ks)f1,
k = 1, 2,3
(4.8)
Speed
Conditions
(rpm)
Slip
K=1
K=2
K=3
LSB
USB
LSB
USB
LSB
USB
(Hz)
(Hz)
(Hz)
(Hz)
(Hz)
(Hz)
No load
1485
0.01
49
51
48
52
47
53
Half Load
1440
0.04
46
54
42
58
38
62
Full Load
1380
0.08
42
58
34
66
26
74
61
Data
Power
0.5 hp
Frequency
50 Hz
Number of phases
Speed
1500 r.p.m
Volt
415 V
Current
1.05 Amp
0.4 mm (approximately)
36
Efficiency(FL)
86%
LabVIEW 8.2 software is used to analyze the signals. It is easy to take any
measurement with NI LabVIEW. The measurements can be automated from several devices
and data can be analyzed spontaneously with this software. Data acquisition card PCI-6251
and acquisition board ELVIS are used to acquire the current samples from the motor under
load. NI M Series high-speed multifunction data acquisition (DAQ) device can measure the
signal with superior accuracy at fast sampling rates. This device has NI-MCal calibration
62
technology for improved measurement accuracy and six DMA channels for high-speed data
throughput. It has an onboard NI-PGIA2 amplifier designed for fast settling times at high
scanning rates, ensuring 16-bit accuracy even when measuring all channels at maximum
speeds. This device has a minimum of 16 analog inputs, 24 digital I/O lines, seven
programmable input ranges, analog and digital triggering and two counter/timers. Figure 4.3
shows the PCI-6251 data acquisition card which is used in experiment. The specifications of
the DAQ card are shown in Table 4.3.
Sr. no.
Specification
Analog Inputs
16
AI Resolution (bits)
16
Analog Outputs
AO Resolution
16
2.8
AO Range (V)
Digital I/O
24
8, up to 10 MHz
The Figure 4.4 show the Data acquisition board ELVIS. The NI ELVIS integrates 12
of the most commonly used instruments including the oscilloscope, DMM, function
generator, and Bode analyzer into a compact form factor ideal for the hardware lab. based
on NI LabVIEW graphical system design software NI ELVIS offers the flexibility of virtual
instrumentation and allows for quick and easy measurement acquisition and display. In the
experiment, the speed of the motor is measured by digital tachometer. The virtual instrument
(VIs) was built up with programming in LabVIEW 8.2. The VIs was used both for
controlling the test measurements and data acquisition, and for the data processing. In order
to test the system in practical cases, several measurements were made to read the stator
current of a motor.
63
64
65
and the desired frequency resolution, several thousand frequency components are generated
by the processing section.
Step III: The last step of this technique is identification of fault frequencies. The fault
frequencies are search out in the spectrum to diagnose the different faults of induction motor.
FFT
MOTOR
A/D converter
Current transducer
Anti-Aliasing filter
66
Number of
samples
Signal
Frequency
resolution
Window
Power
spectrum
Data acquisition
Scan rate
Channel
info
X-Y/Waveform
graph
FFT
analysis
Figure 4.6: Block diagram for obtaining power spectrum using LabVIEW
programming
Data
Scan rate
25000 S/s
Number of samples
2,00,000
0.12
8000
Window
Hanning
Sensor sensitivity
67
The rotor faults were provoked interrupting the rotor bars by drilling into the rotor.
The slip was 0.01, 0.04 and 0.08 at no load, half load and full load respectively. The power
spectrum of the measured phase currents was plotted. The results obtained for the healthy
motor and those having rotor faults were compared, especially looking for the sideband
components having frequencies given by equation (4.8). The data acquisition parameters for
the experiment are given in Table 4.4.
The power spectrums obtained from the current signal for one broken bar at no load,
half load and full load are given in Figures 4.8, 4.12, 4.16. At no load the side bands
frequency is very close to fundamental frequency and the amplitudes of the sidebands is quite
smaller or negligible as shown in Figure 4.8. It can be observed that the detection of the
searched slip frequency sideband at no load is too difficult, since the current in the rotor bars
is small. It is also observed from Figure 4.12 that even at half load side band fault
frequencies are not visible because again their magnitude is low. Thus, it is slightly difficult
to detect broken rotor bar fault at half loaded conditions also.
It is observed that fault frequency side bands for one broken bar are visible only at
full load as shown in Figure 4.16. These frequencies are marked as FF (Fault frequency). The
magnitude of the fault frequencies is approximately -68dB. The complete observations from
power spectrum analysis for one broken bar is given in Table 4.5
68
Table 4.5: Power spectrum analysis of one broken bar at various loading conditions
Fault Frequencies
Load
Figure no. Condition
Slip
4.8
No Load
0.01
4.12
Half Load
0.04
46
54
Visible
42
58
Not visible
4.16
Full Load
0.08
42
58
Visible
34
66
Not Visible
K=1
LSB USB
(Hz) (Hz)
49
51
K=2
Observations LSB USB
Observations
(Hz) (Hz)
Not visible
48
52 Not Visible
The power spectrums obtained from the current signal for five broken bars at no load,
half load and full load are given in Figures 4.9, 4.13, 4.17. At no load, fault frequencies are
not clearly visible because these frequencies are very close to fundamental frequency and
their amplitudes are quite smaller or negligible as shown in Figure 4.9. It can be observed
from the figure that it may not be possible to detect the broken rotor fault at no load or light
load due to small current in the rotor bars. Figure 4.13 shows the power spectrum of motor
with five broken bars at half load condition. It is also observed from this figure that even at
half load fault frequencies are not clearly visible because again their magnitude is low. Thus,
it is slightly difficult to detect broken bar fault at half loaded conditions also. The power
spectrum of motor with five broken bar at full load is shown in Figure 4.17. This figure
clearly show fault frequencies at 34 Hz, 42 Hz, 58 Hz and 66 Hz which is the indication of
broken rotor bar fault. The magnitudes of these fault frequencies are in between -78dB to
-60dB. The complete observation from power spectrum analysis for five broken bar are given
in Table 4.6.
Table 4.6: Power spectrum analysis of five broken bars at various loading conditions
Fault Frequencies
Figure no.
Load
Slip
Conditions
K=1
K=2
LSB
USB
(Hz)
(Hz)
(Hz) (Hz)
4.9
No Load
0.01
49
51
Not visible
48
52
Not Visible
4.13
Half Load
0.04
46
54
Visible
42
58
Not visible
4.17
Full Load
0.08
42
58
Visible
34
66
Visible
69
(iii)
The power spectrums obtained from the current signal for twelve broken bar at no
load, half load and full load is given in Figures 4.10, 4.14, 4.18. Figure 4.10 shows power
spectrum of motor with 12 broken bars under no load condition. Again, at no load condition,
the side band frequencies are very close to fundamental frequency and the amplitudes of the
sidebands is quite smaller or negligible. The detection of the searched slip frequency
sideband at no load or light load is too difficult. It is also observed from Figure 4.14 that side
band fault frequencies are visible at half load condition. The fault frequencies appear at
42Hz, 46Hz, 54 Hz and 58 Hz in the power spectrum which is indication of broken rotor bar
fault. Figure 4.18 show the power spectrum of motor with 12 broken bars at full load
condition. It is observed from the figure that broken bar fault detection at full load may be
performed in more reliable way. The frequency components related to broken bar can be
clearly recognized in the current spectrum. The complete observations from power spectrum
analysis for 12 broken bars are given in Table 4.7
Table 4.7: Power spectrum analysis of twelve broken bars at various loading conditions
Figure no.
Load
Fault Frequencies
Slip
Condition
K=1
K=2
LSB
USB
(Hz)
(Hz)
(Hz) (Hz)
4.10
No Load
0.01
49
51
Not visible
48
52
Not Visible
4.14
Half Load
0.04
46
54
Visible
42
58
Visible
4.18
Full Load
0.08
42
58
Visible
34
66
Visible
The results obtained from the experiments show that the magnitude of the frequency
components increases when the number of broken bars increases. Based on the results
obtained with the systems it can be stated that this method proven to be adequate for the
cases and load conditions considered, as the system was capable to detect the broken rotor
bars faults.
70
Figure 4.8: Power spectrum of faulty motor with 1 broken bar under no load condition
71
Figure 4.9: Power spectrum of faulty motor with 5 broken bars under no load condition
Figure 4.10: Power spectrum of faulty motor with 12 broken bars under no load
condition
72
FF [46 Hz ]
FF [54 Hz ]
Figure 4.12: Power spectrum of faulty motor with 1 broken bar under half load
73
FF [54 Hz ]
FF [46 Hz ]
Figure 4.13: Power spectrum of faulty motor with 5 broken bars under half load
FF [54 Hz ]
FF [46 Hz ]
Figure 4.14: Power spectrum of faulty motor with 12 broken bars under half load
74
FF [58 Hz ]
FF [42 Hz ]
Figure 4.16: Power spectrum of faulty motor with 1 broken bar under full load
75
FF [42 Hz ]
FF [58 Hz ]
Figure 4.17: Power spectrum of faulty motor with 5 broken bars under full load
FF [42 Hz ]
FF [58 Hz ]
Figure 4.18: Power spectrum of faulty motor with 12 broken bars under full load
76
Data
Number of samples
150
Sampling rate
1kHz
Frequency bins
512
77
Number of
samples
Signal
Frequency
resolution
Window
Data acquisition
Scan rate
Channel
info
Plot style
STFT
spectrogram
3D surface
graph
Frequency
bins
XY
projection
Figure 4.19: Block diagram for obtaining STFT spectrogram using LabVIEW
programming
78
Figure 4.21: STFT spectrogram for faulty induction motor with broken bars
79
Conventional FFT analysis is not suitable for analyzing transient signals. Although
Short-Time Fourier Transform (STFT) can be used for analyzing transient signals using a
time-frequency representation, it can only analyze the signal with a fixed sized window for
all frequencies, which leads to poor frequency resolution. However, wavelet techniques can
overcome this problem by using a variable sized window.
80
frequency spectrum with different bandwidth. Table 4.9 shows the frequency bands covered
by the seven details obtained in the performed experiment.
Table 4.9: Decomposition details
Sr. no.
Decomposition
Details
Detail at level1
3200-1600 Hz
Detail at level 2
1600-800 Hz
Detail at level 3
800-400 Hz
Detail at level 4
400-200 Hz
Detail at level 5
200-100 Hz
Detail at level 6
100-50 Hz
Detail at level 7
50-25 Hz
81
MRA
Approximation (Level 1)
MRA
Detail (Level 1)
MRA
Approximation (Level 2)
MRA
Detail (Level 2)
MRA
Approximation (Level 3)
MRA
Detail (Level 3)
MRA
Approximation (Level 4)
MRA
Detail (Level 4)
MRA
Approximation (Level 5)
MRA
Detail (Level 5)
MRA
Approximation (Level 6)
MRA
Detail (Level 6)
MRA
Approximation (Level 7)
MRA
Detail (Level 7)
Signal
Data
acquisition
Figure 4.22: Block diagram for Multiresolution analysis using LabVIEW programming
82
83
Figure 4.24: Multiresolution analysis for faulty motor with broken bars
84
85
FF [33Hz ]
FF [66 Hz ]
Figure 4.27: Power spectrum of motor (Bolts placed on inner position of slotted disc)
86
FF [66 Hz ]
FF [33Hz ]
Figure 4.28: Power spectrum of motor (Bolts placed in outer position of slotted disc)
87
the fault frequency. The wavelet details at level 7 can be used for fault detection because
amplitude at this level significantly increases which is clear indication of fault. The
results of experiment show that wavelet decomposition is the right technique for non
stationary signals.
5. The effect of unbalance rotor is also studied in this research. A slotted disc is mounted on
the shaft of motor to unbalance the rotor. Experimental results show that magnitude of
sidebands increases as unbalanced force increases. Based on the results obtained from the
experiments, it can be concluded that FFT, STFT and Wavelet transform are efficient
techniques to diagnose the rotor faults.
88
CHAPTER 5
Diagnosis Of Stator
Winding Fault In
Induction Motor
5.1 Introduction
The objective of this chapter is to propose condition monitoring of three phase
induction motor using advanced signal processing techniques for detection of stator winding
faults. The stator current contains unique fault frequency components that can be used for
stator winding fault detection. The proposed methods allows continuous real time tracking of
stator winding faults in induction motors operating under steady state and transient (variable
load) conditions. Thus, these methods may be used for continuous monitoring of the motor
health.
89
n
(1 s )
p
(5.1)
The frequencies revealing the presence of short-circuit of winding are in some cases very
close to frequencies related to other kinds of defect, as for example eccentricities. It is very
important to distinguish one frequency from the other. The expected fault frequencies at
various load conditions are shown in Table 5.1.
Speed
Conditions
(rpm)
Slip
LSB
USB
No load
1485
0.01
25 Hz
75 Hz
Full Load
1380
0.08
27 Hz
73 Hz
K=1
91
92
experiments were performed on motor under no load and full load condition. Initially, the
motor was damaged with 5% short circuit of winding. Then, severity of fault was increased
to 15% and 30%. Table 5.2 show the severity of short winding faults and load conditions for
various experiments conducted to diagnose the short winding fault.
Load conditions
5% shortened
No Load
15% shortened
No Load
30% shortened
No Load
5% shortened
Full Load
15% shortened
Full Load
30% shortened
Full Load
The power spectrum of faulty motor with 5% short circuit at no load is given in
Figure 5.3. The fault frequencies appear at 25 Hz and 75 Hz. At full load, fault frequencies
appear at 27 Hz and 73 Hz as shown in Figure 5.7. It is observed from Figure 5.3 that at no
load magnitude of fault frequency is -80dB whereas at full load magnitude is -77dB as shown
in Figure 5.7. It gives an indication that magnitude of fault frequency increases with
93
increases in load. It is also observed from the figures that fault frequencies are clearly visible
which indicates the short circuit winding fault in induction motor.
ii) 15% Short-circuit of winding:
The power spectrums of induction motor are also plotted for no load and full load
operating condition with increased severity of fault (15%). The Figure 5.4 shows the power
spectrum of faulty motor with 15% short circuit of winding at no load. The fault frequencies
appear at 25 Hz and 75 Hz. It justifies the calculated and experimental results. The
magnitude of fault frequencies were found in between -77 dB to -75 dB for LSB and USB.
Magnitude of fault frequencies has been increased if compared with magnitude of 5%
severity of fault. Increases the magnitude of fault frequency with respect to increases in
severity of fault is observed. Increase in magnitude of current component is undesirable
aspect for the performance of induction machine. The same outcome has been observed for
full load condition as shown in Figure 5.8. The fault frequencies appear at 27 Hz and 73 Hz
which is also a calculated value at full load condition. However, the magnitudes of these fault
frequencies have been significantly increased due to increased loading condition and severity
of fault.
iii) 30% Short-circuit stator winding:
The severity of fault is increased by 30% and power spectrums for faulty motor for
no load and full load conditions are shown in the Figures 5.5 and 5.9 respectively. Virtual
Instrument (VI) predicted the current components with increased magnitude which are
obtained at position 25 Hz and 75 Hz for no load condition and 27Hz and 73Hz at full load
condition. The components are distributed symmetrically around fundamental frequencies as
expected. It is observed from the figures that the magnitudes of fault frequencies have been
significantly increased up to -60dB with increase of load and severity of fault.
The condition monitoring of the induction motor with help of Fast Fourier Transform
(FFT) for finding the stator winding faults may give better results on line. Above
observations can be summarized that with increase in load and percentage of short circuit
winding the fault current magnitude increases. The fault frequencies obtained by
mathematical derivation and experimentally are same for all the above cases. The complete
observation from power spectrum analysis for short winding fault is given in Table 5.3.
94
Table 5.3: Power spectrum analysis for short circuited winding fault
5.3
Short
circuited
stator
winding
5%
5.7
Load
Condition
Fault Frequencies
Lower side band
Upper side band
FF
Mag.
FF
Mag.
Observati
ons
No Load
25 Hz
-80 dB
75 Hz
-80 dB
Visible
5%
Full Load
27 Hz
-77 dB
73 Hz
-77 dB
Visible
5.4
15%
No Load
25 Hz
-77 dB
75 Hz
-75 dB
Visible
5.8
15%
Full Load
27 Hz
-72 dB
73 Hz
-72 dB
Visible
5.5
30%
No Load
25 Hz
-71 dB
75 Hz
-62 dB
Visible
5.9
30%
Full Load
27 Hz
-60 dB
73 Hz
-60 dB
Visible
Fig.
No.
95
FF (75 Hz )
FF (25 Hz )
Figure 5.3: Power spectrum of faulty motor with 5% shortened under no load condition
FF (25 Hz )
FF (75 Hz )
Figure 5.4: Power spectrum of faulty motor with 15% shortened under no load
condition
96
FF (75 Hz )
FF (25 Hz )
Figure 5.5: Power spectrum of faulty motor with 30% shortened under no load
condition
97
FF (73Hz )
FF (27 Hz )
Figure 5.7: Power spectrum of faulty motor (5% shortened) under full load
FF (73Hz )
FF (27 Hz )
Figure 5.8: Power spectrum of faulty motor (15% shortened) under full load
98
FF (73Hz )
FF (27 Hz )
Figure 5.9: Power spectrum of faulty motor (30% shortened) under full load
99
Number of
samples
Signal
Frequency
resolution
Data acquisition
Scan rate
Channel
info
Order of
spectrogram
Plot style
Gabor
spectrogram
3D surface
graph
Frequency
bins
XY
projection
Figure 5.10: Block diagram for obtaining Gabor spectrogram using LabVIEW
programming
100
Fault frequency
The data acquisition parameters for this experiment are given in Table 5.4.
Table 5.4: Data acquisition parameters
Parameters
Data
Sampling rate
1000Hz
Number of samples
150
Frequency bins
512
Order of spectrogram
101
Figure 5.11 shows the Gabor spectrogram for a healthy induction motor. The motor
with 30% short winding analyzed in the experiment. The resulting spectrogram is shown in
Figure 5.12. The spectrogram shows the harmonic nearest to main frequency which is the
result of short-circuited winding. The spectrogram which is observed with the Gabor
Transform also consist the time variation. It gives the fault frequencies from the perspective
of time variation and could, therefore, be useful tool for diagnosis of stator winding faults.
Frequency resolution
The equation (5.2) gives the relationship between number of samples (Ns) frequency
resolution (R) and sampling frequency (fs).
Ns =
fs
R
.(5.2)
In this experiment, the sample frequency (fs) is 6400, and number of samples are taken 12600.
The block diagram for Multiresolution analysis using LabVIEW programming is shown in
Figure 5.13.
102
MRA
Approximation (Level 1)
MRA
Detail (Level 1)
MRA
Approximation (Level 2)
MRA
Detail (Level 2)
MRA
Approximation (Level 3)
MRA
Detail (Level 3)
MRA
Approximation (Level 4)
MRA
Detail (Level 4)
MRA
Approximation (Level 5)
MRA
Detail (Level 5)
MRA
Approximation (Level 6)
MRA
Detail (Level 6)
MRA
Approximation (Level 7)
MRA
Detail (Level 7)
Signal
Data
acquisition
Figure 5.13: Block diagram for Multiresolution analysis using LabVIEW programming
103
104
Figure 5.15: Multi resolution analysis for 30% short circuited induction motor
105
[ fs
decreases as (Article 3.8.2). In this case, the sample frequency (fs) is 6400, and 12600
samples were acquired. Thus, the band varies from 0 to 3.2 kHz. When a short circuit
produced between the turns in a stator phase, not only an unbalance appears in currents but
also fault harmonics due to it. The harmonic variation can be noticed in the expected bands
for this kind of fault in range of low frequencies from 25 to 200 Hz. The higher levels of
multiresolution analysis (MRA) do not provide useful information. The fault may be detected
by comparing the lower levels of MRA of motor under healthy and faulty conditions. It can
be clearly observed from the MRA of faulty motor (Figure 5.15) that amplitude at level 7 is
significantly increases which indicates the presence of short circuit fault in induction motor.
The experiments performed and the results obtained show that wavelet analysis achieves
better results in field of short circuit winding faults of the induction motor.
106
programming is shown in Figure 5.17. The induction motor has been initially tested, in the
absence of faults in order to determine the reference current Parks vector pattern
corresponding to the supposed healthy motor. Afterward, short circuited motor was tested. A
time window of 175ms was used for all data acquisition in order to get simple and sufficient
detailed pattern. The sample rate was 2000 sample/second. The number of samples was taken
350.
Load
Data Acquisition
Data treatment
Signal
Frequency
resolution
IQ data
Data acquisition
Scan rate
Park's vector
IQ
Graph
Channel
info
Figure 5.17: Block diagram for obtaining Current Park's vector pattern using
LabVIEW programming
107
Figure 5.19: Current Parks vector pattern for short circuited motor
108
109
results obtained from the experiments, it can be concluded that mutiresolution analysis a
comparatively better technique to diagnose short circuit winding faults of the induction
motor.
5. The Park's vector approach is also introduced for detecting the short winding faults. The
unbalance is created by short circuited winding fault and can be easily detected by Park's
vector approach. An undamaged machine shows a perfect circle in Parks vector
representation whereas an unbalance due to winding faults results in an elliptic
representation of the Parks vector. Thus, Short winding fault can be easily detected by
comparing both patterns.
6. The implemented and tested methods showed their efficiency in fault diagnosis and
condition monitoring of induction motor. The results obtained present a great degree of
reliability, which enables the proposed methods as monitoring tools for diagnosis of short
winding fault of similar motors.
110
CHAPTER 6
111
Centre of rotation +
centre of bore
Air gap
Motor shaft
Stator bore
Rotor
Non-invasive methods can be used to detect the air gap eccentricity in induction machines.
These methods utilize the monitored stator current.
dynamic eccentricity, too. Assuming that the rotor shaft assembly is sufficient stiff, the level
of static eccentricity does not change. Due to the air gap asymmetry, the stator currents will
contain well defined components, and these can be detected.
Dynamic eccentricity means that the rotor is rotating on the stator bore axis but not on
its own axis. The off-center axis of rotation spin along a circular path with the same speed as
the rotor does (first-order dynamic eccentricity). This kind of eccentricity may be caused by a
bent shaft, mechanical resonances, bearing wear or movement, or even static eccentricity.
Therefore, the non-uniform air-gap of a certain spatial position is sinusoidally modulated,
and results in an asymmetric magnetic field. This accordingly gives rise to revolving UMP
[97]. Due to dynamic eccentricity, side band components appear around the slot harmonics in
the stator line current frequency spectra. Figure 6.2 shows an illustration of how the rotor
would rotate in the presence of each type of air-gap eccentricity.
Dynamic eccentricity
Static eccentricity
113
( nrt R nd )
(1 s ) n
p
f1
.(6.1)
where
fag = frequency components in a current spectrum due to rotor slotting and air gap
eccentricity, Hz
nrt = any integer, 0, 1, 2, 3, ...
R = number of rotor bars
nd = eccentricity order number; any integer, 0, 1, 2, 3, ...
nd = 0 for static eccentricity (principal slot harmonics)
nd = 1, 2, 3, ... for dynamic eccentricity
s = nondimensional slip ratio
p = pole-pairs, which is half the number of poles (P), i.e. p = P/2
nws = order number of stator MMF time harmonic or stator current time harmonic;
odd integer, 1, 3, 5, ...
f1 = supply line frequency, Hz
In general, this equation can be used to predict the frequency content for the current signal.
There are three ns in the equation and, therefore, three sets of harmonics: nrt is rotor related,
nws stator related and nd eccentricity related. For static eccentricity variations nd = 0 and for
dynamic eccentricity variations nd = 1, 2, 3, ....
The expected fault frequencies at various load conditions are shown in Table 6.1.
114
Speed
Conditions
(rpm)
Slip
nd=-1
nd=0
nd=1
No load
1485
0.01
916 Hz
941 Hz
965 Hz
Full Load
1380
0.08
855 Hz
878 Hz
901 Hz
nws=1
Dynamic eccentricity can be expressed as percent (%) dynamic eccentricity and defined by:
% dynamic eccentricity =
(6.2)
where
Nominal gap= Total air gap/2
1.mm
Stator
Rotor
0.6mm
115
Figure 6.4: Parts of motor machined for implementing air gap eccentricity
116
Number of
samples
Signal
Frequency
resolution
Window
Power
spectrum
Data acquisition
Scan rate
Channel
info
X-Y/Waveform
graph
FFT
analysis
Figure 6.5: Block diagram for obtaining power spectrum using LabVIEW
programming
frequency is difficult to identify because its magnitude is very less. When motor is tested
under full load condition, the fault frequency appears at 878 Hz as shown in Figure 6.10. It
can be observed that the magnitude of this fault frequency is slightly greater than the fault
frequency (941 Hz) which was appeared at no load condition. This frequency (878 Hz) can
be clearly identified in power spectrum and indicates the presence of static eccentricity. Thus,
the lower level of static eccentricity can be clearly detected at full load condition but the
same is slightly difficult to identify at no load and light load conditions. Table 6.2 shows
Power spectrum analysis for 25% static eccentricity fault.
Load
Fault Frequencies
no.
Conditions Slip
Magnitude
Observations
experimentally)
6.7
No Load
0.01
941 Hz
-82 dB
FF Not Visible
6.10
Full Load
0.08
878 Hz
-80 dB
FF Visible
The motor was also tested for increased level of air gap eccentricity. The air gap
eccentricity was increased up to 50% by machining the housing motor again.
Virtual
Instrument (VI) predicted current components due to abnormal level of static eccentricity at
no load conditions. Figure 6.8 shows the current spectra of motor after its housing was
machined and installed again with 50% air gap setting at no load. The fault frequency again
appears 941 Hz in power spectrum but the magnitude of this frequency could not find
because it become merge into associated frequency. The similar results have been obtained
from the experiments, when motor was test for full load condition with same level of air gap
eccentricity. At full load, the motor was operating at 1.05 amp. The full load speed is 1380
rpm yielding a frequency at 778 and 878 Hz for detection of air gap eccentricity. The Figure
6.11 shows the fault frequency again appears at 878 Hz. This fault frequency can be clearly
observed in the power spectrum. The Table 6.3 shows the complete power spectrum analysis
of induction motor with 50% air gap eccentricity.
118
Table 6.3: Power spectrum analysis for 50% air gap eccentricity
Figure
Load
no.
Conditions Slip
6.8
No Load
Fault Frequencies
Magnitude
941 Hz
--
0.01
Observations
Magnitude Could
not measured
6.11
Full Load
0.08
878 Hz
-75 dB
FF Visible
The motor was tested again with mixed eccentricity to study its effect on current
components. First, the power spectrum between the frequency 900 Hz to 990 Hz was
obtained at no load condition so that this spectrum can be compared with power spectrum of
faulty motor for same loading condition. The Figure 6.12 shows the power spectrum (900Hz990 Hz) of healthy motor. The reading was taken again to obtain the power spectrum of
motor with mixed eccentricity under no load condition as shown in Figure 6.13. This power
spectrum shows the fault frequencies at 916 Hz, 941 Hz and 965 Hz. The similar results have
been obtained from the experiments, when motor was tested at full load condition with same
level of air gap eccentricity. In this case, the fault frequency again appears at 855 Hz, 878 Hz
and 901 Hz in power spectrum but with increased magnitude (Figure 6.15). Due to increased
magnitude, this fault frequency is easy to identify. It is observed from the figures that
magnitude of fault frequencies increases with increase of severity of fault. The results also
indicate that a unique pattern occurred in the power spectrum due to presence of mixed
eccentricity. Table 6.4 shows the analysis of power spectrums of induction motor with mixed
eccentricity.
Table 6.4: Power spectrum analysis for mixed eccentricity
Figure Load
no.
Slip
Fault Frequencies
Magnitude
(Hz)
(dB)
Conditions
Observations
6.13
No Load
0.01
916
941
965
-69
-68
-69
Magnitude
6.15
Full Load
0.08
855
878
901
-66
-63
-66
increases
with
increase of load.
119
FF [941Hz ]
Figure 6.7: Power spectrum of faulty motor with 25% static eccentricity under no load
condition
120
FF [941Hz ]
Figure 6.8: Power spectrum of faulty motor with 50% static eccentricity under no Load
condition
Figure 6.9: Power spectrum of healthy motor under full load condition
121
FF [878 Hz ]
Figure 6.10: Power spectrum of faulty motor with 25% static eccentricity under full
load
FF[878Hz]
Figure 6.11: Power spectrum of faulty motor with 50% eccentricity under full load
122
123
FF (916 Hz )
FF (941Hz )
FF (965 Hz )
Figure 6.13: Power spectrum of faulty motor with mixed eccentricity under no load condition
124
125
FF (855 Hz )
FF (878 Hz )
FF (901Hz )
Figure 6.15: Power spectrum of faulty motor with mixed eccentricity under full Load
126
127
CHAPTER 7
Experimental Study Of
Bearing And Gear Box
Faults Of Induction
Motor
7.1 Introduction
A very important aspect of condition monitoring of induction motor is to detect the
mechanical faults. The reliability of an induction motor is of paramount importance in
industrial, commercial, aerospace and military applications. Bearing play an important role in
the reliability and performance of all motor systems. Due to close relationship between motor
system development and bearing assembly performance, it is difficult to imagine the progress
of modern rotating machinery without consideration of the wide application of bearing. In
128
addition, most faults arising in motors are often linked to bearing faults. The result of many
studies show that bearing problems account for over 40% of all machine failure [12]. In
present chapter, investigations have been done to find the application of advanced signal
processing techniques for detection of bearing faults. As bearing faults are critical to the
functioning of any electromechanical system, they form the main topic of discussion in this
chapter. In some applications such as aircrafts, the reliability of gears may be critical in
safeguarding human lives. For this reason, the detection of gear box faults has been an
important research area. Therefore, the effects of gear box fault on motor terminal current are
also studied in this chapter.
Ball defect
The relationship of bearing vibration to the stator current spectra can be determined
by remembering that any air gap eccentricity produces anomalies in the air gap flux density.
Since ball bearings support the rotors, any bearing defect will produce a radial motion
between the rotor and stator of the machine. The mechanical displacement resulting from
damaged bearing causes the machine air gap to vary in a manner that can be described by a
combination of rotating eccentricities moving in both directions. Due to rotating
eccentricities, the vibrations generate stator currents at frequencies given by [5, 29, 53, 118,]:
fbearing = f1 m. fi ,o
.(7.1)
where m=1,2,3,4..and fi,o is one of the characteristic frequencies which are based upon
the bearing dimensions shown in Figure 7.1
129
Outer.race : f i ,0 =
Nb
D
f r 1 b cos ......(7.2)
2
Dc
where
Nb=number of bearing balls
fr = mechanical rotor speed in hertz
Db = Ball diameter
Dc = Bearing pitch diameter
= Contact angle of the balls on the races
It should be noted from (7.2) that specific information concerning the bearing construction is
required to calculate the exact characteristic frequencies. However, these characteristics race
frequencies can be approximated for most bearings with between six and twelve balls [3].
f 0 = 0.4 N b f r
(7.3)
fi = 0.6 N b f r
.(7.4)
The expected fault frequencies for inner race fault and outer race fault at various load
conditions are given in Tables 7.1 and 7.2
130
Table 7.1: Expected fault frequencies for inner race fault at various load conditions
Load
Speed
Conditions
(rpm)
Slip
m=1
m=2
m=3
LSB
USB
LSB
USB
LSB
USB
(Hz)
(Hz)
(Hz)
(Hz)
(Hz)
(Hz)
No load
1485
0.01
68
168
187
287
306
406
Full Load
1380
0.08
60
160
170
270
282
382
Table 7.2: Expected fault frequencies for outer race fault at various load conditions
m=1
m=2
m=3
Load
Speed
Slip
Conditions
(rpm)
LSB
USB
LSB
USB
LSB
USB
(Hz)
(Hz)
(Hz)
(Hz)
(Hz)
(Hz)
No load
1485
0.01
29
129
108
208
187
287
Full Load
1380
0.08
23
123
97
197
170
270
131
132
133
Experiments
Case 1
Load
conditions
No Load
Full load
No load
Full load
No Load
Full load
No load
Full load
Case 2
Case 3
Case 4
The motor is tested under no load condition with faulty bearing. The fault in bearing
was made by drilling a hole of 2mm diameter in its inner race. It is observed from the power
spectrums of motor that fault frequencies are not clearly visible at no load condition because
their magnitude is less. The power spectrums of faulty motor with 2mm hole in inner race of
bearing under no load condition is shown in Figures 7.5.and 7.6. When the motor is tested
again with same bearing under full load condition, it is observed that magnitude of fault
frequencies are increases but these are slightly difficult to identify in the power spectrum.
The power spectrum of faulty motor with 2mm hole in inner race of bearing under full load
condition is shown in Figures 7.10. The power spectrums for 2mm inner race fault of motor
are shown in Figures 7.5, 7.6 and 7.10 and their analysis for no load and full load is
summarized in Table 7.4.
Table 7.4: Power spectrum analysis for inner race fault of motor with 2mm hole
Figure
no.
7.5 and
7.6
7.10
Load
Conditions
No Load
Slip
0.01
FF
(Hz)
68
Full Load
0.08
60
Fault frequencies
Mag FF Mag FF
(dB) (Hz) (dB) (Hz)
-78
168
-77
187
-76
134
160
-74
170
Mag
(dB)
-78
-76
Observations
FF difficult to
identify
FF difficult to
identify
To observe the effect of severity of bearing fault on current components, the 4mm
hole was drilled in the inner race and then bearing was installed in the motor. The motor was
tested under both no load and full load conditions. The power spectrum of faulty motor with
4mm hole in inner race of bearing under no load condition is shown in Figures 7.7 and 7.8.
These figures clearly show that the fault frequencies appear at 68 Hz, 168 Hz, 187 Hz and
287 Hz in the spectrum which indicates the inner race fault in the bearing of motor. The
magnitudes of these frequencies are between -74 dB to -76 dB.
The motor with same fault was also tested under full load condition. The power
spectrum of faulty motor with 4mm hole in inner race of bearing under full load condition is
shown in Figure 7.11. In this case, the fault frequencies appeared at 60 Hz, 160 Hz, and 170
Hz. It is observed that the magnitude of fault frequencies have been increased significantly. It
is due to increase of load and severity of fault. The power spectrum analysis for 4mm inner
race fault is given in Table 7.5.
Table 7.5: Power spectrum analysis for induction motor with 4mm inner race fault
Fault frequencies
Figure
no.
Load
Condition
Slip
0.01
0.08
FF Mag.
(Hz) (dB)
68
-75
60
-68
FF
(Hz)
168
Mag
(dB)
-74
FF
(Hz)
187
160
-68
170
Mag. Observations
(dB)
-76 FF
clearly
identified
-70 FF
clearly
identified
The motor was also tested with outer race fault of bearing. Initially, the 2mm
diameter of hole was drilled in the outer race of bearing and then it was installed in the motor.
The power spectrum of faulty motor with 2mm hole in outer race of bearing under no load
condition is shown in Figure 7.13. This figure shows that fault frequencies can be clearly
identified in power spectrum at 29 Hz, 108 Hz and 129 Hz. Similar results are obtained from
the experiment when the motor was tested with same fault under full load conditions. In this
case, the fault frequencies are appearing at 23 Hz, 97 Hz and 123 Hz which is indication of
135
outer race fault of bearing. Such frequencies are shown in Figure 7.16. Table 7.6 gives power
spectrum analysis for induction motor with 2mm outer race fault.
Table 7.6: Power spectrum analysis for induction motor with 2mm outer race fault
Figure
no.
7.13
Load
Conditions
No Load
Slip
0.01
FF
(Hz)
29
7.16
Full Load
0.08
23
Fault frequencies
Mag FF Mag FF
(dB) (Hz) (dB) (Hz)
-76
108
-74
129
-72
97
-68
123
Mag
(dB)
-73
-73
Observations
FF
clearly
identified
FF
clearly
identified
The motor was again tested with highly defective bearing. In this case, severity of
fault was increased to 4mm by drilling the hole into outer race of bearing. The results show
that fault frequencies can be clearly identified in power spectrum, when motor is tested under
no load condition and full load conditions. The power spectrum of faulty motor with 4mm
hole in outer race of bearing under no load condition is shown in Figure 7.14. It is observed
that the fault frequencies with increased magnitude have been appeared in the power
spectrum. The similar results have been obtained, when motor was tested under full load
condition. The power spectrum of faulty motor with 4mm hole in outer race of bearing under
full load condition is shown in Figure 7.17. The power spectrum analysis for 4mm outer race
fault is given in Table 7.7.
Table 7.7: Power spectrum analysis for induction motor with 4mm outer race fault
Figure
no.
7.14
Load
Conditions
No Load
Slip
0.01
FF
(Hz)
29
7.17
Full Load
0.08
23
Fault frequencies
Mag FF Mag FF
(dB) (Hz) (dB) (Hz)
-69
108
-69
129
-67
136
97
-65
123
Mag
(dB)
-70
-69
Observations
FF
clearly
identified
FF
clearly
identified
FF (168Hz )
FF (68Hz )
Figure 7.5: Power spectrum of faulty motor with 2mm hole in inner race of bearing under no load condition (m=1)
137
FF (187 Hz )
FF (287 Hz )
Figure 7.6: Power spectrum of faulty motor with 2mm hole in inner race of bearing under no load condition (m=2)
FF (168Hz )
FF (68Hz )
Figure 7.7: Power spectrum of faulty motor with 4mm hole in inner race of bearing under no load condition (m=1)
138
FF (187 Hz )
FF (287 Hz )
Figure 7.8: Power spectrum of faulty motor with 4mm hole in inner race of bearing under no load condition (m=2)
Figure 7.9: Power spectrum of healthy motor under full load condition
139
FF (160 Hz )
FF (170 Hz )
FF (60 Hz )
Figure 7.10: Power spectrum of faulty motor with 2mm hole in inner race of bearing under full load condition
FF (160 Hz ) FF (170 Hz )
FF (60 Hz )
Figure 7.11: Power spectrum of faulty motor with 4mm hole in inner race of bearing under full load condition
140
FF (108Hz )
FF (29 Hz )
FF (129 Hz )
Figure 7.13: Power spectrum of faulty motor with 2mm hole in outer race of bearing under no load condition
141
FF (108Hz ) FF (129 Hz )
FF (29 Hz )
Figure 7.14: Power spectrum of faulty motor with 4mm hole in outer race of bearing under no load condition
Figure 7.15: Power spectrum of healthy motor under full load condition
142
FF (97 Hz )
FF (23Hz )
FF (123Hz )
Figure 7.16: Power spectrum of faulty motor with 2mm hole in outer race of bearing under full load condition
FF (23Hz )
FF (97 Hz )
FF (123Hz )
Figure 7.17: Power spectrum of faulty motor with 4mm hole in outer race of bearing under full load condition
143
Signal
Frequency
resolution
Data acquisition
Scan rate
Frequency
bins
Wigner-Ville
Distribution
Plot style
3D surface
graph
XY
projection
Channel
info
Figure 7.19: Wigner -Ville Distribution (WVD) representation of motor with healthy
bearing
Fault frequencies
145
The block diagram for obtaining Wigner-Ville Distribution representation using LabVIEW
programming is shown in Figure 7.18. The number of sample 150 and sample frequency
1000 Hz have been chosen, which is enough to cover the significant current band of
induction motor. The frequency bins was taken 1024.
146
samples was taken 350. The Figure 7.21 shows the block diagram for obtaining Current
Park's vector pattern using LabVIEW programming.
Number of
samples
Signal
Frequency
resolution
IQ data
Data acquisition
Scan rate
Park's vector
IQ
Graph
Channel
info
Figure 7.21: Block diagram for obtaining Current Park's vector pattern using
LabVIEW programming
147
Figure 7.23: Current Parks vector pattern for faulty bearing with 4 mm diameter hole
in inner race
Figure 7.24: Current Parks vector pattern for faulty bearing with 4 mm diameter hole
in outer race
148
149
f e = f1 m. f r (7.5)
where
f1 = supply frequency
fr = rotational speed frequency of the rotor
m = 1,2,3.............harmonic number
f e = current components due to airgap changes
As seen from above, mechanical oscillations will give rise to additional current
components in the frequency spectrum. Gearboxes may also give rise to current components
of frequencies close to or similar to those of broken bar components. Specifically, slow
revolving shafts will give rise to current components around the main supply frequency
components as prescribed by equation (7.5) where the rotational speed frequency of the shaft,
rotating with Nr rpm, may be calculated as
fr =
f1
n( p2 )
Where
n = gear ratio and
p = the number of poles pairs.
7.7
Gear fault
Transform
detection
using
Fast
Fourier
To detect the gear fault in gear box, FFT based power spectrum was used. The power
spectrum was obtained by programming in LabVIEW. The experimental set up consist of
induction motor, worm and worm gear box, data acquisition card, data acquisition board, and
a transformer. The detail of experimental set up is given below:
150
applied by rope brake dynamometer. The electrical and the mechanical parameters of the
experimental system for a typical supply frequency of 50 Hz are:
151
Damaged .tooth
152
A localized damaged tooth fault is implemented by removing one tooth as shown in Figure
7.27. The Figures 7.28 and 7.29 show the experimental set up for diagnosis of gear box fault.
Initially, motor current was recorded with healthy gear box under load condition. The load is
applied on gear box with rope brake dynamometer. Afterward, the tooth of gear was removed
by grinding operation. Then faulty gear box was coupled to the motor and load is applied on
gear box. Then, motor current was recorded again to get the power spectrum.
153
Figure 7.30: Power spectrum for motor with healthy gear box
Fault. frequencies
154
induction
motor. The signal processing techniques (FFT, Wigner Distribution, Parks Vector) are used
to detect the mechanical faults of motor. The following conclusions may be drawn from the
observations of results obtained by the experiments:
1. Visibility of characteristic frequencies depends upon the severity of bearing fault.
2. If load increases, the magnitude of fault frequencies increases.
3. The Wigner-Ville Distribution (WVD) is an efficient time frequency technique for
detection of bearing faults. The result of WVD representation clearly shows fault
frequencies in spectrum due to presence of bearing fault.
4. Parks vector approach may be used for detecting the bearing faults. The result of this
approach shows that Parks vector current spectrum of healthy motor is different from the
current spectrum of the motor having faulty bearing.
5. To detect the gear box fault, FFT based power spectrum can be used effectively. Results
obtained from the experiment verified that any fault in either the pinion or the driven
wheel generates harmonic components in the motor current spectrum. These components
can be detected in power spectrum of motor.
6. It is further concluded that the selection of current frequency for gear box (broken tooth)
and for rotor broken bar should be considered very carefully because these current
frequencies may be very close to each other.
155
CHAPTER 8
Conclusions,
Contributions, And
Recommendations
8.1 Introduction
The aim of this research is to advance the field of condition monitoring and fault
diagnosis in induction motor operating in variety of operating conditions. The fast growth in
applications of the induction motor in sensitive areas such as nuclear power plants has
increased the need for continuous condition monitoring of motors. Therefore, diagnostic of
various faults of induction motor such as rotor fault, stator winding fault, air gap eccentricity,
bearing failure and gear box fault is the focus of this research.
156
Multiresolution analysis has also been conducted to diagnose the rotor bar fault under
varying load conditions. This method is tested for healthy and faulty condition of motor. The
results obtained from the experiments show that low frequency details are much more
relevant for fault detection because they cover the frequency band corresponding to the
supply and the fault frequency. The seven detail of the described wavelet, which is in the
frequency band of 25-50 Hz was the most significant for the diagnosis of broken bars. There
is no useful information about signal variation available for the decomposition levels from
one to five. The wavelet details at level seven can be easily used for fault detection because
amplitude at this level significantly increases which indicates the fault. This experimental
study reveals that wavelet decomposition is the right technique for non stationary signals.
The effect of unbalance rotor is also studied in this research work. A slotted disc was
mounted on the shaft of motor to unbalance the rotor. The unbalance forces were produced
by fixing the weights to the slotted disc. The experimental results show that magnitude of
sidebands increases as unbalanced force increases. Based on the results obtained from the
experiments, it can be concluded that FFT, STFT and Wavelet Transform are efficient
techniques to diagnose the broken bar faults.
The second phase of this research consists of experimental work related to diagnosis
of short winding faults in induction motors operating under different load conditions. To
diagnose the short winding fault, four current based fault detection techniques such as FFT,
Gabor Transform, Wavelet Transform and Parks vector approach are implemented. Results
show the all expected bands which are the due to short winding fault. Several experiments
were performed on motor under no load condition and with load coupled to shaft of motor.
Initially, the motor was damaged with 5% short circuit of winding. Then, severity of fault
was increased to 15% and 30%. It can be seen that the magnitude of short circuit related
frequencies increases with the severity of short-circuit. Gabor Transform is also applied to
detect the short winding fault. This clearly shows the fault frequency in color map which is
result of short circuit winding fault. Likewise, multiresolution analysis was conducted for
both healthy and faulty motor. Then results were compared to make conclusions. In case of a
short circuit produced between the windings in a stator phase, not only an unbalance appears
in currents but also fault harmonics due to it. The harmonic variation can be noticed in the
expected bands for this kind of fault in range of low frequencies from 25 to 200 Hz.
158
Significant variations have been observed in detail seven where faulty frequency appears.
Based on the results obtained with the system, it can be stated that mutiresolution analysis is
a good technique to diagnose short circuit winding faults of the induction motor. The Parks
vector approach is also introduced for detecting the short winding faults. An undamaged
machine shows a perfect circle in Parks vector representation whereas an unbalance due to
winding faults results in an elliptic representation of the Parks vector approach. The results
obtained from the experiments present a great degree of reliability, which enable these
methods to be used as monitoring tool for similar motors.
The subject of on-line detection of air-gap eccentricity in three phase induction motor
is discussed in the third phase of research work. To detect the eccentricity fault, the non
invasive approach based on the computer aided monitoring of stator current; Fast Fourier
Transform (FFT) is implemented. Experimental results obtained by using a special fault
producing test rig. The results demonstrate the effectiveness of the proposed technique for
detecting presence of air gap eccentricity in operating three phase induction machine. This
study demonstrates that it is possible to detect the presence of air-gap eccentricity in
operating three phase induction motor, by computer aided monitoring of stator current.
Qualitative information about severity of fault can be easily obtained by using FFT. By
comparing the healthy machine with air gap eccentricity cases, it is concluded that magnitude
of static eccentricity related frequencies increases with severity of air gap eccentricity fault.
The fourth phase of research work investigates the feasibility of detecting mechanical
faults such as bearing failure and gear box faults using the spectrum of single phase of the
stator current of an induction motor. Defective rolling element bearings generate eccentricity
in the air gap with mechanical vibrations. The air gap eccentricities cause vibrations in the air
gap flux density that produces visible changes in the stator current spectrum. The signal
processing techniques (FFT, Wigner-Ville Distribution, Parks vector) are applied to detect
the bearing fault and gear box faults of motor.
characteristic frequencies could not seen in the power spectrum if outer race fault and inner
race fault is small in size. As severity of fault increases, the characteristic frequencies
become visible. The results also show that, for defective bearing having 2 mm diameter hole,
the inner race and outer race fault frequencies are slightly difficult to identify in power
spectrum at no load condition. As load is increased, fault frequencies become visible. The
159
8.3 Contributions
The main aim of the research work is to diagnose the common electrical and
mechanical faults experimentally with help of suitable signal processing techniques. In order
to perform accurate and reliable analysis on induction motors, an experimental set up is
designed that can accurately repeat the measurements of signals and can introduce a
particular fault to the motor in isolation of other faults. In the present research work,
LabVIEW environment is used to diagnose the faults with direct online condition monitoring.
The contributions of this research are summarized as follows:
1. Survey in following areas have been performed:
160
3. Motor current signature analysis based methods are applied to diagnose the rotor faults.
Rotor bar fault has been detected with help of Fast Fourier Transform.
Short Time Fourier Transform (STFT) has been successfully implemented for
detection of rotor bar fault.
An experimental study for rotor bar fault diagnosis with help of wavelet analysis
under non stationary load has been presented in this research.
4. The various methods for detection and diagnosis of stator windings faults of induction
motor have been proposed.
Short winding fault has been diagnosed using Fast Fourier Transform in the
research work.
Wavelet analysis was implemented to detect the short winding fault under nonstationary condition.
Short winding fault detection was also achieved with Parks vector Approach.
Gabor spectrogram has been presented for detection of short winding fault.
5. An experimental study of air gap eccentricity faults has been presented with help of FFT.
6. Bearing failure and gear box fault have been experimentally detected with help of current
based monitoring techniques.
Bearing fault analysis is presented in the thesis and bearing failure was detected
experimentally using FFT technique.
Parks vector approach is also presented and used for detection of inner race fault
and outer race fault of motors bearing.
An experimental study to detect the gear box fault has been presented in the
research work.
work helps in understanding the applications and limitations of fault detecting techniques.
The various fault detecting methods proposed in this work are able to diagnose motors faults
more sensitively and more reliably. The present work in the thesis has been published in
journals and proceeding of conferences.
Required
measure
ment
FFT
One
stator
current
STFT
One
stator
current
Faults diagnosed
Advantages
Disadvantages
Lost time
information
Not effective at
light load
condition
Fast speed
Suitable for
varying load
condition
Analyze signal
with fixed sized
window
Poor frequency
resolution
Fine frequency
resolution
Moderate speed
Suitable for
varying load and
light load
conditions
Fine frequency
resolution,
Fast speed
Require
expertise
Easy to diagnose
the fault
One
Gabor
Transform stator
current
One
Wavelet
Transform stator
current
Wigner
Ville
distributio
n
Park
Vector
Approach
One
stator
current
Bearing fault
Three
stator
current
Short
winding
faults
Bearing faults
162
Strong cross
term interference
The investigations can be expanded by introducing multiple stator and rotor fault types
into a motor.
For large size motors, new challenges may exist for current based fault detection.
Therefore, proposed techniques may be applied for fault diagnosis of large size motors.
Additional work is needed to investigate the applicability of other signal processing tools
in characterizing the fault signature.
There is need to study the effect of electric drives because these may change the current
spectrum.
The effects of non-stationary operations on the stator current need to be investigated for
fault detection purposes.
163
References
[1] Peter Vas, Parameter estimation, condition monitoring, and diagnosis of electrical
machines, Clarendon Press Oxford., 1993.
[2]
[3] W.T. Thomson and R.J. Gilmore, Motor current signature analysis to detect faults in
induction motor derives-Fundamentals, data interpretation, and industrial case histories,
proceedings of 32nd Turbomachinery symposium, Texas, A&M university, USA, 2003.
[4]
[5] Benbouzid, M. E. H., A review of induction motors signature analysis as a medium for
faults detection, IEEE Transactions on Industrial Electronics, Vol. 47, October, No.5,
pp. 984-993, 2000.
[6] Boashash B., Time frequency signal analysis in: Advances in spectrum estimation and
array processing, Ed. S. Haykin, Printice Hall, 1990.
[7] M. E. H. Benbouzidi , M. Viera, and C. Theys, Induction motors faults detection and
localization using stator current advanced signal processing techniques, IEEE
Transactions on Power Electronics, Vol. 14, No. 1, pp14-22, Jan. 1999.
[8]
[9]
P. C. Sen, Principles of electric machines and power electronics, John Wiley and
Sons, 1989.
164
[14] O. I. Okoro, Steady and transient states thermal analysis of a 7.5-kW squirrel-cage
induction machine at rated-load operation, IEEE Transactions on Energy Conversion,
Vol. 20, No. 4, pp. 730-736, December, 2005.
[15] Z. Gao, T. G. Habetler, and R. G. Harley, An online adaptive stator winding
temperature estimator based on a hybrid thermal model for induction machines, in
Conf. Rec. IEEE IEMDC05, pp. 754-761, San Antonio, Texas, USA, May, 2005.
[16] J. F. Moreno, F. P. Hidalgo, and M. D. Martinez, Realization of tests to determine the
parameters of the thermal model of an induction machine, IEEE Proc. Electric Power
Application, Vol. 148, No. 5, pp.393-397, September, 2001.
[17] P. Milanfar and J. H. Lang, Monitoring the thermal condition of permanent-magnet
synchronous motors, IEEE Transactions on Aerospace and Electronic Systems, Vol.
32, No. 4, pp. 1421-1429, October, 1996.
[18]
[19] A. J. Ellison and S. J. Yang, Effects of rotor eccentricity on acoustic noise from
induction machines, Proceedings of IEE, 118, (1), pp. 174-184, 1971.
[20] Dorrell, D. G. and Smith, A. C, Calculation and measurements of unbalance magnetic
pull in cage induction motors with eccentric rotors, part 2: experimental investigation,
IEEE Proceedings Electric Power Applications, Vol. 143, No. 3, May, pp. 202-210,
1996.
[21] M.E.H. Benbouzid, "Bibliography on induction motors faults detection and diagnosis",
IEEE Transactions on Energy Conversion, Vol. 14, No. 4, pp. 1065-1074, December
1999.
[22] Belahcen, A., Arkkio, A., Klinge, P., Linjama, J., Voutilainen and V., Westerlund, J.,
1999, Radial forces calculation in a synchronous generator for noise analysis,
Proceeding of the Third Chinese International Conference on Electrical Machines,
Xian, China, pp. 199-122, August 1999.
[23] Li, B., Chow, M. Y., Tipsuwan and Y., Hung, J. C., Neural-Network-Based Motor
Rolling Bearing Fault Diagnosis, IEEE Transactions on Industrial Electronics, Vol.
47, No. 5, October, pp. 1060-1069, 2000.
[24] Jack, L. B., and Nandi, A. K., Genetic algorithm for feature selection in machine
condition monitoring with vibration signals, IEE Proceedings Vision, Image and
Signal Processing, Vol. 147, June, pp. 205-212, 2000.
[25] Dorrell, D. G., Thomson W. T., and Roach, S., Analysis of air-gap flux, current, and
vibration signals as function of a combination of static and dynamic eccentricity in 3-
165
166
167
[48] Milrtic A. and Cettolo M., Frequency converter influence on induction motor rotor
faults detection using motor current signature analysis-Experimental research,
Symposium on Diagnostic for electric machines, Power Electronics and Derives,
Atlanta, GA, USA, 24-26 march, pp. 124-128, Aug.2003.
[49] R. R. Obaid and T. G. Habetler, Current-based algorithm for mechanical fault
detection in induction motors with arbitrary load conditions, IEEE Industry
Applications Society Annual Meeting, pp. 1347-1351, 2003
[50] M.E.H. Benbouzid and G.B. Kliman, What stator current processing based technique to
use for induction motor rotor faults diagnosis?, IEEE Transactions on Energy
Conversion, Vol.18, pp. 238-244, June 2003.
[51] Szab L., Br K.., Dobai J.B., "On the Rotor Bar Faults Detection in Induction
Machines", Proceedings of the International Scientific Conference MicroCAD '2003,
Miskolc (Hungary), Section J (Electrotehnics and Electronics), pp. 81-86.
[52] Lorand S., Barna D., Agoston, Rotor faults detection in squirrel cage induction motors
by current signature analysis, International Conference on Automation, Quality and
Testing, Robotics, May 13 15, Cluj-Napoca, Romania, 2004.
[53] Jason R. Stack, Thomas G. Habetler, Ronald G. Harley, Fault classification and fault
signature production for rolling element bearings, IEEE Transactions on Industry
Applications, Vol. 40, No. 3, pp.735-739, 2004.
[54] Jason R. Stack, Thomas G. Habetler, Ronald G. Harley, Bearing fault detection via AR
stator current modeling, IEEE Transactions on Industry Applications, Vol. 40, No. 3,
pp.740-747, 2004.
[55] Srgio M. A. Cruz and A. J. Marques Cardoso, Diagnosis of stator inter-turn short
circuits in DTC induction motor drives, IEEE Transactions on Industry Applications,
Vol. 40, No. 5, pp.1249-1360, 2004.
[56] Humberto Henao, Claudia Martis, and Grard-Andr Capolino, an equivalent internal
circuit of the induction machine for advanced spectral analysis, IEEE Transactions on
Industry Applications, Vol. 40, No. 3, pp.726-734, 2004.
[57] Lyubomir, Dimitrov V., and Chobanov V.J., Diagnosis of rotor faults of induction
motors, operated in non-rated condition, 27th international spring seminar on
Electronics Technology, 2004 IEEE, No. 0-7803-8422-9/04, pp.110-113, 2004.
[58] Jung J.H., Lee J.J., and Kwon B.H., Online diagnosis of induction motor using
MCSA, IEEE Transactions on Industrial Electronics, Vol. 53, No. 6, pp. 1842-1852,
Dec. 2008.
[59] Szab L., Tth F., Kovcs E., Fekete G, "An overview on induction machine's diagnosis
methods:, Journal of Computer Science and Control Systems, Oradea, 2008, pp.
229-234. ISSN: 1844-6043.
168
[60] Chidong Qiu, Yue Tan, Ren Guang, "An online detection method for incipient motor
faults based on multitaper spectrum estimates", 9th International Conference on
Electronic Measurement & Instruments, ICEMI '09, pp. 460 464, 2009.
[61] Frosini L. and Bassi E., Stator current and motor efficiency as indicators for different
types of bearing faults in induction motors, IEEE Transactions on Industrial
Electronics, Vol. 57, No. 1, pp. 244 251, 2010.
[62] Faiz, J., Ebrahimi, B.M., Akin, B., Toliyat H.A., Dynamic analysis of mixed
eccentricity signatures at various operating points and scrutiny of related indices for
induction motors, Electric Power Applications, IET, Vol.4 , No.1, pp. 1 16, 2010.
[63] L. Eren, and M. J. Devaney, Bearing damage detection via wavelet packet
decomposition of the stator current, IEEE Transactions on Instrumentation and
Measurement, Vol. 53, No. 2, pp. 431 436, April 2004.
[64] Szab L., Br K.., Dobai B.J., Fodor D., Vass J., Wound rotor induction machine's
rotor faults detection method based on wavelet transform, Oradea University Annals,
Electrotechnical Section, 2004, pp. 127-133.
[65] Szab L., Dobai B.J., Br K.., "Discrete Wavelet Transform based rotor faults
detection method for induction machines", Intelligent Systems at the Service of
Mankind, Vol. 2., (eds: Elmenreich, W., Machado, J.T., Rudas, I.J.), Ubooks, Augsburg
(Germany), 2005, pp. 63-74.
[66] Jose A. Antonino-Daviu, Martin Riera-Guasp, Jos Roger Folch, and M. Pilar Molina
Palomares, A method for the diagnosis of rotor bar failures in induction machines,
IEEE Transactions on Industry Applications, Vol. 42, No. 4, pp: 990-996,2006.
[67] J. Cusido; A.Jornet, L. Romeral, J.A. Ortega, A.Garcia, Wavelet and PSD as a fault
detection techniques, IMTC 2006-Instrumentation and Measurement Technology
Conference Sorrento, Italy 24-27 April 2006, pp. 1397-1400.
[68] A. J. M. Cardoso et al., Computer-aided detection of airgap eccentricity in operating
three-phase induction motors by Parks vector approach, IEEE Transactions on
Industrial Applications, Vol. 29, pp. 897901, Sept./Oct. 1993.
[69] A. J. M. Cardoso, E. S. Saraiva, M. L. Sousa Mateus, and A. L. Ramalho, Online
detection of airgap eccentricity in 3-phase induction motors, using Parks vector
approach, IEEE Industry Applications Society Annual Meeting, pp.94-98,1991.
[70] A. M. S. Mendes and A. J. M. Cardoso, Voltage source inverter fault diagnosis in
variable speed AC drives, by the average current Parks vector approach,IEEE
International Electric Machines and Drives Conference, pp.704-706, 1999.
[71] Hamid Nejjari and Mohamed El Hachemi Benbouzid, monitoring and diagnosis of
induction motors electrical faults, IEEE Transactions on Industry Applications, Vol.
36, No. 3, pp.730-735, 2000.
169
[72] Douglas H. , Pillay P., and Barendse P., The detection of inturn stator faults in Doubly
fed induction generators 2005IEEE, No. 0-7803-9208-6/05, 2005.
[73] Izzet Y O Nel, K Burak Dalci and IBrahim Senol, Detection of bearing defects in
three-phase induction motors using Parks transform and radial basis function neural
networks Sadhana Vol. 31, No. 3, pp. 235244, June 2006.
[74] Szab L., Kovcs E., Tth F., Fekete G., "Rotor faults detection method for squirrel
cage induction machines based on the park's vector approach, Oradea University
Annals, Electrotechnical Fascicle, Computer Science and Control Systems Session,
2007, pp. 234-239.
[75] Izzet Yilmaz O Nel and Mohamed El Hachemi Benbouzid, Induction motor bearing
failure detection and diagnosis: park and concordia transform approaches comparative
study, IEEE/ASME Transactions on Mechatronics, Vol. 13, No. 2, pp. 257-262, April
2008.
[76] K. Abbaszadeh, et al., Broken bar detection in induction motor via wavelet
transformation, Twenty-seventh annual conference of the IEEE, IECON 2001, Vol.1,
pp. 95-99, 2001
[77] G. B. Kliman et al., Non-invasive detection of broken rotor bars in operating induction
motors, IEEE Transactions on Energy Conversion, Vol. EC-3, no. 4, pp. 873-879,
1988.
[78] IAS Motor Reliability Working Group, Report of large motor reliability survey of
industrial and commercial installation, part I, IEEE Transactions on Industry
Applications, vol. IA-21, pp. 853-864, July/Aug., 1985.
[79] J. Sottile and J. L. Kohler, An on-line method to detect incipient failure of turn
insulation in random-wound motors, IEEE Transactions on Energy Conversion, vol. 8,
no. 4, pp. 762-768, December, 1993.
[80] S. B. Lee, R. M. Tallam, and T. G. Habetler, A robust, on-line turn-fault detection
technique for induction machines based on monitoring the sequence component
impedance matrix, IEEE Transactions on Power Electronics, Vol. 18, No. 3, pp. 865872, May, 2003.
[81] A. H. Bonnet and G. C. Soukup, Cause and analysis of stator and rotor failures in threephase squirrel case induction motors, IEEE Transactions Industry Application, Vol.,
28, No.4, pp. 921-937, July/August, 1992.
[82] T. A. Lipo, Introduction of AC machine design, Wisconsin Power Electronics Research
Center, 2nd edition, 2004.
[83] S. F. Farag, R. G. Bartheld, and W. E. May, Electronically enhanced low voltage motor
protection and control, IEEE Transactions on Industry Applications, Vol. 29, No. 1,
pp. 45-51, Jan./Feb., 1994.
170
[84] J. T. Boys and M. J. Miles, Empirical thermal model for inverter-driven cage induction
machines, IEE Proc., Electric Power Application, Vol. 141, pp. 360-372, 1994.
[85] K. D. Hurst and T. G. Habetler, A thermal monitoring and parameter tuning scheme for
induction machines, in Conf. Rec. IEEE IAS97, pp. 136-142, New Orleans, LA, USA,
October, 1997.
[86] S. B. Lee and T. G. Habetler, An online stator winding resistance estimation technique
for temperature monitoring of line-connected induction machines, IEEE Transactions
on Industry Applications, Vol. 39, No. 3, pp. 685-694, May/June, 2003.
[87] W. L. Roux, R. G. Harley, and T. G. Habetler, Detecting rotor faults in permanent
magnet synchronous machines, in Conf. Rec. SDEMPED03, pp. 198-203, Atlanta,
GA, USA, August, 2003.
[88] J. R. Stack, T. G. Habetler, and R. G. Harley, Bearing fault detection via autoregressive
stator current modeling, IEEE Transactions Industry Applications, Vol. 40, No. 3, pp.
740-747, May/June, 2004.
[89] M. A. Cash, Detection of turn faults arising from insulation failure in the stator
windings of AC machines, Doctoral Dissertation, Dept. of Electrical and Computer
Engineering, Georgia Institute of Technology, USA, 1998.
[90] D. E. Schump, Predict motor failure with insulation testing, Pulp and Paper Industry
Technical Conference97, pp.48-50, Cincinnati, OH, USA, June, 1997.
[91] D. E. Schump, Reliability testing of electric motors, IEEE Transactions on Industry
Applications, Vol. 25, No. 3, pp. 386-390, May/June, 1989.
[92] R. Maier, Protection of squirrel-cage induction motor utilizing instantaneous power and
phase information, IEEE Transactions on Industry Applications, Vol. 28, No. 2, pp.
376-380, March/April, 1992.
[93] J. Cros and P. Viarouge, Synthesis of high performance PM motors with concentrated
windings, IEEE Transactions on Energy Conversion, Vol. 17, No. 2, pp. 248- 253,
June, 2002.
[94] G. B. Kliman, W. J. Premerland, R. A. Koegl, and D. Hoeweler, A new approach to online turn fault detection in AC motors, in Conf. Rec. IEEE IAS96, Vol. 1, pp.687-693,
San Diego, CA, USA, October, 1996.
[95] F. C. Trutt, J. Sottile, and J. L. Kohler, Online condition monitoring of induction
motors, IEEE Transactions on Industry Applications, Vol. 38, No. 6, pp. 1627-1632,
Nov./Dec., 2002.
[96] J. Penman, H. G. Sedding, B. A. Lloyd, and W. T. Fink, Detection and location of
interturn short circuits in the stator windings of operating motors, IEEE Transactions
on Energy Conversion, Vol. 9, No. 4, pp. 652-658, December, 1994.
171
172
173
[127] W. Wang, Early detection of gear tooth cracking using the resonance demodulation
technique, Mechanical Systems and Signal Processing, Vol. 15, No.5, pp.887-903,
2001.
[128] W. J. Wang and P. D. Mcfadden, Early detection of gear failure by vibration
analysis-Interpretation of the time-frequency distribution using image processing
techniques, Mechanical Systems and Signal Processing, Vol. 7, No. 3, pp.205-215,
1993.
[129] Cusido, J.; Garcia, A.; Navarro, L.M.; Delgado, M.; Romeral, L.; Ortega, A.; On-line
measurement device to detect bearing faults on electric motors IEEE Instrumentation
and Measurement Technology Conference, I2MTC '09. pp. 749 - 752 , 2009.
[130] Riera-Guasp, M.; Cabanas, M.F.; Antonino-Daviu, J.A.; Pineda-Sanchez, M.; Garcia,
C.H.R., Influence of nonconsecutive bar breakages in motor current signature analysis
for the diagnosis of rotor faults in induction motors, IEEE Transactions on Energy
Conversion, , Vol. 25 , No.1, pp. 80 89, 2010.
[131] Masrur, M.A.; Chen, Z.; Murphey, Y., Intelligent diagnosis of open and short circuit
faults in electric drive inverters for real-time applications Power Electronics, IET ,
Vol.3 , No. 2 pp. 279 291, 2010.
[132] Morinigo-Sotelo, D.; Garcia-Escudero, L.A.; Duque-Perez, O.; Perez-Alonso, M.,
Practical Aspects of Mixed-Eccentricity Detection in PWM Voltage-Source-InverterFed Induction Motors, IEEE Transactions on Industrial Electronics, Vol. 57, No. pp.
252 262, 2010.
[133] Boqiang Xu; Liling Sun; Hui Ren, A New Criterion for the Quantification of Broken
Rotor Bars in Induction Motors, IEEE Transactions on Energy Conversion, Vol. 25,
No. 1, pp. 100 106, 2010.
174
175