Académique Documents
Professionnel Documents
Culture Documents
Introduction
CHAPTER-1
INTRODUCTION
1.1 Overview
The electric motor is an electromechanical device that converts electrical energy
into mechanical energy. Being a very prominent part of modern industry, the induction
motors play a vital role in major applications like pumping systems, fans, elevating
systems, electric-powered vehicles, crushers, cement plants and many more industrial
segments. An asynchronous motor, which is in fact an AC motor, where the current
required for generating torque, is induced by electromagnetic induction from the magnetic
field of the stator winding. Therefore, in general, induction motors do not need any
external mechanical commutation, individual excitation or even any self-excitation for a
part of the energy transmitted towards rotor from the stator. The rotor of an induction
motor could be of any kind like squirrel-cage type or wound type.
These induction motors are mainly employed as a critical form of devices for
major industrial activities and are in general, integrated with diverse commercially
available apparatus and applications. All the equipments based on this motor drive,
generally facilitate core competences, significant for making certain business success as
well as operator safety. Numerous electrical components of the operational induction
machines are highly prone to the system failure. For instance, the insulation breakdown in
induction motor causes the stator winding fault, which is immensely influenced by factors
like mechanical vibration, heat, functional duration or age of motor, damage occurred
during installation. Similarly, in some cases, the contamination with oil material also
causes faults in induction motors, that results in system failure. In case of a squirrel-cage
rotor, its bars might be damaged by mechanical stresses that could arise in the machine.
Meanwhile, the bearings in induction motor can be affected due to extreme wear and
smash up, caused due to improper lubrication, unbalanced loading on motor, misalignment
of bearing components with rotor, etc.
1
optimized system monitoring and supervision for induction machines. Even the reduction
of the human-machine interface also demands for an on-line detection requirements,
which can effectively diagnose the faults in induction motor without any hazards and
process disruption.
The low-cost of the induction motor and its miniaturized size, low maintenance
cost, ruggedness and flexible operation with minimum power supply makes this system
highly efficient and useful in the modern industrial process. Although these are very
reliable, they are subjected to different modes of failures/faults. While exhibiting its
operation, this device does suffer from a number of faults. These faults may be inherent to
the machine itself or due to operating conditions. The inherent faults may be due to the
mechanical or electrical forces acting on the machine enclosure. If a fault is not detected in
its incipient stage, or if it is allowed to develop further, it may lead to system failure.
There are variety of machine faults like, winding faults, disturbed parameters of stator and
rotor, broken rotor bars, eccentricity and bearing faults.
The prior detection of faults in induction motor and its optimum diagnosis
facilitates the industry to operate with the least unexpected maintenance mechanism or
industrial shutdown. This minimizes the production loss, financial wastages and even it
prohibits the catastrophic penalties. The condition monitoring and fault diagnosis
mechanisms are required to formulate a well-defined and skilled map in between motor
signals as well as indications of the fault state of the induction motor. Classifying the
conditions of motor and then evaluating the corresponding frequency of faults occurrences
from the signals have never been a simple work and in fact, these are influenced by
numerous factors in induction machine.
Several fault identification methods have been developed and effectively applied to
detect machine faults at different stages by using different machine variables, such as
current, voltage, speed, efficiency, temperature and vibrations. Thus, for safety and
economic considerations, it is essential to monitor the behaviour of motors of different
sizes. Traditionally, maintenance procedures in the industry follow two approaches. The
first one is to perform fixed time interval maintenance, where the engineers take advantage
of slower production cycles to fully inspect all aspects of the machinery. The second is to
simply act in response to the failure of the industrial plants, whenever it takes place. Then,
while making use of today’s technology, the new scientific approach was becoming
possible for maintenance management. One of the key elements of this new approach is
2
predictive maintenance through condition monitoring, which depends upon the condition
of the plant.
Considering the requirement of a highly efficient and robust system for fault
detection in induction motor, the approaches based on Fourier transform and wavelet
transform can play a vital role. The accuracy and spontaneous diagnosis potential of these
signal processing approaches makes them robust and efficient candidates for employing in
majority of applications for fault detection in induction motor. These Approaches have
exhibited better results as stand-alone applications and also with the VLSI based systems.
The presented thesis work, takes into consideration digital signal processing approaches
including DWT technique for accomplishing the goal of fault detection in induction motor.
A number of approaches and systems are there for monitoring the induction motor
functions for ensuring the higher consistency. Few leading approaches are as follows:
3
• Approaches considering the speed and torque of rotor,
• System based on chemical analysis,
• Acoustic noise measurement based techniques,
• Motor current signature analysis (MCSA) based systems,
• Artificial intelligence and neural network based fault diagnosis systems.
In spite of these above mentioned approaches and tools, a number of companies are
there which suffer from unexpected failures of the systems that ultimately results into
industrial lower productivity. Different issues like environment, functions and system
installations might cause failure of the system in its combined form. Therefore, any kind
of system optimization and enhancements could be of significant interest for all.
The extensive utilization of the automation approaches and the reduction in the
direct HMI opportunities in industries have ignited the demand for optimum condition
monitoring and fault diagnosis. On the other hand, now a day the industries emphasize
more on quality and reliable operation of systems throughout the life span of the
machineries.
The faults can be in various places like in the stator of the induction motor, its
rotor, motor bearing parts, or the external peripheral systems associated with the induction
machine or motor. Various fault detection approaches for Induction Motor are shown in
Figure 1.1. The approaches are broadly classified into three categories viz. model based,
signal processing based and artificial intelligence approaches.
4
has been widely used as a tool for the fault detection. Several AI techniques have been
developed and applied in the monitoring processes of faults, among them, the Artificial
Neural Network (ANN), Fuzzy Logic (FL) and Support Vector Machines (SVM).
Typically, expert systems are trained using a database that correlates measurement and
corresponding fault. In practical applications, the severity levels of faults may vary and not
exactly match the database used for training. This can lead to false diagnosis.
5
motor. Here, various kinds of faults have been taken into consideration for induction
motor and the respective fault diagnosis has been done.
These prominent faults in induction motor can cause single or even multiple effects, like,
(1) Unstable voltages and line currents
(2) Mechanical Vibration and noise
(3) Pulsations in torques
(4) Excessive heating of the machineries
(5) Leakage of electromagnetic flux; etc
Details of the faults and their characteristics have been discussed in chapter-2.
1. Early and accurate detection: Early and accurate detection is a prominent and
exceptionally desirable characteristic. However, the challenge in accomplishing
it depends on the fact that quick reaction to fault diagnosis and tolerable
behaviour of the system during normal operation are two conflicting goals.
6
3. Robustness: The goal of robustness is to provide rapid and reliable detection of
faults when the motor is developing a fault.
7
7. Computational requirements: There is a tradeoff between the computational
complexity and system performance. Fast on-line conclusions would need
algorithms and implementations, which are not so much computationally
complex. A sensible compromise between these two contending demands is
desirable.
This is the matter of fact that the process of induction motor condition monitoring
and its potential fault diagnosis is as old as the induction motor itself. Initially the fault
detection in induction motor was based on the conventional approaches of estimation of
over-current and over-voltage for ensuring the optimum operation. Regardless of the
availability of these tools, numerous organizations are still facing problems due to
unexpected failure of machineries with minimized motor life span [1]. Since, the works
delivered by induction motors progressed more and more complicated, the enhancements
for system was realized for condition monitoring and its fault diagnosis [2]. A number of
enhanced diagnostic approaches were advocated and control mechanisms were devised for
ensuring the fault resilient control devices.
The highly efficient and conventional design approaches have been developed for
enhancing the trustworthiness of the induction motor drive systems against a number of
faults that might take place. Unfortunately, those approaches are pricey for realization.
The development of on-line condition monitoring approaches for fault detection in
induction motors has influenced researchers in the last few decades.
8
of the affected phenomenon. They can minimize the losses incurred on finance and it
could avoid catastrophic resultants. A typical fault diagnosis system for Induction Motor is
shown in Figure 1.3.
An ideal circumstance for monitoring and fault diagnosis approach must be taking
place with the minimum measurements from an induction motor based on its signals
characteristic. A motor condition can be stated to provide a clear indication of incipient
fault modes in a minimum time. Condition monitoring has to ascertain a chart between
input as well as output results of the induction motor circumstance. In fact, the
classification of faults category and estimating the occurrence of faults are not simple,
because these are influenced by multiple factors [3].
Wavelet transform based approach has come up as a potential candidate for fault
detection and analysis in operational or on-line induction motors. Discrete Wavelet
transform (DWT), which is a vital technique of signal processing, exhibits an easy, precise
and efficient fault detection and localization approach as compared to other approaches
such as CWT. This presented work, takes into consideration the discrete wavelet transform
and fast Fourier transform (FFT) approach for fault detection and the system models have
been developed for numerous faults detection and localization. The complete research
work can be stated as a hybrid scheme for on-line fault detection in induction motor. This
system would be a kind of unique technique.
9
1.5 Fault Diagnosis Methods
A number of mechanisms and approaches have been developed for diagnosing the
faults in induction motor. Numerous approaches have come out with impressive results.
However, there persists a huge gap for further development and enhancement. This
research studies a few of the essential techniques and it has been found that the signal
processing based approach can be the optimum one that could meet all the associated
requirements for multiple faults analysis and diagnosis simultaneously.
Some important approaches being considered in this work have been presented
below.
(1.1)
(1.2)
(1.3)
10
(1.4)
In case the reference is confined in the stator of the induction motor machine
the above-mentioned expression becomes,
(1.5)
(1.6)
(1.8)
(1.9)
(1.11)
Considering the retrieved matrices of induction motor the associated curve for
faulty motor can be obtained. In case, there exists the rise in fault severity, then the graph
would be getting distorted. In this implementation scheme, there is a major issue, as it
suffers limitations for exhibiting isolation of varied faults in the squirrel-cage induction
machine. This happens due to multiple faults and problems that might create similar
deviation in the Park’s vector results.
11
1.7 Fast Fourier Transformation
In this section, FFT and its implementation with problem of fault detection in
induction motor has been presented. Fast Fourier Transformation is one of the main
algorithms used for computing the Discrete Fourier Transform (DFT) and similarly for its
inverse operation. In general, the Fourier transform converts the time domain into
equivalent frequency domain or vice versa. FFT is the potential and highly efficient
approach for performing aforementioned conversion by factorization of DFT matrix into a
product of sparse factors. This efficiency makes FFT applicable for various applications in
science and engineering.
In practical application, the execution time can be reduced with higher degree and
in such scenario, the enhancement is approximately in relation with N / log (N). Such huge
enhancements cause strengthening of the estimation of DFT pragmatic and thus FFTs
technique becomes more useful in a wide variety of applications, from digital signal
processing to algorithms for quick multiplication of large integers.
On the other hand, the FFT approach is sufficient for computing the similar DFT in
just operations. There could be great differences, specifically with respect to
huge data sets having N as in millions. Practically, the implementation of FFT causes the
reduction in computation time many folds as compared to other approaches and the
enhancements are approximately in proportion to Such potential
12
enhancements facilitate FFT to be practical and thus the FFT becomes a potential solution
for numerous applications in digital signal processing.
An FFT estimates and comes out with the similar results as is retrieved from the
discrete Fourier transformation (DFT) but there exists certain differences in terms of the
execution rate of FFT which is much faster as compared to generic DFT.
(1.12)
Estimating, the mentioned expressions in direct form, there is the need of
operations, which would come out with outputs and the individual result would need a
summation of N terms. FFT is one of the potential approaches for estimating same results
in operations. Precisely, with FFTs, algorithms need only upper
bound operations.
(1.14)
13
Eccentricity is also one of the major kinds of induction motor faults in which the
frequency analysis is made based on following expression:
(1.15)
Bearing failures
One of the most frequent failures in industrial applications is caused due to bearing faults.
The frequency analysis expression for bearing fault is given as follows,
(1.16)
(1.17)
Where, fbb0= lower frequency, fbb1= upper frequency, n = number of balls, = rotor
mechanical frequency
With MCSA, taking into consideration of various parameters as well as the slip
speed of induction motor, it is possible to estimate the frequencies for each failure.
These frequency constituents will be partially caused by different faults that may
exist and the characteristics of the motor load. The stator current has to be sampled and the
different frequency harmonics and their amplitudes should be obtained. Then, with these
parameters, the failure (the frequency) and its intensity (amplitude) can be detected.
14
components such as local, transient or intermittent. In fact, the wavelet transform might be
one of two kinds, discrete or continuous type.
A function in DWT known as the dilation function can be expressed in terms of the
tree of low and high pass filters where every individual step transforms the low pass filter
into further lower and higher frequency signal components. The illustration of such
transformation has been given in Figure 1.4. The process of signal decomposition takes
place with the input original signal by successive decomposition into several signal
components of lower resolution. In other words, the components with higher frequency are
not processed for analysis in further steps. In DWT, the components with lower frequency
of the signal are stated as approximations, while, on the other hand, the higher frequency
components are referred as details.
15
The time-frequency representation of a digital signal can be accomplished by
employing digital filtering approaches. In order to analyse the signals at varying
frequency, the filters of dissimilar cutoff frequency are employed. The input signal is
processed or passed through a succession or series of high pass filters for analyzing high
frequency signals and similarly, is passed through low frequency filters for analyzing
lower frequency signals. The overall mechanism of signal decomposition in DWT has
been illustrated in Figure 1.5.
In this process, the original signal x[n] is initially processed with a half band high
pass filter g[n] and a filter for low pass, stated as h[n]. Once the filtering is accomplished,
the samples generated are eliminated as per Nyquist’s criteria. In general, eliminating
other samples, it would further be decomposed into sub-samples of signal by two and then
the signal would be divided into half of the number of points. Now, the scale of the signal
is doubled. Here it must be noted that the filtering process makes the removal of a section
of the frequency information, but the scale of components remains unchanged.
In fact, the above discussion stated about only one level decomposition, which is in
general termed as sub-band coding. It can be repeated for further decomposition. At every
level, the filtering and then subsequent sub-sampling would result into half the number of
previous samples and therefore the time resolution is halved. In the same way, the
16
frequency band also is reduced to half that makes frequency resolution to be doubled. This
phenomenon continues until only two samples are remaining.
The frequency component, which is the most important in its original signal, would
come out to be in high amplitude in the region of DWT signal. This signal is comprised of
the specific frequencies. One of the significant benefits of DWT over Fourier transform is
that in DWT there is no loss of time localization of considered frequencies.
17
requirement, as this scheme is highly robust for performing analysis of the faults in the
induction motors.
18
1.10 VLSI Platform for Fault Diagnosis
Considering the research objective of on-line fault detection of induction motor,
the approach of digital signal processing offers robust techniques for diagnosing numerous
faults in induction motor. On the other hand, the use of Field Programmable Gate Array
can play a vital role in overall system optimization and efficiency enrichment. Considering
the vital factors viz. time, speed, memory and thus resulting performance, the FPGA
implementation in the fault diagnosis applications could be a better step to optimize fault
diagnosis in induction motor.
The important issues in existing approaches are the higher complexity and lower
performance. On the other hand, majority of new systems cannot be implemented with
conventional techniques. Therefore, a highly optimum and effective system is desired,
which can perform on a VLSI platform and can be effective with real-time embedded
system implementation. In spite of these facts, it is also required that the system must be
functional for multiple faults as the detection and diagnosis with individual faults cannot
be a better or optimum solution.
Considering all these factors and constraints, here in this research work, a number
of approaches have been explored for fault detection and diagnosis in induction motor.
The diagnosis system setups have been developed with faults like stator faults, inter-turn
faults, bearing faults, rotor bar faults and many more with different approaches.
Approaches such as FFT, Park's vector, DWT enabled MCSA system and FPGA based
systems have been explored for fault identification.
In this work, initially, virtual instrumentation based test rig has been employed for
examining system performance with various faults at different load conditions. At last, the
CompactRIO FPGA accelerated system architecture has been used for on-line fault
detection and diagnosis, in case of multiple and mixed faults. The motivating factors that
motivate to develop an optimized system setup for fault detection and diagnosis has been
presented in the next section.
1.11 Motivation
Maintenance of induction motors is one of the serious problems faced by many
industries and utilities. A number of researches have been done for the issues of
automatic and on-line fault detection in induction motor. Few of the main research works
19
and recommendations were like, Electric Power Research Institute motor reliability study
as per the reference [4], states that stator faults are responsible for 37% of the induction
motor failures. According to Neale [5], the purchasing and installation costs of the
equipments usually cost less than half of the total expenditure over the life of the machine
for maintenance. According to Wowk [6], maintenance expenditure typically presents
15% to 40% of the total cost and it can be up to 80% of the total cost.
In this research work, numerous techniques have been employed for performing
fault diagnosis and came out with the conclusion that DSP based techniques are very
effective for diagnosing multiple faults related problems. The implementation of FPGA
could be a great step towards better system architecture and effective processing with
real-time systems and its optimization. The implementation of FPGA platform for system
implementation and the algorithmic approach of signal processing might be a significant
step to optimize the overall system performance. Thus, the high robustness of Field
Programmable Gate Array implementation gave a motivation in developing a highly
efficient and effective system model for fault detection of an induction motor.
By using signal processing based approaches, like discrete wavelet transforms, one
can identify the particular phase of the induction motor, where the fault occurs, using
multiresolution analysis. Subsequently, a quantitative analysis technique could be
exploited for not only the detection of broken rotor bars fault, but also, to know the
severity of such faults in the motor.
20
1. Develop a system architecture and operation setup for generating numerous
induction motor parameters under healthy as well as faulty situations.
3. To simulate system setup in both circumstances for phase currents and other fault
indicating parameters and extract the operational characteristics.
6. To develop an optimum system for fault diagnosis in induction motor using digital
signal processing (DSP) based techniques.
4. To implement the Park’s vector model for fault diagnosis in induction motor.
21
fault diagnosis approach was developed based on Park’s vector analysis, which was
followed by the implementation of motor current signature analysis (MCSA) based
approach with DWT implementation and bearing faults detection on FPGA platform.
As a resulting research goal and ultimate work, a system with CompactRIO model
was developed for detecting numerous kinds of faults in Induction motors. Now
considering the efficiency and system optimization in terms of higher efficacy, advanced
system integration and complexity miniaturization in this research work an FPGA
accelerated system has been developed for multiple and mixed fault diagnosis in induction
motor.
Majority of the existing approaches for condition monitoring and on-line diagnosis
in induction motors employ certain single and specific fault identification approaches. In
this thesis work, a novel investigation and implementation approach was carried out for
fault detection. The developed fault diagnosis approach represents an on-line detection
mechanism for two or more combined faults in an induction motor. On the other hand,
considering the flexible and optimum performance even with miniaturized size
CompactRIO based system development has been done. In research process, a system
development was done with the Laboratory Virtual Instrument Engineering Workbench
(LabVIEW) CompactRIO, while employing FPGA environment for the continuous
monitoring of induction motor. The proposed approach intends to accept multiple sensor
signals simultaneously with the facility of the real-time and continuous monitoring
platform with custom-built fault display module. In spite of these all uniqueness, the fault
display module facilitates visual inspection facilities to the monitoring personnel.
A number of system models have been developed and proposed for detection and
diagnosis of faults in induction motor. These faults also vary with the kind of machines
22
and their operational capability. While considering the frequency of faults in induction
motor, it is found that the stator core problems are not frequent, as compared to stator
winding problems and these are not usually a major concern for small machines. Similarly,
for other faults cases, the cost, as well as efficiency gets influenced a lot.
There are numerous approaches that could be employed for detecting faults in
induction motor. However, it is found that majority of them are functions for individual
cases and even they do fail to provide ultimate solution. On the other hand, the
implementation ability of those systems with the real-time systems gives error in general
that could not be suggested for industrial process. This research work/thesis has explored
the optimization of faults detection by means of numerous technologies and system setup.
Considering the robustness of the approaches like Park’s vector analysis, characteristics
plot analysis, motor current signature analysis (MCSA), etc., a highly responsive and
effective system model has been designed for fault detection in induction motor.
The Field Programmable Gate Array (FPGA) is one of the most effective and
productive platforms to be implemented with the real-time systems that could play a
significant role in performance optimization and miniaturization. Considering this factor,
here in this work a system model has been designed with FPGA platform, with real-time
data acquisition devices.
In initial phase of fault diagnosis for stator or rotor faults, the Park’s analysis could
be a foundation for further implementation as it would deliver a responsive and highly
effective plot characterization based detection system. However, this approach could not
deliver the optimum solution. Therefore, it is required to have a system model that could
implement different fault detection and realization technique with FPGA.
23
Realizing the significant potential of the virtual model for real-time parameters, the
software tools like LabVIEW have been taken into consideration and implemented with
some other accessories like DAQ (data acquisition devices) for real-time parametric
realization. This effort made this research more valuable for coming up with better results
and performance. On the other hand, the implementation of DWT, which can deliver
better solution for fault detection in minimum time, also made the diagnosis system more
robust. The implementation of multiresolution analysis also performed well for DWT
implementation and higher accuracy but in fact, all these approaches came out with better
hope.
Now, as per the introduction of numerous techniques and approaches for fault
detection, the overall system complexity gets increased and even the overall system
outcomes get confined. The ultimate solution for this problem might be the development
of such a system that could detect multiple faults simultaneously and can diagnose the
problem on-line without causing any extra overhead on the system with enriched accuracy
and optimum performances. This became the biggest motivation for research enrichment,
and thus a system model was proposed that could retrieve the multiple data simultaneously
and can effectively diagnose the problem at an instant.
Here a novel fault detection technique has been developed that can effectively
optimize the detection or identification of multiple faults simultaneously. In order to make
this system suitable with real-time hardware, interfacing of induction motor setup has been
done with CompactRIO system that exhibits higher accuracy. The system was
incorporated with various faults in induction motor such as broken bar faults, stator faults;
vibration caused faults (Eccentricity) etc. For making it ensure that the system setup is
functional with proper load configuration, a system operating on a 2 kW induction motor
was employed that exhibited higher accuracy and efficiency in a laboratory. On the other
hand, the implementation of this system with FPGA exhibited higher performance
assurance and system setup with optimized speed, memory, computing capability and thus
overall optimum productivity.
24
bench has been developed with real-time applicable platforms like FPGA and NI
CompactRIO. This is a matter of fact that a broad research domain has been covered in
this thesis and in order to accomplish the optimum objective numerous approaches have
been implemented.
In this work, the motor current signature analysis (MCSA) approach was employed
to perform fault detection and then based on advanced techniques with various workload
configurations the research was carried out in an optimum way. System architecture has
been developed for different kinds of faults in induction motor, and accordingly, the
developed schemes have been implemented for diagnosing the faults. Like in stator
winding fault detection a computer-aided monitoring and stator current Park’s vector
approach have been developed which have been realized with the help of virtual
instrumentation. In this implementation, the research employed the data acquisition
followed by signal conditioning and analysis, which was eventually culminated in result
visualization. The same approach was implemented with broken rotor bar fault detection
and diagnosis process.
In the research implementation with stator inter-turn fault detection, a DWT
module was developed, which is based on multiresolution analysis (MRA) of three-phase
induction motor. In system implementation, induction machines were subjected to exhibit
inter-turn fault, which has been processed for diagnosis with the help of characteristics
patterns caused by inter-turn short circuit fault components in the DWT analysis while
using stator phase currents. DWT was also applied for broken rotor bars detection. All
these systems have been implemented for both healthy as well as faulty conditions, and
respective result analysis has been exhibited with operational parameters of three-phase
induction motors. The parameters such as input voltage 415V, 4 poles, 50Hz coupled with
a mechanical load were taken into consideration to accomplish result analysis.
As the ultimate research work for optimum fault identification, a noble system,
which is functional with two or more faults simultaneously, was developed. The system
model developed was constructed while keeping its applicability with the real-time
system. The uniqueness of this system setup was that, this interfaces induction motor with
CompactRIO system for identifying the faults like broken rotor bars, vibration effects
(eccentricities), leakage flux condition, bearing fault and stator winding status. For
ensuring the performance response of proposed methodology, tests were conducted on an
25
induction motor in a laboratory, which has shown highly satisfactory results and proved its
suitability for on-line detection of single and multiple mixed faults in a flexible way
through its hardware implementation in an FPGA environment.
The system setup developed in this research work is capable of handling multiple
faults in induction motor. It makes the system highly efficient for multiple fault
detection and thus it establishes itself as an optimal approach for fault diagnosis in
an induction motor.
26
Figure 1.6 Plan of Research and Thesis structure
In order to represent a crisp idea and foundation of certain research domain a well-
defined and proper introduction of research domain is essential. Thus, considering this
need, a brief of research domain and especially the foundation of induction motor with
different kinds of faults taking place and their detection approaches have been presented.
Primarily, this chapter discusses the measures to counter faults in induction motor,
different approaches, motivations for research work, research objectives, proposed system,
problem definition and research methodologies being implemented in this work have been
mentioned along with the key contributions of this thesis.
27
Chapter-2 Theoretical Background and Literature survey
This chapter provides a brief theoretical background of induction motor, the types
of faults and their effects on performance, different techniques for fault detection. The
importance of this chapter is that, it presents a domain knowledge to the readers, thus
enabling to understand the research in a better perspective.
This chapter discusses the experimental investigations performed for the diagnosis
of winding faults in induction motor. The implementation of Park’s analysis approach has
been discussed where the current Park’s vector patterns have been analyzed for healthy as
well as faulty motor. FFT based spectrum analysis has been presented. In addition, the
implementation of DWT for stator winding faults has also been discussed in this chapter.
28
implementations on an FPGA platform and LabVIEW Virtual Instrumentation have been
discussed and results are presented.
Chapter-7 Conclusions
This chapter presents the conclusions of the thesis and addresses the future scope
to continue with this line of research and development.
Bibliography
The references considered for the thesis preparation and the transparent citation for
others researchers contributions have been given in this section of the thesis.
Appendices
The details and the data sheets of DAQ devices, Sensors, RIO and cRIO are
provided in this section.
Glossary
A concise glossary of terms related to the research domain is provided.
1.18 Summary
In this chapter, the introduction to research domain, research objectives, its
motivation, and research proposal have been discussed. The research methodologies, their
formulation, primary contributions of research have been deliberated. Providing a brief
and precise introduction to the overall research process has been accomplished in this
chapter. Existing limitations in the available technologies have been reported.
Literature survey and the theoretical background of the research domain have been
presented in the next chapter.
29
Chapter 2
Literature Survey
and
Research Background
CHAPTER-2
LITERATURE SURVEY
2.1 Introduction
In order to perform any quality research, it is always highly desirable to conduct
the research objective's oriented literature survey that could facilitate an in-depth
understanding of research work. The literature review is based on the assumption that
knowledge accumulates, and that we learn from and build on what others have done. It can
provide a better platform to formulate the objectives based on gaps existing between
present day demands and existing techniques or approaches.
This chapter presents a literature survey conducted for the intended research topic,
made until the date. The significance of this chapter would be in providing a deep-rooted
understanding and knowledge transfer of existing approaches for on-line fault diagnosis in
induction motor.
Soualhi, A et al. [8] demonstrated a new approach for detection and diagnosis of fault
based on Hidden Markov Models. This approach applies pattern recognition combining
30
motor current signature analysis (MCSA) and multiple features extraction. Feature
extractions are obtained from transformations made on current and voltage signals to build
the representation space. If the representation space is well picked each operating mode
can be represented as a class. A hidden Markov model is then formed for each class and
used as a classifier for the detection of faults. The proposed approach is tested for bearing
failures and broken rotor bars detection of an induction motor of 5.5 KW. Additionally,
the efficacy of this approach is compared to a neural network based approach. The
experimental results prove that the hidden Markov model based approach is more efficient
in condition monitoring of electrical machines.
31
Aktas, M et al. [12] developed one system with IGBT based techniques for open circuit
faults detection in direct torque control of induction motor. Here they employed DWT
with Symlet2 wavelets as the mother wavelets. In their work, the stator current was
considered for identifying the faults.
Das, S et al. [13] developed an approach of Park's Vector Modulus (PVM) based fault
identification for discriminating stator winding inter-turn fault scenarios, from unbalance
voltage supply situations where both circumstances exhibit apparently same type of
current signatures. In this work, the DSP approaches like FFT and DWT have been used
for PSD estimation and have been employed for accomplishing goal of inter-turn fault
identification.
Khan, M. et al. [14] developed two DWT approaches that possess characteristics of
identification as well as its classification in 3-phase induction motor. In the first scheme,
DWT coefficient was considered as a criterion for comparison. While in another approach,
they employed details of faults in various levels to differentiate faults.
Soualhi, A. et al. [15] presented a new approach for fault detection and diagnosis of
Induction Motors using the signal-based method. It is based on signal processing and an
unsupervised categorization technique called the artificial ant clustering. The proposed
approach is experimented on a squirrel-cage Induction Motor of 5.5 kW to detect broken
rotor bars and bearing failures at various load levels.
Amaral, T.G. et al. [16] presented a new algorithm for the detection of a three-phase
induction motor stator fault. Several fault detection methods are based on the analysis of
the input current Park's vector. This diagnostic technique is based on the identification of a
specified current pattern obtained from the transformation of the three-phase stator
currents to an equivalent two-phase system. This new algorithm proposes a pattern
recognition method to identify induction motor stator faults. The proposed fault diagnosis
system is based on the statistic moment method. Simulation and experimental results are
presented in order to verify the effectiveness of the proposed method.
Gritli, Y et al. [17] developed an approach based on DWT for sensitive detection and
diagnosis of faults based on transient parametric monitoring. Here they developed system
for detection in outer cage faults. A diagnosis method based on discrete wavelet transform
(DWT) and optimized for sensitive detection under transient operating conditions is
proposed in this paper. An experimental study on a custom-built fabricated Cu double
32
cage rotor induction motor shows that the proposed method can provide improved
detection of outer cage faults particularly used in time-varying load applications.
Mohammed, O.A. et al. [18] analyzed the characteristics of three-phase induction motors
possessing internal fault conditions under supply conditions of sinusoidal and non-
sinusoidal. This encompasses two kinds of faults first rotor broken bar and second stator
faults. They investigated the terminal characteristics of the induction motor by coupling
the induction motor transient finite element (FE) model and associated external electric
circuit. For overall system development, DWT signal processing techniques was
employed.
Da Silva Gazzana, D et al. [19] demonstrates a system that works with automated system
for detection of incipient faults in induction motors. The system is constructed on the
Motor Current Signature Analysis method (MCSA) and aimed to be implemented in a
thermal electric power plant in south Brazil. First, the mechanism of fault development is
introduced and explained regarding the common faults of induction motor: stator winding
short circuits, cracked and broken rotor bars and eccentricity faults. The influence of the
load condition on the fault indicator is considered based on practical cases, procured
through fault simulations using a prototype. The vital theoretical and conceptual aspects of
the developed system are presented, comprising the signal acquisition and conditioning. In
addition, the motor signals acquired over a time period are stored in the database. Some
results from the practical application of the system are shown to demonstrate the
capabilities of the system.
Eltabach, M et al. [20] presents a comparative evaluation of different methods for non-
invasive diagnosis of mechanical faults in induction motors. Stator voltages and stator
currents, as well as noise on these signals, are simulated first for a healthy motor and then
for a motor with mechanical faults. These signals are thereafter used to calculate the
instantaneous powers ‘Pab’, ‘Pbc’ and ‘Pca’. Another method to be investigated is the
extended Park vector approach (EPVA) method. Waveforms of the simulated signals were
analyzed using the power spectral density transformation. The amplitude magnification of
these characteristic frequencies is used as a criterion to investigate noise-immunity of the
five diagnosis procedures thus enabling their classification.
Chetwani, S.H. et al. [21] describes the practicality of on-line monitoring technique for
detection of various faults that can be implemented on existing motors without
33
dismantling or shut down. The technique presented here is based on the monitoring of the
current when the machine is normally operated and analyzing the same in frequency
domain for detection of the faults. The technique can detect on-line the presence of various
faults such as broken bar in the rotor cage of induction motor, bearing faults, eccentricity
faults and stator turn-to-turn short circuit, by monitoring and analyzing the stator current.
These various faults within induction motor are simulated in the laboratory and they are
detected by on-line monitoring of current and analyzing it in frequency domain. This
technique was also applied for condition monitoring of motors at a nuclear power plant.
The technique can be used as a diagnostic tool for condition monitoring of motors.
Moin Siddiqui, K. et al. [22] In order to overcome the shortcomings of FFT and STFT
methods, the most recent powerful mathematical tool i.e. Wavelet Transform (WT) has
been used in the rotor broken bar fault detection purpose at different loading conditions.
This work investigates the detection of rotor faults in induction machines by analyzing the
starting current using a newly developed quantification technique based on the wavelet
transform. The technique applies the wavelet transform to the envelope of the motor
starting current. The envelope extraction is used to eliminate the strong fundamental
component that overshadows the characteristic features between a healthy motor and a
faulty motor with broken rotor bars.
Ghate, V.N. et al. [23] proposed a simple, reliable and economical Neural Network (NN)
based fault classifier, in which stator current is used as input signal from motor. Thirteen
statistical parameters are extracted from the stator current and PCA is used to select proper
input. Data is generated from the experimentation on specially designed 2 HP, 4 pole 50
Hz. three-phase induction motor. For classification, NNs like MLP, SVM and statistical
classifiers based on CART and Discriminate Analysis are verified.
Akin, B. et al. [24] presented a signal processing based system for fault diagnosis of rotor
asymmetry at start-up and idle mode. Regular rotor asymmetry tests are performed when
the motor is running at certain speed under load with assumption of stationary stator
current signal. The proposed method achieves the diagnosis simply by testing the rotor
asymmetry at zero speed. This test can be conducted and repeated during start-up and idle
modes. It is experimentally shown that the proposed method detects the fault harmonics at
start-up and standstill to determine the existence and the severity of faults in HEV power
trains.
34
Verucchi et al. [25] the purpose of this article comprises in reviewing the main
alternatives in the field of fault diagnosis for induction machines and to rationally compare
their performance taking into account the required diagnostic data, the number and
importance of the faults they can detect, their swiftness to anticipate a fault and the
certainty level in the final diagnosis.
Xiang Dongyang et al. [26] When the stator winding fault and the roller bearing fault
occur simultaneously, only the simultaneous acquisition of the stator current signal and the
vibration signal is effective for the diagnosis of the composite faults because the
mechanism of fault is unique. A mixed fault diagnosis method based on blind Least Mean
Square was proposed to extract the fault signal from the low frequency vibration signal of
the induction motor. Utilizing this method the fault signal was denoised and the
characteristic components for fault detection can be recognised effectively. The stator
turns short circuit fault and the bearing outer raceway fault were simulated in experiment
platform. Experimental results show that the proposed method is capable of extracting the
fault signal and improving the signal-to-noise ratio. It demonstrates the feasibility of the
composite fault diagnosis by using the vibration signal.
Osheiba, A.M. et al. [27] presents an accurate mathematical model for diagnosis of stator
winding faults in induction motors. This model is based on the axis theory and valid
for both the transient and steady-state conditions. The proposed analysis is tested for
incipient faults and the 3-phase motor currents are recorded using a DSP technique. Since
the stator turn-to-turn shorts are one of the more prevalent and potentially destructive
electrical faults in induction motors, a turn-to-turn fault is considered in such a study. An
experimental motor setup is implemented and the fault is applied at typical percentages of
a single-phase stator winding.
Zouzou, S.E. et al. [28] propose the use of Partial Relative Indexes (PRI) as new fault
indicators. Many experimental tests which are carried out on 3kW three-phase induction
motors with broken bars and inter-turn short-circuit endorse our theoretical ideas and
confirm the effectiveness of the proposed technique to detect such faults under different
load levels.
Onel, I.Y. et al. [29] presented a paper that deals with the problem of bearing failure
detection and diagnosis in squirrel-cage induction motors. Indeed, bearings deterioration is
now the main cause of induction motor failures. In this context, two fault detection
35
techniques, namely the Park transform approach and the Concordia transform, are briefly
presented and compared. Experimental tests, on a 0.75-kW two-pole induction motor with
artificial bearing damages, summarize the main features for small and medium size
induction motors bearing failure detection.
Diallo, D. et al. [30] proposed an approach that is a sensor-based technique using the
stator current measurement. A localization domain made with seven patterns is built with
the stator Concordia mean current vector. One is dedicated to the healthy domain and the
last six are to each inverter switch. A probabilistic approach for the definition of the
boundaries enhances the robustness of the method against the uncertainties due to
measurements and to the PWM. In high-power equipment where it is crucial to detect and
diagnose the inverter faulty switch, a simple algorithm compares the patterns and
generates a Boolean identity indicating the faulty device. In low power applications (less
than 1 kW) where only fault detection is required, a radial basis function (RBF) evolving
artificial neural network is used to build the healthy operation area.
Rafimanzelat, M.R. et al. [31] addresses the developing of a fault diagnosis system for
detection of broken rotor bars, a common mechanical fault in cage induction machines,
using efficient feature extraction techniques and a neural network classifier. The proposed
algorithm uses the stator current and motor speed as inputs. FFT is utilized to obtain the
frequency spectrum of the current signal. An efficient algorithm is then used to extract
suitable features out of the frequency spectrum of the signal. The relevance of the features
for the purpose of fault detection is investigated and verified. A neural network classifier
is then developed and applied to distinguish different motor conditions. A series of data
collected from experiments on a three-phase 3 HP Squirrel-cage induction machine
performed in different load and fault conditions are used to provide data for training and
then testing the classifier.
Saidi L. et al. [32] demonstrates the stator current bi-spectrum analysis to detect two types
of faults mixed, which may occur in three-phase induction motors. Based on real
experimental data, the detection study involves isolated rotor broken bars and damage in
the bearing's inner race rolling element. For multiple faults detection and for lack of
experimental data, only simulated data are used. To deal with the frequency analysis, a
mathematical model of the stator current has been derived and used into the bi-spectrum
formulas. The main contribution of this paper is the development of a theoretical method,
36
which may help the user to assess the presence of fault frequencies in induction motors in
both settings namely single or multiple mixed faults.
Bikfalvi, P. and Imecs, M. [33] stated that fault detection, diagnosis and on-line condition
monitoring of induction machines have received considerable attention in the last twenty
years. Despite various, more or less questionable practical results obtained, the topic
increased in interest especially concerning those methods and techniques, which are
related to rotor fault detection case. Although the recent trend is toward the non-invasive,
potentially sensorless methods (that uses more and more sophisticated mathematical
models in order to avoid false alarms), the present paper takes also into consideration the
traditional, dedicated techniques in use.
Prieto, M.D. et al. [35] presents a novel monitoring scheme carried out to detect bearing
faults. Apart from detecting local deficiencies i.e., single-point ball and raceway faults, it
also takes into consideration the detection of spread out defects, such as roughness. The
development of diagnosis methodologies considering both kinds of bearing faults is,
nowadays, subject of interest in fault diagnosis of electrical machines. First, the method
analyzes the most significant statistical-time features calculated from the vibration signal.
Then, it uses a variant of the curvilinear component analysis, a nonlinear manifold
learning technique, for compression and visualization of the feature behaviour. It allows
interpreting the underlying physical phenomenon. This technique has demonstrated to be
an immensely powerful and promising tool in the diagnosis area. Finally, a hierarchical
neural network structure is used to accomplish the classification stage.
Pusca, R. et al. [36] analyzes the consequences of the eccentricity and one broken bar
faults in the squirrel-cage induction motors and related the rotor speed-time variation of
the radial unbalanced force on the rotor, supported by the motor bearings. This study is
based on finite element models, which performs analyzing the independent or the
combined influence of each fault through the specific signature of faults in the magnetic
37
field outside the motor. The detection of the incipient faults based on these fault signatures
is indicated in order to predict the driving system shutdown.
Santos, F.M.C. et al. [37] describes the performance of two topologies of neural networks
for identification of short circuit in the stator windings and bearing failures. The input data
to the neural networks are statistical parameters extracted from power supplies of
induction motor. Thus, the intelligent system proposed in this paper proved to be efficient
and able to be implemented in monitoring systems failures in induction motors.
Filippetti, F. et al. [38] presented a neural network that can give, together with the rotor
fault diagnosis, the combined rotor-load inertia momentum of an induction machine. The
inputs of the diagnosis system are the spectral components of stator currents, speed and
torque. A specific neural network design containing new fast spline-based neurons with
improved generalization capabilities has been used. The training set is procured from a
faulty machine dynamic model as simulator.
Puscasu, G. et al. [39] propose a neural network based procedure for the recognition of
faults in a DC motor drive system. The analysis procedure of the faults can be used to
warn the human user when a new class of faults is detected.
Gandhi, A. et al. [40] presented a review of existing techniques available for the on-line
stator inter-turn fault detection and diagnosis (FDD) in electrical machines. Special
attention is given to short circuit fault diagnosis in permanent-magnet machines, which are
fast replacing traditional machines in a large variety of applications. As motor current is
the most commonly analyzed signal for fault diagnosis, MCSA is a topic of elaborate
discussion in this paper. Additionally, parametric and finite element models that were
designed to simulate inter-turn fault conditions are critically reviewed.
Martins, J.F. et al. [41] propose a learning algorithm for on-line generation of the
grammar productions, this formulation, then being used for modeling, control and
anomaly detection. Practical applications are described for electromechanical drives.
Grammatical interpolation techniques yield accurate results and the pattern detection
capabilities of the language-based formulation make it a promising technique for the early
detection of faulty behaviour.
Obaid Mustafa, M et al. [42] The aim of this article is to present a fault diagnosis scheme
for the case of squirrel-cage Induction Motors based on uncertainty bounds violation
38
conditions. The suggested scheme has the capability to diagnose two types of faults:
broken rotor bar and short circuit in stator winding. The fault diagnosis is being performed
through a two steps procedure. In the first step, the parameters of the healthy induction
motor are being picked out by utilizing a Set Membership Identification approach, where
corresponding uncertainty bounds are also being furnished. In the second step, proposed
bound violation conditions are being evaluated on-line during a sliding time window.
Corcoran, A.L. et al. [43] use a genetic algorithm to evolve a set of classification rules
with real-valued attributes. They show how real-valued attribute ranges could be encoded
with real-valued genes and present a new uniform method for representing ‘don't cares’ in
the rules. They view supervised classification as an optimization problem and evolve rule
sets that maximize the number of correct classifications of input instances. They use a
variant of the Pitt approach to the genetic-based machine learning system with a novel
conflict resolution mechanism between computing rules within the same rule set.
Experimental results demonstrate the effectiveness of the proposed approach on a
benchmark wine classifier system.
Yu-Seop Park et al. [44] deal with the operating characteristics of a squirrel-cage
induction motor with a broken rotor bar, stator winding inter-turn short and their complex
fault conditions. The considered operating characteristics are phase current, torque and
speed. In addition, differently from conventional fault detecting method using Park's
transformation, this paper applies the MCSA (motor current signal method) to distinguish
the fault categories more precisely. In addition, this paper constructed the experimental
setup to adjust the speed of the induction motor with a permanent-magnet synchronous
generator connected to a load bank to confirm the fault characteristics of the induction
motor based on its slip conditions.
Wegiel, T. et al. [45] presented the realization of a diagnostic system. The so-called
‘distributed system for induction motors diagnostics’ has been developed. The pilot
version of this system is installed in one of the Polish power plants. This system consists
of three layers: devices, network and management. As device and network layers, the
authors propose a tele-measuring system (TMS). This TMS supporting management layer,
which include diagnostic services (DS), is located in a specialized diagnostic center
(SDC).
39
Abul Masrur, M. et al. [46] presented a machine learning technique for fault diagnostics
in induction motor drives. A normal model and an extensive range of faulted models for
the inverter-motor combination were developed and implemented using a generic
commercial simulation tool to generate voltages and current signals at a broad range of
operating points selected by a machine-learning algorithm. A structured neural network
system has been designed, developed and trained to detect and isolate the most common
types of faults: single switch open circuit faults, short-circuits and the unknown faults.
Refaat, S.S. et al. [47] A method for fault detection and diagnosis of stator inter-turn short
circuits and unbalanced supply voltages for three-phase induction machines is presented.
The method is based on the analysis of the ratio of third harmonic to fundamental FFT
magnitude component of the three-phase stator line current and supply voltage to detect
different insulation failure percentages at different load conditions using neural network,
tested on motors with different ratings. The presented method yields a high degree of
accuracy in fault detection and diagnosis between the effects of inter-turn short circuits
and those due to unbalanced supply voltages a more significant and reliable indicator for
detection and diagnosis of stator inter-turn short-circuit faults under unbalanced supply
voltage conditions using ANN.
Somaya A. M. Shehata et al. [48] introduced a robust induction motor faults detection
using electrical signature analysis and the advantages and disadvantages of these
techniques are described. Stator phase current signature, stator current locus diagram,
stator current vector signature, partial power signature and total power signature are
selected for the simulation study applied for 4KW squirrel-cage induction motor.
Neelam Mehala et al. [49] The Motor Current Signature Analysis can easily detect the
common motor faults such as stator winding short circuit, broken rotor bars, bearing
degradation etc. Hence, MCSA is considered as the most popular fault detection method
now a day. The present paper deals with the fundamentals MCSA and condition
monitoring of the induction motor using MCSA. This paper also presents four case studies
of induction motor fault diagnosis. The results show that MCSA can effectively diagnose
abnormal functioning conditions in induction motor applications.
F. Ponci et al. [50] proposed the analysis of ac line current for the purpose of detection of
faults in the stator phase winding of an ac induction motor drive. This method depends on
40
the correlation between the wavelet decomposition coefficients of the current in healthy
and faulty conditions. The results highlight that the fault causes significant waveform
modifications that are localized at certain decomposition levels. This approach may open
the way to effective training for fault detection systems.
Sourabh Dash et al. [51] stated that the Process Fault Diagnosis (PFD) involved
interpreting the status of the plant given sensor readings and process knowledge. Early
diagnosis of process faults while the plant is still operating in a controllable zone can help
prevent event progression and lower the amount of productivity loss during an abnormal
event. PFD forms the first step in Abnormal Situation Management that aims at timely
detection, diagnosis and correction of abnormal conditions.
Ordaz-Moreno et al. [52] proposed an automatic on-line diagnosis algorithm for broken
rotor bar detection, for single low-cost field programmable gate array (FPGA)
implementation, which ensures the development of economical self-operated system. The
proposed algorithm requires less computation load than the previously reported
algorithms, and it is mainly based on the DWT application to the start-up current transient;
a further single mean-square computation determines a weighting function that, according
to its value, clearly indicates the motor condition as either healthy or damaged.
Cabal-Yepez et al. [54] This work introduces a low-cost, real-time FPGA based hardware
processing unit designed for multiple-combined fault detection utilizing information
entropy and artificial neural networks, for analyzing the information of the 3-axis vibration
signals from the machine during the start-up transient. Results show great performance of
the entropy neural system on accurately identifying in an automatic way a healthy
condition, half, one and two broken rotor bars, outer race bearing defect, unbalance and
their combinations.
41
In order to provide a better understanding of the research domain and its nature, it
is desirable to have a panoramic view of the research foundation, its key concepts and its
need for society. In this research work, the focus is on the on-line fault detection and
condition monitoring of three-phase squirrel-cage induction motor. As is common in the
literature, a three-phase squirrel-cage induction motor is referred to as ‘induction motor’
throughout this thesis. In the next sections of this chapter, a brief of the research domain
and its theoretical background have been presented.
The induction motor is the biggest consumer of electricity in the world; the term
induction motor generally refers to the squirrel-cage type. Squirrel-cage induction motor
(SCIM) is the most popularly used motor in the industry. It shares two-thirds of the
industrial power consumption and 30% of the global electricity [56] (International Energy
Agency Report).
There are three hundred millions of SCIMs installed and operating in industry,
driving various processes and mechanisms. Each year over three million induction motors
are being added to the industrial operations. Hence, induction motors are aptly called as
the 'workhorses of the industry.' This motor is the ‘numero UNO’ among all the electrical
motor types.
42
2.4 Induction Motor: ‘ Inside out’
This research work focuses on the condition monitoring, fault detection and
diagnosis of induction motors. Hence, the principle and construction of induction motors
are discussed in this section.
43
The induction motor stator is constructed from insulated laminations of electrical
steel. The laminated structure of the stator must have the cohesion necessary to transmit
the load torque and have low levels of vibration when carrying the magnetic flux. It must
be firmly clamped in cast or fabricated endplates, which are secured to be a cylindrical
frame. The core is constructed within the frame and compressed before the clamping
plates are fitted. On smaller motor, interlocking keys or even welds are used to secure the
clamping plates and on larger motor, the clamping plates are tightened by large bolts.
The stator windings of induction motor comprise conductor bars made up of hard-
drawn, higher strength copper sub-conductors that may be connected in series or parallel.
In the portion of the conductor bar embedded in the stator slot, the insulation system is
compacted by being heated and pressed, or it may be impregnated under vacuum and
pressure. In the end-winding portion, where one coil is connected to another, the insulation
system is not compacted and may be flexible.
The stator component of induction motor is composed of three parts viz., frame,
lamination core and stator winding. The first component, frame supports the stator as well
as the bearings of the rotor mechanically. Another component, the stator windings
comprise three uniformly distributed coils alongside the stator lamination core, which are
connected to the 3-phase power supply input and thus the stator is the only component that
is connected with the power input or supply voltage.
44
Figure 2.3 Rotor of a squirrel-cage induction motor
The induction motor has a laminated rotor where the laminations are clamped
together and shrunk onto the steel shaft. The rotor windings of induction motor are
constructed of hard-drawn copper and are insulated with rigid epoxy or formaldehyde
resin impregnated into woven material. On squirrel-cage induction motors the windings
may consist of insulated copper bars driven into the slots in the laminated rotor or of
aluminum bars cast directly into the rotor (Figure 2.3).
The rotor is in general, aligned on two stator bearings, which could be either
placed on individual pedestals, or it might be mounted into the enclosure of the induction
motor. The bearing components are employed on smaller-size induction motors where the
operational speeds of shaft are comparatively lower and bearing are used with
hydrodynamic oil films only for large size motors. The rotor gets its energy by induction
caused due to the synchronous rotation of the magnetic field of the stator. There is a space
remaining between stator and rotor terminal and this space is called as ‘air gap.’
45
Whenever there is the flow of current, I, through a coil, it generates a magnetic
field across the two poles, North Pole and South Pole in the winding coil. The created
magnetic field H is always in proportion to the induced current I. The magnetic field H
possesses sinusoidal characteristics of spatial distribution, which is then followed by the
inversion of polarity at the interval of half period by 180°e. Consequently, the induced
three-magnetic fields, presented as HA, HB and HC, are created whenever the 3-phase stator
current, IA, IB and IC, are flowing in the stator windings. The phase shift of 120ºe of the 3-
phase stator currents generate a respective 120ºe phase shift on the three-magnetic fields,
HA, HB and HC. The path of these magnetic fluxes remains across the laminations of the
stator and the rotor. The consequential magnetic field generated at every time instant is the
sum of the magnetic fields, HA, HB and HC, at that precise time instant.
(a)
(b) (c)
The resultant of these three magnetic fields is a synchronously rotating field with
constant peak amplitude. Consequently, this revolving magnetic field created by the 3-
phase currents implemented towards the stator windings causes the induction of electrical
currents in the rotor bars, whenever the magnetic flux from the stator windings is cut
across the rotor bars. The generated rotor currents create a magnetic field on the rotor
terminal having reverse polarity with respect to the stator. As, there is an attraction
between the poles with opposite polarity, therefore, the rotor terminal follows the rotating
magnetic field of the stator, which ultimately results into the rotation of rotor with lower
speed as compared to that of the magnetic field of stator. This difference in the speed of
rotation between the rotating field of the stator and rotor is known as the slip speed. In
order to generate the needed torque value, only a little slip speed is essential to induce the
46
essential rotor current because of the small resistance of the shorted rotor bars and
therefore, the rotor part generates a torque, which is in proportion to the product of the
currents flowing in stator as well as rotor.
In induction motor, the 3-phase AC power supply is provided where the 3-phase
currents are shifted by the phase of 120ºe or 2π/3 electrical radians [58].
(2.1)
In the presented expressions, the variable i refers to the current in phase A, i
a b
refers to the current in phase B and similarly i represents the current in phase C. I
c m
represents maximum value of individual phase current at fundamental frequency value and
ω refers to the fundamental electrical angular frequency given in unit (rad/s), φ refers to
the lagging power factor angle in radian and ‘t’ indicates the time(s). Because of the
symmetric phase shift characteristics with 120ºe interval in phase currents, the addition of
all these current components would deliver a zero value as given in the following
equation.
(2.2)
Meanwhile, the phase voltages are also in the phase shift of 120ºe. Taking into
account of the phase voltage v , as reference voltage, the 3-phase voltages can be stated
a
mathematically as follows:
(2.3)
A, v voltage of phase B and similarly v the C phase voltage. Here V refers to the value
b c m
47
of peak value of the phase voltage at fundamental frequency. In terms of polar
presentation, the 3-phase voltages can also be presented as:
(2.4)
Furthermore, because of the symmetric phase shift (120ºe) in voltages, the addition
of voltage components would be zero,
(2.5)
The 3-phase voltage system can be stated either in terms of the phase voltage
or in terms of line voltage . A relation between these variables can be stated as:
(2.6)
Whenever 3-Phase voltage is supplied to an induction motor, the phase currents are
shifted by a power factor angle, φ of 30 ºe lagging, approximately with respect to the
phase voltages, for the classes of motors up to 5 HP, used for experimentation in this
research. In the following phasor diagram (Figure 2.5) the voltage and current components
are presented.
Figure 2.5 The phasor diagram showing 3-phase stator currents and voltages
In this case the individual voltage components , and are presented as follows:
(2.7)
48
While considering these operational components, it should be noticed that the peak
voltage ( ) is proportionate to the root mean square voltage ( ) by a factor of .
Mathematically, it can be given as follows:
(2.8)
(2.9)
(2.10)
The speed with which the magnetic field rotates is generally stated as the
synchronous speed. In case of an induction motor with P number of poles, its synchronous
speed is given by
(2.11)
In the above presented expression, ‘f’ represents the frequency of stator current and
represents synchronous speed in terms of r/min.
Conversely, in real-time operation, the rotor rotates at the asynchronous speed that
is lower than the synchronous speed and thus this small difference in synchronous and
asynchronous speed of rotation is known as the ‘slip speed’. The slip speed can be
expressed as follows:
(2.12)
Where, n is the actual speed of the motor. In other terms, the slip speed can be
expressed in terms of per unit system ( ) and mathematically it is given as,
(2.13)
49
Consequently, higher the values of load torque, higher would be slip and therefore slower
will be the speed of induction motor.
Flux linkage
(2.14)
In this expression, the variable N refers to the number of turns in a coil and φ
represents the magnetic flux in Weber (Wb). Therefore, the flux linkage can be presented
in terms of Weber-turns. Considering the laws of Faraday, an electromagnetic force (EMF,
e), can be induced in an electrical circuit because of the variation of flux linkage with time
[59, 60]:
(2.15)
Here, the flux linkage is also proportional to the current i and therefore
mathematically it can be presented by,
(2.16)
(2.17)
or,
(2.18)
Now, observing the RL Circuit shown in Figure 2.6. the voltage equation can be
obtained as [61],
(2.19)
50
Figure 2.6 RL series circuit
Considering above stated mathematical relations, the following equation can be
obtained for flux linkage.
(2.20)
(2.21)
or
(2.22)
Hence,
(2.23)
Considering a load of RL series circuit operated with a two-phase AC voltage
supply as given in Figure 2.7, the voltages and flux linkages of the phases a and b can be
represented as:
(2.24)
or,
(2.25)
(2.26)
(2.27)
or
(2.28)
51
Consequently,
(2.29)
Similar approach can be implemented for accomplishing the line-to-line flux linkages in
3-phase system, for ψ and ψ . These flux linkage values can be presented as:
bc ca
(2.30)
(2.31)
The above-mentioned expressions can be employed for computing the torque.
Magneto Motive Force (MMF)
The magneto motive force (MMF) represents the strength of a magnetic field.
is proportional to the number of turns in an electric coil and the current that flows
through that coil. Therefore, the determination of the in a coil refers to the ampere-
turn of the considered coil. Consequently, single ampere-turn (AT) indicates the current of
1A that is flowing in a single-turn of coil. As a result, higher current causes higher and of
course, stronger magnetic field.
The magneto motive force in an induction motor can be presented as follows [62]:
(2.32)
In the above expression the variable stands for time, ω represents the angular
frequency or velocity, θ denotes the angular displacement of the rotor in terms of radians
and the variable F indicates the peak value of the fundamental component of the MMF
max
(2.33)
In this equation, K represents the winding factor obtained from the design of an
w
electrical motor, N refers to the number of turns connected in series per phase, P refers to
ph
the number of poles and I denotes the root mean square (rms) value of the phase current.
Torque
Torque is the force that is required for turning a mechanical shaft or machinery
multiplied by its radius (r).
Therefore, the torque (T) can be presented as follows:
52
(2.34)
Here, the variable F presents the force applied to a shaft and r refers to the radius
of the shaft as presented in following Figure 2.8.
(2.35)
and
(2.36)
In the presented expression, P refers to the number of poles and R represents line-
to-line resistance for a star (Y) connected motor. The initial integral represents the flux
linkage parameter, ψ as shown in equation 2.31 and another integral represents the flux
ca
linkage, ψ .
ab
Thus, the torque-speed characteristic plot for an induction motor can be drawn as
shown in Figure 2.9.
53
2.5 Different Types of Faults in an Induction Motor
“A fault is an unpermitted divergence of at least one characteristic property of the
system from the conventional, usual standard condition.” A fault may cause a failure or a
malfunction frequently. Faults are hard to detect, particularly if they are hidden or small
faults. The fault may evolve abruptly, like a step function, or incipiently, like a drift or
ramp-like function as shown in Figure 2.10.
Figure 2.10 Fault evolutions in motor (a) abrupt fault and (b) incipient fault
The presence of incipient faults in a motor may not necessarily deteriorate its
performance. However, they may point out that the component needs replacement before
the likelihood of major malfunction escalations. The successful detection of incipient
faults can therefore be considered as a significant challenge for the design and
development of condition monitoring algorithms. The early detection of these
deteriorating conditions in incipient phase and its elimination is needed for the prevention
of failure of the induction motors, minimizing repair costs and motor outage time.
Some of the motor drives are used in critical process control and cannot be
interrupted under the penalty of stopping the whole process. The reliability of an
adjustable speed induction machine drive is extremely important, especially if it is used
for remote or safety critical applications. For this reason, it is required to have induction
motors with higher degree of fault tolerance in numerous applications and industrial
process, where it is very important to ensure that the continuous operation of the drive
system is maintained.
54
Even though the motors are reliable electric devices, they might be susceptible to
numerous kinds of faults. These electrical faults include the inter-turn short circuit in
stator windings, open-circuit caused faults in the stator windings, faults in broken rotor
bars and some other faults like broken end rings, while mechanical faults comprise failure
of bearing and the eccentricities of rotor [63]. The Figure 2.11 shows the classification of
faults; they may be broadly classified as electrical and mechanical faults. Electrical faults
are further divided into stator and rotor faults.
Several studies have shown that 30-40% of induction motor failures are due to stator
winding insulation breakdown [64]. The materials employed for insulation purposes
generally suffer from the deterioration, because of system overloading and continuous
cycling, momentary voltage on it, stresses caused due to mechanical as well as
environmental situations. From all those possible reasons, the thermal stresses are the
foremost that cause the degradation of insulation material of the stator winding. Still the
most effective insulation also get failed hastily in case the system operates above the
threshold temperature and as per a research [65] for every 10ºC centigrade increase in
temperature, hassles the life span of the employed insulation reduced by 50%.
Due to all the above-mentioned reasons, the failures related to the stator winding could
be classified into five main groups (Figure 2.12):
1. Turn-to-turn,
2. Coil-to-coil,
3. Line-to-line,
55
4. Line-to-ground and
5. Single or multi-phase windings open circuit faults.
The open circuit fault can be caused by other reasons such as mechanical failure of
a machine terminal connector, an internal winding rupture or by an electrical failure in one
of the inverter phase legs [67]. Therefore, an open circuit fault can be considered as one of
the most common faults in induction motor drives [68]. Then while, the most unwanted
outcome of an open circuit fault issue would be a severe accident that can cause damage to
human life and an immediate shutdown of the function of drive.
The presence of these kinds of faults in the induction motors result in unstable
stator voltages and currents, oscillation in torque, reduction in performance efficiency,
system overheating, extreme vibration and the reduction in torque produced. Moreover,
these kinds of motor faults might increase the enormity of definite harmonic components.
The Figure 2.13 illustrates survey conducted in [69] that presents the approximate
percentage of different types of faults occurring in induction motors. As presented in the
56
figure, bearing faults are the frequent faults that cause most of the system failures. The
reason could be highly rugged applications and its industrial operations. Fluctuating load
parameters and respective control mechanism cause great stress on bearing that ultimately
result into system failure. In fact, majority of the industrial processes run on bearing based
infrastructure and its failure causes huge loss in terms of reduction in productivity.
Similarly, the second most effective and frequent fault is the stator fault that is followed
by faults caused in rotor bars of induction motor. Other kinds of faults comprised of
eccentricity, vibration, etc., sum up to 12% of all faults.
57
In general, the stator winding of an induction machine suffers from the stresses
caused due to a number of factors like thermal overloads, mechanical vibrations and
voltage spikes created due to frequency drives. Few of the largest causes of failures in
stator winding are:
Due to high stator core or winding temperatures
Due to Loose bracing for end windings
Due to Contaminations caused by oil, moisture, etc.
Due to short circuits
Due to starting stresses
Due to electrical discharges
Due to leakage of cooling systems
58
Figure 2.14 General damages causing inter-turn short circuit of the stator windings
in three-phase induction motors: (a) Inter-turn short circuits caused in between turns of the
same phase (b) Winding short-circuited (c) Short circuits between winding and stator core
at the end of the stator slot (d) Short circuits caused in between winding and the stator core
in the middle of the stator slot (e) Short circuit caused at the leads (f) Short circuit between
different phases.
Figure 2.15 Inter-turn short circuit of the stator winding: (a) Overhead caused short
circuits in one phase due to motor overload (b) Short circuits in one phase due to the
blocked rotor (c) Voltage transients caused inter-turn short circuits (d) Short circuits in one
phase due to a phase loss in a Y-connected motor (e) Short circuits in one phase because
of phase loss (f) Short circuits in one phase due to an unbalanced stator voltage.
[Figures 2.14 and 2.15: Courtesy of Electromotors™ WEG SA, Brazil]
59
The main reasons of the failure of the stator core have been given as follows:
Core end-region heating caused due to axial flux generated in the end-winding
region
Melting of core due to ground fault currents
Vibrations in laminations due to not secure core-clamping
Manufacturing defects in laminations
Inter-laminar insulation failure
Stator-rotor rubs during assembly and operation
Arcing from winding failure
The squirrel-cage of an induction motor is comprised of rotor bars and end rings. A
broken bar can be incompletely or completely broken. These kinds of bars might be due to
manufacturing mistakes, recurrent starts at rated voltage, stresses caused due to thermal
irregularity and mechanical stresses.
60
A broken bar might cause a number of effects in induction motors. The
manifestation of the sideband components [70-72] is one of the frequent effects of a
broken bar. The sideband components can be found in the power spectrum of the stator
current on left and right sides of the fundamental frequency component. The left side band
constituent is caused due to electromagnetic asymmetries in the rotor cage [71] and on the
other hand, the right sideband component is generated because of resulting speed ripples
caused due to resultant pulsations in torque [70, 73]. The sideband frequencies can be
expressed as:
s (2.37)
Where fbr fault frequency components due to broken bars and fs, the supply frequency.
Other effects of broken rotor bars, for classification purpose in induction motor are
the speed oscillations [73] torque ripples, instantaneous stator power oscillations and stator
current envelopes [74].
61
mechanical vibration that causes noise and ultimately results in reduced magnetic strength.
Because the air gap of an induction machine is significantly smaller as compared to other
categories of electrical machines used for similar applications, these kinds of machineries
are more sensitive to changes caused in the length of the air gap. Generally, there are two
kinds of air gap eccentricity:
1. Static air gap eccentricity and
2. Dynamic air gap eccentricity.
In the first case of eccentricity, the arrangement of the nominal radial air gap length is
permanent in space while in second scenario, the centre of the rotor does not exist at the
centre of the revolving and the arrangement of the least air gap rotates with the rotor. Until
it is detected in early stage, the eccentricity becomes much larger to generate higher
unbalances in radial forces that could cause the stator-to-rotor rub, and ultimately, it
results in a major breakdown of the electrical machinery.
Stator Current: Usually, the stator current is measured using a Hall effect
transducer. Stator phase current holds in it, the frequency components that can be
associated to a diversity of faults such as eccentricity, shorted turns in the stator
windings and broken rotor bars.
Machine Vibrations: Vibrations based monitoring is extensively used to detect
mechanical faults in motors such as bearing faults, eccentricity or mechanical
imbalance. An accelerometer is used as sensor that produces a voltage signal
proportional to the vibrations. Vibration-time signal can be used to plot FFT and
find the fault and its severity.
Axial Magnetic Flux: The axial flux of an induction motor is easily measured
using a circular inductive search coil that is placed on the non-drive end (NDE) of
the motor, placed around the shaft. A voltage proportional to the rate of change of
the axial leakage flux is induced in the coil. This signal comprises of similar
62
frequency components, which are available in the stator current. Axial flux sensing
is useful for assessing the speed as it consists of a potent component at the slip
frequency.
Stator Voltage: This can be measured by clipping a differential voltage probe to
the phase windings. It is useful in calculating the instantaneous torque and
instantaneous power.
Analyzing the spectral components of the motor induced signals is the most
common signal processing based method employed for fault diagnosis and condition
monitoring of induction motors. The fault ‘finger prints’ are unique and can be easily
recognised in the spectrum analysis. Many mechanical and electrical faults produce fault
frequencies in spectrum and fault can be diagnosed based on the knowledge of a healthy
motor parameters. The fault signals show up in a variety of sensor signals including
current, vibration and axial flux. Spectrum analysis can offer information about different
kinds of faults. Some faults induce similar fault frequencies and so necessitate other
information to discriminate them.
The fault detection algorithm uses a threshold for a particular parameter. For
example, there are standard values, which show the allowable levels of vibration
magnitudes, based on the motor size. They are empirically developed values for various
fault frequencies and their magnitudes in the spectrum analysis procedures. For broken
rotor bars, in current spectrum, if sidebands are less than -54 dB with respect to the main
peak then the motor is considered healthy, if they are larger than -45 dB then the motor
would be indicated as faulty, else it comes in marginal band.
63
1) Redundancy.
2) Fault diagnosis and fault tolerant operating strategies by taking remedial actions.
64
Different approaches for motor incipient fault detection and diagnosis have been
successfully proposed by researchers. It is important to emphasize that the use of various
techniques to diagnose different types of faults, as it appears that, no technique can cope
easily with all types of faults. In addition, some of the diagnostic techniques proposed so
far, rely on a detailed knowledge about the motor, usually difficult to obtain when the
motors are already installed and in operation.
2.12 Summary
In this chapter, an objective oriented survey has been provided to get an in-depth
knowledge transfer about the existing approaches and the significant gap between the
present needs and available methodologies. In the initial section of this thesis, a few
research works done for fault detection in induction motor have been referred, many of
them, implementing signal processing and AI technologies. In the literature survey
conducted, it was emphasized on those researches that have been developed by means of
DSP based methods. In the later section of this chapter, a brief of theoretical review about
the induction motor and incipient faults while in the running mode and signature based
fault detection strategies have been discussed.
The next chapter (Chapter-3) would discuss the research conducted and the
experimental models for detecting and diagnosing stator winding faults.
65
Chapter 3
Detection of
Stator Winding Faults
CHAPTER-3
3.1 Introduction
There is a consensus that 35-40 % of induction motor breakdowns are attributed to
the stator winding insulation [77]. Further, it is normally assumed that significant parts of
stator winding-connected failures are triggered by insulation failures in multiple turns of a
stator winding within one phase. This type of fault is known as a 'turn fault' [78]. A stator
turns fault in an AC machine sets off the flow of high circulating current and therefore,
produces undue heat in the shorted turns. If the heat proportional to the square of the
circulating current exceeds the restricting magnitude, an absolute motor failure may
emerge [79]. Nevertheless, the worst outcome of a stator turn fault may be a grave
accident causing loss of life. The organic materials employed for insulation in electric
machines are prone to significant deterioration from a combination of factors viz., thermal
overloading and cycling, transient voltage stresses on the insulation material and
mechanical stresses.
Among the possible causes, thermal stresses are mainly responsible for the
deterioration of the stator winding insulation. Thermal stresses causing stator winding
insulation failure are classified into three categories: aging, overloading and cycling [80].
Even the distinctly superior insulation may fall short if motor is run beyond its
temperature limit. As a convention, the life span of the insulation is reduced by 50% for
every 10ºC raise over the stator winding temperature limit [81]. Therefore, it is necessary
to check the stator winding temperature so that an electrical machine will not operate past
its thermal capacity. For this reason, numerous methods have been put forward [82, 83].
Nevertheless, the inherent constriction of these methods is their incapacity to perceive a
restricted hot spot at its primitive stage. Some mechanical difficulties that hasten
insulation degradation incorporates movement of a coil, vibration due to rotor unbalance,
slack or worn bearings, air gap peculiarity and damaged rotor bars.
The current in the stator winding results in a force on the coils that is proportional
to the square of the current. This force is at its greatest under transient overloads, resulting
in the coils to pulsate at twice the synchronous frequency with shifting in both the radial
66
and the tangential direction. This movement deteriorates the reliability of the insulation
system. Mechanical faults, such as damaged rotor bar, worn bearings and air gap
eccentricity, may be a cause why the rotor hits the stator windings. Hence, such
mechanical failures should be identified before they breakdown the stator winding
insulation [82, 83].
Figure 3.1 illustrates the stator winding faults in induction motor. In Figure 3.2,
photos in upper section, show a healthy winding status, while the photos shown in lower
section, exhibit the faults caused due to winding irregularity and burn outs.
67
Figure 3.2 Stator winding coils
In Figure 3.3, the stator winding fault created artificially has been shown for
demonstration. Such faults of winding irregularity and their effect on induction motor
function, have been investigated experimentally in this study.
68
3.2 Detection of Stator turns fault by Spectrum Analysis
(3.1)
The experimental setup has been shown in Figure 3.4. The 3-phase induction
motor being implemented is having parameters: 415V, 1 HP, f=50Hz, 1440 rpm, 2 pole
pairs.
The system analysis and its measurement can be processed with multiple
operational devices and the acquired data can be retrieved with the help of a highly robust
software tool LabVIEWTM. In order to obtain the sample current value in, the data
acquisition card myDAQ was employed that gets the value from the motor under load.
NI myDAQ is a data acquisition (DAQ) device that gives the ability to measure
and analyse live signals. myDAQ includes two analog inputs and two analog outputs at
200 kS/s and 16 bits, allowing for applications such as sampling a sensor output signal;
eight digital inputs and output lines. In this experiment a scan rate of 2kS/s was selected.
Fault diagnosis flow diagram is shown in Figure 3.5. In this work, speed was
computed by using the tachometer. The virtual instrument (VI) panel (Figure 3.6) was
69
developed while employing the algorithms developed in LabVIEW 8.2. The VI was
functional for the purpose of measurement as well as the processing along with continuous
data acquisition. For assessing the system while considering the practical conditions,
frequent measurements were carried out to get the stator current of the motor.
The VI data acquisition block needs the number of samples, scan rate frequency
resolution and channel information to be specified. Power spectrum subVI performs FFT
analysis of the stator current signal. Spectrum graph is displayed on the front panel of the
software. A time window of 175 ms was used for all data acquisitions for accomplishing a
straightforward and comprehensive pattern. In this experiment a scan rate of 25kS/s was
selected. Power spectrum program of LabVIEW advanced signal processing software was
used for the analysis.
70
Figure 3.6 Block diagram of the Virtual Instrument Panel
(3.2)
(3.3)
For four pole motors (p = 2), the equation will be revealed in the following series
of components:
(3.4)
These components are visible in the frequency spectrum of the stator current (as
sidebands to the fundamental peak). It can be used as a good indicator of the stator short
circuit fault. However, putting n = 1 in following equation, we can get
(3.5)
71
The Table 3.1 and the corresponding Figure 3.7 demonstrate the variation of the
motor slip value for different turn-to-turn short circuits. The slip value rises while the load
level increases. At the same load level, the slip value for 20% turns fault is smaller than
the value of 5% short turn fault.
Table 3.1 Slip value variation with load and stator winding fault
Figure 3.7 Slip value vs. load and stator winding fault
Here, Typical calculations of fault frequencies is shown for a case; when a healthy
motor is at 80% load, the measured rotor speed is 1471 rpm. Therefore, the slip,
s=(1500−1471)/1500≈0.0186, and the fault component frequency can be computed as,
72
Thus, the left and right side band fault frequencies on the frequency spectrum are
fst0= |24.53 − 50| = 25.46 Hz and (3.8)
fst1= |24.53 + 50| = 74.53Hz, respectively. (3.9)
Figure 3.8 demonstrates the frequency spectrum of the stator current data at no-
load. In this figure, the fault frequencies are not found in the side band, as the motor stator
windings are healthy. While the healthy motor operates at 100% load level, as the motor is
healthy, fault frequencies are not found in the frequency spectrum. This is demonstrated in
Figure 3.9.
73
When a turn-to-turn fault happens, the variations in the current signal are depicted
in Figure 3.10. They are the test results for 5% turns short circuit. Although there is a
slight difference with the frequencies computed and experimentally plotted on graph, the
error is within tolerance because the difference is produced by the noise of the load level.
The magnitude at fault frequency 25.36 Hz increases to -90dB. Therefore, we can
conclude that when the motor operates with a turn-to-turn fault, the amplitude of specific
fault frequencies will increase, as expected. The test results for 5%, 10% and 20% short
circuit fault of stator winding are presented in Tables 3.2, 3.3 and 3.4 respectively.
In each test, the harmonic frequencies and the corresponding amplitudes are
recorded at different loading conditions. In addition, the slip values are computed and the
variations in fault frequency amplitudes are plotted for different load levels. It may be
observed from the results that, the magnitude of the fault frequency is a function of the
load level. At higher load levels, the fault frequency components are prominently visible;
whereas at low loads, they are less noticeable
3.2.3 Stator windings shorted at 5%
Table 3.2 Variations of fault frequency components and their magnitudes
with load at 5% stator winding fault
Slip LSB Magnitude USB Magnitude
Load
(pu) (Hz) (dB) (Hz) (dB)
NL 0 1.44 25.36 -90 74.64 -90
HL 50 3.9 25.98 -88 74.03 -88
FL 100 5.5 26.38 -86 73.63 -86
74
Figure 3.11 FFT of stator current at 5% Short circuit No-Load
75
Figure 3.13 FFT of stator current at 10% Short circuit Full-Load
76
LSB [Hz] USB
26.40 74.80
26.20 74.60
26.00 74.40
25.80 74.20
25.60 74.00
25.40 73.80
73.60
25.20
0 20 40 60 80 100
0 50 100
(a) (b)
Figure 3.15 Fault Frequency variation with load at 20% shorted windings:
(a) in Lower Side Band (b) in Upper Side Band
Figure 3.16 FFT of stator current at Full-load and 20% turns shorted
3.2.6 Discussion
MCSA needs only one stator current and non-invasive. This is a simple and also an
accurate method. Harmonics in the stator current of the motor are studied by MCSA.
These harmonics are produced by new revolving flux components due to a fault. This
77
study needs only one current sensor and is based on FFT. This methodology and the
results shown above give reliable results. It is possible from this technique to detect the
fault and to assess its severity.
(3.10)
Now, substituting the equations, a final expression representing the two orthogonal
components of Park's current vector for a healthy electrical machine are given below:
(3.11)
(3.12)
Figure 3.18 depicts the Park’s transformation concept with quadrature and direct
axes in the induction machine. Figure illustrates the conventional three axes a, b and c,
transformed to a new reference frame with two axes, direct and quadrature. Stator Current
Park’s vector for ideal condition is illustrated in Figure 3.19.
78
Figure 3.18 Park’s Transformation
79
of the ellipse, the category of the fault can be identified. Thus, the presence of ellipticity
signifies with an increase of the acuteness of the fault.
The test bench consists of the induction motor and a load, National instrument's
product for data acquisition (myDAQ) and the personal computer arrangement. LabVIEW
8.2 simulation framework was employed for analyzing the system model. The 3-phase
induction motor being implemented is having parameters 415V, 1 HP, f=50Hz and 1440
rpm, 2 pole pairs and 36 rotor slots.
In order to obtain the Park’s vector pattern, the algorithms have been developed
and were employed with a signal processing module of LabVIEW virtual instrumentation
software. Figure 3.21 shows the block diagram of the Virtual Instrument Panel.
80
Figure 3.21 Block diagram of the Virtual Instrument Panel
The VI data acquisition block needs the number of samples, scan rate frequency
resolution and channel information to be specified. The induction motor was initially
scrutinized, in the absence of faults so as to find the reference current Park’s vector pattern
relative to the healthy motor. Later, short circuited motor was evaluated. A time window
of 175ms was employed for all data acquisition so as to obtain a simple and detailed
pattern. The sampling rate was 5kS/s. The block diagram of Virtual Instrumentation
graphical source code is sown in Figure 3.22.
Figure 3.22 VI developed for the detection of Faults Using Park’s Vector Method
81
3.3.2 Experimental Observations
Figure 3.23 demonstrates the Current Park’s vector pattern for a healthy motor that
is believed to be an ideal case.
A fault in the stator windings due to short circuit of turns results in varied
presentation of the Park’s vector. Here it must be noted that the current pattern of faulty
motor is significantly different from that of a healthy motor. It can be inferred from Figure
3.24.
.
Figure 3.24 Plot of Park’s current vector for a faulty machine
82
The shape of the current's phasor, as presented in Figure 3.24, is not of a perfect
circle. The elliptical shape of current’s phasor illustrates the presence of short winding
fault in the squirrel-cage induction motor.
Hence, by analyzing the patterns of the current for both healthy as well as faulty
motor, the short winding fault can be successfully detected. Experimental results
conducted in the laboratory, show the merits of Park’s vector approach on novel virtual
instruments platform. Based on Park's vector approach, different pattern recognition
techniques can be applied in order to automate the fault detection of the induction motor.
83
recognition of patterns in signal. Hence, MRA is well suited for detection as well as the
classification of faults in an induction motor.
The main concept that underlies in the application of DWT is the involvement of
the successive pairs of high and low pass filters at each scaling stage of wavelet transform.
Providing a certain sampled signal, DWT decomposes it into several wavelet signals, an
approximation signal and a detail signal (Figure 3.25). A certain frequency band is
associated with each wavelet signal. The wavelet signal reflects the time evolution of the
frequency components of the original signal, which are contained within its associated
frequency band.
84
Data acquisition of Stator current is accomplished through myDAQ and the
current samples were collected. The 1-D current signal is subjected to wavelet
multiresplution analysis.
1 t b
W [ x(t )] W (a, b)
a x(t ) .
a
dt (3.13)
where b is a translation factor indicating the locality, a is a dilation or scale parameter, ψ(t)
is known as the analyzing or mother wavelet, is an energy normalization term that
makes wavelets of different scale has the same amount of energy. In the present study, the
wavelet transform has been used to analyze the data. db44 wavelet is considered for
multiresolution analysis and it has been found that this is the optimal mother wavelet for
induction motor stator winding fault detection (Figure 3.27)
85
3.4.4 Phase Currents under Healthy and Faulty Condition
The first step to be carried out for the discrete wavelet based fault diagnosis
consists of the capturing of the currents both under healthy as well as inter-turn short
circuit condition of an induction motor. It may be mentioned that, owing to the non-ideal
filtering by the wavelet signals, it is desirable not to set the limits of the band of the
wavelet signal comprising the fundamental frequency f, very close to this frequency; or
else, this component could be filtered partially, within the neighbouring bands, masking
the development of other components within these bands due to its much higher
amplitude.
If a low order wavelet is used, the frequency response gets crappy and the overlap
between adjacent frequency bands increases. Daubechies wavelet with orders higher than
20 has demonstrated acceptable results. In this case, Daubechies-44 (Figure 3.27 ) has
been used as the mother wavelet for the DWT analysis. Flowchart of Fault detection has
been presented in Figure 3.28. Phase current of Stator A in case of healthy Condition is
shown in Figure 3.29 and in case of faulty Condition in Figure 3.30.
86
Figure 3.27 db44 mother wavelet is the most optimal wavelet for Induction Motor stator
fault detection
87
Figure 3.29 Phase current of Stator A in case of healthy Condition
The next step to be performed comprises of the analysis of the wavelet signals,
resulting from the DWT.
88
Table 3.5 Frequency Band representations for the discrete wavelet signal
Frequency band for
Level Signal
fs=2.5kS/sec
d1 1250-2500 Hz
d2 625-1250 Hz
d3 312.5-625 Hz
d4 156.25-312.5 Hz
Detail Signal
d5 78.12-156.25 Hz
d6 39.06-78.12
a6 Approximation 0-39.06 Hz
Quantitative Evaluation: Once the condition of the motor has primarily been
diagnosed, by the use of the qualitative interpretation of characteristic patterns, it is
89
possible to compute the quantification parameters defined for the corresponding fault in
order to assess to what extent the failure has occurred in the machine.
Figure 3.31 shows the sampled healthy phase current (signal s, at the top) and the
signals resulting from the DWT i.e. approximation signal (a6) and detail signals (d6… d1).
These graphs are explained below.
1) The approximation signal a6 does not show any relevant pattern once the initial
oscillation due to electromagnetic transient is extinguished. This means that there are
no significant low- frequency components (below 39.06 Hz) within the signal.
2) The detail signal d6 effectively reproduces the examined healthy current. This is on
account of the sampling frequency used. The frequency band corresponding to this
signal is [39.06, 78.12] Hz (see Table 3.5). It includes the fundamental component of
the current, which is more than 30 times larger than the rest of the components.
90
Figure 3.31 DWT analysis of stator current of a healthy Induction Motor
For the detail signal pattern d4, d3, d2 and d1, as shown in Figure 3.31, It can be seen that
no oscillations are there after the initial electromagnetic transient. This is owing to the fact
that the machine is in healthy condition and therefore, no significant patterns will be
present after the transient condition.
The previous evaluation procedure was repeated but using a motor in which the stator
winding turns are shorted artificially in different phases. Figure 3.32 shows the sampled
stator inter-turn faulty current (signal s, at the top) and the signals resulting from the DWT
i.e. approximation signal (a6) and detail signals (d6… d1).
1) From Figs. 3.31 and 3.32, it is clearly observed that the inter-turn fault can be
detected through the significant change of the approximation signal a6, we see an
obvious disturbance present after the initial electromagnetic transient was
extinguished.
91
Figure 3.32 DWT analysis of stator current of a Faulty Induction Motor
2) The detail signal d6 practically reproduces the analyzed induction motor stator
inter-turn short circuit fault current. This is because, for the sampling frequency used,
the frequency band corresponding to this signal is [39.06, 78.12] Hz (see Table 3.5)
and so includes the fundamental component of the current.
3) From the detail signals d5, d4, d3, d2 and d1, a clear oscillation was observed after
the initial transient. This oscillation is produced by a component with frequency
increasing with time. Comparing figures, in case of Figure 3.31 no significant
oscillation is appeared after the initial transient due to the healthy condition but from
the Figure 3.32, a clearly significant oscillation has appeared after the initial transient
due to winding short circuit fault in the induction motor. This is because the machine
is under inter-turn faulty condition and therefore, some significant inter-turn fault
components exist.
From the results presented and the analysis made, it is evident that the motors
subjected to stator inter-turn short circuit fault can be diagnosed by the typical patterns
92
caused by the fault components in the DWT analysis by the use of the stator phase
currents. Thus, the experimental study, analysis and results presented in this work proves
the efficiency of discrete wavelet transform technique in induction motor fault diagnosis.
3.5 Summary
In this chapter, the system development for stator winding faults detection has been
discussed while employing FFT and discrete wavelet transform as the techniques for
signal analysis and processing. Here, the system models for faulty as well as the healthy
state of induction motor have been developed. Stator inter-turn fault detection has been
discussed with the system implementation for both healthy as well as faulty conditions.
The Park vector based analysis for both the cases and the fault identification has been
performed. The DWT implementation has presented very fast and precise results.
Daubechies wavelet with orders higher than 20 has established acceptable results. In this
experimental investigation, Daubechies-44 has been used as the mother wavelet for the
DWT analysis.
The next chapter describes broken rotor bar fault detection using the various signal
processing schemes.
93
Chapter 4
Detection of
4.1 Introduction
The squirrel-cage rotor design and its manufacturing have undergone small
changes over decades [89]. Faults related to the rotor are generally associated with
magnetic stresses, thermal stresses and residual stresses due to inadequate manufacturing
and environmental stresses that are induced by moisture [90]. The initiation of rotor faults
starts with the higher resistance resulting into the higher heat generation and then develop
cracks or holes in the rotor bars [91]. These faults are more probable to take place at the
end rings.
Different parameters such as pulsations in motor speed, air gap flux, vibration and
motor current can be monitored for the detection of broken rotor bars. Early fault detection
techniques can significantly reduce maintenance costs for these motors. The condition
monitoring aims at fault diagnosis in electric motors. The spectral analysis techniques are
considered as one of the prominent techniques in the literature [92, 93].
The lower sideband is specific to a broken bar, while the upper sideband is a
consequence of the oscillation in speed. The variable represents the frequency of the
phase current and other parameters, s refers to the motor slip with k as constant k = 0, 1,
2,..., n.
(4.1)
94
Figure 4.1 (a) Healthy Rotor (b) Fault created by drilling 3 holes on rotor bars
95
above [97]. Figure 4.2 illustrates the ideal current spectrum for frequency analysis. is the
supply frequency and lower and upper side band components are also shown.
96
NI myDAQ was also employed in the place of USB-6008, for similar test benches
and experiments, which is shown in Figure 4.3(b). NI myDAQ is a data acquisition (DAQ)
device that gives the ability to measure and analyse live signals. myDAQ includes two
analog inputs and two analog outputs at 200 kS/s and 16 bits, allowing for applications
such as sampling a sensor output signal; eight digital inputs and output lines. In this
experiment a scan rate of 2kS/s was selected. The details of DAQ devices is given in
Appendix-A.
(a) (b)
Figure 4.4 shows the framework of data acquisition and analysis system. It
comprises of a current sensor signal given as input to DAQ and PC for data visualization.
A Hall effect sensor WCS2705 was used to isolate the motor phase current signal
with a sensitivity of 255mV/A (details in Appendix-B3). Applied current flowing though
the sensor generates a magnetic field which is sensed by the integrated IC and converted to
a proportional voltage.
97
Tests were carried out for different loads with the healthy motor and with similar
motors having up to 4 broken rotor bars. The rotor faults were created by breaking the
rotor bars by drilling holes into the rotor ensuring that, rotor bar is cut completely. The
measured current signals were processed using the Fast Fourier Transformation (FFT)
through virtual instrumentation. The results obtained for the healthy motor and those
having rotor faults were compared, especially looking for the sideband components.
The motor was coupled with a mechanical load to accomplish the detection of fault
at various loading conditions. In the experimental analysis process, the rated data of the
machine selected for system realization are as given in Table 4.1. Experimental setup is
shown in Figure 4.5, flowchart of LabVIEW based DAQ in Figure 4.6 and VI panel is
presented in Figure 4.7.
98
Figure 4.6 flowchart of LabVIEW based DAQ
No. of Poles 4
99
Figure 4.7 Block diagram of the developed Virtual Instrument Panel
As already stated, the presence of the slip frequency sidebands establishes the
existence of the broken rotor bars. The magnitude is the function of the number of the
broken bars. This can be ascertained by referring to the Table 4.2 in which, amplitudes of
fault frequency components for healthy and faulty motors are given. Figure 4.9 shows the
variation of fault frequencies and their amplitudes with the number of broken rotor bars. It
is evident from the plot that the number of broken bars can be found out through the
measurement of fault frequencies and their amplitudes.
100
(a) Stator current waveform and current spectrum
in healthy motor at No-load
Figure 4.8 Stator current waveform and current spectrum for 4 broken bars
101
Table 4.2 Amplitudes of fault frequency components for healthy and faulty motors
LSB USB Amplitude of Amplitude of
No. of BRBs
Frequency Frequency LSB Frequencies USB Frequencies
0 49 51 -52 -52
1 44 56 -48 -48
2 40 60 -44 -44
3 33 67 -38 -38
4 28 72 -34 -34
Figure 4.9 Variation of fault frequencies and their amplitudes with the number of broken
rotor bars
102
4.3 Wavelet Transform Technique
Wavelet transform establishes itself in applications like fault diagnosis, because of
its feature extracting capability from the raw signals. An assessment of stator current,
mechanical vibration, and auditory schemes in detecting or identifying broken bar and
bearing faults were presented in research work [98]. The MCSA method while employing
Wavelet Transform was illustrated in [99] for broken rotor bar scheme in an induction
motor. In some other work [100] the researchers employed the sensorless induction motor
control approach and they employed the Wavelet-based fault diagnosis. Pragmatic model
decomposition was considered for identifying the nonlinear system variation with the
broken rotor fault using the Wavelet Transform [101]. Meanwhile, a fault linearizing
approach stated as stepping finite element was presented, which could be achieved by
performing filtering, differentiation and transformation of the fault signal using the
Daubechies wavelet for 5 times in [102].
In this phase of the research work, a tradeoff analysis among diverse mother
wavelets has been performed for detecting the fault in rotor bars. Even with getting better
and satisfactory results for performance of DWT, it also possesses some drawbacks that
the consideration of optimal mother wavelet is rather random, not a known priori that
might initiate error in the recognition of analysis parameters. In addition, the overlap
existing between the appeared sub-bands coupled with wavelet signals coming into
existence, principally for lower order wavelet. In this thesis work, a system model which is
focused on performing analysis of the coefficients that have been derived from discrete
wavelet transform have been developed[103, 104].
103
4.3.1 Wavelet functions (Mother Wavelets)
There are a number of general wavelets, that can be used for implementing the
DWT. The characteristics of symmetry establish that, the Haar wavelet possesses linear
phase uniqueness, which states that whenever a filtering process takes place on a signal
with the mother wavelet, it would not cause any deformation of phase in the filtered data.
Furthermore, it is a very simple and basic wavelet having highest time resolution.
Nevertheless, the generated rectangular curve of the Haar wavelet makes its equivalent
spectrum with lower decay rate resulting into low frequency resolution. Another approach
could be a Daubechies wavelet, which is in fact, an orthogonal and asymmetrical wavelet.
The most prominent achievement of the approach of wavelet transform is that for
performing signal analysis, it possesses a large number of mother wavelets. From such
profusion, arises a question, how to select a base wavelet which could be the most
appropriate for performing analysis of signals. Therefore, in the following section, a
general strategy for base wavelet selection is presented.
The characteristic of symmetry ensures that a base wavelet can provide a potential
candidate for filtering linear phases. The wavelet possessing non-zero basis function in a
certain time interval is known as the compact support wavelet. Such processing facilitated
the wavelet to perform the transform in efficiently mapping with the localized features.
There are numerous wavelet families [132]; Some of them are given in Table 4.3.
104
Table 4.3 Some wavelet families
Wavelet Family Name Short Name
Haar wavelet 'haar'
Daubechies wavelets 'db'
Symlets 'sym'
Coiflets 'coif'
Biorthogonal wavelets 'bior'
Reverse biorthogonal wavelets 'rbio'
Meyer wavelet 'meyr'
Daubechies Wavelets (dbN): The dbN wavelets are the Daubechies' extremal
phase wavelets. N denotes the number of vanishing moments. Daubechies
Wavelet filters are also referred to in the literature by the number of filter taps,
i.e., 2N.
Symlet Wavelets (symN): The Symlets or symN wavelets are also known as
Daubechies' least-asymmetric wavelets. The symlets have more symmetric
features than the extremal phase wavelets. In symN, N represents the number of
vanishing moments.
Coiflet Wavelets (coifN): Coifles also exhibit vanishing moments. In coifN, N
denotes the number of vanishing moments for both the wavelet and scaling
functions. These filters are also referred to in the literature by the number of filter
taps, i.e., 2N.
Biorthogonal Wavelet Pairs (bior) Nr.Nd: While the Haar is the only
orthogonal wavelet with linear phase, biorthogonal wavelets with linear phase can
also be designed. The main feature of Biorthogonal wavelets is a pair of scaling
functions and associated scaling filters - one for analysis and one for synthesis.
Haar (haar): The db1 wavelet is also known as the Haar wavelet. The Haar
wavelet is the only orthogonal wavelet with linear phase. The Haar wavelet is also
the simplest wavelet.
105
(a) (b) ( c) (d)
Initially, the transient stator current signals were collected from induction machine
for testing, which was followed by the signal preprocessing. Furthermore, the raw data
were processed with Matlab software using its wavelet toolbox for decomposing the
obtained time domain signal into time-frequency domain. This is then followed by the
extraction of fault features waveband from all proposed wavelet transform levels.
Finally, the entity diagnosis outputs were applied for validating the developed
model. The experiments were conducted with a healthy motor as well as a motor with
drilled bars. The fault in the rotor was created by drilling holes in it. Later, the stator
currents of the motor were sampled by a Hall effect sensing element, that is positioned in
one of the phase line current wires. The stator current was sampled at the frequency of
1.92 KHz rate and was interfaced to a PC by using a data acquisition system.
The observational quantities were measured for its healthy as well as three broken
rotor bars with respect to varying load parameters. The motor was operated with
mechanical loading arrangement. The flowchart showing the wavelet analysis and the
steps for mother wavelet selection are given in Figure 4.11. The frequency levels of the
wavelet decomposition performed in this experiment are shown in Table 4.4 and the
decomposition is illustrated in Figure 4.12.
106
Figure 4.11 Flow chart to determine the best mother wavelet
107
Table 4.4 Frequency levels of wavelet for a sampling frequency of 1.92 KHz
db8 db9
db10 sym7
Figure 4.13 Wavelet candidates considered in the experimentation along with sym8
108
Wavelet function ψ
Scaling function φ
109
evolved when broken rotor bars were present in induction motor. While, these behaviours
are not so simple for operators to analyse, as there exist very small parametric value
changes for performing separation. In order to solve such issues, as per above cited facts,
the two indices were used for diagnosing the broken rotor bars in induction motor.
Wavelet decomposition analysis was performed using Matlab tool for healthy and
faulty motors with 1, 2 and 3 BRBs. Figures 4.16 shows the wavelet analysis performed
for 3 BRBs. Two performance indices, Index 1 and Index 2 are defined as follows, to
determine the optimum wavelet (Eq 4.2 and 4.3).
ST
Index1 = (4.2)
Mean
Variation coefficients
Index 2 = (4.3)
Mean
Tables 4.5 to 4.8 give the results of wavelet decomposition for healthy and faulty
conditions of motors.
Table 4.6 Indices for different wavelets under one broken bar
110
Table 4.7 Indices For Different Wavelets under Two Broken Bars
Table 4.8 Indices for Different Wavelets under Three Broken Bars
111
Figure 4.17 Index 1 for different mother wavelets
The value of index 1 augmented by incidence of the fault in the bars of the rotor
poses rising trend as shown in Figure 4.17. Considering the above-mentioned Tables 4.5 to
4.8, the value of index 1 for a healthy motor is approximately 47% and for the motor with
1 BRB at 50%. Values of Index 2 (Figure 4.18) for different number of broken bars raises
from 2% in 1 BRB to 3% in 2 BRBs. Also it can be observed that this index increases
from 3.1-3.4% in 2 BRBs to 4.6% in 3 BRBs.
112
convenient index for performing diagnosis even with multiple numbers of broken bars in
induction motor.
Results demonstrate that the experimental results for three rotor broken bars, are in
good agreement with the expected results. In order to increase the preciseness or accuracy
of fault detection, the optimum wavelet was considered. However, there exists a lack of
definite rule to state on how to select the precise wavelet. Majority of the inquiries are
depending on the trial-and-error approach. In this research work, for getting better and
more accurate results, the Daubechies db8, db9, db10 and Symlet sym7 and sym8
wavelets were used for identification of broken rotor bars. Results obtained indicate that
sym8 has performed better for wavelet decomposition in fault diagnosis. This is
demonstrated in Figures 4.17 and 4.18.
This research work has come up with certain positive results and analysis, which
could make the system more robust for diagnosing on-line, the broken rotor bar faults.
4.4 Summary
In this chapter, a system for fault identification in rotor bars of induction motors
was developed. Experimental investigations were conducted on 3-phase squirrel-cage
induction motor for the detection of faulty rotor bars. The LabVIEW based spectrum
analysis and Matlab based system models were developed, and performance with both
healthy as well as faulty situations is presented. In wavelet analysis, sym8 was found to be
the optimal mother wavelet for fault detection.
The next chapter describes the mechanical faults and identification of bearing
faults using a reconfigurable system.
113
Chapter 5
Detection of
Bearing Faults
CHAPTER-5
5.1 Introduction
There are a number of faults, which arise due to mechanical abnormal behaviour or
deformation. In this section of the thesis, the system model developed for on-line
diagnosis of mechanical faults in induction motors has been discussed. The most common
fault in induction motor is the bearings fault and the diagnosis of bearings fault has been
presented in this chapter. To experimentally investigate and to accomplish the goal, a
signal processing based model has been developed.
114
5.2.1 Cage and Ball Defect
The bearing cage in a ball bearing bears balls at evenly situated berths and
facilitates the confined rolling of the balls alongside the racetracks. While the motor shaft
is rotating, the bearing cage rotates with a steady angular velocity, which is average of the
inner and outer race angular velocities. The angular velocity of the cage can be utilized to
work out the value of dominant fault frequency because of the cage defect. The fault
frequency due to cage defects, fbbc has been given below:
(5.1)
where,
= Angular speed of the inner race
= Angular speed of the outer race in rpm
= Diameter of Ball
=Contact angle of Ball
= Radius of Inner race
= Radius of Outer race
The outer race is attached to the casing, which is stationary. The shaft and inner
race are mounted together and both revolve at the same angular speed. Consequently, it
can be assumed that:
(5.2)
115
Where, ωr = Rotor angular speed in rpm
Incorporating the above-mentioned assumption, we get
(5.3)
Empirically, the fundamental frequency due to cage defect for a specific ball
bearing containing 6-12 balls in it is given as:
(5.4)
Where, (5.5)
(5.6)
Here, the expression for fbbi of a ball bearing possessing 6-12 balls can be presented as
follows:
(5.7)
Figure 5.3 shows the ball bearing in which fault is created artificially by drilling
hole in the inner race
The frequency of the occurrences of the outer race defect (fbbo) solely relies on the
rate with which bearing balls wrap the point of imperfection on the outer race. The
individual ball moves across the defect point with a rate, which is associated with the
difference of angular speed of the cage and its outer race. The elementary fault frequency
connected to the inner race defect can be expressed mathematically as follows:
(5.8)
Figures 5.4 shows the ball bearing in which fault is created artificially by drilling
holes in the outer race.
116
Figure 5.3 Fault created artificially by drilling hole in the inner race
Figure 5.4 Fault created artificially by drilling hole in the outer race
In general, there are two main kinds of analysis, which are carried out for
accomplishing the goal of fault identification in induction motor. The first is time domain
while another is the frequency domain. In function, the approach based on frequency
analysis presents better results, as it can provide intricate information about the operational
status of the induction motor. On the other hand, the time domain based system analysis
can provide qualitative information about the machine condition.
In general, a fast Fourier transform (FFT) is employed to perform analysis of
machine vibration. In case of higher degree of random vibrations and the noise, the
imprecise information can be generated about the condition of on-line machine. The
factors like noise and random vibrations can be inhibited from the vibration signal using
signal processing tools such as finite impulse response (FIR) filters, averaging, correlation
117
and convolution. Figure 5.5 shows the illustration of vibration signature frequencies in
LabVIEW program for outer race fault, inner race fault and healthy bearing.
In this case study, an experimental setup for the identification of bearing faults was
developed. The virtual instrument system was developed using LabVIEW and this model
was embedded on FPGA using RIO kit.
Field Programmable Gate arrays are the reconfigurable platforms for system
development and realization. Unlike ASIC technique, FPGAs are the tools that can be
reconfigured as per requirement in the future. The programmable logic blocks of
FPGA are called Configurable Logic Block (CLB) and are illustrated in Figure 5.6. The
use of FPGA facilitates reprogramming and customizable system development in various
steps that suits the system development as per the present needs and for further
enhancements. The miniaturization of the system by employing FPGA strengthens the
optimistic view to emerge with real-time application systems. A system-on-chip (SoC)
developed on FPGA platform can be tailor-made to the requirements of a real-time
application development.
118
Figure 5.6 Configurable Logic Blocks of FPGA
119
FPGAs, the development of hardware prototypes is significantly accelerated. In
addition, because of the early availability of hardware prototypes, time-consuming
activities like the start-up and debugging of the hardware are advanced
simultaneously with the overall development.
120
constructs can be used suitably, at lower levels of abstraction. Lower-level constructs are
also productive when there is a need to optimize performance at circuit levels.
The Figure 5.8 illustrates the setup for presenting the test environment employed
for the development of virtual instrumentation and reconfigurable system. Basically, it
consists of an RIO kit, an amplifier, an efficient signal conditioner equipment
arrangement, an accelerometer for vibration measurements and a three-phase induction
motor for system model realization. The experimental setup is shown in Figure 5.9 and
the schematics flow diagram of the fault diagnosis is illustrates in Figure 5.10.
Figure 5.8 Illustration of test rig used for the development of reconfigurable system
121
Figure 5.9 Experimental setup
5.7 Diagnostic System
The diagnostic system has two primary blocks, the virtual instrumentation system
and the RIO based FPGA embedded system. The VI system comprises of an
accelerometer interfaced with the DAS. The system includes an anti-aliasing filter, which
is integrated with the accelerometer. This filter confines the acceleration signal to a
bandwidth of 750 Hz, allowing for a sampling frequency of 1500 Hz. The FFT comprises
of 1024 points, rendering a frequency resolution of 1.46 Hz. The frequency range and the
number of periodograms are selected depending on the type of failure to be analyzed.
122
The FPGA embedded system holds different blocks: drivers for controlling the
communication between the sensor and the FPGA, FFT and periodogram blocks for
obtaining the spectrum of the signals and a decision making unit for postprocessing of the
resulting spectra and furnishing an automatic diagnosis of the motor state. The FFT in the
FPGA has 1024-point resolution. The user can select from a range of averaged
periodograms, which is a feature of the FPGA reconfigurability.
LabVIEW FPGA compiler translates LabVIEW code into VHDL and invokes the
Xilinx compile tools to generate a bitfile, which runs directly on the FPGA (Figure 5.13).
123
Figure 5.12 Program Flow chart
124
Figure 5.13 Code translation with high level abstraction
Figure 5.14 shows a spectrum of vibration signal demodulated with the detection
of the characteristic frequency of outer-race fault. It can be observed that the desired
frequency constituents can be distinguished easily. If the time interval between
periodically happening peaks in the envelope curvature match one among the critical
frequency characteristics of bearing damage, then the matching bearing component will be
indicated as damaged.
125
Figure 5.14 Power spectrum analysis of the vibration signals
Table 5.2 Bearing Frequencies for SKF 6311 rotating at 1485 rpm (24.75 rps) at No-load
and at 1380 rpm (23 rps) at Full-load
Fundamental train
0.382 X 24.75=9.45 Hz 0.382 X 23=9.36 Hz
frequency (cage)
126
Figure 5.15 Characteristic frequencies of bearing faults for half and full load.
127
This main VI acquires data continuously from the designated channel and carries out
the scaling of the data to appropriate engineering units. Then program performs
windowing of the signal and the frequency measurement. Power Spectrum, Power Spectral
Density and FFT averaged periodogram are available frequency measurements. The
resulting spectrum is displayed on a waveform graph as shown in Figure 5.17. The VI
returns the frequency resolution and the period of the acquisition time, based on the scan
rate and the number of samples. There is a display module on which the results of the
diagnosis procedure are displayed.
128
5.18. It is evident from the plots that normal bearing shows no prominent peaks, where as
faulty bearings show fault frequency peaks.
Figure 5.18 Cepstrum plots of a bearing (a) with no fault (b) an outer race fault and (c) an
inner race fault
129
This chapter discussed a comprehensive approach for modeling, simulation and
development of an on-line reconfigurable vibration analysis tool based on FPGA device.
FPGAs offer the maximal DSP performance available on a programmable platform, but
optimizing a DSP algorithms in an FPGA could be challenging. Until recently, the
algorithms necessitated to be ported to HDL and then RTL operational simulation would
be asserted to employing the high-level simulation tests. The set of design tools used, NI
RIO based design abstraction and productivity. This approach employs a high-level
behavioral description of the DSP algorithm. The results obtained were consistent with the
motor faults generated on a bearing, and consequently validate the proposed strategy.
Thus, this implementation, can be replicated and deployed on industrial plants.
5.9 Summary
In this chapter, the NI RIO FPGA based mechanical fault detection of induction
motor was realized and the system was specially developed for Bearing fault detection.
For realization of bearing fault detection, a LabVIEW RIO test setup was developed and
accordingly, a threshold based fault detection system was implemented. The system was
tested with various operational parameters and results were obtained for different faults.
The system exhibits efficient results and establishes the robustness of the strategy.
In the next chapter, a reconfigurable FPGA based embedded system has been
presented based on NI cRIO embedded module. It deals with the multiple and mixed faults
detection of induction motors.
130
Chapter 6
6.1 Introduction
This chapter can be considered as the concluding work and hence the most precise
and advanced system implementation of the research work. In this chapter, the overall
system development, which can diagnose the faults of various kinds, has been discussed
and this approach establishes itself as the hybrid fault diagnostic system that can take care
of multiple kinds of faults. The specific contribution of this work is the embedded system
implementation that ultimately makes the system to be employed with real-time situations.
On the other hand, Field Programmable Gate Array has been used for system modeling,
and thus the optimum system has been realized for simulation as well as the real-time
implementation of the system.
131
avoid random and catastrophic failures in most of the applications. Consequently, the issue
of continuous monitoring and non-invasive diagnosis of these induction machines is of
high concern, and is becoming increasingly significant [108].
In the last two decades, a large amount of research work has been reported on
using the stator current spectrum to sense the rotor faults associated with mechanical
unbalance and rotor broken bars [109]. These techniques require highly skilled
operator/user in order to distinguish a normal operating condition from a potential failure
state. This is so, as the monitored spectral components (either current or vibration), are
based on a number of factors, including normal operating conditions [110, 111]. In [112]
the power supply and motor faults are detected from electrical parameter measurements
through a membership function method based on a pattern recognition solution has been
reported.
The wavelet transform techniques are other known techniques, which have proven
their efficacy in induction motor fault detection, reported in [113, 114]. These techniques
are associated with time-frequency information from non-stationary signals. Artificial
intelligence techniques have also played a significant role in fault detection and diagnosis
of induction motors. For instance, neural networks techniques [115] are employed for the
determination of faults in bearings. In [116], the severity of the broken rotor bars and load
level identification of the system by fuzzy logic and genetic algorithms have been
discussed. In [117] and [118] made an effort in employing information entropy technique
for the detection of rotor broken bars. Unfortunately, these techniques are limited to single
isolated fault identification because of their inherent nature [118].
132
monitoring platform with standard custom-built display module. In addition, this proposed
system has a special feature of visual inspection opportunity for staff in condition
monitoring.
bbo (6.2)
Where, ‘B ’ and ‘P ’ are, respectively, the ball diameter and the pitch diameter, β is the
contact angle between the balls and the ball bearing rings and fr the rotor frequency.
The bearing faults can also be detected using the current, flux and voltage signals
can also be used to detect the bearing faults.
fbb=(fs±m.fv) (6.3)
Where, fbb denotes bearing fault frequencies, m an integer and fv, the vibration
frequencies.
133
6.3.3 Stator Faults:
The stator current related faults can be identified by using the fault frequency
measurement, which can be observed by the FFT analysis of the voltage, current and axial
flux signals [119-120, 122].
(6.5)
Where, ‘p’ is the number of pole pairs of the induction motor and ‘k’ is an integer. It has
been reported that, 2.fs or a combination of rotor frequency fr, 2.fr and 2.fs can be an
indication of stator faults in vibration based monitoring.
(6.6)
Where,
= eccentricity frequency, = supply frequency, r = number of slots
p = pole pairs, n = 1, 3, 5…
The Eccentricity faults can also be detected using the current, flux and voltage
signals.
6.4 CompactRIO based Condition Monitoring System
The proposed condition monitoring of induction motor is implemented in FPGA
environment for obtaining on-line operation with early detection of faults. It is
recommended that the CompactRIO is the most suitable system for real-time processing
units [119,124], which can be easily reconfigured and provides high-speed parallel
operation of tasks. The optimized program has been developed for condition monitoring of
induction motor in the LabVIEW environment, which significantly accelerates the
execution time and provides an open architecture for further modifications [125]. Figure
6.1 shows the arrangements of measurement, data acquisition and analysis modules for
fault detection purpose.
134
Figure 6.1 Measurement, Data Acquisition and Analysis Modules
Figure 6.2 shows the block diagram of various components of fault detection and
condition monitoring system. In this system, signal outputs from sensors are fed to the
signal conditioning blocks. There are eight different signal input channels, which are
multiplexed through ADC for necessary processing and diagnosing. Figure 6.3 shows the
experimental setup of CompactRIO based Induction Motor Diagnosis system.
135
Figure 6.3 Experimental setup of CompactRIO based Induction Motor Diagnosis
Figure 6.4 shows the complete block diagram of CompactRIO based condition
monitoring system of induction motor. In this system, CompactRIO mainly consists of RT
controller 9004, Signal processing module and FPGA 9104 unit. Suitable sensors are
employed for data acquisition from the machine and for further analysis in CompactRIO
system.
136
Figure 6.4 Block diagram of CompactRIO based Induction Motor System with sensor
positions
reconfigurable embedded chassis with FPGA core that can be accessed and
configured using LabVIEW graphical development environment;
In this system (Figure 6.6), CompactRIO performs averaging, fast Fourier transform
(FFT) analysis, windowing, peak and fault detection analysis. In this, the data transfer has
been executed effectively through the network port. However, read/write instruction,
execution time-critical loop tasks, reading data from machine status has been carried out
by FPGA Chassis. The AI module is employed for capturing vibrations, current, voltage
and flux signals using suitable signal conditioning devices and sensors. (Sensor details are
provided in Appendix-C)
137
Figure 6.5 Components of CompactRIO
6.5 Resources
The cRIO platform features a small, rugged, scalable, modular, embedded
architecture comprising of an FPGA, an RT processor and isolated I/O devices [124].
cRIO has industrial class certification for high temperatures, hazardous locations or
potentially explosive environments and shock capability. It is a good candidate to be used
in an industrial environment.
138
The CompactRIO architecture has similarity to that of the desktop systems but
with FPGA plug-in boards. The floating-point processor that can be programmed with
LabVIEW Real-Time is connected to the FPGA on the backplane through an internal PCI
bus. The CompactRIO modules consist of conversion circuitry (for analog modules),
electrical isolation, signal conditioning, permitting direct connection to sensors on the
motor.
Real-time system development is a multistep process that includes (i)
programming, (ii) debugging, (iii) compiling, (iv) downloading and (v) deploying.
LabVIEW Real-Time is a unique facility, as it creates advanced real-time applications but
takes advantage of all the benefits of LabVIEW for Windows graphical programming
environment. While developing real-time software, performance must be ascertained at
each step of process. The execution speed of each component requires to be measured to
ensure that they match their expected performance criteria. During system integration, the
performance of the system needs to be continually checked for the same reason.
All the input channels fulfill the necessity of sensor signals that were applied in the
condition monitoring system. This module is suitable for condition monitoring
applications because it is equipped with eight analog input channels. The analog channels
can capture 10.8V maximum amplitude signals. In addition, this module also contains an
analog-to-digital converter (ADC) and noise-immunity system, which are preferable
features in signal measurements for condition monitoring. This work investigates the
detection of multiple faults (i.e. stator shorted turn faults, eccentricity faults, bearing faults
and broken rotor bar faults) using three types of sensor signals (i.e. current, vibration and
leakage flux) under varied load levels. Appendix-B provides details of the sensor types,
signals and associated bandwidths.
The uniqueness of this work is the advancement of a highly reliable, low-cost,
multichannel multisensory analysis platform, based on the spectrum estimator, capable of
providing an automatic diagnosis of the motor state. This system carries out on-line
continuous monitoring with the help of a post-processing module that automatically
assesses the estimated spectrum, accordingly to the specific failure under diagnosis. All
the approaches mentioned before, require the usage of FPGA technology that allows
developing low-cost, rapid and reconfigurable architectures for real-time analyses [128].
Hence, the primary contribution of this work is the utilization of an FPGA based cRIO
module for developing an on-line multichannel spectrum analysis based continuous
139
monitoring of multiple sensor signals with an embedded postprocessor for an automatic
diagnosis of the machine state.
Figure 6.7 Block diagram of the Virtual Instrument Panel with FPGA
The main virtual instruments panel comprises of subVIs for the detection of
various faults. SubVIs are custom designed for various faults to be detected by the cRIO
real-time monitoring software (Figure 6.8.). The cRIO condition monitoring system is
used to detect broken rotor bar, eccentricity, stator shorted turns and bearing faults (Figure
6.9).
140
Figure 6.8 Block diagrams of SubVI panels
Figure 6.10 shows flowchart of the developed software for fault identification and
detection. As it is depicted in flowchart, FPGA performs multiple operations
simultaneously, such as, acquiring the 8-channel data & performing low pass anti-aliasing
141
filtering and transferring the data via direct memory access (DMA). It also drives the
custom-mode display module via SPI.
The Induction machine data analysis is mainly performed in the real-time controller. In
this execution, the process sequence is as follows: initialization of data, machine data
calibration from relevant sensors, followed by FFT analysis, peak detection of fault
142
frequency and frequency averaging [127]. Once the FFT analysis of induction motor is
monitored, a number of subroutines are executed to identify exact faults based on the valid
fault frequency equations described in section 6.3. With this, the fault level algorithms are
implemented to estimate the specific fault levels based on the peak values of the fault
frequencies. Figure 6.11 shows the part of the program showing different signal evaluation
channels.
Figure 6.11 Part of the program flowchart showing different signal evaluation channels
The fault detection algorithm uses a threshold value for a particular parameter.
There are standard values, which show the allowable levels of vibration magnitudes based
on the motor size. There are empirically developed values for various fault frequencies
and their magnitudes in the spectrum analysis procedures. For broken rotor bars, in current
spectrum, if sidebands are less than -54 dB with respect to the main peak then the motor is
considered healthy, if they are larger than -45 dB then the motor would be indicated as
faulty, else it comes in marginal band. Threshold values for the peaks of side band
frequencies are stored in the program data for comparison and decision-making.
In the spectrum analysis, graphs fed from the vibration sensors, several fault
frequency amplitudes can be identified and the rotor and stator fault frequency amplitudes
can be also be identified using the axial leakage flux sensor. A block diagram of the
diagnosis system, with different sensors and their positions are illustrated in the Figure
6.4. The system monitors the motor continuously, using a set of current, voltage, vibration
and flux sensors (See Appendix-B). If a fault is detected, the display module alerts the
operator and displays the fault type using an LCD panel.
143
To meet the application requirement, three basic modules are embedded in
CompactRIO base unit: (1) CompactRIO chassis (2) CompactRIO RT embedded RT
controller. RT controller cRIO-9004 can perform real-time stand-alone execution. It
provides 512 Mb of nonvolatile storage memory, 64 Mb RAM and a 200 MHz Pentium
processor. The Ethernet port is connected to a PC for reengineering and uploading FPGA
bit files and the developed condition monitoring software. The cRIO RT controller is
deployed to perform real-time implementation of the condition monitoring application,
including windowing, FFT analysis, averaging, peak detection and fault analysis. The
display module uses SPI protocol to communicate with the FPGA.
In this experimentation, FPGA cRIO-9104 (3 million gate sizes, eight I/O module
slots, Xilinx Virtex II XC2V3000-4FG676I) chassis is employed, with the signal
processing routines. Moreover, the FPGA Chassis reads sensor signals directly from the
AI module. The FPGA writes the status of the motor to a custom-design display module
through SPI triggering technique and it gets data and transfers it through the PCI bus to the
RT host controller. Many techniques are implemented for the host synchronization, such
as interrupt request, polling and DMA. These procedures are programmed through the
LabVIEW FPGA environment.
Signal [V] contains the signal generated from a sensor expressed in equivalent
millivolts. The sensors like Hall effect current sensor, flux sensor and accelerometers
outputs are in millivolts. However, the current, flux and vibrations have their own
engineering units. Hence, there is a need for appropriate calibration, mapping and scaling
144
between the sensor output voltages and their real units. For example, vibrations are being
measured, using an accelerometer, select g, m/s^2. If the units needed are not listed, it can
be customized by using the custom label. This process is usually referred to as scaling of
engineering units. This is performed by VEU conversion VI.
Channel info contains data about the measurement system used, ahead of the signal
reaches the DAQ instrument. Sensor sensitivity [mV/EU] indicates the sensitivity of the
sensor in mV/EU.
The Output labeled as scaled signal [EU] gives back the scaled signal expressed in
the preferred engineering units. Scaled signal [EU] also incorporates channel info as a
property of the waveform. Unit label returns the selected engineering units. Error out
contains error information. If 'error in' indicates that an error occurred before this VI, 'error
out' contains the same error data. Otherwise, it reports the error status that this VI or
function generates.
145
This VI reads data from the file specified. The sensor data is scaled to engineering
units as explained in the previous section. Then, the power spectrum of the data is
determined over a subset of the baseband stretch and the spectrum peaks are picked out
and logged in a table. The peaks are recognised using an algorithm that accounts for the
time domain window implemented using the data. The threshold for the peak search can
be entered manually, or an automatic default threshold will be used.
6.6.3 Windowing
Executing a window function to the acquisition that joins the waveform end points
in a smoother shape before computing the FFT, will result in a superior spectral resolution.
This technique is also known as ‘windowing’.
Some are more efficient for specific kinds of signal types. Some types improve the
frequency resolution, i.e., they make it simpler to detect the exact frequency of a peak in
the spectrum plot. Some windows improve the accuracy of the amplitude, i.e., they point
to the level of the peak, with better precision. The best type of window should be selected
for each specific implementation.
146
Figure 6.15 Principle of windowing
Generally, the spectrum is worked out based on a windowed signal input. The
cosine window, such as Hanning, Hamming, Blackman-Harris, and so on, are widely used.
The existence of three primary bins indicates a local maximum on the power spectrum.
Therefore, when the Spectrum Peak Search program locates three predominant bins, a
curve-fitting algorithm maps the window shape onto the bins and approximates the
frequency and amplitude of that specific tone.
147
(a) (b)
( c)
Figure 6.16 Threshold Limit testing process on front panel of VI:
(a) Normal (b) Crossing the threshold (c) Limit crossed
Figure 6.17 shows the voltages and currents of the healthy motor observed on the
LabVIEW front panel screen and Figure 6.18 demonstrates the stator turns fault current.
Figure 6.19 demonstrates real-time vibration signals for the multiple faults of broken rotor
bars and eccentricity.
148
Figure 6.17 Three-phase voltage and current waveforms in CompactRIO based Induction
Motor Diagnosis
149
Figure 6.20 shows the FFT spectrum of for the possible diagnostic
evaluations of the broken rotor bars, in which, the Fast Fourier Transforms (FFT)
amplitude in range of 46Hz-54Hz, broken bar signals are shown. However, the other
spectrum lobes can be detected, in addition to the fundamental signal, i.e., eccentricity
Figure 6.21.
150
In CompactRIO system, an execution Trace ToolkitTM from National Instruments
(NI) Inc. has been used for measurement of execution time. In the time analysis, it has
been noticed that 8-channel acquisition and filtering processes have taken less than 0.8 ms
approximately. Figure 6.22 shows the summary of the execution time durations of the fault
detection and the diagnosis algorithms.
While, the multiple and mixed faults analysis system is in the on-line mode, the
time taken for bearing, eccentricity and rotor fault frequency model calculations is 2 ms
approximately. The program was developed for the detection of four faults. However, in
implementation phase, tests were focused for three types of faults; bearing, eccentricity
and rotor fault. Limit testing of faults has taken 20 ms. From the literature, it has been
noticed that, the execution time duration of the algorithm for the proposed task is in the
standard range for the condition monitoring system of induction motor applications.
Total processing time for 8-channel monitoring, fault analysis and opening/closing
applications and CompactRIO standard operating system took approximately 200 ms.
Through the literature, it has been noticed that, the execution times of the algorithms for
the proposed research work are in the acceptable range for the condition monitoring
system of induction motor.
Four fault types were considered in this work; viz., stator winding fault, broken
rotor bars, bearing (outer race) fault and static eccentricity fault. The combined faults of
broken rotor bars and eccentricity exhibited precise results, as per the expectations.
151
The other fault combinations also have performed better, but needed trial-and-error
series of system tuning. subVIs, which were fed from Flux sensor demanded more efforts
in the calibration of the sensor output. However, diagnostic logic and algorithm have
proved their accuracy and robustness.
Repeatability tests (25 tests for each case) were conducted on the prototype model
of the system to validate the reliability of fault diagnosis procedure. Mainly, three types of
fault detection test cases were assessed statistically. The results demonstrate satisfactory
performance and are given in table 6.20. Overall, More than 90% success has been
recorded in these tests.
152
The implementation of this system has a massive initial learning curve, that is to be
overpowered by the system developer, so that many more VIs with different algorithms
can be developed for using cRIO.
6.8 Summary
In this chapter, the research work and system development for detection of
multiple mixed kinds of faults using a hybrid system model has been discussed. Here a
highly robust system model for on-line diagnosing of multiple and combined faults in
induction motor was developed. An embedded assisted system model was designed based
on NI cRIO FPGA system, which has been tested for varying numerous performance
parameters. The results obtained have exhibited better performance even with varying
operational parameters.
The next chapter presents the research conclusions and its justification for
proposed objectives. The contributions and future scopes would also be discussed in
chapter 7.
153
Chapter 7
Conclusions
CHAPTER-7
CONCLUSIONS
7.1 Conclusions
This research work has explored the different faults taking place in induction
motor and their respective diagnosis with different techniques. Ultimately, considering the
robustness of the signal processing techniques and the Field Programmable Gate Array
(FPGA) approach, these techniques have been incorporated to develop a highly robust and
efficient system setup to identify the multiple faults in the on-line condition monitoring
system of induction motor. The system has been implemented on setups with virtual as
well as real-time operational systems, so that the functionalities of the developed system
can be justified.
The overall research has been accomplished in different phases, where the
individual faults have been diagnosed separately, experimental investigations have been
implemented in different setups and the respective results have been presented. In the
initial phase, for the detection of stator faults, the Motor Current Signature Analysis
(MCSA) has been implemented for plotting the spectrum analysis graphs. The MCSA
based system implementation has been used for stator winding short circuit faults and
broken rotor bar faults in induction motors.
A novel algorithm for detection of broken rotor bars (BRB) using Discrete Wavelet
Transforms has been presented. Statistical analysis was performed to validate the detection
of BRBs. Mother Wavelet Sym8 was found to be a good candidate for wavelet
decomposition in fault diagnosis.
154
the standard range for the condition monitoring system of induction motor applications.
Similarly, for bearing fault diagnosis the FFT based vibration analysis system has been
developed and implemented with FPGA platform.
Fast Fourier Transform and Discrete Wavelet Transform based MCSA have
performed better in this research, but considering the requirement of a higher, effective
and robust system setup for multiple fault detections, the CompactRIO module based
system has been developed. This system combination not only identifies the faults on-line
in the induction motor, but it also facilitates monitoring facilities.
The overall system has been implemented for both cases in healthy as well as
faulty conditions and respective result analysis has been exhibited with operational
parameters of three-phase induction motors of varied ratings 440V, 4 poles, 50Hz coupled
to a load. The comparison of various techniques employed in this research has been
presented in Table 7.1.
155
In operation, the ultimate proposal made in this research has performed well for
faults detection even with enhanced accuracy and higher performance. Thus, considering
all these facts and justifications, it can be concluded that the proposed research work has
been accomplished successfully, while fulfilling all expectations and unique proposal.
The present work investigated the fault detection techniques, taking into
consideration the steady state parameters of the motor. The detection strategies
based on the starting current analysis may be evaluated, as the starting current has
the advantage of less sensitivity to load levels.
The diagnosis of power quality problems, which are related to motor power supply,
may be incorporated in the system.
156
Bibliography
BIBLIOGRAPHY
[1] M. E. H. Benbouzid. “A review of induction motors signature analysis as a
medium for faults detection.” IEEE Transactions on Industrial Electronics, Vol.47,
No.5, pp. 984-993, 2000.
[2] S. Nandi, H. A. Toliyat, and X. Li. “Condition monitoring and fault diagnosis of
electrical motors – a review.” IEEE Transactions on Energy Conversion, Vol.20,
No.4, pp. 719- 729, 2005.
[7] Kato, T., Inoue, K., Yoshida, K., Itokawa, T., "Diagnosis of multi-turn faults of
induction motor by direct detection of asymmetry admittance component," Power
Electronics and Drive Systems (PEDS), 2013 IEEE 10th International Conference
on 22-25 April 2013, pp.778-783.
[8] Soualhi, A, Clerc, G. Razik, H. Lebaroud, A., "Fault detection and diagnosis of
induction motors based on hidden Markov model," Electrical Machines (ICEM),
2012 XXth International Conference on 2-5 Sept. 2012, vol., no., pp.1693-1699,
[9] Ahamed, S. K., Subrata Karmakar, M. Mitra, and S. Sengupta. "Novel diagnosis
technique of mass unbalance in rotor of induction motor by the analysis of motor
starting current at no-load through wavelet transform." Electrical and Computer
Engineering (ICECE), 2010 International Conference on, pp. 474-477. IEEE,
2010.
157
[10] Premrudeepreechacharn, Suttichai, Utthiyoung, T., Kruepengkul, K., Puongkaew,
P., "Induction motor fault detection and diagnosis using supervised and
unsupervised neural networks." Industrial Technology, 2002. IEEE ICIT'02. 2002
IEEE International Conference on. Vol. 1. IEEE, 2002.
[11] Trabelsi, M., M. Boussak, P. Mestre, and M. Gossa. "Pole voltage based approach
for IGBTs open circuit fault detection and diagnosis in PWM-VSI-Fed induction
motor drives." In Power Engineering, Energy and Electrical Drives
(POWERENG), 2011 International Conference on, pp. 1-6. IEEE, 2011.
[12] Aktas, M., Turkmenoglu, V., "Wavelet-based switching faults detection in direct
torque control induction motor drives," Science, Measurement & Technology, IET,
vol.4, no.6, pp.303-310, November 2010
[13] Das, S., Purkait, P., Chakravorti, S., "Separating Induction Motor Current
Signature for stator winding faults from that due to supply voltage
unbalances," Power and Energy in NERIST (ICPEN), 2012 1st International
Conference on, vol., no., pp. 1-6, 28-29 Dec. 2012
[14] Khan, M., Rahman, M.A., "Discrete Wavelet Transform Based Detection of
Disturbances in Induction Motors," Electrical and Computer Engineering, 2006;
International Conference on 19-21 Dec. 2006, pp.462-465.
[15] Soualhi, Abdenour, Guy Clerc, and Hubert Razik. "Detection and diagnosis of
faults in induction motor using an improved artificial ant clustering
technique."Industrial Electronics, IEEE Transactions on 60, no. 9 (2013): 4053-
4062.
[16] Amaral, T.G., Pires, V.Fernao, Martins, J.F., Pires, A.J., Crisostomo, Manuel M.,
"Statistic Moment Based Method for the Detection and Diagnosis of Induction
Motor Stator Fault," Power Engineering, Energy and Electrical Drives, 2007.
POWERENG 2007. International Conference on, 12-14 April 2007, pp.106-110,
[17] Gritli, Yasser, Sang Bin Lee, Fiorenzo Filippetti, and Luca Zarri. "Advanced
diagnosis of outer cage damage in double squirrel-cage induction motors under
time-varying condition based on wavelet analysis." In Energy Conversion
Congress and Exposition (ECCE), 2012 IEEE, pp. 1284-1290. IEEE, 2012.
158
[18] Mohammed, O.A., Abed, N.Y., Ganu, S., "Modeling and Characterization of
Induction Motor Internal Faults Using Finite Element and Discrete Wavelet
Transforms," Magnetics, IEEE Transactions on, vol.42, no.10, pp. 3434-3436, Oct.
2006
[19] Da Silva Gazzana, D., Pereira, L.A., Fernandes, D., "An automated system for
incipient fault detection and diagnosis in induction motors based on MCSA,"
Industrial Technology (ICIT), 2010 IEEE International Conference on, 14-17
March 2010, pp.1227-1232.
[20] Eltabach, M., Charara, A., Antoni, J., "Comparative investigation of non-invasive
diagnosis methods for mechanical fault detection in induction motors," Industrial
Electronics, 2004 IEEE International Symposium on, 4-7 May 2004, vol.1, pp.365-
370.
[22] Moin Siddiqui, K., & Giri, V. K. (2012, March). Broken rotor bar fault detection in
induction motors using Wavelet Transform. In Computing, Electronics and
Electrical Technologies (ICCEET), 2012 International Conference on, pp. 1-6.
IEEE 2012.
[23] Ghate, Vilas N., and Sanjay V. Dudul. "Fault diagnosis of three-phase induction
motor using neural network techniques." In Emerging Trends in Engineering and
Technology (ICETET), 2009 2nd International Conference on, pp. 922-928. IEEE,
2009.
[24] Akin, Bilal, Salih Baris Ozturk, and Hamid A. Toliyat. "On-board Fault Diagnosis
of HEV Induction Motor Drive at Start-up and During Idle Mode." Vehicle Power
and Propulsion Conference, 2007. VPPC 2007. IEEE, pp. 140-147. IEEE, 2007.
[25] Verucchi, Carlos Javier, Acosta, Gerardo Gabriel, "Fault Detection and Diagnosis
Techniques in Induction Electrical Machines," Latin America Transactions, IEEE
(Revista IEEE America Latina) on, March 2007, vol.5, no.1, pp.41-49.
159
[26] Dongyang, Xiang, Wu Zhengguo, Hu Wen Biao, and Hou Xinguo. "A composite
fault diagnosis method of induction motor based on blind signal processing." In
Computer, Mechatronics, Control and Electronic Engineering (CMCE), 2010
International Conference on, vol. 3, pp. 511-514. IEEE, 2010.
[27] Osheiba, Ali M., Hatem A. Darwish, A-MI Taalab, and Aly A. Afify. "Transient
modeling of stator winding inter-turn short-circuits in induction motors." In Power
Systems Conference, 2006. MEPCON 2006. Eleventh International Middle East,
vol. 2, pp. 643-647. IEEE, 2006.
[29] Onel, Izzet Yilmaz, and M. El Hachemi Benbouzid. "Induction motor bearing
failure detection and diagnosis: Park and concordia transform approaches
comparative study." Mechatronics, IEEE/ASME Transactions on, 13, no. 2 (2008):
pp.257-262.
[30] Diallo, D., Benbouzid, M.E.H., Hamad, D., Pierre, X., "Fault Detection and
Diagnosis in an Induction Machine Drive: A Pattern Recognition Approach Based
on Concordia Stator Mean Current Vector," Energy Conversion, IEEE
Transactions on Sept. 2005, vol.20, no.3, pp.512-519.
[31] Rafimanzelat, M.R., Araabi, B.N., Khosroshahli, E., "Induction machine broken
bar detection using neural networks based classification," Cybernetics and
Intelligent Systems, 2004 IEEE Conference on 1-3 Dec. 2004, vol.2, pp.1362-1367.
[32] Saidi, L., Fnaiech, F., Capolino, G., Henao, H., "Stator current bi-spectrum
patterns for induction machines multiple faults detection," IECON 2012 - 38th
Annual Conference on IEEE Industrial Electronics Society Oct. 2012, pp.5132-
5137.
[33] Bikfalvi, Peter, and Mária Imecs. "Rotor Fault Detection in Induction Machines:
Methods and Techniques-State-of-the-Art." In Automation, Quality and Testing,
Robotics, 2006 IEEE International Conference on, vol. 1, pp. 199-204. IEEE,
2006.
160
[34] Seshadrinath, J., Singh, B., Panigrahi, B.K., "Single-Turn Fault Detection in
Induction Machine Using Complex-Wavelet-Based Method," Industry
Applications, IEEE Transactions on Nov.-Dec. 2012, vol.48, no.6, pp.1846-1854.
[35] Prieto, M.D., Cirrincione, G., Espinosa, A.G., Ortega, J.A., Henao, H., "Bearing
Fault Detection by a Novel Condition Monitoring Scheme Based on Statistical-
Time Features and Neural Networks," Industrial Electronics, IEEE Transactions
on Aug. 2013, vol.60, no.8, pp.3398-3407.
[36] Puska R., "A learning pattern recognition system using neural network for
diagnosis and monitoring of aging of electrical motor," Industrial Electronics,
Control, Instrumentation, and Automation, 1992. Power Electronics and Motion
Control., Proceedings of the 1992 International Conference on 9-13 Nov 1992,
vol.2, pp.1074-1077.
[37] Santos, F.M.C., da Silva, I.N., Suetake, M., "Neural network classifier for faults
detection in induction motors," Computer Applications Technology (ICCAT), 2013
International Conference on 20-22 Jan. 2013, pp.1-5.
[38] Filippetti, F., Uncini, A., Piazza, C., Campolucci, P., Tassoni, C., Franceschini, G.,
"Neural network architectures for fault diagnosis and parameter recognition in
induction machines," Electrotechnical Conference on 13-16 May 1996 MELECON
'96., 8th Mediterranean, vol.1, pp.289-293.
[39] Puscasu, G., Palade, V., Stancu, A., "A procedure of fault detection for driving
systems using self-organizing neural networks," Electrotechnical Conference,
2000, MELECON 2000. 10th Mediterranean 2000, vol.2, pp.446-448.
[40] Gandhi, A., Corrigan, T., Parsa, L., "Recent Advances in Modeling and On-line
Detection of Stator Inter-turn Faults in Electrical Motors," Industrial Electronics,
IEEE Transactions on May 2011, vol.58, no.5, pp.1564-1575,
[41] Martins, J.F., Dente, J.A., Pires, A.J., Mendes, R.V., "Language identification of
controlled systems: modeling, control, and anomaly detection," Systems, Man, and
Cybernetics, Part C: Applications and Reviews, IEEE Transactions on May 2001,
vol.31, no.2, pp.234-242.
161
[42] Obaid Mustafa, M., Nikolakopoulos, G., Gustafsson, T., "A Fault diagnosis
scheme for three-phase induction motors based on uncertainty bounds," IECON
2012 - 38th Annual Conference on 25-28 Oct. 2012, pp.1606-1611.
[43] Corcoran, A.L., Sen, S., "Using real-valued genetic algorithms to evolve rule sets
for classification," Evolutionary Computation, 1994. IEEE World Congress on
Computational Intelligence., Proceedings of the First IEEE Conference on 27-29
Jun 1994, vol.1, pp.120-124.
[44] Yu-Seop Park, Seok-Myeong Jang, Jang-Young Choi, Cheol-Soo Goo, "Fault
detection of squirrel-cage induction motor by analyzing motor current signals,"
Vehicle Power and Propulsion Conference (VPPC) on 9-12 Oct. 2012, pp.219-
222.
[45] Wegiel, T., Sulowicz, M., Borkowski, D., "A Distributed System of Signal
Acquisition for Induction Motors Diagnostic," Diagnostics for Electric Machines,
Power Electronics and Drives, 2007, IEEE International Symposium on 6-8 Sept.
2007, pp.261-265.
[46] Abul Masrur, M., Chen, Z., Murphey, Y., "Intelligent diagnosis of open and short
circuit faults in electric drive inverters for real-time applications," Power
Electronics, IET March 2010, vol.3, no.2, pp.279-291.
[47] Refaat, S.S., Abu-Rub, H., Saad, M.S., Aboul-Zahab, E.M., Iqbal, A., "Detection,
diagnoses and discrimination of stator turn-to-turn fault and unbalanced supply
voltage fault for three-phase induction motors," Power and Energy (PECon), 2012
IEEE International Conference on 2-5 Dec. 2012, pp.910-915.
[49] Mehala, Neelam, and Ratna Dahiya. "Motor current signature analysis and its
applications in induction motor fault diagnosis." International journal of systems
applications, engineering & development vol.2, no. 1, pp.29-35, 2007.
162
[50] F. Ponci, A. Monti, L. Cristaldi, and M. Lazzaroni, “ iagnostic of faulty induction
motor drive via wavelet decomposition,” IEEE Trans. Instrum. Meas., vol. 56, no.
6, pp. 2606–2615, Dec. 2007.
[55] Chow, Mo-Yuen. "Guest editorial special section on motor fault detection and
diagnosis." Industrial Electronics, IEEE Transactions on 47, no. 5, 982-983, 2000.
[56] Waide, Paul, and Conrad U. Brunner. Energy-efficiency policy opportunities for
electric motor-driven systems. No. 2011/7. OECD and International Energy
Agency Publishing, 2011.
[57] WEG, General Purpose Motors. Brazil: WEG Electric Motors Corporation, 2004.
[58] B. K. Bose, Modern power electronics and AC drives. Upper Saddle River, NJ:
Prentice Hall, 2002.
[59] E. W. Kimbark, Power System Stability, vol. III. New York: IEEE Press, 1995.
163
[60] P. L. Cochran, Polyphase Induction Motors: Analysis, Design, and Applications.
New York: Marcel Dekker, 1989.
[65] A. Siddique, G. S. Yadava, and B. Singh, "A review of stator fault monitoring
techniques of induction motors," Energy Conversion, IEEE Transactions on, vol.
20, pp. 106-114, 2005.
[66] A. H. Bonnett, A. H. Bonnett, and G. C. Soukup, "Cause and analysis of stator and
rotor failures in three-phase squirrel-cage induction motors," Industry Applications,
IEEE Transactions on, vol. 28, pp. 921-937, 1992.
[69] Thomson, W.T., Fenger, M.; “Current Signature Analysis to etect Induction
Motor Faults,” In IEEE Industrial Applications Magazine, 2011.
[70] S. Nandi, H. A. Toliyat, and X. Li, "Condition Monitoring and Fault Diagnosis of
Electrical Motors - A Review," IEEE Transactions on Energy Conversion, vol. 20,
pp. 719-729, 2005
[71] A. Bellini, F. Filippetti, G. Franceschini, C. Tassoni, and G. B. Kliman,
"Quantitative evaluation of induction motor broken bars by means of electrical
signature analysis," IEEE Transactions on Industry Applications, vol. 37, pp.
1248-1255, 2001.
164
[72] M. E. H. Benbouzid, "A review of induction motors signature analysis as a
medium for faults detection," IEEE Transactions on Industrial Electronics, vol. 47,
pp. 984-993, 2000.
[75] Electric Power Research Institute, “Improved Motors for Utility Applications,”
Publication EL-2678, Vol. 1, 1763-1, final report, October, 1982.
[77] IAS MR Working Group, “Report of large motor reliability survey of industrial
and commercial installation, IEEE Trans. Industry Applications, vol. I-A-21, pp.
853-864, Jul/Aug, 1985.
[78] Sottile Jr, Joseph, and Jeffery L. Kohler. "An on-line method to detect incipient
failure of turn insulation in random-wound motors." Energy Conversion, IEEE
Transactions on 8, no. 4, 762-768, 1993.
[79] 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.
865-872, May, 2003.
[80] A. H. Bonnet and G. C. Soukup, “Cause and analysis of stator and rotor failures in
three- phase squirrel case induction motors,” IEEE Trans. Industry Application,
vol., 28, no. 4, pp.921-937, 1992.
165
[82] S. F. Farag, R. G. Bartheld, and W. E. May, “Electronically enhanced low voltage
motor protection and control,” IEEE Trans. Industry Applications, vol. 29, no.1,
pp. 45-51, 1994.
[83] S.B.Lee and T.G.Habetler, “An on-line stator winding resistance estimation
technique for temperature monitoring of line-connected induction machines,” IEEE
Trans. Industry Applications, vol. 39, no.3, pp. 685-694, May/Jun, 2003.
[86] M.A.Cash, “ etection 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.
[87] .E.Schump, “Predict motor failure with insulation testing,” Pulp and Paper
Industry Tech Conference 1997, pp.48-50, Cincinnati, OH, USA, Jun, 1997.
[90] S. Nandi, H.A. Toliyat, X. Li, “Condition monitoring and fault diagnosis of
electrical motors -a review,” IEEE Transactions on Energy Conversion, Vol. 20,
pp. 719-729, 2005.
[91] Çalış, Hakan, and Abdülkadir Çakır. "Rotor bar fault diagnosis in three-phase
induction motors by monitoring fluctuations of motor current zero crossing
instants." Electric power systems Research, no. 5, 385-392, 2007.
[92] S. Seker, “ etermination of air gap eccentricity in electric motors using coherence
analysis,” IEEE Power Engineering Review, Vol. 20, pp.48-50, 2000.
166
[93] S. Seker, E. Ayaz, “Feature extraction related to bearing damage in electric motors
by wavelet analysis,” Journal of the Franklin Institute, Vol. 340, pp. 125-134,
2003.
[94] G.B. Kliman, R.A. Koegl, J. Stein, R.D. Endicott, M.W. Madden, “Non-invasive
detection of broken rotor bars in operating induction motors,” IEEE Transactions
on Energy Conversion, Vol. 3, pp. 873-879, 1988.
[96] Siddiqui, Khadim Moin, Kuldeep Sahay, and V. K. Giri. "Health Monitoring and
Fault Diagnosis in Induction Motor-A Review," International Journal
of Advanced Research in Electrical, Electronics and Instrumentation Engineering,
Vol. 3, Issue 1, January 2014
[97] M. Haji, H.A. Toliyat, “Pattern recognition - a technique for induction machines
rotor broken bar detection,” IEEE Transactions on Energy Conversion, Vol. 16,
pp. 312-317, 2001.
[98] Li, Weidong, and Chris K. Mechefske. "Detection of induction motor faults: a
comparison of stator current, vibration and acoustic methods." Journal of vibration
and Control, no. 2 pp.165-188, 2006.
[99] Patton, R J and Chen, J, “Robustness in model based fault diagnosis,” In Concise
Encyclopedia of Simulation and Modeling, (D Atherton, P Borne eds), Pergamon
Press, pp. 379-392, 1992.
167
[102] Ahmed, Intesar, Nesimi Ertugrul, and Wen Liang Soong. "A study on the detection
of fault frequencies for condition monitoring of induction machines," School of
Electrical and Electronics Engineering University of Adelaide, Australia, 2005.
[103] Shi, Pu, Zheng Chen, Yuriy Vagapov, and Zoubir Zouaoui. "Optimal wavelets for
broken rotor bars fault diagnosis." In Diagnostics for Electric Machines, Power
Electronics and Drives (SDEMPED), 2013 9th IEEE International Symposium on,
pp. 49-56. IEEE, 2013.
[104] Sadri, H, “Induction Motor Broken Rotor Bar Fault etection Using Wavelet
Analysis,” Master's Thesis, Iran University of Science & Technology, 2004.
[105] Ponci, Ferdinanda, Antonello Monti, Loredana Cristaldi, and Massimo Lazzaroni.
"Diagnostic of a faulty induction motor drive via wavelet decomposition,"
Instrumentation and Measurement, IEEE Transactions on, no. 6, pp. 2606-2615,
2007.
[107] Hachemi, B. M. EI., and Kliman, G. B., “What stator current processing-based
technique to use for induction motor rotor faults diagnosis?” IEEE Trans. on
Energy Conversion, Vol. 18, No. 2, pp. 238-244, June 2003.
[109] Chow, T. W. S., and Fei, G., “Three-phase induction machines asymmetrical faults
identification using bi-spectrum,” IEEE Trans. Energy Conversion, Vol. 10, pp.
688-693, 1995.
[110] Kilman, G. B., and Stein, J., “Methods of motor current signature analysis,” Elec.
Mach. Power Syst., Vol. 20, pp. 463-474, 1992.
[111] Benbouzid, M. E. H., Vieira, M., and Theys, C., “Induction motors faults detection
and localization using stator current advanced signal processing techniques,” IEEE
Trans. Power Electron., Vol. 14, No. 1, pp. 14-22, January 1999.
[112] Ondel, Olivier, Guy Clerc, Emmanuel Boutleux, and Eric Blanco. "Fault detection
and diagnosis in a set “inverter–induction machine” through multidimensional
168
membership function and pattern recognition." Energy Conversion, IEEE
Transactions on, Vol. 24, No. 2, pp. 431-441, June 2009.
[113] Cusido, Jordi, Luis Romeral, Juan Antonio Ortega, Javier A. Rosero, and A. Garcia
Espinosa. "Fault detection in induction machines using power spectral density in
wavelet decomposition." Industrial Electronics, IEEE Transactions on, 55, no. 2,
pp. 633-643, 2008.
[114] Kia, S. H., Henao, H., and Capolino, G. A., “ iagnosis of broken bar fault in
induction machines using discrete wavelet transform without slip estimation,”
IEEE Trans. Indl. Appln., Vol. 45, No. 4, pp. 1395-1404, July 2009.
[115] Mahamad, A. K., Saon, S., and Hiyama, T., “Predicting remaining useful life of
rotating machinery based artificial neural network,” Comput. Math. Appl., Vol. 60,
No. 4, pp. 1078-1087, August 2010.
[117] Bouzida, A., Touhami, O., Ibtiouen, R., Fadel, M., Rezzoug, A., and Beloucherani,
A., “Fault diagnosis in industrial induction machines through discrete wavelet
transform,” IEEE Trans. Indl. Electron., Vol. 58, No. 9, pp. 4385-4395, September
2011.
[119] Bakhri, S., Ertugrul, N., Soong, W. L., and AI-Sarawi, S., “Investigation and
development of a real-time on-site condition monitoring system for induction
motors,” In Proceedings Australasian universities Power Engg. Conf., pp. 1-6,
December 2007.
[120] Thomson, W. T., and Fenger, M., “Current signature analysis to detect induction
motor faults,” IEEE Trans. Indl. Appl., Mag., vol. 7, pp. 26-34, 2001.
169
[121] Velmurugan, S., and C. Rajasekaran. "A reconfigurable on-chip multichannel data
acquisition and processing (DAQP) system for multichannel signal processing."
In Pattern Recognition, Informatics and Medical Engineering (PRIME), 2013
International Conference on, pp. 109-114. IEEE, 2013.
[122] Zhiwei, Wang, Hou Shuming, and Xu Zhe. "Inertial Measurement Unit's Rapid
Auto-test System Design Based on cRIO." Missiles and Space Vehicles 006, 2012.
[123] LI, Chao, Feng LI, and Kai WANG. "Real-time Measure and Control Platform of
Missile Electro-hydraulic Servo Mechanism Based on NI cRIO." Hydraulics
Pneumatics & Seals, 14, 2013.
[125] Ian Fairweather, Anne Brumfield, LabVIEW- A developer’s Guide to Real World
Integration Book by CRC, 2012, pp. 133-205.
[126] Jovitha Jerome, Virtual Instrumentation Using LabVIEW, Book by PHI Learning
Private Limited, New Delhi, 2010.
[127] Cory L.Clark, LabVIEW Digital Signal Processing, Book by The McGraw-Hill
Companies, 2005.
[129] Rick Bitter, Taqi Mohiuddin, and Matt Nawrocki, LabVIEW- Advanced
programming Techniques 2nd Edition, Book by CRC Press, Taylor & Francis
Group, 2007.
170
LIST OF RESEARCH PUBLICATIONS
International Journals:
171
International Conferences:
National Conferences:
172
VITA
S.M.SHASHIDHARA
173