TKK Dissertations 51

Espoo 2006

ELECTROMAGNETIC FLUX MONITORING FOR

DETECTING FAULTS IN ELECTRICAL MACHINES
Doctoral Dissertation

Marian Dumitru Negrea

Helsinki University of Technology

Department of Electrical and Communications Engineering
Laboratory of Electromechanics
TKK Dissertations 51
Espoo 2006

ELECTROMAGNETIC FLUX MONITORING FOR

DETECTING FAULTS IN ELECTRICAL MACHINES
Doctoral Dissertation

Marian Dumitru Negrea

Dissertation for the degree of Doctor of Science in Technology to be presented with due permission
of the Department of Electrical and Communications Engineering for public examination and debate
in Auditorium S4 at Helsinki University of Technology (Espoo, Finland) on the 29th of November,
2006, at 12 noon.

Helsinki University of Technology

Department of Electrical and Communications Engineering
Laboratory of Electromechanics

Teknillinen korkeakoulu
Sähkö- ja tietoliikennetekniikan osasto
Sähkömekaniikan laboratorio
Distribution:
Helsinki University of Technology
Department of Electrical and Communications Engineering
Laboratory of Electromechanics
P.O. Box 3000
FI - 02015 TKK
FINLAND
URL: http://www.tkk.fi/Units/Electromechanics/
Tel. +358-9-451 2384
Fax. +358-9-451 2991
E-mail: electromechanics@hut.fi

© 2006 Marian Dumitru Negrea

ISBN-13 978-951-22-8476-4
ISBN-10 951-22-8476-6
ISBN-13 978-951-22-8477-1 (PDF)
ISBN-10 951-22-8477-4 (PDF)
ISSN 1795-2239
ISSN 1795-4584 (PDF)
URL: http://lib.tkk.fi/Diss/2006/isbn9512284774/

TKK-DISS-2217

Picaset Oy
Helsinki 2006
 AB
HELSINKI UNIVERSITY OF TECHNOLOGY ABSTRACT OF DOCTORAL DISSERTATION
P. O. BOX 1000, FI-02015 TKK
http://www.tkk.fi

Author Marian Dumitru Negrea

Name of the dissertation
Electromagnetic flux monitoring for detecting faults in electrical machines

Date of manuscript June 12, 2006 Date of the dissertation November 29, 2006
Monograph Article dissertation (summary + original articles)
Department Electrical and Communications Engineering
Laboratory Electromechanics
Field of research Electrical machines
Opponents Associate Professor Ewen Ritchie, Institute of Energy Technology, Aalborg University, Denmark
Professor Manés Fernández Cabanas, University of Oviedo, Gijon, Spain
Supervisor Professor Antero Arkkio
Abstract
The ability of the electromagnetic flux measured in various locations of a 35-kW cage induction motor to provide useful
information about faults was investigated. The usefulness of this monitoring parameter was assessed in comparison with
some other electrical parameters used for fault detection, such as stator phase current and circulating currents between
the parallel branches of the stator winding.
The following faults were investigated in this thesis: a turn-to-turn short circuit in the stator winding; rotor cage-related
faults (breakage of rotor bars); static and dynamic eccentricity, and bearing fault.
The relevant fault signatures of the studied electrical parameters were obtained from measurements and/or from
numerical electromagnetic field simulations in steady state. These signatures were analysed and compared in order to
deduce the most appropriate quantity for the detection of a specific fault. When and where possible, the accuracy of
different fault signatures issuing from numerical electromagnetic field simulations was validated by experiments. This
investigation is essential since, following a good agreement, it may be assumed that if a monitoring system cannot
detect and diagnose an artificial fault from the virtual measurement signals, it is hardly likely to work with real
electrical machines, either. In this respect, the numerical methods of analysis limited the present study to such faults that
affect the electromagnetic field of a machine.
On the exclusive basis of data obtained from simulations, a study of the modifications brought by various stator winding
designs to some of the asymmetrical air-gap electromagnetic flux density harmonics responsible for the detection of
various faults was carried out.
The analysis of a core fault (insulation fault in the stator lamination) artificially implemented in the numerical
electromagnetic model of the machine in terms of finding a suitable parameter to sense such a fault was also studied in
this work.
Keywords fault diagnostics, condition monitoring, electrical machines, induction motors, electromagnetic flux,
finite element analysis, search coils.
ISBN (printed) 951-22-8476-6 ISSN (printed) 1795-2239
ISBN (pdf) 951-22-8477-4 ISSN (pdf) 1795-4584
ISBN (others) Number of pages 140
Publisher Helsinki University of Technology, Laboratory of Electromechanics
Print distribution Laboratory of Electromechanics
The dissertation can be read at http://lib.tkk.fi/Diss/isbn9512284774
4
PREFACE
Most of the present research work was carried out as part of a collaborative research project
between the Laboratory of Electromechanics and Laboratory of Control Engineering of
Helsinki University of Technology. The research project was entitled “Fault detection and
diagnosis for AC electrical machines”. The industrial partners involved in this project were
ABB Industry Oy, KCI Konecranes International Oyj, KONE Oyj, and Kuppari Mittaus Oy.
The Technology Development Centre − TEKES − made a major contribution to the financing
of this work.

I would like to express my sincere gratitude to Professor Antero Arkkio for his invaluable
support and encouragement throughout the whole work.

I wish to thank the Head of the Laboratory of Electromechanics, Professor Asko Niemenmaa,
and his predecessor, Professor Emeritus Tapani Jokinen, for offering me the opportunity to
work in this laboratory, for their cordiality, and for all their support.

For his co-operation in the “Fault detection and diagnosis for AC electrical machines”
research project, thanks should be addressed to Pedro Vicente Jover Rodriguez. The co-
operation with the Laboratory of Control Engineering, and in particular with Sanna Pöyhönen,
Ph. D., deserves acknowledgement.

I would particularly like to acknowledge the beneficial discussions and help of Marius Rosu,
Ph. D and Mircea Popescu, Ph. D.

I owe special thanks to my colleagues in the Laboratory of Electromechanics for offering a

pleasant atmosphere.
In particular, Professor Ewen Ritchie and Tuomo Lindh, Ph. D. deserve my sincere gratitude
for agreeing to be the pre-examiners of this work, thus bringing fruitful observations and
suggestions to it.
I am obliged to Mr. Simon Gill for the revision of the language.
The financial support of Suomen Kultuurirahasto, Tekniikan Edistämissäätiö, and the Fortum
foundation is gratefully acknowledged.
My deepest emotions and thoughts go to my beloved wife Lidia. Her love, support, and trust
in our dreams gave me confidence in my struggle to carry out and finalise this work.
Last but not least, I am deeply grateful to my wonderful parents, Ion and Maria, and my
brother Ionel.

Espoo, October 2006

Marian Dumitru Negrea

5
6
TABLE OF CONTENTS

PREFACE ............................................................................................................................................... 5
READING INSTRUCTIONS ................................................................................................................. 9
LIST OF SYMBOLS ............................................................................................................................ 11
1 INTRODUCTION............................................................................................................................ 13
1.1 Background and importance of fault diagnosis and condition monitoring ................................. 13
1.2 Aim of the work .......................................................................................................................... 14
1.3 Scientific contribution................................................................................................................. 15
1.4 Structure of the work................................................................................................................... 16
2 FAULTS IN ELECTRICAL MACHINES ..................................................................................... 17
2.1 General......................................................................................................................................... 17
2.2 Fault types in electrical machines ................................................................................................ 18
2.2.1 Winding faults – Stator- and rotor-related............................................................................ 19
2.2.2 Stator core-related faults....................................................................................................... 21
2.2.3 Rotor-related faults ............................................................................................................... 22
2.2.4 Bearing faults........................................................................................................................ 24
3 FAULT INDICATORS FOR ELECTRICAL MACHINES............................................................ 27
3.1 General.......................................................................................................................................... 27
3.2 Potential fault indicators for a technical device ............................................................................ 27
3.3 Potential fault indicators for electrical machines.......................................................................... 28
3.3.1 General.................................................................................................................................. 28
3.3.2 Indicators for detecting winding-related faults..................................................................... 31
3.3.3 Indicators for detecting rotor-related faults .......................................................................... 31
3.3.4 Indicators for detecting bearing faults .................................................................................. 31
3.3.5 Indicators for detecting eccentricity ..................................................................................... 33
3.4 Current monitoring ....................................................................................................................... 34
3.4.1 Phase current......................................................................................................................... 34
3.4.2 Circulating current................................................................................................................ 38
3.4.3 Shaft currents........................................................................................................................ 38
3.4.4 Current monitoring - drawbacks........................................................................................... 39
3.5 Magnetic flux monitoring ............................................................................................................. 40
3.5.1 General principles, comparisons, sensor-measuring technologies ....................................... 40
3.5.2 Sensors used for condition monitoring in electrical machines – the search coils................. 40
3.5.3 Electromagnetic flux regions to be monitored in electrical machines .................................. 41
3.5.4 Electromagnetic flux harmonics of order “p ± 1” used for fault detection…………………44
3.5.5 Turn-to-turn short circuit fault detection in stator windings................................................. 44
3.5.6 Broken rotor bars detection....................................................................................................45

7
 3.5.7 Bearing faults detection.........................................................................................................45
3.5.8 Eccentricity detection ........................................................................................................... 46
4 METHODS OF ANALYSIS............................................................................................................ 47
4.1 Finite element analysis.................................................................................................................. 47
4.1.1 Finite elements methods – general description, areas of application.................................... 47
4.1.2 Numerical electromagnetic field analysis for electrical machines ....................................... 47
4.1.3 Description of the numerical electromagnetic field simulation tool used in this thesis ..… 48
4.1.4 Fault modelling procedure.................................................................................................... 51
4.2 Method of measurements.............................................................................................................. 54
4.3 Signal processing methods............................................................................................................ 56
4.3.1 General.................................................................................................................................. 56
4.3.2 Analysis in steady state......................................................................................................... 56
4.3.3 Transient state analysis ......................................................................................................... 56
4.3.4 Artificial intelligence techniques.......................................................................................... 57
5. RESULTS OBTAINED FROM EXPERIMENTS AND SIMULATIONS..................................... 59
5.1. Current monitoring...................................................................................................................... 59
5.1.1. Stator winding currents ....................................................................................................... 59
5.1.2. Circulating current between parallel branches .................................................................... 63
5.2. Electromagnetic flux monitoring ................................................................................................ 70
5.2.1. Inter-turn short circuit in the stator winding........................................................................ 70
5.2.2. Rotor cage-related faults ..................................................................................................... 82
5.3. Static eccentricity, dynamic eccentricity, bearing fault .............................................................. 92
5.4. Core fault................................................................................................................................... 112
6. FINAL DISCUSSION ................................................................................................................... 115
7 CONCLUSIONS ............................................................................................................................ 119
REFERENCES.................................................................................................................................... 121
Appendix I Information about the measurement devices.................................................................... 133
Appendix II Stator winding configurations......................................................................................... 137

8
READING INSTRUCTIONS

There are various ways of setting out references for scientific publications. In this thesis, the
references to the literature are indicated using the Harvard (author-date) system. In this
system, a textual citation generally requires only the name of the author(s) and the year of
publication.

This may appear at the end of a sentence, before the full stop. Alternatively, the author's
surname may be integrated into the text, followed by the year of publication in parentheses.
The full references are listed at the end of the work in alphabetical order and require the
following details:

For an article
1. name(s) of author(s) of the article (surname and initials or given name)
2. year of publication
3. title of article, in single quotation marks
4. title of periodical (italicised)
5. volume number
6. issue (or part) number
7. page number(s)

For a book
1. name(s) of author(s), editor(s), or compiler(s) (surname and initials or given name) or
the institution responsible
2. year of publication
3. title of publication and subtitle if any (all titles are italicised)
4. series title and individual volume if any
5. edition, if other than first
6. publisher
7. place of publication
8. page number(s) if applicable

For electronic resources

1. name(s) of author(s)
2. date of publication
3. title of publication
4. publisher/organisation
5. edition, if other than first
6. type of medium
7. date item retrieved
8. name or site address on internet (if applicable)

First level chapters are numbered consecutively (1, 2, 3, ...) and subchapters are numbered by
chapters (1.1, 1.2,...,2.1, 2.1.1, 2.1.2, 2.2,...,).
The equations, figures and tables are numbered according to a similar rule:
chapter_number.equation/figure/table number (i.e. 1.1, 1.2…)

9
10
LIST OF SYMBOLS

Symbol overview

A, A magnetic vector potential; cross sectional area

B, B magnetic flux density
Br r - component of the magnetic flux density
Bϕ ϕ - component of the magnetic flux density
a, b points delimiting a short circuited section of the winding
ez unit vector in the z-direction
fs supply frequency
fb mechanical frequency depending on the bearing characteristics
fr rotational frequency
fbng vibration bearing frequency
fv characteristic vibration frequency
fbrb characteristic frequencies for the detection of broken rotor bar/s
fshort characteristic frequencies for the detection of inter-turn short circuits
fecc characteristic frequencies for the detection of eccentricity
H, H magnetic field intensity
Hc coercive field
i electric current
Ic circulating current
J, J electric current density
k constant
l length of the conductor
m constant
n constant
N number of turns
nd eccentricity order
nw order of the stator time harmonics
Nr number of broken rotor bars
p number of pole pairs; electric power
q constant
Q2 number of rotor slots
R DC resistance of the conductor
r, ϕ, z coordinates of a circular cylindrical coordinate system
s slip
S integration surface
t time
u voltage
ν reluctivity
σ conductivity

11
φ electric scalar potential
φp, φw phase angles
Ω electric scalar potential
ω1 fundamental angular frequency of the supply voltage
ωr rotor angular velocity

Abbreviation overview

MCSA motor current signature analysis

MMF magnetomotive force
AC alternating current
DC direct current
FE finite element
rms root mean square value
UMP unbalanced magnetic pull
FFT fast Fourier transformation
brb broken rotor bars
ecc eccentricity
NB stator winding configuration consisting of no parallel branches (see Appendix II)
SB stator winding configuration consisting of two parallel branches – special configuration (see
Appendix II)
2B stator winding configuration consisting of two parallel branches (see Appendix II)
4B stator winding configuration consisting of four parallel branches (see Appendix II)
meas. measurements
simul. simulations
cf core fault
p±1 search coils purposed to sense the air-gap electromagnetic flux

12
1 INTRODUCTION

1.1 Background and importance of fault diagnosis and condition monitoring

Electrical machines have been used extensively for many different industrial applications
since several decades ago. These applications range from intensive care unit pumps, electric
vehicle propulsion systems, and computer-cooling fans to electric pumps used in nuclear
power plants. The electrical energy that is consumed in (induction) motors accounts for
around 60% of the electrical energy that is consumed by industry in developed economies
(Williamson 2004).
The present-day requirement for the ever-increasing reliability of electrical machines is now
more important than ever before and continues to grow. Advances are continually being made
in this area as a result of the consistent demand from the power generation and transportation
industries. Because of the progress made in engineering and materials science, rotating
machinery is becoming both faster and lighter, as well as being required to run for longer
periods of time. All of these factors mean that the detection, location, and analysis of faults
play a vital role in the good operation of the electrical machine and are essential for major
concerns such as the safety, reliability, efficiency, and performance of applications involving
electrical machines. Although continual improvement in design and manufacturing has
become a priority task among contemporary manufacturers of electrical machines, faults still
can and do occur.
Since the analysis and design of rotating machinery is critical in terms of the cost of both
production and maintenance, it is not surprising that the fault diagnosis of rotating machinery
is a crucial aspect of the subject and is receiving ever more attention. As the design of rotating
machinery becomes increasingly complex, as a result of the rapid progress being made in
technology, so machinery condition monitoring strategies must become more advanced in
order to cope with the physical burdens being placed on the individual components of a
machine.
When faults do occur and the machine fails in service, the result could, at best, be the loss of
production and revenue, or, at worst, catastrophic for the industrial process and potentially
dangerous to the operators.
The issues of preventive maintenance, on-line motor fault detection, and condition monitoring
are of increasing importance, taking into consideration essential concerns such as:
- ageing motors,
- lack of redundancy in the event of a machine failure,
- high-reliability requirements,
- cost competitiveness.
During the past twenty years, there has been a substantial amount of research into the creation
of new condition monitoring techniques for electrical machine drives, with new methods
being developed and implemented in commercial products for this purpose (Chow 2000,
Benbouzid 1999, Nandi et al. 2005). The research and development of newer and alternative
diagnostic techniques is continuous, however, since condition monitoring and fault diagnosis
systems should always suit new, specific electric motor drive applications. This continuous
research and development is also supported by the fact that no specific system/technique may
be considered generally the best for all the applications that exist, since an operator must treat
each motor drive as a unique entity. In this respect, the potential failure modes, fundamental
causes, mechanical load characteristics, and operational conditions all have to be carefully

13
taken into consideration when a monitoring system is to be designed or selected for a specific
application (Thomson 1999).
The large amount of previous work carried out in the area of fault diagnosis and condition
monitoring shows that there have been many challenges and opportunities for engineers and
researchers to focus on. Various recommendations and solutions concerning condition
monitoring technologies have been given in this area, mainly depending on the machine type,
size, operating conditions (loading), available instrumentation, cost constraints etc.
In order to allow analysts to correlate different aspects of each technology to troubleshoot
symptoms and determine a course of action to avert failures, several fields of science and
technology, such as electrical, mechanical, thermal, and sometimes chemical engineering
should be closely considered and combined whenever possible. This is also a stringent
requirement when aiming to build a competitive condition monitoring system.

1.2 Aim of the work

The main aim of this thesis is to study the ability of electromagnetic flux to provide useful
information about various faults in an induction machine. The usefulness of this monitoring
parameter will be assessed in comparison with some other electrical parameters used for fault
detection, such as stator phase current and circulating currents between the parallel branches
of the stator winding (if there are such).

Another aim of this thesis is to validate by experiments, when and where possible, the
accuracy of different fault signatures issuing from numerical electromagnetic field
simulations.
On the exclusive basis of data obtained from simulations, a study of the modifications brought
by various stator winding designs to some of the asymmetrical air-gap electromagnetic flux
density harmonics responsible for the detection of various faults will be carried out.
The analysis of a core fault (insulation fault in the stator lamination) artificially implemented
in a numerical electromagnetic model of a machine, in terms of finding a suitable parameter
to sense such a fault, is also studied in this work.
The area of interest for this thesis is restricted to induction machines but, the possibility of
extrapolating the findings to other machine types is discussed at various points.

14
1.3 Scientific contribution

Scientific contribution of the author

First of all, this study represents a detailed analysis of the electromagnetic flux patterns that
are supposed to provide potential useful information about a fault in an induction machine.
For capturing such patterns, six search coils were employed in the measurements and four
search coils in the simulations. The complexity of this sensor network for capturing the
electromagnetic flux in various parts of the electrical machine and the critical comparative
analysis of the indications provided by various fault indicators may be viewed as an important
contribution to the existing state of the art in this research area.
The investigation carried out on finding the ability of electromagnetic flux eccentricity
harmonics of the order “p±1” to detect machine abnormalities other than various types of
eccentricity, both from measurements and simulations, is considered a new contribution.
This study also presents an attempt to implement in numerical simulations an insulation fault
located in the stator lamination and to suggest an electrical fault indicator that may be
confidently used to detect such an abnormality.
The investigations of the modifications brought by various stator winding designs to: 1) some
of the asymmetrical air-gap electromagnetic flux density harmonics responsible for the
detection of various faults, and 2) to the ability of stator branch currents and, also, circulating
currents to sense such faults are considered to be further original contributions.

Contribution of other members of the research team

Most of the present research work was carried out as a part of a collaborative research project
between the Laboratory of Electromechanics and Laboratory of Control Engineering of
Helsinki University of Technology. The research project was entitled “Fault detection and
diagnosis for AC electrical machines”. The industrial partners involved in this project were
ABB Industry Oy, KCI Konecranes International Oyj, KONE Oyj, and Kuppari Mittaus Oy.
The main task of the Laboratory of Electromechanics was to develop simulation models and
to carry out measurements for electrical machines working under both faulty and healthy
conditions. Relying on the data provided by the Laboratory of Electromechanics, the main
task of the Control Engineering Laboratory was to develop advanced signal processing
techniques for fault diagnosis in electrical machines.
Antero Arkkio – played a very active role in providing comprehensive guidance and
supervision throughout this work. Since he is the developer of the finite element
electromagnetic field simulation code used in this work for modelling the electrical
machinery, he made a major contribution by implementing the faults and by giving an insight
into the capabilities and features offered by this tool.
Pedro Jover Rodriguez – made an important contribution in building and tuning the
experimental measuring set-up. As a member of this project affiliated to the Laboratory of
Electromechanics, he also contributed to this work with several comments on various aspects
of the research.
A part of the results presented in this thesis were previously reported in conference and
journal papers. Among the most representative of these papers are the followings: Negrea et
al. (2002, 2004, 2005, 2006) and Pöyhönen et al. (2003). This thesis collects, unifies and
summarises these previous reported results besides presenting the ones that were not yet
reported in a publication.

15
1.4 Structure of the work

The structure of the research work is reflected in the divisions of the thesis:
1. Chapter 2 presents an overview of the faults to be found in various types of electrical
machines together with a description of the possible causes, failure mechanisms and
symptoms produced by each fault.
2. Chapter 3 offers a general view of the fault indicators usually used for detecting faults in
electrical machines, with a special focus on current and electromagnetic flux monitoring,
since this thesis relies only on the indications provided by these parameters.
3. Chapter 4 deals with the description of the methods of analysis used in this work:
numerical electromagnetic field analysis (simulations) and measurements. Following a short
history of the development of finite element methods in the area of electrical engineering, this
chapter then describes the finite element analysis procedure used in the modelling of a cage
induction motor. The faults implemented in the motor’s structures and their modelling
procedures are discussed. A description of the measurement set-up used for the validations of
the results obtained from simulations and a brief presentation of the signal processing
techniques used in the field of fault detection in electrical machines is also to be found in this
chapter.
4. Chapter 5 reports the results obtained from the simulations and measurements.
5. Finally, the results obtained both from the simulations and measurements are discussed in
Chapter 6, while Chapter 7 concludes by considering the most important findings of this
work. The areas of the work which the author believes could be extended and improved in
future research work are also discussed in Chapter 7.

16
2 FAULTS IN ELECTRICAL MACHINES
2.1 General

A fault in a component is usually defined as a condition of reduced capability related to

specified minimal requirements and is the result of normal wear, poor specification or design,
poor mounting (here also including poor alignment), wrong use, or a combination of these. If
a fault is not detected or if it is allowed to develop further it may lead to a failure (Thorsen
and Dalva 1999).

Several surveys have been carried out on the reliability of electrical machines. In such
surveys, a large number of machine operators were usually questioned on the types and
frequency of faults occurring in their plant. The largest of these surveys, carried out by the
General Electric Company, was reported in an EPRI (Electric Power Research Institute)
publication (1982) and covered about 5000 motors, approximately 97% of which were three-
phase cage induction motors. According to this survey, Fig. 2.1 presents the distribution of
faults occurring in the motors surveyed. It must be noted that Fig. 2.1 provides data from
machines working in many different applications and in several different branches of
industry. It is known that the occurrence of any of the fault types will depend heavily on the
specific application of the machine. For example, it has been found that in cage induction
machines, the incidence of rotor cage failures can be at least as high as stator winding failures
in applications where the machine is continuously being stopped and restarted under a heavy
load (drilling machines in the oil and mining industries).

1
Fig. 2.1 Distribution of faults (EPRI 1982).

17
Fig. 2.2 Distribution of failures among failed components for electrical machines working in
the petrochemical industry (Thorsen and Dalva 1999).

Based on the work of Thorsen and Dalva (1999), Fig. 2.2 is intended to highlight the failure
distribution among failed components in 483 high-voltage induction motors working in the
same branch of industry, i.e. the petrochemical industry. Such motors very often operate in
extreme conditions within offshore activities. They are often started directly on-line, which
leads to large starting currents and torque pulsation. These conditions are harmful for the
motor, weakening different machine components in time. Comparing the results of this survey
with the ones presented by EPRI (Fig. 2.1), it becomes clear that the occurrence of a specific
fault type depends considerably on the specific application of the machine and on the
environment the machine is operating in.
Since some electrical machines are subject to different environmental conditions (such as
moisture intrusion in most offshore activities), it is important to have an idea about the
dependence of the failure rate on the environment. In this respect, Thorsen and Dalva (1999)
show that the failure rate for motors situated outdoors in extremely tough conditions (in both
onshore and offshore plants) may be 2.5 times higher than the failure rate for motors situated
indoors.

2.2 Fault types in electrical machines

This section presents a comprehensive description of the most common faults to be found in
electrical machines. For each fault, the possible causes and mechanisms of failure are briefly
outlined.
According to Nandi and Toliyat (1998), the major faults arising in electrical machines may
generally be classified as:
- stator faults resulting in the opening or shorting of the winding,
- turn to ground faults,
- abnormal connection of the stator windings,
- a broken rotor bar or cracked rotor end-rings,
- static and/or dynamic air-gap irregularities,
- a bent shaft which results in rub between the stator and rotor, causing serious damage to
the stator core and windings,
- shorted rotor field winding,
- demagnetisation of permanent magnets,
- bearing and gearbox faults.

18
2.2.1 Winding faults – Stator- and rotor-related
General
Industrial surveys and other studies have shown that a large percentage of failures in an
electrical machine result from a fault related to the stator winding and core (EPRI 1982, IEEE
1985). Many works have indicated that the majority of motor stator winding failures result
from the destruction of the turn insulation. In most cases, this failure starts as a turn-to-turn
fault that finally grows and culminates in major ones such as coil-to-coil, phase-to-phase, or
phase-to-ground failure, ultimately causing motor breakdown (Kliman et al. 1996).
Shorted turns in the stator winding belong to that class of faults that may often have a
negligible effect on the performance of the machine but the presence of which may eventually
lead to a catastrophic failure. Therefore, stringent demands for means to minimise the
occurrence and mitigate the effects of turn insulation breakdown are highly desirable (Stavrou
et al. 2001).

Causes
The stator winding of an electrical machine is subject to stresses induced by a variety of
factors, including, among the chief ones, thermal overload, mechanical vibrations, and voltage
spikes caused by adjustable-frequency drives etc. According to Tavner and Penman (1987),
some of the most frequent causes of stator winding failures are:
- high stator core or winding temperatures,
- slack core lamination, slot wedges, and joints,
- loose bracing for end winding,
- contamination caused by oil, moisture, and dirt,
- short circuits,
- starting stresses,
- electrical discharges,
- leakages in the cooling systems.

Failure mechanisms and symptoms produced by the fault

Early investigations on failure mechanisms in motors (Crawford 1975) concluded that the
great majority of failures seemed to be associated with wire insulation, resulting in low-power
intermittent arcing, which causes erosion of the conductor until enough power is drawn to
weld them. Crawford claims that once the welding has occurred, high induced currents in the
shorted loops lead to rapid stator failure.

Short circuits in stator winding

In large generators and motors in power plants, the stator and rotor winding insulation is
exposed to a combination of thermal, electrical, vibrational, thermo-mechanical, and
environmental stresses during operation. In the long term, the multiple stresses cause ageing,
which finally leads to insulation breakdown. For this reason, it is important to estimate the
remaining insulation integrity of the winding after a period of operating time.
Deterioration of the winding insulation usually begins as an inter-turn fault involving a few
turns of the winding. A turn fault in the stator winding of an electrical machine causes a large
circulating current to flow in the shorted turns. Such a circulating current is of the magnitude
of twice the locked rotor current; it causes severe localised heating and sustains favourable
conditions for the fault to rapidly spread to a larger section of the winding (Kliman et al.
1996). The locked rotor currents are of the order of 6-10 times the rated current (Rotating

19
Electrical Machines Part 15 1995), (Wiedenbrug et al. 2003). If left undetected, turn faults can
propagate, leading to catastrophic phase-ground or phase-phase faults.
Excessive heating caused by turn-to-turn shorts is the reason why motors in this condition will
almost always fail in a matter of minutes, if not seconds. A basic rule of thumb to consider is
that every additional 10°C causes the winding to deteriorate twice as fast as when the
operation takes place in the allowable temperature range. Failure of the insulation between
the winding and ground can cause a large ground current, which would result in irreversible
damage to the core of the machine. This fault may be so severe that the machine might even
have to be removed from service. If the fault is detected at an early stage, the machine can be
put back into service by just re-winding the stator, while, on the other hand, replacing the
whole motor means increased downtime (Tallam et al. 2003).
For high-voltage machines and large low-voltage machines, the development of a time delay
between a direct turn-to-turn short circuit and ground insulation failure is very short, probably
only a few seconds. For these types of machines, regular monitoring of the winding insulation
condition utilizing on-line Partial Discharge analysis was successfully used since the ‘70s
(Nattrass 1993). On-line monitoring of discharge activity in the structure of a stator winding
produces an accurate indication of the deterioration process. Regular monitoring provides the
opportunity for early detection of problems and possible remedial action thereby prolonging
the life of the machine. For smaller machines, the development of a time delay between a
direct turn-to-turn short circuit and ground insulation failure can be from some minutes up to
as much as some hours, depending on the severity of the fault and the loading of the motor.
Another fault associated with the stator winding is called “single-phasing”. In this case, one
supply line or phase winding becomes open-circuited. The resulting motor connection has a
line voltage directly across two phases (assuming a “star” connected machine) which is
equivalent to a single-phase circuit.
The effect of an insulation fault between turns is to eliminate a turn or group of turns from the
stator winding. This will be of little consequence but it will be quantifiable in the flux
distribution in the air-gap (Penman et al. 1994).
Fig. 2.3 shows an inter-turn short circuit between two points, a and b, of a complete stator
winding. The path to the circulating current between these points is closed and the path A–A’
can be expanded into two independent circuits. Fig. 2.3 shows that the two currents, the phase
current and the current which flows in the short circuited part, produce opposite MMFs.
Therefore, inter-turn short circuits have a cumulative effect in decreasing the MMF in the
vicinity of the short-circuited turn(s). Firstly, when a short circuit occurs, the phase winding
has less turns and, therefore, less MMF. Secondly, the short-circuit current MMF is opposite
the MMF of the phase winding. The circulating current Ic is a result of the galvanic contact
between points a and b but also due to the contribution brought by the transformer effect or
mutual induction.

Fig. 2.3 Inter-turn short circuit.

20
2.2.2 Stator core-related faults

General

Stator core problems are rare (1% of all faults, according to Fig. 2.1) compared to stator
winding problems and such problems are not usually a major concern for small machines.
However, the repair/rebuild process is more costly in the case of a stator core failure, since it
usually requires the entire core to be replaced. Therefore, there has been interest in identifying
the causes of core problems and finding ways of monitoring the core in order to detect and
prevent stator core failure, especially in the case of large machines, where the cost of repair
and outage can be significant (Tallam et al. 2003).
Tavner and Anderson (2005) stated that such faults are relatively rare, even for large electrical
machines. The same authors claim that on many occasions the details of such failures assume
major commercial significance, and therefore failure investigations have, of necessity, to be
handled in a confidential manner, touching as they do on the design, manufacture, operation,
and insurance of large electrical plant. This may be one of the reasons why no literature on
core faults has been published but the scientific principles of the mechanisms at work have
been studied in considerable detail and papers published on those principles in the
international literature.

Causes

The stator cores of electrical machines are built from thin insulated steel laminations with the
purpose of minimising the eddy current losses for higher operational efficiency. In the case of
medium/large machines, the core is compressed after the core laminations are stacked in order
to prevent the individual lamination sheets from vibrating and to maximise the thermal
conductance in the core.
According to Tavner and Penman (1987) and Kliman et al. (2004), the main causes of stator
core failure are:
- core end-region heating resulting from axial flux in the end-winding region,
- core melting caused by ground fault currents,
- lamination vibration resulting from core clamping relaxation,
- loosening of core-tightening at the core end resulting from vibration during operation,
- relaxation of lamination material resulting from the compression of insulation material
with time and temperature,
- manufacturing defects in laminations – non-uniform thickness within lamination sheets
causes cumulative non-uniform pressure distribution,
- inter-laminar insulation failure,
- mechanical damage to the inner surface of the stator during assembly, inspection, re-
wind, and re-wedge,
- heat, chemicals, or mechanical force applied when stripping the winding during rewind,
- stator-rotor rub during assembly and operation,
- arcing from winding failure,
- foreign particles introduced during assembly, inspection, or repairing,

Inter-laminar faults are very difficult to monitor on-line since the fault causes only localised
flux re-distribution and heating. The core of a large machine is usually inspected during or
after manufacturing, during regular maintenance, and after repair. Before any thermal or
electromagnetic techniques were developed for detecting inter-laminar insulation failure, the
detection of core faults relied on visual inspection (Lee et al. 2005).

21
Mechanisms of failures and symptoms produced by the fault

If laminations are shorted together for one of the reasons above, a circulating eddy current
larger than that found in normal operation is induced in the fault loop. The circulating fault
current causes additional power loss in the core and results in localised heating, which may
grow in severity and eventually cause the laminations to burn or melt. As a result, the stator
insulation and windings can also be damaged, causing ground current through the stator core,
which may potentially cause machine failure.

2.2.3 Rotor-related faults

General
Because of different rotor constructions, consisting of:
- rotor bars − for cage induction machines,
- rotor windings − for conventional synchronous machines and slip-ring induction machines,
- rotor permanent magnets − for permanent magnet machines,
and constituent materials:
- aluminium (copper, steel) for cage rotor bars,
- permanent magnets for rotors of permanent magnet machines,
- copper wires for the wound rotor of synchronous machines,
rotor faults may be considered to be more complex and various than stator-related ones.
Following the previous description of the rotor configurations and constituent materials, the
most common rotor faults an electrical machine may encounter may be classified as:
- fractures (breakage) of rotor bar and/or end-ring in cage induction motors,
- short-circuits in the field winding occurring in conventional synchronous machines with
a wound rotor,
- demagnetisation of the permanent magnets in permanent magnet machines,
- rotor pole displacements in permanent magnet machines and synchronous machines.

Short circuits in rotor winding - failure mechanisms

Short-circuited turns in power generator rotor windings cause operational problems, such as
high vibration levels; therefore, early detection is essential.
Similarly to the case of stator winding-related faults, inter-turn short circuits usually appear
because of mechanical, electromagnetic, or thermal stress conditions.
Normally, the resistance of the windings on opposite poles is identical. The heat produced by
Joule’s effect is distributed symmetrically about the rotor forging. If the inter-turn insulation
is damaged in such a way that two or more turns of the winding become short-circuited, then
the resistance of the damaged coil diminishes and, if the poles are connected in series, less
heat is generated than in the symmetrical coil on the opposite pole. The rotor body thus
experiences asymmetric heating, which produces a thermal bow in the rotor body, causing
vibration. The unbalanced magnetic forces on the rotor produced by the change in the
magneto-motive force (MMF) from the winding contribute to increased vibration (Ramirez-
Nino and Pascacio 2001).

22
Rotor failures of the induction machines - physical structural damages
General
Unlike stator design, cage rotor design and manufacturing has undergone little change over
the years. As a result, rotor failures now account for around 10% of total induction motor
failures (EPRI 1982, Kliman et al. 1996). However, in the field of fault diagnosis and the
condition monitoring of electrical machines, most of the research presented in the literature
deals with induction motor rotor failures, while bearing-related failures, which account for
40-50% of motor failures, are not so widely discussed. Rotor cage-related faults perhaps
received so much attention in the literature as a result of their well-defined associated fault
frequency components.
Causes
Manufacturing process defects
For a rotor cage, physical damage faults can arise at the manufacturing stage through
defective casting in the case of die-cast aluminium rotors, or through poorly welded or
brazed bar-to-end-ring joints in the case of fabricated rotor cages. A defective cast
aluminium rotor may have air bubbles within the casting, thus increasing the resistances of
the rotor bars and consequently resulting in hot spots in the bars where the resistance is
greatest and which could lead to a complete fracture of the bar (Paterson 1998).
Severe operational conditions
Under normal operating conditions, large mechanical and thermal stresses are present,
especially if the machine is being continually stopped and restarted or if the machine is
heavily loaded. It is well known that the rotor current during starting can be as much as ten
times the normal full load current and that the effects of these large currents are
represented by very large thermal stresses in the rotor circuit. The starting period is also
characterised by minimal cooling and maximum mechanical forces, which over-stress the
rotor bars.

Mechanisms of failures and symptoms produced by the fault

The sequence of events following the cracking of a rotor bar is described as follows: the
cracked bar will increase in resistance and will overheat at the crack. The bar will break
completely and arcing will occur across the break. This arcing will then damage the rotor
laminations around the faulted bar. The neighbouring bars will carry an increased current
and will be subject to increased stresses, eventually causing these bars to fail. Finally, the
broken bars may lift outwards because of centrifugal forces and could catastrophically
damage the stator windings (Paterson 1998).

Eccentricity
General
Machine eccentricity is defined as a condition of the asymmetric air-gap that exists
between the stator and rotor (Vas 1993). The presence of a certain level of eccentricity is
common in rotating electrical machines; some manufacturers and users specify a maximum
permissible level of 5 percent, whereas in other cases, a maximum level of 10 percent of
the air-gap length is allowed by the user (Thomson and Gilmore 2003). However,
manufacturers normally try to keep the total eccentricity level even lower in order to
reduce vibration and noise and minimise unbalanced magnetic pull (UMP) (Nandi and
Toliyat 1998). Since the air-gap of an induction machine is considerably smaller than in
other types of machines with a similar size and performance, this type of machine is more
sensible to changes in the length of the air-gap.

23
 There are two types of air-gap eccentricity: static air-gap eccentricity and dynamic air-gap
eccentricity (Fig. 2.4). In the case of static air-gap eccentricity, the position of the minimal
radial air-gap length is fixed in space, while in the case of dynamic eccentricity, the centre
of the rotor is not at the centre of the rotation and the position of the minimum air-gap
rotates with the rotor. However, the static and dynamic eccentricities are basic
classifications, since varieties and modifications such as unilateral eccentricities, as well as
angular and radial misalignments, are just as possible.

a) concentric b) static eccentricity c) dynamic eccentricity

Fig. 2.4 Eccentricity types.
Causes
Static eccentricity may be caused by the ovality of the stator core or by the incorrect
positioning of the rotor or stator at the commissioning stage. Assuming that the rotor-shaft
assembly is sufficiently stiff, the level of static eccentricity does not change.
The dynamic eccentricity may be caused by several factors, such as manufacturing
tolerances, wear of bearings, or misalignment, mechanical resonance at critical speed, and
incorrect manufacture of the machine components. Rotor “whirl” near a critical speed is
another source of dynamic eccentricity and is an important consideration in larger,
flexible-shaft machines.

Mechanisms of eccentricity production and symptoms produced by the fault

In reality, both static and dynamic eccentricities tend to co-exist. An inherent level of static
eccentricity exists even in newly manufactured machines as a result of manufacturing and
assembly methods. This causes a steady UMP in one direction and with usage this may
lead to a bent rotor shaft, bearing wear and tear etc., resulting in some degree of dynamic
eccentricity. Unless detected early, the eccentricity becomes large enough to develop high
unbalanced radial forces that may cause stator-to-rotor rub, leading to a major breakdown
of the machine (Barbour and Thomson 1997).

2.2.4 Bearing faults

General
Because of the close relationship between motor system development and bearing assembly
performance, it is difficult to imagine the progress of modern rotating machinery without
considering the wide application of bearings. As reported by Kliman et al. (1997) and EPRI
(1982), bearing faults may account for 42%-50% of all motor failures. Motor bearings may
cost between 3 and 10% of the actual cost of the motor, but the hidden costs involved in
downtime and lost production combine to make bearing failure a rather expensive
abnormality (Barker 2000).

24
Bearing faults might manifest themselves as rotor asymmetry faults, which are usually
included in the category of eccentricity-related faults. Otherwise, ball bearing-related defects
can be categorised as outer bearing race defects, inner bearing race defects, ball defects, and
train defects. Figure 2.5 presents the artificially created outer bearing race fault studied in this
work.

Causes, mechanisms of failure, and symptoms produced by faults

Different stresses acting upon a bearing may lead to excessive audible noise, uneven running,
reduced working accuracy, and the development of mechanical vibrations and, as a result,
increased wear. As long as these stresses are kept within the design capabilities of the bearing,
premature failure should not occur. However, if any combination of them exceeds the
capacity of the bearing, then the lifetime may be drastically diminished and a catastrophic
failure could occur.
More than twenty years ago, few bearing failures were electrically induced but at the
beginning of the ’90s a study by Kerszenbaum (1992) showed that bearing failures are about
12 times as common in converter-fed motors as in direct-on-line motors. This relatively high
percentage of electrically induced motor bearing failures is due to the modern high-frequency
switching power devices that were rapidly developing in that period. Such devices,
employing, for instance, bipolar junction transistors (BJTs) and faster (shorter rise time as a
result of fast switching) insulated gate bipolar transistors (IGBTs) produce unintended
consequences for peripheral equipment, generally described as electromagnetic interference
(EMI) (Busse et al. 1997). Concerning this issue, relying on simulations, analytical
expressions, and experiments, Mäki-Ontto presents some methods for the mitigation of shaft
voltages and bearing currents in frequency converter-fed AC motors (Mäki-Ontto 2006).
However, Barker (2000) claims that mechanical issues remain the major cause of bearing
failure. The same author provides a list of reasons and mechanisms that usually cause bearing
failures:
- thermal overloading,
- misalignment of the shaft,
- excessive loading (both static and/or dynamic), (axial/radial combined),
- mechanical overload,
- excessive shock and vibration,
- inappropriate shaft fit,
- machining defects,
- bad handling and/or mounting,
- improper application,
- improper installation,

Fig. 2.5 Artificially created bearing fault studied in this work.

25
- heavy radial and axial stresses caused by shaft deflection,
- lifetime load profile,
- environmental/external problems,
- contamination and corrosion caused by pitting and the sanding action of hard and
abrasive minute particles or corrosive action of water, acid, etc.
- improper lubrication, including both over- and under-lubrication, causing heating and
abrasion
- bearing currents,
- shear stress.

26
3 FAULT INDICATORS FOR ELECTRICAL MACHINES
3.1 General
The history of fault diagnosis, condition monitoring, and protection is as old as technical
devices themselves. Generally, on-line condition monitoring and diagnostics requires the
sensing and analysis of such signals that contain specific information (symptoms) which is
characteristic of the degradation process, problem, or fault to be detected. Various factors
need to be considered when selecting the most appropriate monitoring technique for
application in an industrial environment. The most important factors, according to Thomson
(1999), are listed below:
1. the sensor should be non-invasive,
2. the sensor and instrumentation system must be reliable,
3. the diagnosis must be reliable,
4. the severity of the problem should be quantified,
5. ideally, an estimation of the remaining run-life should be given,
6. ideally, a prediction of the fundamental cause(s) of the fault should be provided via on-
line information from sensors etc.
It is extremely difficult and, in most cases, impossible to satisfy all the above criteria, mainly
because of the complexity of the degradation mechanisms, abnormalities, and nature of the
fault. Because of the world-wide interest in the condition monitoring and diagnostics of drive
systems, substantial advances have been made during the last fifteen years and in many cases
it is possible to achieve Criteria 1 to 4. However, Criteria 5 and 6 are extremely difficult to
achieve.

3.2 Potential fault indicators for a technical device

Because of the wide variety of physical phenomena to be found in a technical device, several
fields of science and technology need to be considered when designing and developing
competitive monitoring and diagnosis systems. As an example, Fig. 3.1 illustrates the
complexity of the interactions of physical phenomena characteristic for a general
electromagnetic energy converter structure.

Fig. 3.1 Interactions of physical phenomena characteristic for a general electromagnetic

energy converter structure (Driesen 2000).

27
Different fields of physics, such as electrical, mechanical, thermal, fluid flow, and motion
interact in a complex manner, as depicted in the previously mentioned picture. Various
parameters belonging to these fields may be found to be suitable potential fault indicators for
the technical device. The control system that governs the overall system and the power
electronic system specialised in translating its commands into energetic signals supplied to the
device may together be considered the core of a condition monitoring and fault diagnosis
scheme.

3.3 Potential fault indicators for electrical machines

3.3.1 General

It was previously claimed that the monitoring and diagnosis of electrical machines is a very
popular topic, since it corresponds with industrial requests for an increasing number of
applications for which reliability is the key point. In this respect, sensor implementation is
fundamental for the development of industrial methods at a cost compatible with the
applications in which monitoring and diagnosis are necessary.
In order to ensure safe and reliable operation, the manufacturers and users of electrical
machines initially relied on simple protections such as over-current, over-voltage, or earth-
fault. However, as the tasks performed by electrical machinery grew more complex,
improvements were also sought in the area of fault diagnosis (Nandi and Toliyat 1999).
A number of potential measurement parameters are likely to provide nowadays useful
condition-indicating information for possible failure modes in any electrical machine.
Thorsen and Dalva (1999) categorise mainly these parameters as mechanical, like vibrations,
acoustic, speed fluctuations, and electromechanical, like currents, electromagnetic leakage
fluxes, surges, and partial discharges. In addition, come temperature, oil particle and gas
analysis.
Following, a short description of the most popular mechanical parameters, temperature, oil
particle and gas analysis based methods is presented. The electromechanical parameters used
in the field of fault diagnosis and condition monitoring of electrical machines will be
discussed in more detail in the following sections of this thesis.
Acoustic emission monitoring works with ultrasonic but also with audible frequencies, and is
particularly promising for detecting bearing faults. The contact between rolling elements with
and without cracks generates waves that propagate through the machine with the speed of
sound. The waves have little energy, but their high frequencies can be detected by
piezoelectric transducers. Rienstra and Hall (2004) present a study on the basic principles of
the acoustic monitoring and on the industrial use of this technique for bearing fault detection.
However, the acoustic monitoring has also proven feasible in the attempt to identify electrical
faults, such as loose stator coils (Gaylard et al. 1995). In this case, the identification procedure
was performed automatically by a neural network.
Nau and Mello (2000) indicate the causes of acoustic noise in induction motors and present
methods to reduce the noise before and after the motors are manufactured.
Vibration monitoring uses vibration transducers such as measuring accelerometers of
piezoresistive types with linear frequency spectrum. Vibrations are caused by magnetic,
mechanical and/or aerodynamic forces. The measuring parameters can be displacement,
velocity and acceleration. Vibrations are measured directionally, e.g. radially or axially, and
the transducers are often placed on the bearings for detecting mechanical faults (Wang and
Gao 2000). However, by placing probes on the stator, it is also possible to detect non-even air
gap (Cameron et al. 1986), stator-winding or rotor faults (Trutt et al. 2002), asymmetrical

28
power supply (Chow 1996) and unbalances in the driven load (Obaid and Habetler 2003),
(Leonard and Thomson 1986).
The users of electrical machinery pay a special attention to the temperature monitoring since
the basic rule of thumb claiming that ‘every additional 10°C causes a winding to deteriorate
twice as fast as when the operation takes place in the allowable temperature range’
represents a very serious concern for the good operation of their machinery. Other
components of electrical machinery may be irreversibly affected by higher temperatures; the
case of permanent magnets found in the structure of a permanent magnet machine rotor is a
good example in this respect. The magnetic characteristics of permanent magnets are
temperature-dependent and a high temperature may lead to irreversible demagnetisation of
the permanent magnets. Negrea et al. (2001) present a study dealing with the temperature
distribution in a permanent magnet synchronous motor used for ship propulsion drives. More
specifically, the transient thermal behaviour of the motor under fault conditions, i.e. during
short circuits occurring in the terminals of the test motor is studied, and various
considerations on the issues of stator winding and permanent magnets protection are
presented.
Temperature monitoring devices installed to monitor bearing temperatures are also often used.
Remembering that bearing failures cause the majority of motor failures, many maintenance
departments utilize thermal image scanning to look for abnormally hot spots on in-plant
inspections (Malinowski and McCormick 2002).
For figuring out ventilation malfunctions, the coolant bulk outlet temperature is monitored,
particularly when the machine is stressed beyond its rated data.
Gas in oil analysis is the traditional way to monitor insulation condition. Dissolved gases in
the oil produced by thermal ageing can provide an early indication of an incipient fault. Gases
normally analyzed are hydrogen, oxygen, carbon monoxide, carbon dioxide, methane, ethane
ethylene, and acetylene. The gas in oil analysis together with the oil particle, and other
methods relying on chemical analysis, are extensively presented by Tavner and Penman
(1987).
Table 3.1, based on Payne et al. (2002), briefly presents:
- what type of instrumentation is required to monitor some of the most popular machine
parameters used in fault detection in the case of a transverse flux motor,
- the degree of accuracy of fault indication that may be obtained when relying on a
specific parameter,
- the level of expertise an operator needs in order to interpret the recorded data,
- how invasive a dedicated sensor for each fault indicator would be,
- possible means of analysis (signal processing techniques).
Even though the content of Table 3.1 refers to a transverse flux motor, which is not typical of
industrial drives, this may also be considered as being representative of any type of electrical
machine, since the same potential measurement parameters are presented by other research as
likely to provide useful condition-indicating information for possible fault modes.
From the multitude of previously-mentioned machine parameters, some authors claimed that
on-line technologies based on temperature, axial flux, and shaft current/voltage
measurements present enough ability to protect the motor against fault conditions which were
previously difficult or impossible to detect (Bowers and Piety 2001). These measurements are
known as proactive. A proactive concept may be defined as the capability of controlling a
situation by controlling a situation capable of causing a fault rather than waiting to respond to
it after it happens.

29
Table 3.1 Fault indicators in electrical machines (Payne et al. 2002),

30
3.3.2 Indicators for detecting winding-related faults

The detection of stator winding faults in low-voltage motors during operation has been a
problem, mainly because such a fault is not always distinguishable when the stator current is
monitored (Joksimovic and Penman 2000). A large amount of work has been carried out on
developing condition monitoring techniques based on other machine parameters such as:
- axial leakage component of the electromagnetic flux (Penman et al. 1994),
- electrically excited vibrations (Trutt et al. 2002),
- negative-sequence impedance (Kohler et al. 2002), (Melero et al. 2003),
- instantaneous power (Legowski et al. 1996),
- partial discharge testing (Green et al. 2005),
- electromagnetic torque (Hsu 1995).
Frequent changes in the temporal behaviour of the power supply cause unbalance, which in
turn obscures the fault signature, resulting in false alarms. Such a false alarm could lead one
to detect the presence of a stator fault, even though the root cause of the problem is supply
unbalance. Similar arguments could be made regarding the impact of low-frequency load
variations and load changes on mechanical fault detection and the effectiveness of various
methods in detecting such problems (Parlos and Kim 2001).
To detect shorted turns in the rotor windings, several methods have been used, such as the air-
gap search coil technique (Connolly et al. 1985, Ramirez-Nino and Pascacio 2001), the
monitoring of the circulating stator current in double-circuit machines (Muhlhaus et al. 1985),
measurement of the rotor shaft voltage, or monitoring the harmonic components present in the
generator excitation current for synchronous generators (Penman and Jiang 1996).

3.3.3 Indicators for detecting rotor-related faults

The consequences of faulty rotors are poor starting performance, excessive vibration, and
higher thermal stresses. All these contribute to the further deterioration of the rotor, as well as
to secondary failures of the stator.
Various methods have been proposed for the detection of rotor faults, relying on monitoring
motor torque (Thomas et al. 2003), rotor speed (Hargis et al. 1982), electromagnetic flux
(Elkasabgy et al. 1992, Penman et al. 1994), vibrations in the machine stator housing
(Cameron et al. 1986), and stator current (Kliman 1988, Benbouzid 2003). The analysis of the
current spectrum is the most popular, because of its simple way of recording the stator input
current while the motor is running under load hence no interruption to the machine operation
being required.
Relying on extensive experiments carried out on an induction motor with various rotor faults
and under various load conditions, Trzynadlowski and Ritchie (2000) have shown the
usefulness of the instantaneous power as diagnostic media for rotor faults in induction
machines. The stator current, which is regularly used in motor signature analysis, yielded
inferior results.

3.3.4 Indicators for detecting bearing faults

Bearing faults may lead to increased vibration and noise levels. Considering these symptoms,
it is easy to understand why previous research has found out that bearing faults are best
detected by monitoring vibrations, shock pulses, and acoustic emissions. However, many of
the previous investigations and methods proposed to detect bearing faults are based on the
spectral analysis of electrical quantities (Schoen et al. 1995, Cardoso and Saraiva 1993, Obaid

31
et al. 2000). They have the advantage that they work with standard current sensors that are
already available in most drive applications and may provide the same indications without
requiring access to the motor by correlating the characteristic bearing frequencies to the
spectral components of the stator currents (Kliman and Stein 1990). The fault spectral lines in
the current can be determined on the basis of the bearing geometry and the rotational speed
(Schoen et al. 1994).
However, a fault signature is detectable in the stator current only if the bearing fault causes a
displacement of the rotor within the air-gap, resulting in a distortion of the air-gap field.
Hence, it is hard to obtain a reasonable signal-to-noise ratio at an early stage of bearing fault.
For a 15 kW four pole induction motor, Lindh et al. (2003) have investigated the use of stator
current signal for the detection of an outer ring defect of a ball bearing with normal radial
clearance. It was found that the stator current measurement as a bearing fault indicator is not
adequate for this motor type since the modification produced by the radial movement of the
rotor was found very small if the radial movement was restricted with bearing with small
radial clearance. An outer race defect was clearly indicated only in the case of the large
internal radial clearance of the bearing.
Obaid et al. (2003) claim that the main disadvantage of using the current for monitoring the
condition of bearings comes from the difficulty of distinguishing bearing fault signatures from
non-fault components or noise in the stator current.
Stack et al. (2004) also claim that, the main disadvantage of stator-current-based bearing fault
detection is that the effects of a bearing fault are often subtle and difficult to predict. This is
the reason why they propose a modeling technique where the changes in the stator current
spectrum are compared to a baseline spectrum, rather than searching for specific fault
components. These changes in spectral content are then used to identify developing faults.
Before this modeling technique is applied, the stator current should be filtered to remove the
significant non-bearing fault components such that only changes in bearing health are tracked.
Stack et al. (2004) claim and there are no bearing fault detection techniques in industry that
are current-based.
Additional methods based on special sensors placed in the machine in order to detect shaft
voltages or axial magnetic flux caused by a bearing defect have also been reported (Ong et al.
2000). Besides an increase in costs, the mounting of additional sensors is, however, also a
practical problem in terms of motor design and approval by the manufacturer, operator, or
safety legislation authorities.
Kliman et al. (1997) claim that bearing faults may also produce small torque perturbations.
These perturbations are major for some types of faults (such as a pit in the race), resulting in
predictable frequencies, but very small for other types, such as a dented cage or damaged
balls. Therefore the necessity arises that a monitoring system must be able to resolve very
small components of the spectrum, even when their frequencies are not definitely known, and
recognise when they differ from a nominal or healthy motor. In this respect, Schoen et al.
(1994) use neural network techniques with the purpose of characterising the spectra of the
stator current that are associated with the normal state of a motor and load and then
determining when the spectra have changed significantly from the nominal to indicate a fault.
Li et al. (2000) used the commercially available Machinery Fault Simulator (SpectraQuest) to
obtain vibration data from bad bearings to train a neural network with the purpose to identify
fault severity. Both simulation and real-world testing results obtained indicate that neural
networks can be effective agents in the diagnosis of various motor bearing faults through the
measurement and interpretation of motor bearing vibration signatures

32
3.3.5 Indicators for detecting eccentricity

If a certain level of eccentricity between the rotor and the stator occurs in an electrical motor,
whatever its origin, it causes new air-gap field harmonics to appear and/or an increase in the
amplitude of the previously existing harmonics. Then, it produces a global effect that
stimulates the development of the following side effects (Duque et al. 2004):
- unbalanced magnetic pull,
- parasite torque,
- intensification of vibration and noise levels,
- decrease in the rotor speed,
- electric current flowing through the bearings.
Many surveillance schemes determine the Fourier spectrum of a single line current in order to
monitor the condition of the motor (Thomson 1994, Nandi and Toliyat 2002). These schemes
evaluate additional fault specific harmonics that are due to rotor misalignment. The location
of these harmonic waves is given by the number of rotor bars and the actual slip (Hirvonen
1994, Barbour and Thomson 1997, Dorrell 1997). It is very important to sense the
misalignment between the motor and mechanical load, since this may initiate a radial force,
which pushes the rotor to the side and can produce higher eccentricity ranges.

Vibration signals can also be monitored in order to detect eccentricity-related faults. Cameron
(1986) gives the high-frequency vibration components to be monitored for detecting static or
dynamic eccentricity. The relationship of the vibration of the bearing to the stator current
spectra can be determined by remembering that any air-gap eccentricity produces anomalies
in the air-gap flux density. Most bearing defects produce a small radial motion between the
rotor and stator of the machine that may be perceived as a form of eccentricity.

Mechanical unbalances give rise to two first-order current harmonics. Because of the
interaction of the currents and voltages, both these current harmonics are also reflected by a
single harmonic component in the frequency spectrum of the electric power. It is claimed that
this single component is easier to assess than both the current harmonics (Kral et al. 2004, Liu
et al. 2004).

Eccentricity-specific signals are also present in the electromagnetic flux, which can be
measured by search coils which sense the axial leakage flux and the electromagnetic flux
from the air gap (Früchtenicht et al. 1982, Tenhunen et al. 2003, Dorrell et al. 1995).

Other schemes evaluate mechanical (Dorrell et al. 1997) or acoustic vibrations caused by
mechanical unbalances, torque signatures (Wieser et al. 1999), and radial forces (Smith and
Dorrell 1996, Arkkio 1996).

Cardoso et al. (1997) use a detection technique that evaluates space phasors calculated from
two or three measured currents. This technique analyse the characteristic patterns of the locus
diagram of the current phasor.

33
3.4 Current monitoring
This section presents the state of the art for different monitoring techniques based on motor
current signature analysis (MCSA) that are used to detect and, eventually, localise different
faults in electrical machines.
Why current monitoring?
Vibration measurements have historically been the foundation of most on-line condition
monitoring programs, but new techniques such as those involving spectral analysis of the
electric line current powering the motor are becoming of significant interest. The main
problem concerning the monitoring methods based on, for instance, measurement of the rotor
speed, vibration, and fluxes is that they are essentially invasive, requiring transducers to be
fitted in or around the machine, with an obvious interruption to operation. Besides the
increase in costs, the mounting of additional sensors is also a practical problem in terms of
motor design and approval by the manufacturer, operator, or safety legislation authorities.
The condition monitoring schemes that rely on the analysis of the motor current are the most
attractive, as the current sensors are usually installed by default in the motor control centre for
other control or protection purposes. As a consequence, a variety of methods are applied to
today's cost-effective microprocessor hardware platforms in order to accurately diagnose
impending failures of electrical machines and many commercial products rely on integrated
packages incorporating analysis techniques based only on the voltages and currents available.
The MCSA technique is applied to machines operating under steady state conditions and has a
basic requirement that for reliable diagnosis a substantial current must flow; i.e. the motor
must be operating at or near full load conditions. There are certain situations, however, when
this requirement is impractical, i.e. if the motor has been removed from service and taken to a
workshop for repair. Under such circumstances, the motor can only be tested under no-load
conditions (Burnett et al. 1995).
The current monitoring method is highly likely to meet Criteria 1-4, which were previously
mentioned in Section 3.1 and depicted again in Fig. 3.2.

Fig. 3.2 Factors to be considered when selecting the most appropriate monitoring technique
for application in an industrial environment (Thomson 1999).

3.4.1 Phase current

Stator winding faults

Some diagnostic methodologies have been proposed to detect stator short circuits that rely on
the fact that stator asymmetry causes an increase in the space harmonics. The interaction

34
between the electrical quantities at the supply frequency and the different space harmonics
produces additional time harmonic components in the stator and rotor currents. As a
consequence, the space harmonics of the air-gap flux density may be considered useful to
obtain reliable results. Thomson et al. (1998) and Filippetti et al. (2000a) claim that the
harmonic components of the stator currents caused by the rotor slotting can be used for
diagnostic purposes.
On the other hand, Joksimovic and Penman (2000) state that as a result of the nature of the
rotor, no new frequency components of the line current spectra can appear as a consequence
of the fault in the stator winding of an electrical machine. For the case of a turn-to-turn short
circuit in the stator winding, only a rise in some of the frequency components which already
exist in the line current spectra of a healthy machine was observed both from their
experiments and simulations.
It is well known that degradation of the insulation between stator winding turns introduces a
dissymmetry in the first harmonic components of the currents and consequently a so-called
negative sequence component. It was suggested that this effect may be used for the detection
of an insulation failure (Vas 1993).

Rotor related faults

When used properly, the analysis of the electric current signature of the motor has proven to
be very reliable for quantifying broken rotor bars.
Supposing that the current in a healthy motor has the fundamental supply frequency fs,
changes in the load of the motor modulate the amplitude of the current to produce sidebands
besides this fundamental frequency component. Broken or fractured rotor bars generate a
sideband below the supply frequency, which is displaced by twice the slip frequency (2sfs)
from the supply frequency. This cyclic variation in the current reacts back to the rotor to
produce a torque variation at twice the slip frequency, giving rise to a speed variation twice
the slip frequency. This speed effect reduces the lower sideband fs(1-2s) and produces an
upper sideband at the frequency fs(1+2s). It was found that the magnitudes of these sidebands
are affected by the motor-load inertia (Thomson 1999, Filippetti et al. 1996). Concluding, the
characteristic fault sideband components fbrb around the fundamental for detecting broken bar
faults are given by:

f brb = (1 ± 2ks ) f s k = 1, 2, 3... (3.1)

Deleroi (1982) suggests that additional spectral components should be observed in the stator
line current with the purpose of detecting rotor-cage-related faults:

 k  
f brb = f s   (1 − s ) ± s  (3.2)
 p  
where k has the integer values 1, 2, 3..., p is the number of pole-pairs, s is the per unit slip and
k/p=1, 3, 5…
The most prominent sideband frequency is the one that appears at twice the slip frequency
below the line frequency. The ratio of this lower sideband amplitude to the main supply
frequency component gives an estimation of the severity of the fault, indicating the amount of
broken or fractured bars (Hirvonen 1994, Kliman et al. 1988). Besides the sideband
components at twice the slip frequency, current components near the rotor-slot harmonic
frequency may be found. In the case of broken rotor bars, the modulation of the stator current
at twice the slip frequency can be found on the spectrum by analysing the slot frequencies:

35
  N (1 − s ) 
f brb = f s ⋅  r ± 2n  ± 2 sf s
 p  (3.3)
where n = 0, 1, 2…and Nr is the number of broken rotor bars.
Other techniques, such as those based on monitoring the torque and instantaneous power,
have been shown to be sensitive to rotor bar faults, but have not been proved to be as reliable
or to provide a quantitative estimate of broken bars. In this respect, a comparison and
performance evaluation of different diagnostic procedures that use input electric signals to
detect and quantify rotor breakage in induction machines supplied by the mains was achieved
by Bellini et al. (2000). This work proved that the current signature analysis provides the
information to diagnose the rotor breakage accurately and quantitatively by retrieving an
effective diagnostic index that sums the amplitudes of the two sideband components in the
current spectrum given by Eq. 3.1. Alternative diagnostic methods that use current space
vector modulus, as well as instantaneous power, or instantaneous torque, lose information and
thus do not allow an accurate quantification of the breakage, since they are affected twice by
the speed reaction.
Siau et al. (2003) claim that it is difficult to detect partial rotor bar breakages or make
accurate estimates of the number of broken rotor bars using the current spectrum alone.
A possible solution for improving the accuracy of rotor fault detection and estimation of
severity would be to combine the information from the current and flux signals.
Burnett and Watson (1995) presented a methodology to find the location of broken bars
within a rotor of an induction motor. The technique used to detect the location of broken bars
involved using the stator windings as reference locations within the motor from which the
distance travelled by the broken bar can be computed. This technique was found to be
dependent on the successful detection of the frequency components given by Eq. 3.2.
A drawback of the diagnostic procedure based on MCSA is the possible confusion with the
motor current modulation produced by other events, such as pulsating load, undesired
interaction between the motor and the train equipment, and particular rotor design, which also
cause sideband current components. If the load variation frequency is near, the resulting
current spectrum is similar to that of a faulted rotor (reflected as two anomalous lines in the
current spectrum 2sf from the supply frequency line), but previous experiences showed that
the two causes can still be distinguished (Filippetti et al. 1998). However, a particular rotor
structure design, such as a spidered structure with the same number of legs and poles as are
found in large motors, may produce a magnetic asymmetry whose effect is the same as rotor
electrical asymmetry (Thomson 1992).

Bearing faults
Generally, the bearing condition monitoring techniques rely on the indications provided by
the motor current and vibration. The efficacy of current monitoring for bearing fault detection
by correlating the relationship between vibration and current frequencies caused by incipient
bearing faults was investigated by Schoen et al. (1994). In this respect, it was experimentally
proven that there is a relationship between the vibrational and current frequencies, since the
stator current signature can be used to identify the presence of a bearing fault. This combined
analysis is supported by the fact that the mechanical vibrations are associated with variations
in the physical air-gap of the machine. When ball bearings support the rotor, any bearing
defect will produce a radial motion between the rotor and stator of the machine. Such
variations cause the air-gap flux density to be modulated and the stator currents to be

36
generated at predictable frequencies related to the electrical supply and vibrational
frequencies.
Vas (1993) and Schoen et al. (1994) suggest that the mechanical damages referred to bearing
faults introduce harmonic components in the current spectrum at the frequencies
f bng = f s ± f b (3.4)
where fs is the supply frequency and fb is a mechanical frequency depending on the type of
fault and on the bearing characteristics. Other components at different frequencies can be
introduced by load anomalies and it should be noted that they can cause misunderstandings,
because these components can be confused with those caused by machine faults.
Schoen and Habetler (1997) present a method for removing the load effects from the
monitored quantity of the machine. The methods compares the actual stator current to a model
reference value which includes the load effects and the difference between these two signals
provides a filtered quantity, independent of variations of the load, that allows continuous on-
line condition monitoring to be conducted without concern for the load condition. Simulation
and test results presented the effects on the spectrum of the monitored quantity for both
constant and eccentric air gaps when in the presence of an oscillating load.

Eccentricity

Variation in the air-gap length because of static or dynamic eccentricity leads to variations in
the air-gap flux density. As induced stator currents are affected, many of their frequency
components can help to identify both static and dynamic eccentricity.
Vas (1993) claims that, generally, the presence of static and dynamic eccentricity can be
detected using MCSA. Cameron et al. (1986) found that static eccentricity can be identified
by changes in the mechanical vibrations caused by the rotor-slot harmonics, and dynamic
eccentricity is characterised by the appearance of unique frequency components in both the
current and frame vibration signals. The same authors claim that static eccentricity variations
result in the introduction of dynamic eccentricity components in the current and vibration,
thus indicating that dynamic eccentricity is a by-product of static eccentricity.
According to Cameron (1986), the frequencies of the harmonics resulting from the
asymmetries caused by the slotting and eccentricity can be calculated according to:

f ecc = fs ( kQ2 ± nd )
(1 − s ) ± nw  (3.5)

 p 
where fs denotes the machine supply frequency and Q2 gives the number of rotor slots, nd = 0
in the case of static eccentricity, and nd = 1, 2, 3… in the case of dynamic eccentricity (nd is
known as eccentricity order), s is the slip, p is the number of pole-pairs, k is any integer, and
nw is the order of the stator time harmonics that are present in the power supply driving the
motor (nw = ±1, ±2, ±3, ±5, etc.). In the event that one of these harmonics is a multiple of
three, it cannot exist in the line current of a balanced three-phase machine. However, it has
been shown by Nandi and Toliyat (1998) that only a particular combination of a machine
pole-pair p and a rotor slot number Q2 will give rise to significant components related only to
static or only to dynamic eccentricity. This relationship for a three-phase machine is given by

Q2 = 2 p 3 ( m ± q ) ± r  ± k (3.6)

where m±q = 0, 1, 2, 3, and r = 0 or 1, k = 1.

37
According to Obaid et al. (2000), Dorrell et al. (1995), and Nandi and Toliyat (1998), the
effects of combined static and dynamic eccentricity cause characteristic sideband currents in
the current spectrum given by
1 s
f ecc ,i fs 1 k (3.7)
p
where k is the order number and is an integer. The interaction of these harmonics with the
mainly sinusoidal supply voltage causes eccentricity-specific harmonics in the power and
torque spectrum at:
1 s
f ecc, p fsk (3.8)
p
These low-frequency components also give rise to high-frequency components, as described
by Penman et al. (1994). However, these components are strong only for machines whose
pole-pairs and rotor slot numbers are given by Eq. 3.6, where k = 1, while for machines with k
= 2 these frequency components are rather weak. Since a changing load torque may also result
in current harmonics similar to those calculated with the above equations, a constant load is
usually assumed. Schoen and Habetler (1995 and 1997) found that the magnitudes of the
frequency components caused by load changes are always larger than those of eccentricity
harmonics.
Another motor current-related approach for the detection of eccentricity in induction motors
was suggested by Cardoso and Saraiva (1993) and consists of monitoring the Current Park’s
Vector.

3.4.2 Circulating current

In many electrical machines (usually, but not only, in large synchronous generators), the
stator/rotor windings are parallel-connected in order to deliver the rated required
characteristics at the terminals of the machine.
The additional (parallel) connections serve to generate an MMF which counteracts the
asymmetrically distributed magnetic field in the air-gap resulting from the asymmetrical
construction or assembly of the machine, especially the off-centre position of the rotor in the
stator bore. Such an asymmetrical magnetic field induces mechanical vibration in the rotor
and stator (with the attendant increased load on the bearings), as well as an unbearable noise.
Equalising currents between winding coils are responsible for the drastic reduction of radial
forces and ensure almost vibration-free operation (Berman and Erlicki 1991).
Concerning the use of circulating currents in fault diagnostics, it was claimed that for stator
winding short circuits in double-circuit machines, the measurement of the differential current
between the parallel connected half-phases represents an accurate way to detect a fault
(Penman and Jiang 1996, Tavner and Penman 1987). Single-winding stators present a more
intractable problem, although it has been observed that there are some harmonic changes in
line current under such conditions. The computation of the circulating current under faulty
conditions such as eccentricity and short circuits in the rotor winding of large synchronous
generators was also discussed by Foggia et al. (1999).

3.4.3 Shaft currents

Irregularities in the magnetic circuits of electrical machines may result in unwanted voltages
that lead to shaft currents through the shaft, bearings, bearing supports, and closing through

38
the machine framework. The IEEE Standard Test Procedure for Polyphase Induction Motors
and Generators (IEEE 2004) discusses the shaft current and presents a measurement method
for recording either the voltage across the ends of the shaft or the current.
Ong et al. (1997) claim that only the Rogowski coil measurement yields accurate
measurements of shaft currents, whereas the other method either yields inaccurate
measurements or may result in other problems for the machine's integrity.

3.4.4 Current monitoring - drawbacks

Taking into consideration the following potential drawbacks of a condition monitoring system
relying only on current monitoring, the motivation for approaching and relying on additional
fault indicators seems to be clear and realistic:
- For multiple faults or different varieties of drive schemes, MCSA may not provide
enough ability to discriminate between different faults, since such abnormalities and time
harmonics may end up generating similar signatures.
- The current spectrum is influenced not only by fault conditions but also by other factors,
including the supply, static and dynamic load conditions, noise, machine geometry, and
fault conditions, and these conditions may lead to errors in fault detection.
- When a commercial diagnostic system has to operate between a few kW up to MW range
of different designs and driving a wide range of mechanical loads, providing a diagnosis
from sensing only the current and its subsequent analysis becomes much more complex
than is at first apparent (Thomson 1999).
- In order to facilitate the successful detection of rotor faults during steady state operation,
a large supply current needs to flow. This large current is usually obtained by monitoring
the motor while it is running under full load conditions. The necessity of requiring a large
current may not be appropriate nor achievable in some cases if, for example, the motor
has been taken off-line or removed to a workshop environment. In addition to requiring a
large current, it was found that these monitoring techniques also have difficulty in
detecting other common rotor faults over and above simple broken bars. These faults
include damaged end-rings or broken bars within a double-bar rotor machine (Burnett
and Watson 1995).
- It is very important to bear in mind that the MCSA techniques are not able to detect all of
the possible faults in an induction machine. In this respect, many researches, including
this work, were aimed to transfer a part of the MCSA knowledge to techniques that
monitors the signatures produced by the fault to the electromagnetic flux signals.

39
3.5 Magnetic flux monitoring
Why flux monitoring?
As previously shown, electrical motors experience a wide range of mechanical problems
common to most machinery, such as unbalance, misalignment, bearing faults, and resonance,
but electrical motors also experience their own specific set of problems, which are a result of
electro-magnetically generated fields in the stator and rotor. In this respect, monitoring
devices relying on the information provided by the electromagnetic fluxes produced by any
small unbalance in the magnetic or electric circuit of motors may be efficiently used in
addition to or as alternatives to the widely-used current monitoring. An example of such a
combined current-magnetic flux analysis was carried out by Kliman et al. (1988), where,
based on an evaluation of numerous test runs, the stator current and external leakage flux
were selected as the most practical signals containing the needed information for the detection
of broken bars.
Generally, it may be claimed that a monitoring technique based on magnetic flux
measurement is highly likely to meet Criteria 1-4 (Fig. 3.2).

3.5.1 General principles, comparisons, sensor-measuring technologies

There are many ways to sense magnetic fields, most of them based on the strong interaction
between magnetic and electric phenomena. Magnetic sensing techniques exploit a broad range
of physics and chemistry disciplines. In this respect, Lenz (1990) presents some of the most
common magnetic sensor technologies used in different fields of science (Table 3.2). This
table compares approximate sensitivity ranges; in some cases the projected ranges based on
further improvements are indicated by dashed lines. It is very important to bear in mind that
the electronic measurement devices considerably influence the sensitivity range for each
concept. There are many other factors to be considered when deciding what type of sensor is
best suited for a specific application. Among these factors, the frequency response, size, and
power should be carefully taken into account.
More information on the general characteristics of magnetic fields and magnetic field meters
that are typically used in electrical machines, on calibration methods for magnetic field
meters, and on the sources of measurement uncertainty can be found in the specifications for
magnetic flux density and electric field strength meters provided by IEEE (IEEE 1994).

3.5.2 Sensors used for condition monitoring in electrical machines – the search coils
General
When the stator phase current is used for the monitoring of rotor faults (such as broken rotor
bars in cage induction machines), the stator winding is used as a search coil for problems
associated with the rotor. The converse of this concept is that the rotor may be used as a
search coil for stator faults. For example, the axial magnetic leakage flux sensing is a similar
technique but detects changes in the line current harmonics indirectly via the axial flux
spectrum.
Search coils are employed to capture flux signals from inside and outside the machine. Such
coils are able to provide electrical “quality” signatures sensitive to conditions which alter the
electrical characteristics of the motor, such as broken rotor bars, eccentricity, unbalance
between phases, and stator faults. The voltage pick-up of such coils is directly proportional to
the rate of change of the flux.

40
 Table 3.2 Magnetic sensor comparison (Lenz 1990)

The occurrence of a fault in an electrical machine results in a change in the air-gap space
harmonic distribution. A search coil is able to detect the time harmonics but is not able to
capture the space harmonics. Space harmonics in the stator cause time harmonics in a rotating
rotor. By choosing the search coil properly, the number of space harmonics to be monitored
may be restricted.

3.5.3 Electromagnetic flux regions to be monitored in electrical machines

General
In the industrial environment, the flux measurements are preferably taken with the flux coil
placed outboard in the axial direction, since it is easier to take readings in the axial direction
in order to ensure the repeatability of the flux coil position. On the other hand, the
recommended mounting location for some monitoring units is radial for the optimal location
to detect temperature and vibration (Burnett 2002).
The use of internally mounted search coils is highly invasive and is not a practical option for
machines already installed. However, when the fault information is more valuable than that
provided by only one fault indicator (for example, current monitoring), such search coils have
to be confidently implemented. For example, the technique of using air-gap search coils has
been proved to be more sensitive than the stator current in detecting faults such as broken
damper bars or short-circuited turns in power generator rotor windings (Ramirez-Nino and
Pascacio 2001, Connolly et al. 1985, Karmaker 2003).

41
Fig. 3.3 presents some of the most comprehensively studied magnetic fluxes for the
identification of particular fault components produced by different stator- or rotor-related
faults.
The axial flux appears in all electrical machines and is produced because no machine can be
constructed perfectly symmetrically; it will be always present in electrical machines as a
result of the small asymmetries on both the rotor and stator sides. Since asymmetries are
coming from both the rotor and stator sides, the axial leakage flux contains a component
caused by the stator currents and a component caused by the rotor currents.
The axial flux measurements are a potential method for monitoring the condition of the stator
and rotor winding of induction motors during operation. The axial flux analysis has also been
found to be able to identify rotor asymmetries, shaft misalignments, and bearing faults
(Tavner and Penman 1987, Vas 1993).

Fig. 3.3 Schematic of flux regions used for fault detection and condition monitoring of
electrical machines (Thomson 1999).
Increments in the amplitudes of specific fault sideband frequencies are indicative of such
abnormalities. The axial leakage flux trajectory is not clearly defined in electrical machines,
but usually this is associated with the shaft. In practice, the detection of axial leakage flux is
relatively straightforward. The technique of measuring the axial leakage flux is simple and
non-invasive. All that is required is a search coil placed concentrically with the drive shaft of
the motor. Some authors claim that, typically, there is sufficient axial leakage flux that it is
possible to mount the coil, or coils, external to the machine case (Kokko 2003).
End winding leakage fluxes are the main cause of the axial leakage flux, which is measured
using axial leakage flux measurements.
A source of troublesome eddy currents in the stator core end regions of large machines is
caused by the "back-of-core leakage flux", which is the small component of armature flux
that is not contained by the core and which permeates the space behind the core and tends to
be drawn into circumferential members of the core frame. The axial members of the core
frame are exposed to this leakage flux and act as a squirrel cage, with the circumferential
members of the core frame at the ends of the machine providing return paths (Anderson).

42
Electromagnetic flux regions to be monitored for the machine under study in this work

In this thesis, the search coils presented in Figs. 3.4−3.5 were studied both by means of
measurements (carried out for a 35-kW cage induction motor) and simulations (when
possible) in order to find out their ability to sense specific machine faults:
- “1” internal search coil mounted around one stator tooth; wave winding formed of a
conductor 1.7 mm in diameter,
- “2” external search coil formed of 200 turns, with its length equal to the active axial
length of the machine and width equal to 2 pole pitches; the conductor used is 0.2 mm in
diameter,
- “3” search coil formed of 200 turns mounted in the non-drive end (ventilator side); the
conductor used is 0.2 mm in diameter,
- “4” search coil formed of approximately 300 turns mounted around the motor shaft
(internal), the conductor used is 0.2 mm in diameter,
- search coils sensing the electromagnetic flux harmonics of the order “p±1”; wave
windings formed of a conductor 1.7 mm in diameter and placed in the slot openings of the
stator slots.
When the end connections of the coils are spread apart a wave or series winding is formed. In
a wave winding there are only two paths regardless of the number of poles.
The external search coil, with its width equal to 2 pole pitches and the search coils purposed
to sense the electromagnetic flux harmonics of the order “p±1”, are chosen with the purpose
of eliminating the main symmetrical air-gap field. In this way, only those spatial harmonics
that appear during a working condition that affects the symmetry of the machine are to be
displayed. In the simulations, the indications provided by these search coils should be ideally
zero for healthy operation as a result of the perfect machine symmetry, which will not be the
case for the measurements obtained with tests on a real machine because of inherent
constructional asymmetries. The search coils intended to sense the axial flux were chosen
with it being borne in mind that any asymmetry, whether coming from the rotor or from the
stator side, in an electrical machine is reflected in this flux.

Fig. 3.4 Search coils used for the electromagnetic flux monitoring.

43
3.5.4 Electromagnetic flux harmonics of order “p ± 1” used for fault detection
As will be shown in the following sections of this work, there is a stringent need to look for a
fault indicator that is able to point out eccentricity-related abnormalities and bearing faults. In
this respect, the electromagnetic flux eccentricity harmonics of the order “p±1” that interact
with the fundamental harmonic of the machine and generate the forces will be studied to find
out their ability to detect machine abnormalities other than various types of eccentricity; this
approach is a novelty in the field of fault diagnosis in electrical machines. The
electromagnetic flux eccentricity harmonics of the order “p±1” were determined by analytical
means and are of the following form (Früchtenicht et al. 1982, Tenhunen et al. 2003):

bp ±1 ( x, t ) = B p ±1 cos ( p ± 1) x − (ω1 ± ω w ) t − (ϕ p ± ϕ w )  (3.9)

where p is the pole-pair number of the motor, ω1 is the fundamental frequency, ωw is the
whirling frequency, and φp and φw are phase angles.
Dorrell et al. (1995) have shown that dynamic eccentricity produces air-gap field components
rotating at fs± fr Hz (where fs is the supply frequency and fr is the rotational frequency of the
rotor) with corresponding p±1 pole-pairs in addition to the usual p pole-pair field. It can be
seen that both static and dynamic eccentricity produce their own “p±1” field components and
these field components may be used for monitoring purposes.
Fig. 3.5 presents the positioning of the search coils used for the detection of the “p±1”
eccentricity harmonics in the electromagnetic flux. Basically, the balanced pair(s) of the
search coils in the air-gap are so arranged that the main symmetrical air-gap field is
eliminated and only the lack of symmetry resulting from the flux associated with a fault is to
be displayed. The three search coil pairs placed 120° apart in the air-gap (Fig. 3.5.b) are
connected in series.

a) search coil “p-1” b) search coil “p+1”

Fig. 3.5 Search coils positioning for measuring the “p±1” harmonics (the test motor has p=2).

3.5.5 Turn-to-turn (inter-turn) short circuit fault detection in stator windings

In the case of a stator winding-related fault, there is a strong need for techniques that would
not only detect the occurrence of the fault but would also locate its position in the winding
while the motor is operating. Since the effect of a turn-to-turn short circuit will be to increase
asymmetry and consequently modify axial flux, a possible method to locate the position of the
shorted turn is based upon localised measurement of the magnetic field symmetry disruption
in the end-winding region of the motor.
Penman et al. (1994) tested a technique based on axial leakage flux. A large coil wound
concentrically around the shaft of the machine has the ability to sense a fault in the winding
while the motor is operating. The set-up needed with this method for the fault location
procedure consists of four coils symmetrically mounted in the four quadrants of the motor at a

44
radius of about half the distance from the shaft to the stator end winding (Fig. 3.6). The
frequency components to be detected in the axial flux component are given by


f short = f s  k ± n
(1 − s ) 
 (3.10)
 p 
where p is the number of pole-pairs, fs is the supply frequency, k = 1, 3 (order of the supply
time harmonics), s is the slip, and n = 1, 2, 3, …(2p-1).

Fig. 3.6 Coil geometry arrangement for reliable turn fault location (Penman et al. 1994).

3.5.6 Broken rotor bars detection

Elkasabgy et al. (1992) present different techniques for the detection of broken rotor bars in
the cage rotor of an induction machine. Among these techniques, those involving the study of
voltages induced in search coils appeared to provide the most useful, reliable, and cost-
effective diagnostic techniques. It was found that an external search coil placed against the
casing is just as effective as an internal coil mounted around a stator tooth tip for the detection
of a rotor fault. The characteristic frequencies at which the rotor fault would become
distinguishable are given by:
 k  
f brb = f s   (1 − s ) ± s  (3.11)
 p  
with k = 1, 2, 3.
Techniques relying on the indications provided by internally mounted search coils proved to
sense well enough rotor faults such as broken damper bars (Karmaker 2003).
Following the work of Penman et al. (1986), the detection of rotor bar faults is also possible
by frequency domain analysis of the axial leakage flux, which is monitored by using an
external search coil wound around the shaft of a machine. The frequency components are still
given by (3.11) with k = 1, 2, 3.
Other characteristic frequencies at which rotor faults would become distinguishable are given
by Vas (1993):
f brb = sf s (3.12)
f brb = f s (2 s ± 1) (3.13)

3.5.7 Bearing fault detection

The voltages resulting from the irregularities in the magnetic circuits of motors lead to shaft
currents through the shaft, bearings, bearing supports, and closing through the machine
framework. The procedure of measuring the shaft current has already been discussed in
Section 3.4.3.

45
3.5.8 Eccentricity detection
Air-gap eccentricity induces harmonics in the air-gap flux. These harmonics give rise to
axially directed flux and the amplitude of axially directed flux harmonics is proportional to
the amount of air-gap eccentricity.
Rankin et al. (1995) have demonstrated that the shaft voltage's AC component can be used as
an indicator to static eccentricity in the stator to rotor air-gap of a diesel driven salient pole
alternator.

Summary

Chapter 3 has presented a general view of the fault indicators used for electrical machines,
with a special focus on the ones of an electrical nature – current in the stator winding and
electromagnetic flux captured in various machine locations.
The possible signatures to be produced by various faults were presented and will be resorted
to in Chapter 5 that deals with the presentation of the results obtained form experiments and
simulations.
However, it is very important to bear in mind that the MCSA techniques are not able to detect
all of the possible faults in an induction machine and many research, including this work, are
aimed to transfer a part of the MCSA knowledge to techniques that monitors the signatures
produced by the fault to the electromagnetic flux signals.

46
4 METHODS OF ANALYSIS
4.1 Finite element analysis
4.1.1 Finite elements methods – general description, areas of application
The finite element method is a technique used to solve complex problems, which are
represented by differential equations. It effectively transforms the problem into a series of
algebraic problems, which are much easier to compute. Electromagnetic problems are
described by Maxwell’s equations (Maxwell 1865), which relate electric fields to magnetic
fields.
The finite element method splits the problem domain into a large number of small elements.
The elements are typically triangular or quadrilateral in shape. Different techniques exist for
the derivation of the algebraic equations from the initial problem region. The two most widely
used methods are the Variational Method (Binns et al. 1992, Silvester and Ferrari 1996,
Jiamning 1993), such as the Rayleigh Ritz method, and the method of Weighted Residuals,
such as the Galerkin Method (Binns et al. 1992, Salon 1995). Both types have advantages,
depending on the particular problem to be solved.
In order to solve a problem, the following information is required:
- the problem geometry must be specified and split into many small regions (elements),
- the material properties are required, i.e. the conductivity, permeability, and so on,
- the excitations in the problem region must be known (normally current densities in con-
ducting regions),
- any constraints which include boundary conditions must be specified.
The first published paper to use the term “Finite Elements” was published by Clough in 1960.
However, similar techniques were used as far back as 1943 by Courant.
The technique was initially used in civil engineering and aeronautical problems, especially by
the large aircraft companies, which could afford the powerful computers required for carrying
out stress analysis work on aircraft designs. Since then, the method has increased in
popularity, mainly because of rapid advances in computing power and speed. Finite element
analysis techniques are now used in a wide variety of engineering problems, including
mechanical stresses, electric and magnetic fields, heat transfers, and fluid flows. The
advantages offered by the finite element method over more traditional methods are that
complex geometries can be represented accurately, material properties and forcing functions
can be modelled accurately, and material non-linearity can be accounted for.

4.1.2 Numerical electromagnetic field analysis for electrical machines

History
The potential of using finite element methods for analysing the magnetic field distribution in
electrical machines was identified in the early ’70s (Chari and Silvester 1971). As well as
fixed mesh models, where the actual model is stationary and the rotating fields and current
distributions are modelled by complex number representation, time-stepping techniques have
been developed (Arkkio 1988), where the rotor is physically rotated inside the stator to give a
more realistic representation of machine operation.
With the finite element method of modelling an electrical machine, the complex geometry of
the machine and the magnetic non-linearity of the core regions can be accurately represented,
allowing the effects of saturation to be modelled. The induced currents in the rotor cage, for
instance, allowing for skin effects, can also be modelled accurately.

47
Three-dimensional finite element modelling has been discussed since the early ’80s (Chari et
al. 1982), but because of the even greater computational burden this involves, the extent of its
use was rather limited until the mid-2000s.

Application to fault diagnostics and condition monitoring

The basis of any reliable diagnostic method is an understanding of the electric, magnetic, and
mechanical behaviour of the machine in a healthy state and under fault conditions. As a
consequence of electrical or mechanical faults which may arise during the operating period,
many electrical machines are operating under asymmetrical conditions. In this respect, the
study of the asymmetrical magnetic field of the electrical machine is an inseparable part of the
diagnostic procedure.
The magnetic field distributions and operating characteristics provided by computer
simulations are essential in foreseeing the changes of motor performance resulting from the
parameter changes that appear as the consequence of different faults. Computer simulations
based on mathematical models represent an effective and inexpensive method for studying the
influence of different motor faults on drive performance. Simulation results represent a
contribution to the correct evaluation of the measured data in diagnostic procedures, which
are an important part of a supervision system based on expert systems and artificial
intelligence methods (Vas 1999).
Numerical modelling and simulation can provide virtual measurement data from machines
that have faults implemented in their structures. Then, modern signal processing techniques
can be used to convert the simulation results into a form from which faults can easily be
detected. It is very important to bear in mind that if a monitoring system cannot detect and
diagnose an artificial fault from the virtual measurement signals, it is hardly likely to work
with real electrical machines, either.
The electromagnetic model is valid for studying only those faults which modify the magnetic
field distribution in the machine. It is not suited to studying, for instance, partial discharges in
the winding’s insulation or small mechanical vibrations caused by a fault in the bearings.

4.1.3 Description of the numerical electromagnetic field simulation tool used in this
thesis
The numerical electromagnetic field simulation tool used in this thesis was developed in the
Laboratory of Electromechanics of Helsinki University of Technology and was originally
presented by Arkkio (1987).
By using this simulation tool, one may take full advantage of the possibility to monitor a wide
variety of electrical parameters (and not only) and to implement various fault types
characterised by specific severity degrees. This method of analysis is based on the combined
solution of the magnetic field equations and the circuit equations of the windings. The
equations are discretised by the finite element method. In order to keep the amount of
computation at a reasonably low level, several simplifications were made:
- the magnetic field in the core of the machine is assumed to be two-dimensional.
- the three-dimensional end region fields are modelled approximately by constant end-
winding impedances in the circuit equations of the windings.
- the current density in the stator conductors is constant, meaning that the skin effect in
the stator is neglected.

48
 - the laminated iron core is treated as a non-conducting, magnetically non-linear medium,
and the non-linearity is modelled by a single valued magnetisation curve.
The three-dimensional effects of the stator and rotor end winding leakage reactances are
accounted for by their inclusion in equivalent circuits inside the finite element model.
The magnetic vector potential A satisfies the equation
∇ × (ν∇ × A) = J (4.1)
where ν is the reluctivity of the material and J is the current density. The current density can
be expressed as a function of the vector potential and the electric scalar potential φ
∂A
J = −σ − σ ∇φ (4.2)
∂t
where σ is the conductivity of the material. In the two-dimensional model, the vector potential
and the current density have only the z-components
A = A( x, y , t )e z
(4.3)
J = J ( x, y , t )e z

The scalar potential φ has a constant value in the cross-section of a two-dimensional

conductor, and it is a linear function of the z-coordinate. The gradient of the scalar potential
can be expressed with the aid of the potential difference u induced between the ends of the
conductor. By substituting (4.2) in (4.1), the field equation becomes
∂A σ
∇ × ( ν∇ × A) + σ = ue z (4.4)
∂t l
where l is the length of the conductor. A relation between the total current i and the potential
difference u is obtained by integrating the current density (4.2) over the cross-section of the
conductor
∂A
u = R i + R∫ σ dS (4.5)
S ∂t
where R is the DC resistance of the conductor. The circuit equations for the rotor cage are
constructed by applying Kirchhoff’s laws and (4.5) for the potential difference. The details of
the construction of the circuit equations have been presented by Arkkio (1987).

Time dependence
A time-dependent field is solved by discretising the time at short time intervals ∆t and
evaluating the field at the times t1, t2, t3, (tk+1 = tk + ∆t). The time-dependence of the field is
modelled by the Crank-Nicholson time-stepping method. In this method, the vector potential
at time tk+1 is approximated as following

1  ∂A ∂A 
Ak +1 =  + ∆t + Ak (4.6)
2  ∂t k +1 ∂t k 

By adding the field equations written at the times tk and tk+1 together and substituting the sum
of derivatives from (4.6), the following expression is obtained:

49
 2σ
∇ × (ν k + 1∇ × Ak + 1 ) + Ak + 1
∆t

=
σ
l {
uk + 1ez − ∇ × (ν k ∇ × Ak ) −
2σ
∆t
σ
Ak − uk ek
l } (4.7)

The potential difference equation (4.5) is discretised in the same way as the field equation, the
result being
1 1 A −A
( uk +1 + uk ) = R ( ik +1 + ik ) + R ∫ σ k +1 k ⋅ dS (4.8)
2 2 S
∆t
Eqs. (4.7) and (4.8) form the basic system of equations in the time-stepping formulation.
Starting from the initial values and successively evaluating the potentials and currents of the
next time-steps, the time variation of the quantities is worked out.

Motion of the rotor

In a general time-stepping analysis of a running motor, the equations for the rotor and stator
fields are written in their own co-ordinate systems. The solutions of the two field equations
are matched with each other in the air-gap. The rotor is rotated at each time-step by an angle
corresponding to the mechanical angular frequency. The rotation is accomplished by changing
the finite element mesh in the air-gap.

Numerical solution
The construction of the circuit equations and the details of the numerical solution of the
coupled field and circuit equations have been presented by Arkkio (1987). The finite element
discretisation leads to a non-linear system of equations in which the unknown variables are
the nodal values of the vector potential and the currents or potential differences of the
windings. The system of equations is solved by the Newton-Raphson method.

Evaluation of operating characteristics

The magnetic field, the currents, and the potential differences of the windings are obtained in
the solution of the coupled field and circuit equations as discussed above. Most of the other
machine characteristics can be derived from these quantities.
The calculation of the electromagnetic torque is computed using the method developed by
Coulomb (1983). This method is based on the principle of virtual work, for calculating the
forces from a finite element solution. In this method, the force is calculated as a partial
derivative of the coenergy functional with respect to virtual movement.

Finite element mesh

The magnetic field of a healthy electrical machine is periodic in space, typically from one
pole-pair to the next one. In order to reduce the complexity of the geometry and the number
of nodes of the finite element mesh, the calculations are usually performed over the smallest
symmetrical part of the motor model. However, a fault in the machine disturbs the symmetry,
and the whole machine cross-section has to be modelled.
For fault detection purposes, the interest is focused more on qualitative than exact
quantitative results, i.e. more focused on detecting how the monitored parameter behaves as a
function of time and not so much in magnitude. In this respect the finite element meshes to be
used can be relatively sparse, as long as the geometric symmetry is the same as for the faulty
machine and the time dependence of the parameters is modelled properly. In this study,
triangular first-order finite elements are used, and the finite element meshes typically contain
6000-8000 elements.

50
In the numerical simulations, during healthy motor operation, the symmetry forbids the flux
linkage through the search coils “p±1” and through the one externally wound around two
pole pitches of the test motor. In a faulty machine, a flux linkage through these search coils
may be detected.
The input for the finite element code may be represented by the voltage waveforms measured
via a transient recorder directly from the frequency converter or by various other supply
sources. In this thesis, the simulation models were tested with the voltage waveforms
measured from the frequency converter but also with a sinusoidal supply. During the
simulations, the speed of the rotor was kept constant since the objective of the thesis was to
study only the operation at steady state.

4.1.4 Fault modelling procedure

The test motor used in this thesis is a cage induction motor whose main parameters are
presented in Table 4.1. Table 4.2 presents the faults taken into consideration in this work both
from experiments and simulations.
Table 4.1 Main parameters of the test motor.
Number of poles 4
Parallel branches in stator winding 2
Turns/coil in stator winding 11
Stator connection star
Rated slip [%] 3
Rated power [kW] 35
Rated frequency [Hz] 100
Rated voltage [V] 400
Rated current [A] 64
Number of stator slots 48
Number of rotor slots 40

Table 4.2 Faults studied in the numerical simulations and in the measurements.
Fault Simulations Experiments
Shorted turn in stator winding yes yes
Broken rotor bar/s yes yes
Static rotor eccentricity yes yes
Dynamic rotor eccentricity yes yes
Bearing fault no yes
Core faults yes no

Shorted turns
To model a shorted turn, the sides of this turn in the finite element mesh are substituted by
conductors in a perfect short circuit. There is no galvanic contact between the reduced phase
winding and the new, shorted conductors. A shorted coil is treated in a similar manner. Fig.
4.1 presents a simple description of the turn-to-turn short circuit implemented in the
simulations.
In measurements, the turn-to-turn short circuit was implemented between two adjacent turns
of the stator winding (see Fig. 4.1), was initiated and cleared from outside the motor, and did
not comprise any resistor specialised in limiting the effects of such an operating state. By
using a switcher, the duration of this faulty condition was limited to about one second.
However, the measurements at full load were not carried out, since the main concern was to
keep the test machine healthy for further tests and not to expose it to extreme conditions.

51
Fig. 4.1 Description of the turn-to-turn short circuit implemented in the simulations (left
figure) and measurements (right figure).

Static and dynamic eccentricity

Static eccentricity is obtained by shifting the rotor by 30% of the radial air-gap length and
rotating the rotor around its centre point in this new position. Dynamic eccentricity is obtained
by shifting the rotor by 33% of the air-gap length, but rotating it around the point that is the
centre point of the stator bore.
Experimentally, a dynamic eccentricity of 33% was obtained by fitting non-concentric
support parts between the shaft and bearing (see Fig. 4.2). In order to create an artificial static
eccentricity in the test motor, the inner diameter of the motor end-shields was reduced by
about 0.6 mm (the size of the machine air-gap is 0.8 mm). Both end-shields were then shifted
in the same direction in order to obtain a static eccentricity of around 30%.

Fig. 4.2 Artificially created dynamic eccentricity.

Broken rotor bars
Fig. 4.3 presents how the circuit equations of the cage windings are composed of the
potentials of the bars inside the rotor core, bar ends outside the core, and end-ring segments
connecting the bars. When modelling a broken bar, the resistance of the bar end outside the
core is increased to a value 1000 times the DC resistance of the whole bar. When modelling a
broken end-ring, the resistance of an end-ring segment between two bars is increased to a
value 1000 times the original resistance of the segment. In addition to the description of the
model implemented in the simulations for the study of various rotor cage-related failures, Fig.
4.3 also presents the artificially created rotor cage faults obtained by drilling holes in the rotor
bars.

52
Fig. 4.3 Description of the rotor cage-related faults implemented in simulations (left figure)
and artificially created in measurements (right figure).

Core faults – fault in core insulation

To model a fault in sheet insulation on a tooth-top surface, a current is allowed to flow along
the tooth-top and return back into the frame. A similar fault can also be created in several
adjacent teeth. Fig. 4.4 presents a description of the core fault implemented in the simulations.
Table 4.3 is intended to offer a view of the magnitude of the various fault levels implemented
in the simulations (“cf1” – “cf9” were used to indicate various degrees of core fault; see Fig.
4.4). The various stator winding configurations studied and referred to in this table as “NB”
(no parallel branches), “SB” (special configuration), “2B” (two parallel branches – this is the
original configuration to be found in the motor tested experimentally), and “4B” (four parallel
branches) are described in detail in Appendix II.

Fig. 4.4 Description of the core fault implemented in the simulations.

53
 Table 4.3 Eddy current losses at the fault location and corresponding total
electromagnetic loss following the core fault implementation.
Stator winding configurations – losses [W] Total electromagnetic losses [W]
“NB” “SB” “2B” “4B” “NB” “2B” “4B” “SB”
cf1 31 29 32 31 3106 2830
cf3 94 87 94 93 3260 2900
cf6 188 173 188 187 3347 2980
cf9 281 260 281 279 3426 3068

4.2 Method of measurements

The test cage induction motor described in Table 4.1 was fed from an inverter that had a
switching frequency of 3 kHz. A DC generator was used for loading the motor. The currents,
voltages, power and supply frequency, were measured using a wide band power analyser. The
measurements were carried out for three different load conditions. Figs. 4.5-4.7 present the
measuring set-up. A detailed description of a few of the devices used in this set-up is to be
found in Appendix II. The current and voltage waveforms were recorded with a transient
recorder. Hall sensors (LEM) were used as current transducers, and the voltages were
measured through an isolation amplifier. The sampling frequency was 40 kHz and a typical
number of samples was 20000. The recording system was calibrated using the measurements
from the power analyser. The measurements were carried out only for the original stator
winding topology, consisting of 2 parallel branches.

supply U Isolation Transient

INVERTER amplifier recorder

sensors
wide band LEM DC Load
power Hall
IM
generator resistor
analyzer sensors

Fig. 4.5 Schematic of the measuring set-up.

Fig. 4.6 Test induction motor (left) and DC generator used as load.

54
From the beginning, it is important to specify that the induction machine used in the
measurements has a skewed rotor but the simulated machine has a non-skewed one. Because
of this, some differences between the measured and simulated results should be expected,
especially in the magnitude of the rotor-slot harmonics.
Fig. 4.8 presents the cross-sectional geometry of the machine and the magnetic field
distribution at rated load operating conditions for healthy operation.

Fig. 4.7 Measurement devices. From top to bottom: transient recorder, keyboard, power
analyser, isolation amplifier, and connection box.

Fig. 4.8 Cross-sectional geometry of the test motor and the electromagnetic field distribution
for healthy operation at rated load.

55
4.3 Signal processing methods
4.3.1 General
With advances in digital technology over the last several years, adequate data processing
capability is now available on cost-effective, microprocessor-based, protective-relay
platforms to monitor motors for a variety of abnormalities in addition to the normal protection
functions.
Various electromechanical operational parameters able to provide rich information for
diagnosis are analysed in steady state but also in transient state (especially at no-load starting)
using approaches such as:
- frequency-domain analysis,
- time-domain analysis,
- time-frequency domain analysis,
- neural network,
- model-based techniques.

4.3.2 Analysis in steady state

The frequency-domain analysis for the detection and localisation of abnormal electrical and
mechanical conditions in electrical machines is widely used and accepted in the industrial
environment. This popularity is due to the availability of the Fourier transform technique, as
the characteristics of the studied signals are more easily noticed in the frequency domain than
in the time domain.
Advanced signal processing techniques, such as high-resolution spectral analysis are also
used since they may lead to a better interpretation of the spectra characteristic to specific
machine abnormalities. In this respect, it was experimentally proved that stator current high-
resolution spectral analysis is very sensitive to induction motor faults modifying main spectral
components, such as voltage unbalance and single-phasing effects (Benbouzid et al. 1999).
Bi-coherence spectra were used by Li et al. (1995) to derive features that relate to the
condition of a bearing. The application of bi-spectral and tri-spectral analysis to the vibration
monitoring was also discussed by McCormick and Nandi (1999).

4.3.3 Transient state analysis

Elder et al. (1989) developed a technique which allows the detection and identification of
specific faults within induction motors during the starting transient. During this short period
the machine is under conditions of severe and electrical stress. This type of analysis at
transient state is important especially for the detection of eccentricity since any UMP is a
maximum at starting.
Since under transient conditions the frequency components indicative of faulty operations are
non-stationary in both the time and frequency domains, other signal processing strategies
applicable to time variants need to be applied. Time-frequency domain techniques use both
time and frequency domain information, allowing for the investigation of transient features. A
number of time-frequency domain techniques that have been proposed in literature are
summarised by Ocak and Loparo (2001) as following:
- Short Time Fourier Transform (STFT),
- Spectogram - a representation of frequency components versus time (Burnett et al.
1995),
- Wigner Ville Distribution (Burnett et al. 1995),
- Wavelet Decomposition (Douglas et al. 2005, Yen and Lin 2000).

56
Since in both the measurements and simulations the signals studied as possible fault
indicators were captured in steady state, only the traditional signal processing technique based
on Fourier transformation (FFT) was used in this thesis.

4.3.4 Artificial intelligence techniques

The condition monitoring and fault detection of electrical machines have moved in recent
years from traditional techniques to Artificial Intelligence (AI) techniques. In the AI-based
systems, several quantities are utilized as input signals such as, stator currents and voltages,
electromagnetic fluxes, frame vibrations, etc. The AI techniques have numerous advantages
over conventional fault diagnostic approaches; besides giving improved performance, these
techniques are easy to extend and modify, and can be made adaptive by the incorporation of
new data or information. The AI-based techniques may use expert systems, artificial neural
networks, fuzzy logic, fuzzy-neural networks, genetic algorithms, support vector machines,
etc.
A review of the developments in the field of AI-based diagnostic systems in electrical
machines and drives is provided by Vas (1999) and Filippetti et al. (2000b). Siddique et al.
(2003) focus their review of various AI techniques to the induction motors, aiming more
specifically to the stator winding fault detection.
From the multitude of AI-based diagnostic systems, the neural networks have a wide
industrial applicability (Meireles et al 2003). A representative work dealing with the
application and design of artificial neural networks for electrical motors fault detection is the
one of Chow et al. (1993). Following, there will be presented few references to contributions
dealing with the detection of various faults using the neural network approach.
Salles et al. (2000), show that a neural network approach is able to distinguish between load
anomalies (oscillation torque, repetitive dip of torque) and rotor asymmetries (broken bars) of
cage induction motors. This is a very important issue since a load anomaly causes a sequence
of speed and current spectral components that can be confused with the one produced by a
cage-related fault.
Filippetti et al. (2005) report an induction machine rotor fault diagnosis based on a neural
network approach. After the neural network was trained using data achieved through
experimental tests on healthy machines and through simulation in case of faulted machines,
the diagnostic system was found able to discern between “healthy” and “faulty” machines.
Tallam et al. (2000), present an on-line neural network based diagnostic scheme, for induction
machine stator winding turn fault detection. This scheme is claimed to be insensitive to
unbalanced supply voltages or asymmetries in the machine and instrumentation. In addition, it
is claimed that a turn fault can be detected in the early stage of development.
Huang et al. (2004) propose a scheme to monitor voltage and current space vectors
simultaneously in order to monitor the level of air gap eccentricity in an induction motor. For
the amplitudes of eccentricity related components that change non-monotonically with the
operating conditions, an artificial neural network is used to learn the complicated relationship
and estimate corresponding signature amplitudes over a wide range of operating conditions.
Li et al. (2000) use neural networks to perform motor bearing fault diagnosis based on the
extracted bearing vibration features. Computer-simulated data were first used to study and
design the neural network motor bearing fault diagnosis algorithm. Actual bearing vibration
data collected in real-time were then applied to perform initial testing and validation of the

57
approach. The results show that neural networks can be effectively used in the diagnosis of
various motor bearing faults through appropriate measurement and interpretation of motor
bearing vibration signals.
Support Vector Machine (SVM) is a novel machine learning method introduced in early 90's
and it has been successfully applied to numerous classification and pattern recognition
problems such as text categorization, image recognition and bioinformatics. SVM based
classification scheme were designed for different tasks in cage induction motor fault
diagnostics and for partial discharge analysis of insulation condition monitoring and were
found highly competitive with, e.g., neural networks, which are widely studied also in the
area of condition monitoring and fault diagnosis of electrical machines (Pöyhönen et al.
2002a, Pöyhönen et al. 2002b, Pöyhönen 2004).

58
5. RESULTS OBTAINED FROM EXPERIMENTS AND SIMULATIONS
First of all, the relatively good agreement between the computed and measured total
electromagnetic losses is presented in Fig. 5.1. These machine quantities are responsible for
the correct evaluation of the motor performance and are considered very important for the
procedure of validating the simulated data.
Once more, it should be mentioned that the measured machine has a skewed rotor, while the
rotor of the simulated one is unskewed. The main effect of slot skew that is the filtering of
high frequency harmonics may be observed in the waveforms of various simulated motor
parameters characterised by higher ripple content in comparison with the similar measured
ones (i.e. Fig. 5.2.1, etc.).

Fig. 5.1 Experimental and simulated total electromagnetic losses.

5.1. Current monitoring

5.1.1. Stator winding currents
First of all, the ability of the phase current to detect specific faults will be tested, since it was
shown that monitoring this parameter is the most convenient (non-invasivity) and cheapest
way to sense a fault in terms of expense caused by any additional sensors. For all the current
spectra presented in this section, the supply frequency component (100 Hz) has been
normalised to be set at 0 dB.
On the basis of the measurements, no new frequency components of the line current spectra
appeared as a consequence of the inter-turn short circuit in the stator winding; only a rise of
15 dB in the amplitude of the 3rd harmonic component and of 3 dB in the magnitude of the
fundamental was observed from experiments. This amplitude change is independent of the
machine loading (half load or no load). With reference to the case of healthy operation,
irrespective of the machine loading, an increase of around 6 dB in the amplitudes of the rotor-
slot harmonics was found. The rotor-slot harmonics were calculated according to Eq, 3.5
where k and nw are 1 and nd is 0.
Based on simulations, Fig. 5.1.1 presents the modifications produced in the frequency
spectrum of the branch currents by the inter-turn short circuit. These new components are
according to Eq. 3.10 (Penman et al. 1994, Thomson 2001) and are of about the same
amplitude for all the configurations studied except the “SB” one (Appendix II), where the
amplitudes are 25 dB higher. Minor modifications (increment of about 2-3 dB) in the
magnitude of the rotor-slot harmonics were observed, irrespective of the stator winding
configuration and machine loading. The magnitude of the fundamental frequencies suffers
similar changes to the ones mentioned before (3.5 dB for “4B” and “SB”, 1.6 dB for the “2B”
stator winding type).

59
 2B 4B

Fig. 5.1.1 Frequency spectrum of the branch current – inter-turn short circuit – simulations
at full load.
In the measurements, an inter-turn short circuit in the stator winding was best reflected in the
unbalance of the stator winding currents (Fig. 5.1.2). A work dealing with the detection of
stator-related faults based on current amplitudes has been presented by Jover and Arkkio
(2003).
It came out from both the simulations and experiments that the eccentricities, bearing-related
faults, and major rotor cage faults do not result in any unbalance of the stator currents;
spectral analysis then becomes indispensable when attempting to detect such faults.
From the simulations, it was found that the static eccentricity is not reflected in the indications
of the branch currents in any of the configurations that were studied. The measurements
revealed increases of around 6-8 dB in the amplitudes of the rotor-slot harmonics, irrespective
of the machine loading. Modifications of similar magnitudes are to be found in the case of a
bearing fault. However, these modifications are not very specific and it would be quite
difficult to rely confidently on them for claiming the presence of a fault and, moreover,
distinguishing between various faults.
Fig. 5.1.3 presents the frequency spectra of the stator winding branch current for healthy
operation and for abnormal operations provoked by rotor cage faults. A rotor cage fault is
clearly visible when the sideband components closest to the fundamental given by Eqs. 3.1
and 3.2 are checked. For a rated slip of 3%, according to Eq. 3.1 these side bands are situated
at 94 Hz and 106 Hz, and at 45.5 Hz and 51.5 Hz, etc., according to Eq. 3.2. The results
obtained from simulations stand for the original “2B” winding configuration. They look
similar to the ones obtained from a motor equipped with a “4B” winding configuration that
are not presented here.

Fig. 5.1.2 Stator winding branch currents – inter-turn short circuit (top) and no fault (bottom)
– measurements at half load.

60
Fig. 5.1.3 Frequency spectra of the branch current – rotor cage faults – measurements (top)
and simulations (bottom) at full load.

Fig. 5.1.4 presents the distortions produced by a major rotor cage-related fault (three broken
rotor bars) in the stator winding branch current spectrum for the stator winding configurations
“NB” and “SB”. It is clear that for the stator winding configurations “NB” and “SB” a rotor-
cage-related fault does not produce similar modifications in the indications of the branch
currents as is the case with the “2B” and “4B” winding configurations; the distortions
according to Eq. 3.2 are not found any more.

Fig. 5.1.5 offers a quantitative view of the magnitude of the distortions produced by various
degrees of severity of the fault at full load. The measurements (marked by dots) reveal a good
agreement with the simulations in terms of new frequency components detected as a result of
a rotor cage fault. Simulations and measurements at half load do not cause major changes to
the magnitude of the frequency components characteristic of this fault, but the distortion
content is poorer than the one that is clear at full load (the frequency components around the
fundamental according to Eq. 3.1 are not found any more in the cases of the “2B” and “4B”
configurations). At no load it is almost impossible to detect a rotor cage fault, since most of
the characteristic frequency components responsible for the fault detection are missing or are
characterised by a dramatic change in magnitude (less than 50 dB) with reference to the
fundamental.

Fig. 5.1.4 Frequency spectra of the branch current – rotor cage fault – simulations at full load.

61
Fig. 5.1.5 Characteristic fault frequencies produced by rotor cage-related faults in the stator
winding branch current spectrum – simulations (represented by bars) and measurements
(represented by dots) at full load. One broken rotor bar (top) and three broken rotor bars
(bottom).

On the basis of the simulations, the dynamic eccentricity is not reflected in the indications of
the branch currents for the cases of the “NB” and “SB” stator winding configurations. For the
“2B” stator winding configuration, Fig. 5.1.6 presents the distortion produced by dynamic
eccentricity in the branch current spectrum. For the “2B” and “4B” stator winding
configurations, Table 5.1 presents a summary of the most important modifications considered
useful for the detection of such asymmetry (according to Eq. 3.5). No relevant modifications
in comparison with the case of healthy operation were found in the magnitudes of the rotor-
slot harmonics.

The relatively good agreement between the measurements and simulations in terms of ability
to predict the location of the frequency sidebands responsible for the detection of this fault is
shown by arrows.

Fig. 5.1.6 Spectrum content of the stator winding branch current – dynamic eccentricity –
simulations (top) and measurements (bottom) at half load. Figure corresponding to the
operation of a motor equipped with the original stator winding configuration of type “2B”.

62
Table 5.1 Modifications in the spectrum of the branch current caused by dynamic eccentricity [dB].
Full load Half load No load
Simul. Simul. Simul.
f [Hz] Meas. Meas. Meas.
“2B” “4B” “2B” “4B” “2B” “4B”
51.5 -33 -34 -28
50.7 -27 -28 -22
50 -18 -22 -9
148.5 -23 -23 -26
149.2 -20 -20 -19
150 -12 -13 -5
1991.5 -62 -50 -60
1918.5 -43 -41 -56
1949.5 -44 -44 -30
2088.5 -41 -36 -56
2118.5 -42 -54 -58
2149.5 -32 -46 -18

5.1.2. Circulating current between parallel branches

Since the stator winding phase of the test induction motor consists of two half-phases
(branches) connected in parallel and these phases are accessible from outside the motor, the
analysis of the differential circulating current between the half-phases is also performed with
the purpose of fault detection. Figs. 5.1.7-5.1.8 present the measured and simulated
waveforms of the circulating currents between stator winding parallel branches for healthy
machine operation and for the faults under study.

For a machine which has its stator winding distributed in parallel branches, it is quite clear
that by monitoring the circulating current one could easily identify both a major rotor fault
(such as a rotor bar breakage) and a stator-related fault (such as an inter-turn short circuit).
The circulating current may then be used as an additional diagnostic medium when aiming to
sense a stator-related fault or as a first sign of other abnormal operation, since it would be
easier to sense the current unbalance in the parallel branches than compute the current
frequency spectra.

However, it appears difficult to discriminate between different fault conditions and between a
fault and a healthy motor condition on the sole basis of the magnitudes of the circulating
currents. In this respect, it was found that a static eccentricity level of 30% is reflected in the
circulating current almost identically as a bearing fault and the magnitude of the circulating
current in these cases is almost the same as that which is characteristic of healthy machine
operation.
The relatively clear agreement between the predicted and measured circulating currents is
noteworthy. In simulations, the stator winding branch currents are equal (perfect balance) for
healthy operation and their difference is zero.

63
Fig. 5.1.7 Measured circulating currents.

64
Fig. 5.1.8 Simulated circulating currents for the test motor equipped with the original stator
winding configuration of type “2B”.

Discussion

Inter-turn short circuit

On the basis of the measurements, no new frequency components in the line current spectra
appeared as a consequence of a fault in the stator winding of an electrical machine. For the
case of a turn-to-turn short circuit in the stator winding, only a rise in the amplitude of the 3rd
harmonic component and of the fundamental was observed from the experiments. This
change in amplitude was independent of the machine loading. With reference to the case of
healthy operation, irrespective of the machine loading, an increase of around 6 dB in the
amplitudes of the rotor-slot harmonics was found.
The monitoring of the 3rd harmonic of the motor supply current was also found to be an
effective method for the detection of the presence of stator faults by other authors (Cruz and
Cardoso 2004). The only drawback of a technique relying on sensing this frequency
component is that low-voltage motors, especially random-wound motors, usually have
appreciable levels of inherent asymmetries and it may be possible that the residual
asymmetries of the motor may lead to the appearance of the 3rd harmonic component in the
supply currents, even when no faults are present in the motor.
The simulations were in relatively good agreement with the measured results. In comparison
with similar signatures produced by other faults, the new frequency sidebands produced in the
frequency spectrum of the branch currents by the inter-turn short circuit are so small that they
may be considered non-existent. For the “4B” and “2B” winding configurations, the

65
magnitude of these new frequency components is about 60 dB lower with reference to the
fundamental, while for the “SB” configuration these components are only 35 dB lower with
reference to the fundamental. The simulations also predicted minor modifications (increment
of about 2-3 dB) in the magnitude of the rotor-slot harmonics and in the magnitude at the
fundamental frequency (3.5 dB for “4B” and “SB”, 1.6 dB for the “2B” stator winding type).
The simulated and experimental results obtained for a motor equipped with a stator winding
configuration of type “2B” are in accordance with those presented by Joksimovic and Penman
(2000). They have also shown that no new frequency components of the line current spectra
can appear as a consequence of an inter-turn short circuit, but only a rise in some of the
frequency components which already exist in the line current spectra of a healthy machine can
be observed. These harmonics were the ones caused by rotor slotting.
The inability of the measured stator current to clearly detect a turn-to-turn fault was
mentioned by Henao et al. (2003). In this work, in the case of six turns short-circuited in a
phase of an 11-kW cage induction machine, the stator current was found not to be sensitive to
this fault. However, the stray flux, which was studied in parallel with the stator current, was
found to be affected by the fault being studied.
An increase in the magnitude of the rotor-slot harmonics produced by an inter-turn short
circuit was also observed by Stavrou et al. (2001). In this work, with a 150-kW slip ring
induction motor and a 30-kW cage induction motor, it was shown that the most reliable
indicators of the presence of the fault were the lower sideband of the field rotational
frequency with respect to the fundamental, together with some of the components that are
related to slotting.
However, the experimental results obtained in this section are in contradiction with the results
presented by Thomson (2001). In this work, relying on measurements carried out on a few
low-voltage cage induction motors of 2 and 4 poles respectively, it was found that certain
current components are clearly produced by inter-turn fault. The reasons for differences
between the results presented in this thesis and the ones presented by Thomson (2001) are
unclear. However, one may suspect that the stator winding configuration of the test machine
represents a cause for the pronounced nature of some specific frequency sideband components
produced by an inter-turn short circuit.
Yazidi et al. (2004) observed that the techniques relying on the stator winding current analysis
are not always reliable, especially when the number of shorted turns is small compared to the
total number of turns in a phase winding. Following a comparative investigation, the leakage
flux was found to be a more effective way to detect the stator fault than the stator current in
terms of specific frequency components at 750 Hz and 850 Hz (15th and 17th harmonic) which
are excited increasing their magnitude to a level of 20 dB.
Concerning the use of the rotor-slot harmonics as potential indicators of a fault in the stator
winding, Nandi and Toliyat (2000) claimed that although MCSA can detect these
components, they may be confused with voltage unbalance in machines. In this respect, it was
claimed that these harmonics can be unambiguously detected at the terminal voltages of the
machine just after it has been switched off, since the supply voltage is absent and the voltage
unbalance cannot influence the detection process.
The behaviour of induction machines affected by stator short circuits by assuming the
dominant slot harmonic components of the stator currents as diagnostic indexes was also
investigated by Gentile et al. (2003). The results obtained from the simulations of different
asymmetric machines using the space vector approach confirm that the main harmonic
components of the stator currents resulting from the rotor slotting of eighteen motors may be

66
considered a reliable diagnostic index. However, contrary to the results obtained by Nandi
and Toliyat (2000), it was shown that these harmonic components are almost completely
insensitive to possible supply voltage asymmetries.
In accordance with the results presented in this section, following a study on fault diagnosis in
the stator windings of a low-voltage induction motor fed by PWM inverters, Villada et al.
(2003) also found only a small increase in some components of the stator current spectrum
which are difficult to detect. Nevertheless, the vibration and axial leakage flux frequency
spectra experienced measurable differences at many frequencies. Therefore, for a better
indication of a stator winding fault, a combined analysis of these signals was recommended.
In conclusion, it may be stated that monitoring the stator current does not always provide
useful and sufficient information for the detection of a stator winding fault. Moreover, the
possible reflections of the turn-to-turn short circuit in the stator currents are also characteristic
of other faults and the probability of discriminating clearly between various faults remains
small. The circulating current between parallel branches and the unbalance in the phase
currents are suitable indicators for confidently claiming a turn-to-turn fault, but this is only
true in the case of a symmetrical supply.

Rotor cage fault - broken rotor bars

A rotor cage-related fault appears in a different manner for motors equipped with various
stator winding configurations. The distortions according to Eq. 3.1 are to be found in all of the
cases under study but those according to Eq. 3.2 only hold true for motors equipped with a
stator winding configuration of types “2B” and “4B”. However, the new frequency sideband
components given by Eq. 3.2 provide a precise signature of the rotor cage fault which is not
the case for those given by Eq. 3.1, which may also appear as a result of pulsating loads,
interactions between the motor and the train equipment, and particular rotor designs
(Thomson 1992).
Simulations and measurements at half load do not produce major changes to the magnitude of
the frequency components characteristic of this fault but the distortion content is poorer than
that found at full load. At no load it is almost impossible to detect a rotor cage fault since
most of the characteristic frequency components responsible for the fault detection are
missing or are characterised by a dramatic change in magnitude (less than 50 dB) with
reference to the fundamental. This conclusion is in accordance with those obtained by other
researchers, who also claim that in order to facilitate the successful detection of rotor faults
during steady-state operation a large supply current is required to flow.

Static eccentricity, dynamic eccentricity, and bearing fault.

Previously, it was shown that an inter-turn short circuit in the stator winding was best
reflected in the unbalance of the stator winding currents. It emerged from both simulations
and experiments that the eccentricities under study and bearing-related fault do not result in
any major visible unbalance of the stator phase currents. It appears, then, that spectral analysis
becomes indispensable when attempting to detect such faults.
However, relying on both experiments and simulations, this section concludes that the static
eccentricity being studied does not bring any major modification to the spectral content of the
branch current (in simulations, the same conclusion is valid for the case of static eccentricity,
irrespective of the stator winding configurations). Moreover, not even the circulating current
between the parallel stator winding branches is able to offer useful information about the fault

67
(the magnitude of these currents during such asymmetries is not very different from that
corresponding to healthy operation). A similar conclusion may be drawn for the case of a
bearing fault.

Considering only the experimental data, it was also found that the case of a bearing fault does
not produce any major distortion in the spectral content of the branch current but only
increases of around 6-8 dB in the amplitudes of the rotor-slot harmonics, irrespective of the
machine loading; similar modifications were found for the case of static eccentricity.
However, these are not clear signs to be used for an accurate detection of the implemented
static eccentricity and bearing fault.
The difficulty to sense a bearing fault from the stator current was also mentioned by other
authors.
Kim et al. (2003) found it very difficult to sense any important distinction between the
spectrum of a healthy motor current and the spectrum of the current corresponding to a
motor operating with a damaged bearing. The same conclusion was valid for two cases of
an eccentric air-gap (1 - condition of moving the rotating centre 25% upward at the end of
the inboard shaft and 2 - condition of moving the rotating centre at the end of the outboard
shaft 20 % downward and 10% right).
Lindh et al. (2003) have investigated the use of the stator current signal for detection of an
outer ring defect of the ball bearing with normal radial clearance of a 15 kW four pole
induction motor. It was found that the stator current measurement as a bearing fault
indicator is not adequate for this motor type since the modification produced by the radial
movement of the rotor was found very small if the radial movement was restricted with
bearing with small radial clearance. The outer race defect was clearly indicated only in the
case of a large internal radial clearance of the bearing.
Obaid et al. (2003) also showed that detecting the characteristic frequencies of a bearing
fault in the motor current is difficult, even when the corresponding fault frequencies are
prominent in the motor vibration. Only a very big window (a drilled hole) artificially
created in the inner race showed some modifications in the stator current.
The results obtained for the case of dynamic eccentricity both from experiments and
simulations are in accordance with the results obtained by most of the other research in this
field.
Tsoumas et al. (2005) also showed that motor current signature analysis is able to detect an
air-gap eccentricity of 30% for a slip-ring asynchronous induction motor by monitoring the
characteristic sideband frequency components given by Eq. 3.7.
The investigations carried out by Nandi et al. (1997) confirmed the presence of harmonics
as described by Eq. 3.5 in the presence of air-gap eccentricity. The dynamic eccentricity
(obtained by replacing the good bearings with bad ones and running the machine at load)
was reflected in an increase in the magnitude (about 3 times in comparison with healthy
operation) of the components provided by Eq. 3.5. New characteristic frequency
components were also found during the dynamic eccentricity.
Nandi et al. (2002) showed that a link exists between the low- (Eq. 3.7) and high-frequency
(Eq. 3.5) elements of the harmonics in the line-current spectrum in the presence of
eccentricity. For various levels of mixed eccentricity in a four-pole, 3-hp cage-induction
motor, the air-gap flux density components around the fundamental and rotor-slot

68
 harmonics were found in the line current. However, it was proven that not all machines
produce strong high-frequency components.

Studying the stator current spectra for both symmetrical and dynamically eccentric rotors,
Joksimovic et al. (2000) also observed that in the symmetrical condition only the base
frequency exists, while for the dynamic eccentricity condition, sideband frequencies
around the base frequency appear at specific frequencies (Eq. 3.7).

Dorrell et al. (1997) studied the effects of combined dynamic and static rotor eccentricity
on air-gap fields without recourse to slotting effects. Lower-frequency components in the
current and vibration spectra were identified as being suitable for monitoring to assess the
degree and type of rotor eccentricity present.

Another interesting conclusion drawn from the simulations is that the dynamic eccentricity is
not reflected in the indications of the branch currents in the cases of the “NB” and “SB” stator
winding configurations.

69
5.2. Electromagnetic flux monitoring

5.2.1. Inter-turn short circuit in the stator winding

In the following sections, only the most representative figures for fault detection and
discrimination that correspond only to a specific loading of the test motor will be presented.
However, for some specific cases figures will be presented that correspond to more than one
loading point. A summary of the main results obtained from the studies carried out under all
the loading conditions under study will always be presented.

The measurements for the case of the inter-turn short circuit in the stator winding at full load
are not available, since the main concern was to keep the test machine healthy for further tests
and not to expose it to extreme conditions.

For each of the search coils under study, the time-domain and frequency-domain
representation of the voltages induced during the inter-turn short circuit in the stator winding
from both the measurements and simulations will be shown.

In the simulations, the indications provided by all the search coils studied, with the exception
of the one wound around a stator tooth, are ideally zero for healthy operation because of the
perfect machine symmetry, which is not the case for the real machine tested in the
measurements. However, in most of the cases the time representation of the measured signals
for the case of healthy motor operation are not presented here since the changes caused by the
fault are not easy to distinguish from such a representation.

SEARCH COIL AROUND STATOR TOOTH

Fig. 5.2.1 presents, both from the measurements and simulations, the time-domain
representation of the voltages induced in the search coil placed around a stator tooth for the
case of the inter-turn short circuit in the stator winding. The measured waveform
corresponding to the healthy operation was not included in this figure since it looks almost
identical with the one corresponding to the studied fault.

In the beginning, two search coils wound around different stator teeth (situated about 45° left
and right, respectively, of the fault locations) were used, with the aim of verifying if the
position of the shorted turn with respect to the search coil produces different indications about
the fault. This may happen because this search coil is not purposely intended to make use of
the machine symmetry for capturing various flux harmonic components (see the external
search coil around two pole pitches and internal search coils for monitoring the “p±1”
harmonic components). However, the indications provided by both of these coils seemed to
be identical.
From the measurements it was found that the indications provided by the search coil around a
stator tooth are of no major relevance when aiming to detect a stator winding fault. The
paucity of the information provided by this search coil refers mainly to the absence of the new
frequency sideband components characteristic of a stator fault (see Section 3.5.5). Choosing
the fundamental frequency as reference, the only significant changes in comparison with
healthy operation were observed at 500 Hz (10-dB increase at half load and 6-dB at no load)
and at 900 Hz (10-dB increase both at half-load and no load). The changes corresponding to
the half load operation are presented in Fig. 5.2.2. In the high-frequency range, no
modifications were found in the amplitudes of the slot harmonics.

70
Fig. 5.2.1 Time-domain representation of the voltages induced in the search coil placed
around stator tooth – inter-turn short circuit in the stator winding – measurements (top) and
simulations (bottom) taken at half load.
Fig. 5.2.3 presents the frequency-domain representation of the voltages induced in the search
coil wound around a stator tooth for the case of the inter-turn short circuit based on data
gathered from simulations of the test motor equipped with various stator winding
configurations. It may be observed that this type of fault introduces new characteristic fault
sideband frequencies only for the winding configurations of the types “NB” and “SB”. The
new relevant frequencies for detecting this fault are according to Eq. 3.10 (Melero et al. 2003,
Penman et al. 1994) and are indicated by arrows. This is in agreement with the results
obtained from the measurements of the test motor equipped only with a stator winding
consisting of 2 parallel branches, where no new frequency sideband components characteristic
of a stator winding fault were found.

Fig. 5.2.2 Frequency-domain representation of the voltages induced in the search coil wound
around a stator tooth during the inter-turn short circuit – measurements at half load.

Fig. 5.2.3 Frequency-domain representation of the voltages induced in the search coil wound
around a stator tooth during the inter-turn short circuit – simulations at half load.

71
Relying on simulations at full load, Fig. 5.2.4 presents the magnitudes of the new frequency
sideband components found to be characteristic of the inter-turn short circuit. The
fundamental frequency of 100 Hz is the reference at which these components are reported.
This type of fault remains relatively easy to identify at half load (the magnitudes of the new
frequency sideband components do not vary much from the ones presented in Fig. 5.2.3), but
at no-load operation there are no longer any clear signs of such new components (most of the
magnitudes of the sidebands of interest fall below the level of 40 dB).

Fig. 5.2.4 Characteristic frequency sidebands produced by the inter-turn short circuit in the
indications provided by a search coil wound around a stator tooth – simulations at full load.
Table 5.2.1 presents the changes to the rotor-slot harmonics caused by the inter-turn short
circuit. This type of analysis will not be repeated for the other search coils implemented in the
simulations, since this is the only case where the indications for healthy operation are non-
zero. First, this table presents the magnitude of the frequencies of interest with reference to
the fundamental for the case of both healthy and abnormal operation, respectively. After that,
the difference between these two modes of operation that are supposed to give an indication
of the condition of the motor is presented. From these tables it can be seen that the inter-turn
short circuit is reflected more clearly for a motor equipped with stator windings of the types
“4B” and “SB”, respectively.

Table 5.2.1 Changes produced by the inter-turn short circuit to the rotor-slot harmonics
captured by a search coil wound around stator tooth [dB].
1840 Hz 2040 Hz 2240 Hz
“4B”,”SB” “NB”, “2B” “4B”,”SB” “NB”, “2B” “4B”,”SB” “NB”, “2B”
fault no fault no fault no fault fault no fault no fault no
fault fault fault fault fault
4 -3 1 -3 7 1 3 1 -9 -23 -16 -23
7 4 6 2 14 7
Full load
1870 Hz 2070 Hz 2270 Hz
“4B”,”SB” “NB”, “2B” “4B”,”SB” “NB”, “2B” “4B”,”SB” “NB”, “2B”
fault no fault no fault no fault no fault no fault no
fault fault fault fault fault fault
4 -3 -1 -3 7 -1 2 -1 -12 -27 -20 -27
7 2 8 3 15 7
Half load
1900 Hz 2100 Hz 2300 Hz
“4B”,”SB” “NB, “2B” “4B”,”SB” “NB”, “2B” “4B”,”SB” “NB”, “2B”
fault no fault no fault no fault fault no fault no fault no
fault fault fault fault fault
3 -4 -1 -4 5 -3 1 -3 -12 -31 -20 -31
7 3 8 4 19 11
No load

72
SEARCH COIL PLACED AROUND TWO POLE PITCHES
Fig. 5.2.5 presents, both from the measurements and simulations, the time-domain
representation of the voltages induced in the search coil placed around two pole pitches of the
test motor in the case of the inter-turn short circuit in the stator winding. The measured
waveform corresponding to the healthy operation was not included in this figure since it looks
almost identical with the one corresponding to the studied fault. It is observed that the inter-
turn short circuit in the stator winding is reflected in a different manner for various stator
winding configurations.

Fig. 5.2.5 Time-domain representation of the voltages induced in the search coil placed
around two pole pitches of the test motor – inter-turn short circuit – measurements (topmost
figure) and simulations taken at half load.

Fig. 5.2.6 shows the frequency-domain representation of the voltages induced in the search
coil wound around two pole pitches following the inter-turn short circuit. These
representations correspond to various stator winding configurations implemented in the
simulations.

Fig. 5.2.6 Frequency-domain representation of the voltages induced in the search coil wound
around two pole pitches – inter-turn short circuit – simulations at full load (top) and no load
(bottom).

73
The relevant frequencies for detecting this fault are according to Eq. 3.10 (Melero et al. 2003,
Penman et al. 1994). The frequency components of 250 Hz and 500 Hz for operation at full
load and no load, respectively, were chosen as references for the dB representation (these
frequencies were characterised by the highest amplitude).
For the winding configurations of the types “2B” and “4B”, the inter-turn short circuit does
not produce any new characteristic frequency sideband components to be captured by a search
coil wound around two pole pitches. This is in accordance with the results extracted from the
measurements, where the only sign of a possible fault was represented by an increment in the
fundamental frequency of about 9 dB and 14 dB in half-load and no-load testing, respectively.
New characteristic fault sideband frequency components are found in the cases of stator
winding configurations of the “NB” and “SB” types. Of these two winding configurations, the
“SB” type is observed to be slightly more sensitive to the inter-turn short circuit than the
“NB”. At no load, the potential information about the fault is found to be considerably
diminished.

In the high-frequency range, the experimental turn-to-turn short circuit that was implemented
is reflected in a relatively slight increment in the rotor-slot harmonics of about 4 dB in
comparison with healthy operation. It was not possible to analyse such modifications from the
simulations, since in healthy operation, because of the perfect machine symmetry, the signal
captured by this coil is ideally zero.

SEARCH COIL “p-1”

Fig. 5.2.7 presents, both from the measurements and simulations, the time-domain
representation of the voltages induced in the search coil “p-1” for the case of the inter-turn
short circuit in the stator winding.
Fig. 5.2.8 shows the frequency-domain representation of the voltages induced in the “p-1”
search coil following the inter-turn short circuit. These representations correspond to various
stator winding configurations implemented in the simulations.
Similarly to the previous cases, new characteristic fault sideband frequency components are
introduced by the fault to the stator winding of the “NB” and “SB” types. The “SB”
configuration is also reflected by the indication of this search coil as being more sensitive to
the inter-turn short circuit than the “NB” one.

Fig. 5.2.9 presents the magnitudes of the new characteristic fault sideband frequency
components for the cases of stator winding configurations of the “NB” and “SB” types. These
figures are with reference to the rotor-slot harmonics (2040 and 2100 Hz for operation at full
load and no load, respectively). The relevant frequencies for detecting this fault are according
to Eq. 3.10.

74
Fig. 5.2.7 Time-domain representation of the voltages induced in search coil “p-1” – inter-
turn short circuit – measurements (topmost figure) and simulations taken at half load.

2B
4B

Fig. 5.2.8 Frequency-domain representation of the voltages induced in the “p-1” search coil
– inter-turn short circuit – simulations at half load.

Fig. 5.2.9 Characteristic frequency sidebands produced by the inter-turn short circuit in the
indications provided by “p-1” search coil – simulations at full load (top) and no load
(bottom).

75
Using a “p-1” search coil, for a motor equipped with winding configurations of the “NB” and
“SB” types, the inter-turn short circuit in the stator winding is relatively easy to identify at
half load (the magnitudes of the new frequency components are of the same magnitude as the
ones presented in Fig. 5.2.9) but not in no load operation, where there is a major reduction of
35 dB in the magnitude of these new components.
Relying on the measurements, Table 5.2.2 presents the main changes produced by the inter-
turn short circuit to those frequency components also found during healthy operation. The
only modification to be noticed is the increase in the magnitude of the 3rd harmonic, and
therefore this is the only one shown in the table.

Table 5.2.2 Changes produced by the inter-turn short circuit – measurements [dB].
Half load No load
“p-1” “p-1”
f [Hz]
fault - no fault fault no fault fault - no fault fault no fault
300 Hz 19 -6 -25 19 -1 -20

SEARCH COILS “p+1”

Fig. 5.2.10 presents, both from the measurements and simulations, the time-domain
representation of the voltages induced in the search coil “p+1” for the case of the inter-turn
short circuit in the stator winding.
Fig. 5.2.11 presents the frequency-domain representation of the voltages induced in the
studied search coils from the measurements.
Relying on the measurements, the main changes produced by the inter-turn short circuit in
comparison with healthy conditions may be found in:
- the magnitude of the fundamental frequency: increments of 12 dB and 16 dB for operation at
half load and no load, respectively.
- the magnitude of the 3rd harmonic: increments of about 24 dB, irrespective of the motor
operation point.

Fig. 5.2.12 shows the frequency-domain representation of the voltages induced in the “p+1”
search coil following the inter-turn short circuit. These representations correspond to various
stator winding configurations implemented in the simulations at half load.

76
 2B
4B

Fig. 5.2.10 Time-domain representation of the voltages induced in search coil “p+1” during
the inter-turn short circuit – measurements (topmost figure) and simulations taken at half
load.

Fig. 5.2.11 Frequency-domain representation of the voltages induced in search coil “p+1”
during the inter-turn short circuit – measurements taken at half load.

Fig. 5.2.12 Frequency-domain representation of the voltages induced in the “p+1” search
coil during the inter-turn short circuit – simulations at half load.

The new characteristic fault sideband frequency components introduced by the stator winding
configurations of the “NB” and “SB” types are presented in Fig. 5.2.13. These representations
are with reference to the rotor-slot harmonics. The relevant frequencies for detecting this fault
are according to Eq. 3.10.

77
Fig. 5.2.13 Characteristic frequency sidebands produced by the inter-turn short circuit in the
indications provided by “p+1” search coil – simulations at full load (topmost figure), half
load (middle figure), and no load (bottom figure).

Using a “p+1” search coil for winding configurations of the “NB” and “SB” types, the inter-
turn short circuit in the stator winding is relatively easy to identify, even at half load. In no
load operation, however, the potential fault information becomes very poor (major reduction
in the magnitude of some frequency side-band components).

SEARCH COILS AROUND THE SHAFT AND AT THE NON-DRIVE END (NEAR VENTILATOR)

Fig. 5.2.14 presents the time-domain representation of the voltages induced in the search coils
placed around the motor shaft and at the non-drive end of the motor for the case of the inter-
turn short circuit in the stator winding. Fig. 5.2.15 shows the frequency-domain representation
of the voltage induced during the inter-turn short circuit in the search coil positioned around
the motor shaft at a half load motor operation. Simulations are not available in these cases as
the 2D simulation tool is not able to take into account the 3D effects.
The inter-turn short circuit in the stator winding, irrespective of the motor loading, does not
produce any new components in the spectral content of the voltage induced in a search coil
placed at the non-drive end of the motor. In the current work, the only sign of a fault is
represented by a major increase in the magnitude of the fundamental frequency
(approximately 10 times, irrespective of load). For the case of a search coil wound around the
shaft this increase was found to be about 7 times and 4 times for the operations at half-load
and no load, respectively. Important information about the fault may also be extracted from
the magnitude of the 3rd harmonic.
For the search coil placed at the non-drive end of the test motor, the rotor slotting harmonics
increased by only 4 dB in half-load operation, in comparison with healthy operation, while in
no-load operations no change in the magnitude of this frequency component was found.
In the high-frequency range, a turn-to-turn short circuit in the stator winding is reflected only
in the indications provided by the search coil around the shaft by an increment of around 6-8
dB in the magnitude of the rotor-slot harmonics at half-load and no load, respectively.

78
Fig. 5.2.14 Time-domain representation of the voltages induced during the inter-turn short
circuit in search coils positioned around the motor shaft and at the non-drive end of the motor
(topmost figure) – measurements at half load.

Fig. 5.2.15 Frequency-domain representation of the voltage induced during the inter-turn
short circuit in the search coil positioned around the motor shaft – measurements at half load.

Discussion
First of all, it is important to notice from the simulations that for a machine equipped with
winding configurations of the types “2B” and “4B”, the inter-turn short circuit does not
produce any new characteristic frequency sideband components to be captured by the search
coils being studied. The measurements that were carried out only for the case of the motor
equipped with a stator winding consisting of 2 parallel branches (“2B”) pointed out that the
new relevant frequency signatures for detecting shorted turns from electromagnetic flux are
not observed in the search coils being studied.
Search coil around stator tooth: From the measurements, the only significant changes in
comparison with healthy operation were observed at 500 Hz (10-dB increase at half load and
6-dB at no load) and at 900 Hz (10-dB increase both at half load and no load). In the high-
frequency range, no modifications were found in the amplitudes of the rotor-slot harmonics,
as opposed to the modifications of around 6 dB that were observed in the branch current
(Section 5.1.1). The simulations indicated the difficulty of finding clear fault signatures in a
motor equipped with stator winding configurations of the types “2B” and “4B”. However, the
changes caused by the inter-turn short circuit to the rotor-slot harmonics that are captured by a

79
search coil wound around a stator tooth were found to be more important (7-15 dB) than the
ones found in the branch current (2-3 dB, according to the results presented in Section 5.1.1).
Search coil wound around two pole pitches: From the measurements, the only sign of a
possible fault was represented by an increment of the fundamental frequency of about 9 dB
and 14 dB in half-load and no-load testing, respectively. This is in agreement with the
simulations where, for the winding configurations of the types “2B” and “4B”, the inter-turn
short circuit did not produce any new characteristic frequency sideband components. In the
high-frequency range, the experimental implemented turn-to-turn short circuit is reflected in a
relatively slight increment in the rotor-slot harmonics of about 4 dB in comparison with
healthy operation.
Search coil “p+1”. From the measurements, the major increments in the amplitude of the
fundamental frequency (12-16 dB) and in the magnitude of the 3rd harmonic (24 dB) are
higher than the ones produced by this fault for other fault indicators. It is also very probable
that other faults will not be reflected in a similarly major manner in the indications of this
coil. For the winding configuration of the “2B” and “4B” types, the simulations (supposed to
point out only those frequency components caused by the fault) also indicate the fundamental
and the 3rd harmonic as being responsible for providing important information about the inter-
turn short circuit.
Search coil “p-1”. Relying on the measurements, the major changes produced by the inter-
turn short circuit to the 3rd harmonic (19 dB both for half load and no load operation) are
noteworthy. This fault signature may also be considered important since it is expected that
other faults will not be reflected in a similarly major manner in the indications of this coil. For
the winding configurations of the “2B” and “4B” types, the simulations also indicate the 3rd
harmonic as a carrier of important information about the inter-turn short circuit.
The search coils around the shaft and at the non-drive end of the machine offer useful
information about a winding fault, especially by capturing the major modification produced in
the magnitude of the fundamental frequency component. The severity of these modifications
may be considered a unique signature for this type of fault.
Following the previous discussion, it becomes clear that most of the search coils that were
studied are able to sense an inter-turn fault, relying mostly on the indications provided by the
behaviour of the fundamental and 3rd harmonic frequency component. Major modifications in
the magnitudes of the frequency components, as suggested by Eq. 3.10, were identified in the
simulations only for those machines equipped with a stator winding configuration of the
“NB” and “SB” types.
It is important to mention that it is quite difficult to compare the results presented in this
thesis with those presented by other authors, since most of their contributions do not provide
additional information about the stator winding configurations of the motors being studied,
i.e. parallel/not parallel branches (Melero et al. 2003, Villada et al. 2003, Yazidi et al. 2004,
Assaf et al. 2004, Thailly et al. 2005). However, Henao et al. (2003, 2004) and Penman et al.
(1994) clearly specify the parallel winding configurations used in their works.
Melero et al. (2003) concluded that the harmonic component at the frequency fs-fr (according
to Eq. 3.10) may provide useful information about a minimal stator winding turn-to-turn short
circuit and the greatest change (28 dB) may be found in the final state of the fault, when the
motor is close to catastrophic failure. This final condition was found after 33 cycles (1 cycle =
starting-up – running a full load for ten minutes – stopping; T = 220°C). In comparison with
the investigations carried out on the behaviour of the air-gap torque harmonics and negative-
sequence impedance during such a fault, the increases of the axial leakage flux showed a

80
more linear trend. Moreover, the total increase between the healthy state and the final phase of
the failure was the largest, so this method could be more suitable for predicting how far the
motor was from the end of its useful life.
The advantage of using a detection technique relying on flux monitoring was also suggested
by Henao et al. (2003). This detection technique was found to be more reliable than the
MCSA, which provided no useful information about the fault being studied. The case of an
induction machine with six turns short-circuited in phase tested at standstill and when the
machine was rotating without load torque and at the rated torque, both under sinusoidal power
supply and voltage source inverter was investigated. For the induction machine at standstill,
the stray flux was affected by an increase in the magnitude of harmonics 3 and 9, with more
than 20 dB for the third and 10 dB for the ninth. For the machine rotating without load torque,
the effect that made possible the detection of the fault was localised in harmonics 5 and 7,
which were excited, increasing their magnitude to around 10 dB. When the induction machine
rotated at the rated speed with short-circuited turns, the only frequency component showing
sensitivity to the short circuit was at 450 Hz (9th harmonic) with an increase of around 10 dB
in its magnitude.
Penman et al. (1994) detected changes in the harmonics with the lowest nk product (Eq. 3.10)
for a motor in which two of the copper turns were short-circuited. This motor was equipped
with a wound rotor and tests were carried out at no load and at 48 kW, respectively. The
maximum ratio between the induced voltages in the search coil for the unshorted and shorted
conditions, respectively, was around 1.2, irrespective of the motor loading
For an 11-kW cage induction motor, as a consequence of a short-circuited section of 12.5% of
one phase full winding, Thailly et al. (2005) found considerable increments in the magnitudes
of the rotor-slot harmonics. It is claimed that by monitoring the amplitude of these harmonics
one may obtain useful information for the detection of an inter-turn short circuit in the stator
winding. The results of the simulations carried out in this work for the search coil wound
around a stator tooth are in accordance with these conclusions, but the results obtained from
the measurements are not.
Relying on the experimental results obtained from the testing of an 11-kW induction motor,
Assaf et al. (2004) presented a sensitivity analysis of the recorded axial leakage flux in the
presence of 1% and 6% short circuits of the total turn number for various operating
conditions. In full-load operation, the harmonics components affected the most by the fault
(15-22 dB) were the ones corresponding to the frequencies 3fs-3fr, fs+3fs, 5fs+fr, and 5fs+3fr.
Villada et al. (2003) conducted a series of tests on 5-hp and 3-hp induction motors in order to
verify the ability of the axial leakage flux to detect shorted turns. Once more, Eq.3.10 was
found suitable to provide a good agreement between the spectra of the flux leakage signal and
the portion of the winding under fault.

81
5.2.2. Rotor cage-related faults
SEARCH COIL AROUND STATOR TOOTH

Relying on both the measurements and simulations, Fig. 5.2.16 presents the time-domain
representation of the voltages induced in a search coil placed around the stator tooth for the
case of a major rotor cage fault, such as three broken rotor bars.

Fig. 5.2.16 Time-domain representation of the voltages induced in the search coils placed
around stator tooth – 3 broken rotor bars – measurements (top) and simulations (bottom)
taken at half load.

Fig. 5.2.17 shows the frequency-domain representation of the voltages induced in the search
coil around stator tooth during a rotor cage-related fault. These representations are based on
measurements at half load and stand only for the original stator-winding configuration
consisting of 2 parallel branches. From the measurements repeated with no load, the breakage
of three rotor bars remains easily visible (the results corresponding to this test are very similar
to those presented in Fig. 5.2.17), but the breakage of one rotor bar is harder to detect (based
on the presence of the fault-specific frequency sideband components used, to be indicated by
arrows in the following figures).

Fig. 5.2.18 shows the frequency-domain representation of the voltages induced during a rotor
cage fault. These representations correspond to various stator-winding configurations
implemented in the simulations and are with reference to the rotor-slot harmonics. At half
load, it is possible to visualise signs characteristic of a rotor cage fault (indicated by the
arrows). In average, these signals are around 10 dB lower in comparison with those
corresponding to the full load. However, at no load it is more difficult to detect this type of
fault, on average, the signals are around 15 dB lower than the ones corresponding to the full
load. The minimum rotor cage-related fault that was studied, 1 broken rotor bar, is almost
impossible to detect at no load.

A comparison based on Figs. 5.2.17 and 5.2.18 points out a good agreement in terms of new
characteristic fault frequency sideband components produced by a rotor cage fault and
obtained from the measurements and simulations respectively (see the arrows indicating the
frequencies of interest for the detection of a rotor cage-related fault according to the
predictions given by Eq. 3.2: 47.7, 50.7, 146.3, 149.3 Hz, and so on for various values of k).

82
 1

Fig. 5.2.17 Frequency-domain representation of the voltages induced in the search coil
around stator tooth during rotor cage-related faults – 3 broken rotor bars (top) and 1 broken
rotor bar (bottom) – measurements at half load.

Fig. 5.2.18 Frequency-domain representation of the voltages induced in the search coil
wound around a stator tooth – 1 broken rotor bar (top) and 3 broken rotor bars (bottom) –
simulations taken at half load.

SEARCH COIL PLACED AROUND TWO POLE PITCHES

Relying on both the measurements and simulations, Fig. 5.2.19 presents the time-domain
representation of the voltages induced in a search coil placed around two pole pitches of the
test motor for the case of a motor operating with three broken rotor bars. Fig. 5.2.20 shows
the frequency-domain representation of the voltages induced in the search coil wound around
two pole pitches from the simulations. The representation of the characteristic fault sidebands
is with reference to the rotor-slot harmonics. The results from all the load operation points
that were studied are included in order to point out the capability of this search coil to sense
various degrees of a fault at various loads.

83
Fig. 5.2.19 Time-domain representation of the voltages induced in the search coil placed
around two pole pitches – 3 broken rotor bars – measurements (topmost figure) and
simulations taken at half load.
SB

Fig. 5.2.20 Frequency-domain representation of the voltages induced in the search coil
wound around two pole pitches – 3 broken rotor bars – simulations at full load and half load
(topmost three figures) and no load (bottom figure).
In terms of new frequency sideband components, the rotor cage-related fault is reflected in a
similar manner for a motor equipped with any of the stator winding configurations that were
studied. In terms of distortion magnitudes, for all the loading points under study, but
especially at lower loadings, a motor equipped with stator windings of the types “NB” and
“SB” is observed to exhibit more important changes.

84
Figs. 5.2.21 and 5.2.22 show the frequency-domain representation and the characteristic
frequency sidebands produced by rotor cage-related faults of various levels of severity in the
indications provided by the search coil wound around two pole pitches from measurements at
full load. The ability of this search coil to sense the degree of severity of the rotor cage fault is
noteworthy.

Fig. 5.2.21 Frequency-domain representation of the voltages induced in the search coil
wound around two pole pitches – 3 broken rotor bars – measurements at full load.

Fig. 5.2.22 Characteristic frequency sidebands produced by various rotor cage-related faults
in the indications provided by the search coil placed around two pole pitches – measurements
at full load.

A rotor cage-related fault may also be detected from measurements at half load; on average,
the magnitude of the previous exposed frequencies is reduced by about 10-15 dB in
comparison with those corresponding to the full load operation point. However, the sideband
component situated in the immediate vicinity of the fundamental is no longer detectable at
this operation point.
From a no load measurement, the rotor cage related-faults become relatively difficult to detect
since some of the previously mentioned sideband components disappear and the remaining
ones are characterised by magnitudes 20-25 dB lower than the ones corresponding to the full
load operation point.
In addition to the frequency components observed in Fig. 5.2.21 and summarised in Fig.
5.2.22, it is important to notice the very low-frequency component situated at around 3 and 2
Hz for the full-load and half-load operation points, respectively. The origin of this sideband
may be one of those suggested by Kliman et al. (1988). In this contribution, a frequency
component given by sfs, was associated with the signature produced by a broken end-ring
fault, but the same component was also found in the cases of rotor and voltage source
asymmetry. The conclusion is that this frequency can be used for condition monitoring, but
the possibility of other fault modes has to be taken into account.

85
SEARCH COIL “p-1”

Fig. 5.2.23 presents, both from the measurements and simulations, the time-domain
representation of the voltages induced in the search coil “p-1” for the case of a motor
operating with 3 broken rotor bars.

Relying on the simulations, Fig. 5.2.24 shows the frequency-domain representation of the
voltages induced in the “p-1” search coil. The magnitude of the new frequency sideband
components characteristic of a rotor cage fault are found to be higher for the motor equipped
with windings of the “NB” and “SB” types. At no load, important signs of cage fault are lost
and the condition of the motor may become difficult to assess, even in the case with 3 broken
rotor bars. The fault signatures found from a motor equipped with the “SB” and “4B” winding
configurations were not included in this figure, since they look similar to the ones
corresponding to the “NB” case. The dB representations are with reference to the 150-Hz
component, since this was found to have the highest amplitude.

Fig. 5.2.23 Time-domain representation of the voltages induced in the search coil “p-1” – 3
broken rotor bars – measurements (top) and simulations taken at half load.

Based on the measurements, Fig. 5.2.25 shows the frequency-domain representation of the
voltages induced in the search coil “p-1” for the case of a broken rotor bar. The clear
agreement between the measurements and simulations in terms of characteristic frequency
sidebands components found responsible for the detection of such abnormalities is obvious.
This picture also points out the reduced ability to detect faults of measurements at lower
motor loadings; a broken bar is undetected. The dB representation is with reference to the
150-Hz component.

86
Fig. 5.2.24 Frequency-domain representation of the voltages induced in the “p-1” search coil
during various cage-related faults – simulations at various loads.

Fig. 5.2.25.a Frequency-domain representation of the voltages induced in the “p-1” search
coil – one broken rotor bar – measurements at full load.

87
Fig. 5.2.25.b Frequency-domain representation of the voltages induced in the “p-1” search
coil – measurements at no load.

SEARCH COIL “p+1”

Fig. 5.2.26 presents, both from the measurements and simulations, the time-domain
representation of the voltages induced in the search coil “p+1” for the case of a motor
operating with three broken rotor bars.

Fig. 5.2.26 Time-domain representation of the voltages induced in the search coil “p+1” –
three broken rotor bars – measurements (top) and simulations at half load.

Relying on the simulations, Fig. 5.2.27 shows the frequency-domain representation of the
voltages induced in the “p+1” search coil. The magnitude of the new frequency sideband
components characteristic of a rotor cage fault are found to be higher for the motor equipped
with a stator winding of the types “NB” or “SB”. Similarly to the case of the “p-1” search
coil, at no load, important signs of cage fault are lost and the condition of the motor may
become difficult to assess, even in the case of three broken rotor bars.
Based on the measurements, Fig. 5.2.28 shows the frequency-domain representation of the
voltages induced in the search coil “p+1” for various rotor cage faults. The agreement
between the measurements and simulations in terms of the characteristic frequency sideband
components found responsible for the detection of such abnormalities is very clear.

88
Fig. 5.2.27 Frequency-domain representation of the voltages induced in the “p+1” search
coil during various cage-related faults – simulations at various loads.

Fig. 5.2.28.a Frequency-domain representation of the voltages induced in search coil “p+1”
during various cage-related faults – measurements at full load.

89
Fig. 5.2.28.b Frequency-domain representation of the voltages induced in search coil “p+1”
during various cage-related faults – measurements at no load.

SEARCH COILS AT THE NON-DRIVE END AND AROUND THE MOTOR SHAFT

Fig. 5.2.29 presents the time-domain representation of the voltages induced in the search coils
located at the non-drive end of the motor and around the motor shaft, respectively for an
operation of the test motor characterized by 3 broken rotor bars. For these cases, simulations
are not available. The dB representations are with reference to the fundamental frequency.
Fig. 5.2.30 points out the characteristic frequency sideband components produced by rotor
cage-related faults in the indications provided by a search coil placed at the non-drive end of
the test motor.

Fig. 5.2.29 Time-domain representation of the voltages induced during a cage-related fault in
the search coils located at the non-drive end of the motor near the ventilator (top) and around
the motor shaft (bottom) – measurements at full load.

90
Fig. 5.2.30 Characteristic frequency sidebands produced by a rotor cage-related fault in the
indications provided by the search coil placed at the non-drive end of the motor –
measurements at full load.

For lower loadings of the test motor, the useful fault indications decrease drastically. The only
frequency sidebands to provide potential information about fault remain the ones at a very
low frequency and in the vicinity of the fundamental (diminished by approx. 5 dB for the case
of three broken rotor bars). At half load and no load, only the case with three broken rotor
bars is detectable.
Fig. 5.2.31 presents the frequency-domain representation of the voltages induced in the search
coil placed around the motor shaft. For a healthy machine operation, these components are
non-existent. In contrast to the search coil intended to sense the flux in the non-drive area of
the test motor, this search coil is able to provide useful information about a fault with the
same accuracy at half load. However, at no load only the case with three broken bars remains
relatively visible, but with a lower sensitivity (on average the magnitudes of the characteristic
fault sidebands are about 10 dB lower).

Fig. 5.2.31 Characteristic frequency sidebands produced by various rotor cage-related faults
in the indications provided by the search coil placed around the motor shaft – measurements
at full load.

Discussions
According to the initial expectations, a large supply current was required to flow in order to
facilitate successful detection of rotor faults during steady-state operation. This large current
is obtained whilst the motor is running under full load and half load conditions.
At least for a motor operating at full and half loads, the rotor cage-related faults that were
studied were reflected in well-defined associated fault frequency components (given by Eq.
3.11) in all of the search coils under study. This is in opposition to the results obtained from
the case of the branch current monitoring, where a cage-related fault was reflected in this
manner (Eq. 3.2) only for those motors equipped with a stator winding configuration of the
types “2B” and “4B”.

91
The distortions in the branch current resulting from a cage fault given by Eq. 3.1 were found
in all of the cases that were studied, but such signatures may not be confidently used for rotor
cage fault detection purposes since they may also appear as a consequence of some non-faulty
operating conditions, such as pulsating loads, interactions between the motor and the train
equipment, and particular rotor designs.

Following the discussion previously presented, the benefits of using search coils for detecting
rotor cage-related faults become obvious.

5.3. Static eccentricity, dynamic eccentricity, bearing fault

SEARCH COIL AROUND STATOR TOOTH

Based on the measurements and simulations, respectively, Figs. 5.3.1 and 5.3.2 present the
time-domain representation of the voltages induced in the search coil located around a stator
tooth of the test motor under various operating conditions. For this search coil, the dB
representations shown in Figs. 5.3.3 and 5.3.4 are with reference to the fundamental
frequency.
From the measurements, it was found that a bearing fault and static eccentricity could not be
detected from the indications provided by this search coil. For the case of static eccentricity,
the same conclusion was drawn after analysing the simulation data corresponding to the
motor equipped with the original winding configuration of type “2B” and “4B” (see Fig.
5.3.4).
The most important changes produced by the dynamic eccentricity are shown in Figs. 5.3.3
and 5.3.4 and, for both the measurements and simulations, the changes corresponding to
various motor loadings are summarised in Table 5.3.1. In contrast to the case of static
eccentricity, the dynamic eccentricity is observable for a motor equipped with all the stator
winding configurations. It is observed that the magnitude of the new frequency component
indications provided by this search coil are higher at a lower loading for a motor equipped
with stator winding configurations of the “NB” and “SB” types. This conclusion is in
agreement with the one drawn by Dorrell (1993b), who claimed that the UMP resulting from
dynamic eccentricity decreases as the motor is loaded from no-load.

Fig. 5.3.1 Time-domain representation of the voltages induced in the search coil placed
around a stator tooth – bearing failure, dynamic eccentricity – measurements at half load.

92
 NB SB 2B 4B

Fig. 5.3.2 Time-domain representation of the voltages induced in the search coil placed
around a stator tooth – static eccentricity – simulations at half load.

Fig. 5.3.3 Frequency-domain representation of the voltages induced in a search coil placed
around a stator tooth – dynamic eccentricity – measurements at half load.

Fig. 5.3.4 Frequency-domain representation of the voltages induced in the search coil placed
around a stator tooth – simulations at various loads.

93
 Table 5.3.1 Amplitudes of the sideband frequencies characteristic of dynamic eccentricity [dB].
Full load Half load No load
Simul. Meas. Simul. Meas. Simul. Meas.
f [Hz] ‘NB’ ‘NB’ ‘NB’
‘4B’ ‘2B’ ‘2B’ ‘4B’ ‘2B’ ‘2B’ ‘4B’ ‘2B’ ‘2B’
‘SB’ ‘SB’ ‘SB’
51,5 -47 -57 -52 -43
50,7 -45 -65 -56 -50
50 -35 -65 -54 -47
148,5 -25 -38 -36 -24
149,2 -24 -39 -36 -23
150 -22 -37 -34 -23
1991.5 -26 -27 -28 -28
1918.5 -26 -24 -24 -28
1949.5 -21 -21 -20 -39
2088.5 -19 -19 -18 -27
2118.5 -31 -31 -33 -28
2149.5 -31 -33 -31 -37

Fig. 5.3.5 presents the characteristic frequency sideband components produced by the static
eccentricity in the indications provided by a search coil placed around a stator tooth of a
motor equipped with stator windings of the types “NB” and “SB”. At a lower loading of the
machine, some of the sidebands important for assessing the condition of the motor are lost
and, on average, the level of magnitude for all the remaining characteristic fault sidebands
diminishes by about 10-15 dB. This behaviour is in agreement with the results presented by
other researchers (Dorrell 1993a), stating that as a result of static eccentricity the UMP
increases rapidly as the motor is loaded from no load.

full load

no load

Fig. 5.3.5 Characteristic frequency sidebands produced by static eccentricity in the

indications provided by the search coil placed around a stator tooth – simulations.

In the high-frequency range, the experimentally implemented dynamic eccentricity was

reflected in the rotor-slot harmonics by increasing their magnitudes by around 3 dB
irrespective of the machine loading in comparison with healthy operation. Such modifications
are not found in the simulations in any of the stator winding configurations.

94
Slightly higher changes (5-7 dB) are to be found for the experimentally implemented static
eccentricity, while no changes in the amplitude of the rotor-slot harmonics are to be found
from the simulations.
In the high-frequency range, the experimental implemented bearing fault was very poorly
reflected in the rotor-slot harmonics, increasing their magnitudes only by around 1-2 dB,
irrespective of the machine loading in comparison with healthy operation.

SEARCH COIL PLACED AROUND TWO POLE PITCHES

Based on the measurements and simulations, respectively, Figs. 5.3.6 and 5.3.7 present the
time-domain representation of the voltages induced in the search coil located around two pole
pitches of the test motor under various operating conditions.

Fig. 5.3.6 Time-domain representation of the voltages induced in the search coil placed
around two pole pitches – bearing fault, dynamic eccentricity – measurements at half load.

Fig. 5.3.7 Time-domain representation of the voltages induced in the search coil placed
around two pole pitches – simulations at half load.

95
According to the measurements, only the dynamic eccentricity may be detected via an
external search coil placed around two pole pitches (Fig. 5.3.8). The frequencies that offer
possible information about this fault are marked by arrows and are according to Eq. 3.7.
It is relatively clear that at a lower motor loading, the dynamic eccentricity appears clearer in
the flux spectrum provided by this search coil.
From the simulations it was found that new frequency components characteristic of static
eccentricity appear only in the cases of a motor equipped with stator winding configurations
of the “NB” and “SB” types (Figs. 5.3.9 and 5.3.10).
For dynamic eccentricity, it was found that new frequency sideband components appear,
irrespective of the stator winding configuration, and the asymmetry becomes clearer for a
motor operating at a lower loading (Fig. 5.3.9). This is in accordance with the results obtained
from the measurements and presented in Fig. 5.3.8.
The magnitudes of these new fault frequency sideband components are presented with
reference to the rotor slot frequencies (2040 Hz and 2100 Hz for operation at full load and no
load, respectively). It is important to notice that even at a lower loading, the magnitudes of the
new frequency sidebands resulting from static eccentricity do not diminish.

Fig. 5.3.8 Frequency-domain representation of the voltages induced in a search coil placed
around two pole pitches – dynamic eccentricity – measurements at various loads.

The simulation tool in which the dynamic eccentricity was implemented is able to predict the
same new frequency sideband components as the ones found in the measurements. The
signatures produced by the dynamic eccentricity in this search coil are identical to the ones
found in the current spectrum (Section 5.1) and summarised in Table 5.3.2.

96
Fig. 5.3.9 Frequency-domain representation of the voltages induced in the search coil placed
around two pole pitches – static eccentricity, dynamic eccentricity – simulations at various
loads.

97
 NB
SB
full load

NB
SB
no load

Fig. 5.3.10 Characteristic frequency sidebands produced by static eccentricity in the

indications provided by the search coil placed around two pole pitches of the test motor –
simulations at various loads.

Table 5.3.2 Amplitudes of the sideband frequencies characteristic of dynamic eccentricity [dB].
Full load Half load No load
Me- Me- Me-
Simul. Simul. Simul.
f [Hz] as. as. as.
‘NB’ ‘SB’ ‘4B’ ‘2B’ ‘2B’ ‘NB’ ‘SB’ ‘4B’ ‘2B’ ‘2B’ ‘NB’ ‘SB’ ‘4B’ ‘2B’ ‘2B’
51,5 -21 -21 -28 -21 -21
50,7 -10 -10 -18 -18 -15
50 -8 -8 -16 -12 -15
148,5 -8 -8 -16 -8 -18
149,2 -1 -1 -6 -3 -11
150 -5 -1 -3 2 -12
1991.5 -11 -11 -8 -2 -20
1918.5 -18 -18 -14 -10 -16
1949.5 -12 -30 -12 -2 -20
2088.5 -4 -4 -1 -9 -35
2118.5 -5 -5 -2 -20 -14
2149.5 -9 -5 -4 -12 -13

In the high-frequency range, the experimentally implemented dynamic eccentricity was

reflected in the rotor-slot harmonics by increasing their magnitudes by around 2 dB in full-
load motor operation in comparison with healthy operation. No such signs were found for the
half-load and no-load motor operation points. From the simulations, such modifications could
not be analysed, since in healthy operation the signal captured by this coil is ideally zero.
Slightly higher modifications (3-5 dB) are to be found for the experimental implemented
static eccentricity, irrespective of the motor loading, while from the simulations such
modifications could not be analysed, since in healthy operation the signal captured by this coil
is ideally zero.
In the high-frequency range, the experimentally implemented bearing fault was very poorly
reflected in the rotor-slot harmonics, increasing their magnitudes by around only 2 dB only at
half-load in comparison with healthy operation.

98
SEARCH COIL “p+1”

Based on the measurements and simulations, respectively, Figs. 5.3.11 and 5.3.12 present the
time-domain representation of the voltages induced in the “p+1” search coil under various
operating conditions.

Fig. 5.3.11 Time-domain representation of the voltages induced in search coil “p+1” –
bearing fault, dynamic eccentricity – measurements at half load.

static eccentricity

NB, SB

static eccentricity

2B, 4B

dynamic eccentricity NB, SB

dynamic eccentricity 2B, 4B

Fig. 5.3.12 Time-domain representation of the voltages induced in search coil “p+1” – static
eccentricity, dynamic eccentricity – simulations at half load.

99
Based on the measurements and simulations at half load, Figs. 5.3.13 and 5.3.14 show the
frequency-domain representation of the voltages induced in the search coil “p+1” for the case
of static eccentricity. A summary of the most important changes offering potential
information about the static eccentricity for all of the motor loadings that were studied may be
found in Tables 5.3.3 and 5.3.4. The dB representation is with reference to the 150-Hz
component in the case of the measurements and with reference to the rotor-slot harmonics for
the simulations (full load 1840 Hz, half load 1870 Hz, no load 1900 Hz).

Fig. 5.3.13 Frequency-domain representation of the voltage induced in the “p+1” search
coil – static eccentricity – measurements at half load.

Table 5.3.3 Main changes produced by the static eccentricity in the

indications provided by the “p+1” search coil – measurements [dB].
Ustatic eccentricity-Uhealthy
f [Hz] 100 300 rotor-slot harmonics
full load 11 1 7-15
half load 13 10 5-9
no load 14 2-4 5-7

Fig. 5.3.14 Frequency-domain representation of the voltage induced in the “p+1” search coil
– static eccentricity – simulations at half load.

Table 5.3.4 Main changes produced by the static eccentricity in the indications provided
by the “p+1” search coil – simulations at various loads [dB].
f [Hz] 100 150 50 300
winding config. ‘NB’ ‘SB’ ‘2B’ ‘4B’ ‘NB’ ‘SB’ ‘2B’ ‘4B’ ‘2B’ ‘2B’
full load -5 -5 -8 -11 -10 -10 -30 -45 -42 -30
half load -5 -5 -7 -11 -7 -7 -26 -45 -40 -30
no load -8 -8 -8 -12 -12 -12 -21 -46 -38 -30

It is important to notice that the magnitude of the frequency components of interest for the
detection of static eccentricity is inversely proportional to the number of parallel branches to
be found in the stator winding of the test motor.
Based on the measurements and simulations at half load, Figs. 5.3.15 and 5.3.16 show the
frequency-domain representation of the voltages induced in the search coil “p+1” for the case
of dynamic eccentricity. A summary of the most important modifications offering potential
information about the dynamic eccentricity for all of the motor loadings that were studied
may be found in Tables 5.3.5 and 5.3.6. The dB representation is with reference to the rotor-
slot harmonics for both the simulations and measurements.

100
Fig. 5.3.15 Frequency-domain representation of the voltage induced in the “p+1” search coil
– dynamic eccentricity – measurements at full load.

Fig. 5.3.16 Frequency-domain representation of the voltages induced in the “p+1” search
coil – dynamic eccentricity – simulations at full load.

Table 5.3.5 Main changes produced by the dynamic

eccentricity – measurements at various loads [dB].
f [Hz] 50 150 250
full load 6 8 8
half load 2 4 11
no load 12 16 21

Table 5.3.6 Main changes produced by the dynamic eccentricity in the indications provided
by the “p+1” search coil – simulations at various loads [dB].

It is observed that at lower levels of motor loading, the magnitude of the frequency
component that exhibits the most important change (150 Hz) is higher. This behaviour was
also observed from the simulations (Table 5.3.6).
Concerning the behaviour of the sideband components responsible for the detection of the
dynamic eccentricity in the case of a motor equipped with various stator winding
configurations, a similar effect to the one observed with static eccentricity is found.
A comparative analysis between the measurements and simulations points out a very good
agreement and the ability of the simulation tool to predict the same new frequency sideband
components.

101
A bearing fault may not be detected from the indications provided by this search coil. In the
high-frequency range, the maximum modifications of the rotor-slot harmonic magnitudes do
not exceed the level of 2-3 dB, irrespective of the motor loading.
In comparison with the previous search coils, both static and dynamic eccentricity may be
detected more easily with the use of a search coil purposely intended to capture the harmonics
of the order “p+1” (Table 5.3.7).

Table 5.3.7 Changes to be found in the magnitude of the rotor-slot

harmonics – measurements at various loads [dB].
Static eccentricity Dynamic eccentricity
Ustatic eccentricity – Uhealthy Udynamic eccentricity – Uhealthy
Full load Half-load No load Full load Half load No load
6-13 4-11 6-12 12-24 8-22 8-17

SEARCH COIL “p-1”

Based on the measurements and simulations, respectively, Figs. 5.3.17 and 5.3.18 present the
time-domain representation of the voltages induced in the “p-1” search coil under various
operating conditions.

Fig. 5.3.17 Time-domain representation of the voltages induced in search coil “p-1” –
bearing fault, dynamic eccentricity – measurements at half load.

NB, SB

2B, 4B

Fig. 5.3.18.a Time-domain representation of the voltages induced in search coil “p-1” – static
eccentricity – simulations at half load.

102
 NB, SB

2B, 4B

Fig. 5.3.18.b Time-domain representation of the voltages induced in search coil “p-1” –
dynamic eccentricity – simulations at half load.
Based on the measurements and simulations, respectively, Figs. 5.3.19 and 5.3.20 show the
frequency-domain representations of the voltages induced in the search coil “p-1” for the
cases of static and dynamic eccentricity. The dB representations are presented with reference
to the rotor slot frequencies for the simulations and with reference to the 150-Hz component
for the measurements. Tables 5.3.8-5.3.12 present a summary of the most important
modifications offering potential information about the static eccentricity for all of the loadings
under study.

Fig. 5.3.19 Frequency-domain representation of the voltages induced in the “p-1” search coil
– healthy and static eccentricity – measurements (topmost figure) and simulations with
various loads.

103
Fig. 5.3.20 Frequency-domain representation of the voltages induced in the “p-1” search coil
– healthy and dynamic eccentricity – measurements (two topmost figures) and simulations at
full load.

Table 5.3.8 Main changes produced by the static eccentricity in the

indications provided by the “p-1” search coil – measurements [dB].
U´static eccentricity – Uhealthy
f [Hz] 100 300
full load 15 10
half load 14 10
no load 15 20

Table 5.3.9 Main changes produced by the static eccentricity in the

indications provided by the “p-1” search coil – simulations [dB].
f [Hz] 50 150
winding
‘NB’ ‘SB’ ‘2B’ ‘4B’ ‘NB’ ‘SB’ ‘2B’ ‘4B’
config.
full load -27 -27 -49 -60 -22 -24 -41 -60
half load -22 -22 -44 -61 -21 -21 -36 -54
no load -10 -10 -36 -60 -22 -22 -28 -52

104
 Table 5.3.10 Behaviour of the 3rd harmonic for the case of
dynamic eccentricity in the indications provided by the “p-1”
search coil – measurements.
Ustatic eccentricity / Uhealthy
full load 2
half load 1.4
no load 0.25

Table 5.3.11 Main changes produced by the dynamic eccentricity in the

indications provided by the “p-1” search coil – simulations [dB].
f [Hz] 50 150
winding config. ‘NB’ ‘SB’ ‘2B’ ‘4B’ ‘NB’ ‘SB’ ‘2B’ ‘4B’
full load -16 -16 -28 -31 -1 -1 -9 -14
half load -10 -10 -26 -29 2 2 -7 -12
no load 0 0 -23 -28 3 3 -4 -10

Table 5.3.12 Changes in the magnitude of the rotor-slot harmonics observed in the
indications provided by the “p-1” search coil – measurements [dB].
Static eccentricity Dynamic eccentricity
Ustatic eccentricity – Uhealthy Udynamic eccentricity – Uhealthy
Search coil Full load Half load No load Full load Half load No load
“p-1” 4-6 5-6 4-6 14-26 12-20 8-20

A comparative analysis between the measurements and simulations in terms of measured and
predicted specific frequency components points out a very good agreement for the case of
static eccentricity.

In the high-frequency range, it is almost impossible to detect bearing fault with this search
coil. The maximum modifications of the rotor-slot harmonic magnitudes do not exceed the
level of 3-4 dB, irrespective of the motor loading.

In comparison with the previous search coils, both static and dynamic eccentricity may be
detected more easily via the use of a search coil purposely intended to capture the harmonics
of the order “p-1” (Table 5.3.12).

105
SEARCH COILS LOCATED AROUND THE SHAFT AND AT THE NON-DRIVE END OF THE
MOTOR

Figs. 5.3.21 and 5.3.22 present the time-domain representation of the voltages induced in the
search coil located around the motor shaft and at the non-drive end of the motor, respectively.
For this search coils, simulations are not available and the dB representations are with
reference to the fundamental frequency.

From the frequency-domain representation of the voltages induced in these search coils (Figs.
5.3.23 and 5.3.24) it was found that:
- a search coil wound around the shaft and purposely intended to sense the axial flux is able
to indicate a fault in the bearing. The measurements carried out for this work show that this
was the clearest means to observe new frequency components resulting from this fault.
However, this signature is almost similar to the one obtained from the static eccentricity;
- a dynamic eccentricity is easily detected by a search coil positioned around the motor shaft
at any motor loading
- the search coil placed externally at the non-drive end of the motor is able to indicate the
dynamic eccentricity even from measurements at no load but is not able to sense the cases of
static eccentricity and bearing fault.

Fig. 5.3.21 Time-domain representation of the voltages induced in the search coil located
around the shaft – bearing fault, dynamic eccentricity – measurements at half-load.

Fig. 5.3.22 Time-domain representation of the voltages induced in the search coil positioned
at the non-drive end of the motor – bearing fault, dynamic eccentricity – measurements at
half-load.

106
Figs. 5.3.23 Frequency-domain representation of the voltages induced in a search coil placed
around the motor shaft – static eccentricity, dynamic eccentricity, bearing fault –
measurements at half load.

Fig. 5.3.24 Frequency-domain representation of the voltage induced in a search coil placed
at the non-drive end of the test motor – dynamic eccentricity – measurements at half load.

107
Figs. 5.3.25-5.3.28 summarise the characteristic frequency sideband components produced by
the faults presented in Figs. 5.3.23 and 5.3.24 and also present similar results obtained from
other motor loadings.

Fig. 5.3.25 Characteristic frequency sideband components produced by the static eccentricity
in the indications provided by the search coil wound around the motor shaft – measurements.

Fig. 5.3.26 Characteristic frequency sideband components produced by bearing fault in the
indications provided by the search coil wound around the motor shaft – measurements at
various loads.

Fig. 5.3.27 Characteristic frequency sideband components produced by the dynamic

eccentricity in the indications provided by the search coil wound around the motor shaft –
measurements at various loads.

108
Fig. 5.3.28 Characteristic frequency sideband components produced by the dynamic
eccentricity in the indications provided by the search coil placed at the non-drive end of the
motor – measurements at various loads.

In the high-frequency range, the bearing fault is relatively more clearly reflected in the
indications provided by the search coil around the shaft. An increment of around 10 dB in the
magnitude of the rotor-slot harmonics at full load seems to be more important than the similar
modifications produced by the same fault and by various faults in other search coil
indications. For a half-load and no load, the same modifications are about 7 and 4 dB,
respectively.

For the search coil around the shaft, in the high-frequency range, the experimental
implemented static eccentricity was reflected in the rotor-slot harmonics by increasing their
magnitudes with around 8-10 dB in full load motor operation in comparison with healthy
operation. For the half-load and no-load motor operating points, these indications decrease to
levels of 4 and 3 dB, respectively.

The dynamic eccentricity is not reflected in any modifications of the rotor-slot harmonics,
irrespective of the motor loading for both of these search coils.

109
Discussion
Since it has been shown that the dynamic eccentricity may be easily observed in all of the
fault indicators that were studied, in this section the focus will be mainly on a critical analysis
of the means to detect the static eccentricity and a bearing fault.

Search coil around stator tooth. From the measurements, it was found that it was not
possible to detect a bearing fault or the static eccentricity from the indications provided by
this search coil. For the case of static eccentricity, the same conclusion was drawn after
analysing the simulation data corresponding to the motor equipped with the original winding
configuration (type “2B”). Since not even the rotor-slot harmonics offer an important hint
about the static eccentricity, it becomes clear that other means of detecting this type of
eccentricity are needed.
Search coil around two pole pitches. On the basis of the measurements and simulations,
only the dynamic eccentricity may be detected via an external search coil placed around two
pole pitches. In the high-frequency range, the experimental implemented dynamic eccentricity
was not reflected very clearly in the magnitude of the rotor-slot harmonics (the increase in the
magnitude of the rotor-slot harmonics was only 2 dB in full-load motor operation, in
comparison with healthy operation, and smaller for lower loadings). From the simulations, it
was not possible to analyse such modifications, since in healthy operation the signal captured
by this coil is ideally zero. Slightly higher modifications (3-5 dB) are to be found for the
experimental implemented static eccentricity, irrespective of the motor loading, while from
the simulations such modifications could not be analysed, since in healthy operation the signal
captured by this coil is ideally zero. However, one may not confidently rely on such
modifications to claim the ability to detect a static eccentricity, since such modifications are
common to other faults/asymmetries (i.e. turn-to-turn short circuit in the stator winding). In
the high-frequency range, the experimental implemented bearing fault was very poorly
reflected in the rotor-slot harmonics, increasing their magnitudes by around only 2 dB only at
half-load in comparison with healthy operation.
Search coil “p+1”. From the measurements, the static eccentricity is clearly revealed in the
information provided by the changes in the magnitude of the following frequency components
in comparison with healthy operation:
- fundamental: increments between 11-14 dB,
- 3rd harmonic: increments between 2-10 dB,
- rotor-slot harmonics: increments between 5-15 dB (irrespective of the loading)
Search coil “p-1”. Relying on the measurements, the static eccentricity is clearly revealed in
the information provided by the changes in the magnitude of the following frequency
components in comparison with the healthy operation:
- fundamental: increments of around 15 dB (irrespective of the loading),
- 3rd harmonic: increments between 10-20 dB (getting higher at lower loads),
- rotor-slot harmonics: increments between 4-6 dB (irrespective of the loading).

It was not possible to detect a bearing fault from the “p±1” search coils, since only a small
change in the magnitude of the rotor-slot harmonics (increment of 2-3 dB) in comparison with
healthy operation was found. Moreover, such minor changes in the magnitude of the rotor-slot
harmonics were also found for other types of faults, and therefore, such an indication is not
even suitable for achieving discrimination between them.

110
Search coil wound around the shaft. A search coil purposely intended to sense the axial flux
was found to be able to indicate a fault in the bearing, this being the clearest means to observe
a signature produced by this fault. However, this signature was found to be almost similar to
the one obtained from the static eccentricity, which, in terms of the appearance of new
frequency components, was also invisible in the indications of the search coils previously
mentioned. In the high-frequency range, the bearing fault was relatively more clearly reflected
in the indications provided by this search coil; an increment of around 10 dB in the magnitude
of the rotor-slot harmonics at full load seems to be more important than the similar
modifications produced by the same fault and by various faults in other search coil
indications. For the half-load and no load, the same modifications were about 7 and 4 dB,
respectively.
The static eccentricity was also reflected in the rotor-slot harmonics, increasing their
magnitudes by 8-10 dB in full-load motor operation in comparison with healthy operation.
For the half-load and no-load motor operating points, these indications decrease to levels of 4
and 3 dB, respectively.

Search coil placed externally at the non-drive end of the motor. This search coil was not
able to sense the cases of static eccentricity and bearing fault. Potential indicators about these
faults, if there were any, were very small and were probably covered by the noise floor.
However, even with this poor performance in detecting static eccentricity and a bearing fault,
such an external search coil is able to indicate the dynamic eccentricity even from
measurements at no load.

111
5.4. Core fault

The study of the stator winding currents and branch currents does not provide any useful sign
of the core faults under study in any of the stator winding configurations that were
investigated.

Electromagnetic flux monitoring

Since, in the simulations, the test motor is considered to be perfect symmetrical, in healthy
operation the only non-zero indications are provided by the search coil around a stator tooth.
The indications provided by this coil for the healthy condition remain unchanged, irrespective
of both the various stator winding configurations that were studied and the core faults that
were implemented. The same findings were also found to be valid for the search coil placed
around two pole pitches of the test motor.
The only signs found to be suitable for the detection of such types of fault were the ones
provided by the search coils “p-1” and “p+1”. However, the indications provided by these
search coils do not show important changes for the various degrees of the faults that were
studied (core fault “cf” 1-9; see Figs. 4.2-4.3).
Fig. 5.4.1 presents the time-domain representation of the voltages induced in the “p-1” and
“p+1” search coils in motor operation with the most severe core fault that was studied.

Fig. 5.4.1 Time-domain representation of the voltages induced in the “p-1” (top) and “p+1”
(bottom) search coils in motor operation with a core fault – simulations.
Next, the possibility of sensing core fault in each of the winding configurations that were
studied is presented.

Stator winding configuration of type “4B”

This fault is not characterised by any new characteristic fault frequency sidebands. The
magnitude of the fundamental frequency, which decreases with the progress of the fault
degree, is the only sign of a core fault problem.
Table 5.4.1 Modifications in the magnitudes of the fundamental
frequency captured by search coil “p-1” (left) and “p+1” (right).

112
Stator winding configuration of type “2B”
No considerable change in the magnitude of the fundamental component was found for the
motor equipped with a stator winding consisting of two parallel branches.

Table 5.4.2 Modifications in the magnitudes of the fundamental

frequency captured by the search coils “p-1” (left) and “p+1” (right).

Stator winding configuration of types “SB” and “NB”

Table 5.4.3 Modifications in the magnitudes of the fundamental frequency
captured by the search coils “p-1” (left) and “p+1” (right) – winding
configuration of “SB” type.

Table 5.4.4 Modifications in the magnitudes of the fundamental

frequency captured by the search coils “p-1” (left) and “p+1” (right) –
winding configuration of “NB” type.

Some new frequency sideband components are found at the frequencies of 50 Hz and 150 Hz,
respectively. Their magnitudes with reference to the one at the fundamental frequency are
represented in Table 5.4.5. New sideband components at these frequencies were also found as
a result of the simulated inter-turn short circuit in the stator winding.

Table 5.4.5 Magnitudes of the frequency components considered potential fault

indicators for the detection of core fault [dB].

113
114
6. FINAL DISCUSSION

This section is intended to gather and concentrate all the important conclusions presented at
the end of each subsection dealing with the analysis of the experimental and simulated data.
In this respect, the reader should expect from this section a more qualitative critical analysis
rather than a quantitative one.
First of all, the ability of the phase current to detect specific faults was tested, since
monitoring this parameter is the most convenient and cheapest way to sense a fault.
In principle, any rotor cage-related fault is clearly seen from stator currents when checking the
specific sideband frequency components extracted from the well-known equations presented
in the literature.
As observed from both the measurements and simulations, a turn-to-turn short circuit in the
stator winding was best reflected in the unbalance of the stator winding phase currents and in
the rise in magnitude of some frequency components which already exist in the line current
spectra of a healthy motor.
Relying on both experiments and simulations, this work concludes that the static eccentricity
does not bring about any major distortion in the spectral content of the branch current. A
similar conclusion was found for the experimentally implemented bearing and core fault,
respectively.
Since the stator winding phase of the test induction motor consists of two half-phases
(branches) connected in parallel and these phases are accessible from outside the motor, the
analysis of the differential circulating current between the half-phases was performed with the
purpose of fault detection. The measured and simulated waveforms of the circulating currents
between stator winding parallel branches for healthy machine operation and for the faults
under study were presented. It was observed that a major rotor fault (such as 3 broken rotor
bars), a stator-related fault (such as a turn-to-turn short circuit), and the dynamic eccentricity
can clearly be seen by monitoring the circulating currents. However, by relying only on the
indications provided by the circulating current, it appears difficult to discriminate between
various fault conditions. Moreover, the cases of static eccentricity and bearing fault are not
possible to detect as abnormal conditions on the sole basis of monitoring only the magnitudes
of the circulating currents. Taking into account the drawbacks characteristic of a monitoring
method relying on the stator current, it was further proceeded to study the ability of the
electromagnetic fluxes captured in various motor locations to sense the same faults.
From the measurements, it was found that the indications provided by the search coil
purposely intended to sense the electromagnetic flux around a stator tooth are not of much use
when aiming to detect a stator winding fault. The inability of the information provided by this
search coil refers both to the absence of new frequency sideband components characteristic of
a stator fault and to major changes in the magnitude of the frequency components produced
by the slotting.
However, an external coil over two pole pitches may provide more useful information about
the inter-turn short circuit, giving an increment in the fundamental frequency of about 3 and 5
times in half load and no load testing.
When aiming to detect a stator winding fault, the search coils purposely intended to sense the
air-gap harmonics of the order “p±1” provided richer information about this fault, in
comparison with the other search coils implemented in the simulations. A “p-1” search coil is
useful in that it is able to detect important changes in the magnitudes of the 3rd, 5th, and rotor-

115
slot harmonics, while the “p+1” search coil is able to sense important changes in the
magnitude evolution of the fundamental frequency. Such important changes in the magnitude
of the fundamental frequency were also found in those search coils implemented only in the
measurements: around the shaft and at the non-drive end.
From the simulations it was important to notice that new characteristic fault sideband
frequency components are introduced by the stator winding configurations “NB” and “SB”
during the inter-turn short circuit. Unfortunately, none of these winding configurations was
studied experimentally in order to validate this effect. Usually, a motor equipped with an
“SB” type stator winding configuration is observed to be more sensitive to the inter-turn short
circuit than one equipped with an “NB” configuration.
A rotor cage-related fault is the most easily detected abnormality in an induction motor and it
was no surprise to find out that all of the search coils that were studied were found to be able
to provide useful information about such faults. A good agreement between the data obtained
from the measurements and simulations was found. This agreement is in terms of the new
characteristic fault frequency sideband components produced by a rotor cage fault.
For a rotor-cage related fault, the richness of the information provided by the search coils that
were studied is proportional to the loading level at which the test motor is operated. At a no-
load operation, for instance, important signs of the cage fault are lost and the motor condition
may become difficult to assess.
A search coil wound around the shaft and purposely intended to sense the axial flux was able
to indicate a fault in the bearing and it appeared that this was the clearest means to observe a
modification introduced by this fault. However, this signature is almost similar to the one
obtained from the static eccentricity.
In terms of new frequency sideband components produced by a static eccentricity, the search
coil around the shaft is the only one to sense such modifications. The static eccentricity is
reflected in the indications of the “p+1” and “p-1” search coil only by increments in the
magnitudes of specific frequency sideband components. The search coils around the stator
tooth and over two pole pitches are of no great use when trying to sense a static eccentricity.
A search coil placed externally at the non-drive end of the motor is able to indicate the
dynamic eccentricity even from measurements at no load, but is not able to sense cases of
static eccentricity and bearing fault.
The case of dynamic eccentricity is detected without any difficulty, irrespective of the motor
loading, by all of the search coils that were studied. It is important to notice that for lower
levels of motor loading, the magnitude of these frequency components is higher.
A comparative analysis between the measurements and simulations for the case of dynamic
eccentricity points out a good agreement in terms of predicted and measured new frequency
sideband components responsible for the detection of this asymmetry (excepting the case of
the “p+1” search coil).
The study of the stator winding currents and branch currents does not provide any useful sign
for the case of a core fault in any of the stator winding configurations that were investigated.
The only signs found suitable for the detection of such types of fault were the ones provided
by the search coils “p-1” and “p+1”. However, the indications provided by these search coils
do not show important changes for the various degrees of the faults under study.
Unfortunately, no measurements with an artificially implemented core fault are available in
order to validate these claims.

116
An interesting conclusion drawn from the simulations is that for a motor equipped with a
stator winding of the “NB” or “SB” type, all of the faults under study were reflected in a
similar way in the indications provided by various fault indicators. It is important to notice
this, since these configurations are different (“NB” stator winding consisting of no parallel
branches, “SB” stator winding consisting of two parallel branches, but with a special
configuration - see Appendix II). Therefore, it is concluded that not only the number of
parallel branches but also the internal connections of the winding affects the fault indications.
Table 6.1 presents a summary of the results concerning the suitability of each of the fault
detection methods that were studied for the detection of specific faults. This table is purposely
intended, mainly, to offer a general view of those faults that are difficult to detect and/or need
special attention because of this.
The criteria used for ranking a method are not general for all of the cases and the reader is
recommended to carefully address each of the previous chapters for a clearer understanding of
those given here. For example, in the case of the inter-turn short circuit in the stator winding,
the phase branch current was ranked as a good fault indicator considering the unbalance
introduced by such a fault to the branch currents. The increase in the amplitudes of the
specific frequency sideband component of the branch current was also considered an
important symptom for ranking this parameter as a good fault indicator. On the other hand,
when ranking the phase current as a good fault indicator for a rotor cage fault, the presence of
new frequency components in its spectrum was relied on.
When claiming that a parameter offers poor/ambiguous information about a fault, the
possibility of finding similar indications for other faults is considered. However, this type of
classification was also used for the case of the core fault in order to distinguish between the
richness of the information provided by the “p±1” search coils for various stator winding
configurations.
In addition, Table 6.1 is intended to offer a starting point and some suggestions for further
work on various issues such as:
- investigations to be carried out using 3D simulation tools, thus making it possible to take
advantage of the important fault signatures (as observed from the measurements), provided by
the electromagnetic fluxes circulating along the axial direction of the machine,
- experimental implementation of the core faults and the subsequent measurements for
verifying the findings obtained from the simulations.
Important advances to this thesis may also be found in the investigations of other cage
induction machines and, finally, different machine types. Such an investigation would allow
some general conclusions to be drawn that would be of great help for the industry, which is
always interested in having a generalised strategy for the condition monitoring and fault
diagnosis of various types of electrical machines.

A challenging task for future research would be, for example, the study of rotor faults in
permanent magnet machines, since it was found that such research is very rare and such
studies are essential nowadays because of the increasing interest in the use of permanent
magnet machines.

117
Table 6.1. Faults versus recommended methods for detecting faults – summary.

118
7 CONCLUSIONS

This research work was mainly concerned with the analysis of the electromagnetic flux
patterns supposed to contain potential useful information about a fault in a cage induction
machine. Other electrical parameters used for fault detection such as stator phase current and
circulating currents between the parallel branches of the stator winding, were used as
comparative means to asses the usefulness of this parameter in fault detection.
The relevant fault signatures of the electrical parameters that were studied were mostly
derived both from the measurements and from numerical electromagnetic field simulations in
a steady state.
When and where possible, the accuracy of different fault signatures derived from numerical
electromagnetic field simulations was verified by experiments. For fault detection purposes,
the interest was focused more on qualitative than exact quantitative results, i.e. more focused
on finding how the monitored parameter behaves as a function of time and not so much in
magnitude. In this respect, despite various differences in the magnitudes of some simulated
and measured parameters, the agreement between the measurements and simulations was
considered as “good” in terms of the ability to detect fault characteristic frequency
components introduced by specific faults.
A rotor cage-related fault is the most easily detected abnormality in an induction motor and it
was no surprise to find out that by monitoring both the stator current and electromagnetic flux
at all the search coil locations one may obtain useful information about such faults. For a
rotor-cage related fault, the richness of the information provided by the search coils that were
studied is proportional to the loading level at which the test motor is operated.
A turn-to-turn short circuit in the stator winding was best reflected in the unbalance of the
stator winding phase currents. In terms of characteristic fault frequency components, observed
from the experiments, a turn-to-turn short circuit in the stator winding was reflected only in a
rise in magnitude of a few frequency components already present in the line current spectra of
a healthy motor. The search coils purposely intended to capture the harmonics of the order
“p±1” provide more useful information about a stator winding fault. No useful information
about static eccentricity and bearing faults was found in the stator current.
It was observed that a major rotor fault (such as a rotor bar breakage), a stator-related fault
(such as a turn-to-turn short circuit), and the dynamic eccentricity are clearly seen by
monitoring the circulating currents but such information is not good enough to discriminate
between various fault conditions. Moreover, it is not possible to detect the cases of static
eccentricity and bearing fault as abnormal conditions on the sole basis of monitoring the
magnitudes of the circulating currents.
The usefulness of the search coils purposely intended to capture the harmonics of the order
“p±1” may be found especially when trying to detect static and dynamic eccentricities.
However, the information provided by these coils is of no great significance when trying to
detect a bearing fault. A search coil intended to sense the axial flux and wound around the
shaft was found to be the only means to provide a relatively clear modification because of this
fault. However, the fault signatures provided by this coil are found to be almost similar to the
ones produced by the static eccentricity.
The case of dynamic eccentricity was also detected without any difficulty, irrespective of the
motor loading, by all the fault indicators that were studied. Moreover, for lower levels of
motor loading, the magnitude of the frequency components responsible for the detection of

119
such asymmetries is higher. This is in contrast with the case of the other faults studied,
excepting the inter-turn short circuit, where when the motor was operated at lower loadings
the sensitivity of the fault detection procedure diminished.
The study of the stator winding currents and branch currents does not provide any useful signs
in the case of a core fault in any of the stator winding configurations investigated. The only
signs found suitable for the detection of such types of fault were the ones provided by the
search coils “p-1” and “p+1”. However, the indications provided by these search coils do not
show important changes in the various degrees of the faults being studied.
The experimental and computational research contained in this thesis was carried out only for
a cage induction machine of 35 kW but, for some faults common to other types of electrical
machines, such as inter-turn short circuits in the stator winding, bearing faults, and various
types of eccentricity, it is possible to generalise these results. However, this thesis also
contains a few references concerning the use of the fault indicators under study for other types
of motor and their specific faults.
Finite element analysis was found, according to the initial expectations, to be a flexible tool
when studying faulty machines, as practically all the electromagnetic faults could be analysed
using the same basic tool.
Other advances that could be made on the basis of this work lie mainly in the area of the
experimental validation of those findings concerning the various effects that a stator winding
may have on the ability of various indicators to detect faults. The experimental
implementation of the core fault and verification of the claims obtained from the simulations
may provide the industrial environment with a new and efficient means of monitoring core
conditions.

ASSESMENT ON THE MATTER OF ACHIEVING THE INITIAL OBJECTIVES OF

THE THESIS
At the very end, it may be claimed that most of the research objectives set in the beginning of
the research were achieved. Especially, the electromagnetic flux was found able to provide
precious information about possible means to detect all of the faults under study; this was not
always the case with the stator current.

When and where possible, the accuracy of different fault signatures issuing from numerical
electromagnetic field simulations was validated by experiments. However, due to the
constraints imposed by the software (lack of a 3D simulation tool) and by the lack of test
motors equipped with various stator winding configurations, the validation procedure could
not be fully completed. This procedure represents a challenging task for a further work.
I believe that:
- the wide variety of the electrical parameters that were studied purposely intended to detect
various faults;
- the study on the ability of the electromagnetic flux eccentricity harmonics of order “p±1” for
detecting faults other than various types of eccentricity, and,
- the comparative analysis of the multitude of results provided by measurements and by a
specific time-stepping finite element approach that allows an unlimited access to the physical
quantities from the test machine,
will give the reader the impression of something new in the broad field of fault diagnosis and
condition monitoring for electrical machines.

120
References
Anderson, A., ‘A note on generator core failures’. Available on web at: http://www.antony-
anderson.com/failure1.htm.
Arkkio, A., 1987, ‘Analysis of induction motors based on the numerical solution of the magnetic field
and circuit equations’, Helsinki, Acta Polytechnica Scandinavica, Electrical Engineering Series No.
59, ISBN 951-666-250-1, 97 p. Available on web at: http://lib.hut.fi/Diss/list.html#1980.
Arkkio, A., 1988, ‘Time stepping finite element analysis of induction motors’, 8th International
Conference on Electrical Machines–ICEM ’88, Pisa, Italy, Sept. 12-14, 1998, pp. 275-281.
Arkkio, A., 1996, ‘Unbalanced magnetic pull in cage induction motors–dynamic and static
eccentricity’, 12th International Conference on Electrical Machines–ICEM’96, Vigo, Spain, Sept. 10-
12, 1996, pp. 192-197.
Arkkio A., Anttila, M., Pokki, K., Lantto, E., 2000, ’Electromagnetic force on a whirling cage rotor’,
IEE Proceedings Electrical Power Applications, Vol. 147, No. 5, pp. 353-360.
Assaf, T., Henao, H., Capolino, G.-A., 2004, ‘Simplified axial flux spectrum method to detect
incipient stator inter-turn short-circuits in induction machine’, IEEE International Symposium on
Industrial Electronics, May 4-7, 2004, Vol. 2, pp. 815-819.
Barbour, A., Thomson, W.T., 1997, ‘Finite element study of rotor slot designs with respect to the
current monitoring for detecting static airgap eccentricity in squirrel-cage induction motor’, The 1997
IEEE Industry Applications Society Conference: The 32nd IAS Annual Meeting, New Orleans,
Louisiana, Oct. 5-8, 1997, pp. 112-119.
Barker, S., 2000, ‘Avoiding premature bearing failure with inverter fed induction motors’, Power
Engineering Journal, Vol. 14, Issue 4, pp. 182-189.
Bellini, A., Filippetti, F., Franceschini, G., Tassoni, C., Kliman, G.B., 2000, ‘Quantitative evaluation
of induction motor broken bars by means of electrical signature analysis’, The 2000 IEEE Industry
Applications Society Conference: The 35th IAS Annual Meeting, Rome, Italy, Oct. 8-12, 2000, Vol. 1,
pp. 484-491.
Benbouzid, M.E.H., Vieira, M., Theys, C., 1999, ‘Induction motors faults detection and localisation
using stator current advanced signal processing techniques’, IEEE Transactions on Power Electronics,
Vol. 14, Issue 1, pp. 14-22.
Benbouzid, M.E.H., 1999, ‘Bibliography on induction motors faults detection and diagnosis’, IEEE
Transactions on Energy Conversion, Vol. 14, Issue 4, pp.1065-1074.
Benbouzid, M. E. H., 2000, ‘A review of induction motors signature analysis as a medium for faults
detection’, IEEE Transactions on Industrial Electronics, Vol. 47, Issue 5, pp. 984-993.
Benbouzid, M. E. H., Kliman, G. B., 2003, ‘What stator current processing-based technique to use for
induction motor rotor faults diagnosis ?’, IEEE Transactions on Energy Conversion, Vol. 18, No. 2,
pp.238-244.
Berman, M., Erlicki, M.S., 1991, ‘Behaviour of an induction motor with equalizing connections in the
stator winding’, Proceedings of the 17th Convention of Electrical and Electronics Engineers in Israel,
March 5-7, 1991, pp. 324-327.
Binns, K.J., Lawrenson, P.J., Trowbridge, C.W., 1992, ‘The analytical and numerical solution of
electric and magnetic fields’, John Wiley & Sons Publishers, 1992, ISBN: 0-471-92460-1, 486 p.
Bowers, S.V., Piety, K.R., 2001, ‘Proactive motor monitoring through temperature, shaft current and
magnetic flux measurements’, Emerson Process Management–CSI. Available on web at:
http://www.compsys.com.

121
Burnett, T., 2002 ‘Principles of flux diagnostics’, Emerson Process Management–CSI MotorView.
Available on web at: http://www.compsys.com/.
Burnett, R., Watson, J. F., Elder, S., 1995, ‘The application of modern signal processing techniques to
rotor fault detection and location within three-phase induction motors’, Conference on Instrumentation
and Measurement Technology, Apr. 24-26, 1995, pp. 426.
Burnett, R., Watson, J.F., 1995, ‘The current analysis program-a software tool for rotor fault detection
in three phase induction motors’, 7th International Conference on Electrical Machines and Drives
(Conference Publication No. 412), Durham, UK, Sept. 11-13, 1995, pp. 156-160.
Busse, D.F., Erdman, J.M., Kerkman, R.J., Schlegel, D.W., Skibinski, G.L., 1997, ‘An evaluation of
the electrostatic shielded induction motor: a solution for rotor shaft voltage buildup and bearing
current’, IEEE Transactions on Industry Applications, Vol. 33, Issue 6, pp. 1563-1570.
Cameron, J.R., Thomson, W.T., Dow, A.B., 1986, ‘Vibration and current monitoring for detecting air-
gap eccentricity in large induction motors’, IEE Proceedings, Vol. 133, pt. B, No. 3, May, pp. 155-
163.
Cardoso, A.J.M., Saraiva, E.S., 1993, ‘Computer-aided detection of air-gap eccentricity in operating
three-phase induction motors by Park’s Vector Approach’, IEEE Transactions on Industry
Applications, Vol. 29, No. 5, pp. 897-901.
Cardoso, A.J.M., Cruz, S.M.A., Fonseca, D.S.B., 1997, ‘Inter-turn stator winding fault diagnosis in
three-phase induction motors by Park'
s Vector approach’, IEEE International Electric Machines and
Drives Conference IEMDC ‘97, Milwaukee-WI, USA, May 18-21, 1997, pp. MB1/5.1-MB1/5.3
Chari, M.V.K., Silvester, P., 1971, ‘Finite element analysis of magnetically saturated d.c machines’,
IEEE Transactions, PAS-90, pp. 2362-2368.
Chari, M.V.K., Konrad, A., Palmo, M.A., Angelo, J.D., 1982, ‘Three dimensional vector potential
analysis for machine field problems’, IEEE Transactions on Magnetics, Vol. MAG-18, No. 2, pp. 436-
446.
Chow, M.-Y.; Sharpe, R.N.; Hung, J.C., 1993, ‘On the application and design of artificial neural
networks for motor fault detection’, IEEE Transactions on Industrial Electronics, Vol. 40, Issue 2, pp.
181-188.
Chow, T.W.S., 1996, ‘Condition monitoring of electric machines using third-order spectrum analysis’,
The 1996 IEEE Industry Applications Society Conference: The 31st IAS Annual Meeting, San Diego-
CA, USA, Oct. 6-10, 1996, Vol. 1, pp. 679-686.
Chow, M-Y., 2000, ‘Guest editorial special section on motor fault detection and diagnosis’ IEEE
Transactions on Industrial Electronics, Vol. 47, issue 5, pp. 982-983.
Clough, R., 1960, ‘Finite elements in plane stress analysis’, Proceedings of the 2nd Conference on
Electronic Computation, ASCE, 1960.
Connolly, H.M., Lodge, L., Jackson, R.J., Roberts, I., 1985, ‘Detection of inter-turns short-circuits in
generator rotor windings using air-gap search coils’, Conference on Electrical Machines Design and
Applications, 1985, pp. 11-15.
Coulomb, J. L., 1983, ‘A methodology for the determination of global electromechanical quantities
from the finite element analysis and its application to the evaluation of magnetic forces, torques and
stiffness’, IEEE Transactions on Magnetics, Vol. 19, Issue 6, pp. 2514-2519.
Courant, R., 1943, ‘Vibrational method for the solution of problems of equilibrium and vibrations’,
Bulletin of the American Mathematical Society, Vol. 49.
Crawford, D. E., 1975, ‘A mechanism of motor failures’, 12th IEEE Electrical and Electronics
Insulation Conference, Institute of Electrical and Electronics Engineers, Inc., New York, NY, pp. 126-
129.

122
Cruz, S.M.A., Cardoso, A.J.M., 2004, ‘Diagnosis of stator inter-turn short circuits in DTC induction
motor drives’, IEEE Transactions on Industry Applications, Vol. 40, Issue 5, pp. 1349-1360.
Deleroi W., 1982, ‘Squirrel cage motor with broken bar in the rotor - physical phenomena and their
experimental assessment’, 5th International Conference on Electrical Machines–ICEM ‘82, Budapest,
Hungary, 1982, pp. 767-770.
Dorrell, D. G., 1993a, ‘Calculation of unbalanced magnetic pull in cage inductlon machines‘, Ph.D.
Dissertation, University of Cambridge, U.K, 1993.
Dorrell, D. G., 1993b, ‚The effects of dynamic rotor eccentricity in cage induction motors‘, in
Proceedings UPEC’94, University College Galway, Galway, Ireland, 1994, pp. 402-405.
Dorrell, D.G., Thomson, W.T., Roach, S., 1995,’ Combined effects of static and dynamic eccentricity
on airgap flux waves and the application of current monitoring to detect dynamic eccentricity in 3-
phase induction motors’, 7th International Conference on Electrical Machines and Drives Conference
–EMD ‘95, Durham, U.K, Sept. 11-13, 1995, pp. 151-155.
Dorrell, D.G. , Thomson, W.T., Roach, S., 1997, ‘Analysis of airgap flux, current, vibration signals as
a function of the combination of static and dynamic eccentricity in 3-phase induction motors’, IEEE
Transactions on Industry Applications, Vol. 33, No. 1, pp. 24-34.
Douglas, H., Pillay, P., Ziarani, A.K., 2005, ‘Broken rotor bar detection in induction machines with
transient operating speeds’, IEEE Transactions on Energy Conversion, Vol. 20, Issue 1, pp. 135-141.
Driesen, J., 2000, ‘Coupled electromagnetic-thermal problems in electrical energy transducers’,
Doctoral Thesis, Leuven, May, 2000, 216 p.
Duque, O., Pérez, M., Moríñigo, D., 2004, ‘Practical application of the spectral analysis of line current
for the detection of mixed eccentricity in cage induction motors fed by frequency converter’, 16th
International Conference on Electrical Machines–ICEM ’04, Sept. 5-8, 2004, Krakow, Poland.
Available on CD, No. 444, 6 p.
Elder, S., Watson, J.F., Thomson, W.T., 1989, ‘Fault detection in induction motors as a result of
transient analysis’, 4th International Conference on Electrical Machines and Drives, London, UK,
Sept. 13-15, 1989, Vol. 1, pp. 182-186.
EPRI, 1982, ‘Improved Motors for Utility Applications’, Publication EL-2678, Vol. 1, 1763-1, final
report, October, 1982.
Elkasabgy, N.M., Eastham, A.R., Dawson, G.E., 1992, ‘Detection of broken bars in the cage rotor on
an induction machine’, IEEE Transactions. on Industry Applications, Vol. IA-22, No. 6, pp. 165-171.
Filippetti, F., Franceschini, G., Tassoni, C., Vas, P., 1996, ‘AI techniques in induction machines
diagnosis including the speed ripple effect’, The 1996 IEEE Industry Applications Society Conference:
The 31st IAS Annual Meeting, San Diego, Oct. 6–10, 1996, pp. 655-662.
Filippetti, F., Grellet, G., Salles, G., Franceschini, G., Tassoni, C., 1998, ‘Monitoring of induction
machines load torque disturbances: an alternative NN-based method‘, The 1998 IEEE Industry
Applications Society Conference: The 33rd IAS Annual Meeting, St. Louis-MO, USA, Oct. 12-16,
1998, pp. 103-110.
Filippetti, F., Franceschini, G., Gentile, G., Meo, S., Ometto, A., Rotondale, N., Tassoni, C., 1999,
‘1989-1998: State of art of model based diagnostic procedures for induction machines stator inter-turn
short circuits‘, IEEE lnternational Symposium on Diagnostics for Electrical Machines, Power
Electronics and Drives - SDEMPED, Gijon, Spain, Sept. l-3, 1999, pp 19-31.
Filippetti, F., Gentile, G., Meo, S., Ometto, A. and Rotondale, N., 2000a, ‘High order current spectral
components utilisation to improve induction machine diagnostics‘, 14th International Conference on
Electrical Machines––ICEM 2000, Helsinki, Finland, Aug. 28-30, 2000, pp. 941-945.

123
Filippetti, F., Franceschini, G., Tassoni, C., Vas, P., 2000b, ‘Recent developments of induction motor
drives fault diagnosis using AI techniques’, IEEE Transactions on Industrial Electronics, Vol. 47,
Issue 5, pp. 994-1004.
Filippetti, F., Franceschini, G., Tassoni, C., 2005, ‘Neural networks aided on-line diagnostics of
induction motor rotor faults’, IEEE Transactions on Industry Applications, Vol. 31, Issue 4, pp. 892-
899.
Foggia, A., Torlay, J-E., Corenwinder, C., Audoli, A., Herigault, J.,1999, ‘Circulating current analysis
in the parallel-connected windings of synchronous generators under abnormal operating conditions’,
International Conference on Electric Machines and Drives–IEMDC ' 99, Seattle-WA, USA, May 9-12,
pp. 634-636.
Früchtenicht, J., Jordan, H., Seinch, H., 1982, ‘Exzentnzitätsfelder als Urache von Laufinstabilitäten
bei Asynchronmachinen’, Parts 1 and 2, Arkiv fur Electroteknik, Vol. 65, pp. 271-292.
Gaylard, A., Meyer, A., Landy, C., 1995, ‘Acoustic evaluation of faults in electrical machines’, 7th
International Conference on Electrical Machines and Drives, Durham, UK, Sept. 11-13, pp. 147-150.
Gentile, G., Meo, S., Ometto, A., 2003, ‘Induction motor current signature analysis to diagnostics, of
stator short circuits’, 4th IEEE International Symposium on Diagnostics for Electric Machines, Power
Electronics and Drives SDEMPED ’03, Atlanta-GA, USA, Aug. 24-26, 2003, pp. 47-51.
Green, V., Sasic, M., Wright, B., 2005, ‘Summary of industry practices, guides and standards related
to on-line partial discharge monitoring of stator insulation condition’, Industry Applications Society
52nd Annual Petroleum and Chemical Industry Conference, Sept. 12-14, pp. 359-365.
Hargis, C., Gaydon, B. G., Kamish, K., 1982, ‘The detection of rotor defects in induction motors’,
Proceedings IEE - EMDA Conference, London, 1982, pp. 216-220.
Henao, H. Demian, C. Capolino, G.-A., 2003, ‘A frequency-domain detection of stator winding
faults in induction machines using an external flux sensor’, IEEE Transactions on Industry
Applications, Vol. 39 , Issue 5, pp. 1272-1279.
Henao, H. Martis, C. Capolino, G.-A., 2004, ‘An equivalent internal circuit of the induction
machine for advanced spectral analysis’, IEEE Transactions on Industry Applications, Vol. 40, Issue
3, pp. 726-734.
Hirvonen, R., 1994, ‘On-line condition monitoring of defects in squirrel cage motor’, 11th
International Conference on Electrical Machines–ICEM ’94, Paris, France, Vol.2, pp. 267-272.
Hsu, J.S., 1995, ‘Monitoring of defects in induction motors through air-gap torque observation’, IEEE
Transactions on Industry Applications, Vol. 31, Issue 5, pp. 1016-1021.
Huang, X., Habetler, T.G.; Harley, R.G., 2004, ‘Detection of rotor eccentricity faults in closed-loop
drive-connected induction motors using an artificial neural network’, IEEE 35th Annual Power
Electronics Specialists Conference–PESC, Aachen, Germany, June 20-25, Vol. 2, pp. 913-918.
IEEE, 1985, ‘Motor reliability working group. Report of large motor reliability survey of industrial
and commercial installations’, IEEE Transactions on Industry Applications, 1985, 1A. (4), Part I & II.
IEEE, 1994, ‘IEEE recommended practice for instrumentation: specifications for magnetic flux
density and electric field strength meters-10 Hz to 3 kHz’, IEEE Std. 1308-1994.
IEEE, 2004, ‘IEEE standard test procedure for polyphase induction motors and generators’, Electric
Machinery Committee of the IEEE Power Engineering Society, IEEE Std. 112-2004 (Revision of IEEE
Std 112-1996), 2004.
Jianming, J., 1993, ‘The finite element method in electromagnetics’, Wiley, New York, 1993, ISBN: 0-
471-43818-9, 784 p.

124
Joksimovic, G.M., Penman, J., 2000, ‘The detection of inter-turn short circuits in the stator windings
of operating motors’, IEEE Transactions on Industrial Electronics, Vol. 47 , Issue 5 , pp. 1078-1084.
Joksimovic, G.M., Durovic, M.D., Penman, J., Arthur, N., 2000, ‘Dynamic simulation of dynamic
eccentricity in induction machines-finding function approach’, IEEE Transactions on Energy
Conversion, Vol. 15, Issue 2, pp. 143-148.
Jover Rodriguez, P., Arkkio, A., 2003, ‘Induction motor stator fault detection using fuzzy logic’,
International Conference on Circuits, Signal and Systems–IASTED ‘ 03, Cancún, Mexico, May 19-21,
2003, pp. 96-100.
Karmaker, H.C., 2003, ‘Broken damper bar detection studies using flux probe measurements and time-
stepping finite element analysis for salient-pole synchronous machines’, 4th IEEE International
Symposium on Diagnostics for Electric Machines, Power Electronics and Drives–SDEMPED ‘03,
Atlanta-GA, US, Aug. 24-26, 2003, pp. 193-197.
Kerszenbaum, I., 1992, ‘Shaft currents in electric machines fed by solid-state drives’, IEEE
Conference on Industrial and Commercial Power Systems, Pittsburgh-PA, USA, May 4-7, 1992, pp.
71-79.
Kim, K., Parlos, A.G., Mohan Bharadwaj, R., 2003, ‘Sensorless fault diagnosis of induction motors’,
IEEE Transactions on Industrial Electronics, Vol. 50, Issue 5, pp. 1038-1051.
Kliman, G.B., Koegl, R. A., Stein, J., Endicott, R. D., Madden, M. W., 1988, ‘Noninvasive detection
of broken rotor bars in operating induction motors’, IEEE Transactions on Energy Conversion, Vol. 3
Issue: 4, pp. 873-879.
Kliman, G.B., Stein, J., 1990, ’Induction motor fault detection via passive current monitoring’, 9th
International Conference in Electrical Machines–ICEM’90, Boston-MA, Aug. 13-15, 1990, pp. 13-17.
Kliman, G.B., Premerlani, W. J., Koegl, R.A., Hoeweler, D., 1996, ‘A new approach to on-line fault
detection in ac motors’, The 1996 IEEE Industry Applications Society Conference: The 30th IAS
Annual Meeting, San Diego-CA, USA, Oct. 6-10, 1996, pp. 687-693.
Kliman, G.B., Premerlani, W.J., Yazici, B., Koegl, R.A., Mazereeuw, J., 1997, ‘Sensorless, online
motor diagnostics’, IEEE Computer Applications in Power, Vol. 10, No.2, pp. 39-43.
Kliman, G.B., Sang Bin Lee, Shah, M.R., Lusted, R.M., Nair, N.K., 2004, ‘A new method for
synchronous generator core quality evaluation’, IEEE Transactions on Energy Conversion, Vol. 19,
Issue 3, pp. 576-582.
Kohler, J.L., Sottile, J., Trutt, F.C., 2002, ‘Condition monitoring of stator windings in induction
motors. I. Experimental investigation of the effective negative-sequence impedance detector’, IEEE
Transactions on Industry Applications, Vol. 38, Issue 5, pp. 1447-1453.
Kokko, V., 2003, ‘Condition monitoring of squirrel-cage motors by axial magnetic flux
measurements’, Doctoral thesis, Oulun Yliopisto, Oulu 2003, ISBN 951-42-6938-1, p. 157. Available
on web at: http://herkules.oulu.fi/isbn9514269381.
Kral, C., Habetler, T.G., Harley, R.G., 2004, ‘Detection of mechanical imbalances of induction
machines without spectral analysis of time-domain signals’, IEEE Transactions on Industry
Applications, Vol. 40, Issue 4, pp. 1101-1106.
Lee, S.B., Kliman, G.B., Shah, M.R., Nair, N.K., Lusted, R.M., 2005, ‘An iron core probe based inter-
laminar core fault detection technique for generator stator cores’, IEEE Transactions on Energy
Conversion, Vol. 20 , Issue 2, pp. 344-351.
Legowski, S.F., Sadrul Ula, A.H.M., Trzynadlowski, A.M., 1996, ‘Instantaneous power as a medium
for the signature analysis of induction motors’, IEEE Transactions on Industry Applications, Vol. 32,
Issue 4, pp. 904-909.

125
Lenz, J.E., 1990, ‘A review of magnetic sensors’, Proceedings of the IEEE, Vol. 78, Issue 6, pp. 973-
989.
Leonard R A , Thomson W. T, 1986, "Vibration and stray flux monitoring for unbalanced supply and
inter-turn winding fault diagnosis in induction motors", British Journal of NDT, July, pp. 211-215.
Li, B., Chow, M-Y., Tipsuwan, Y., Hung, J.C., 2000, ‘Neural-network-based motor rolling bearing
fault diagnosis’, IEEE Transactions on Industrial Electronics, Vol. 47, No. 5, pp. 1060-1069.
Li, J. C., Ma, J., Hwang, B., 1995, ‘Bearing localized defect detection by bicoherence Analysis of
Vibrations’, Journal of Engineering for Industry, American Society of Mechanical Engineers, New
York, USA, Vol. 117, No. 4, pp. 625-629.
Lindh, T., Ahola, J., Kamarainen, J. K., Kyrki, V., Partanen, J., 2003, ’Bearing damage detection
based on statistical discrimination of stator current’, 4th IEEE International Symposium on Diagnostics
for Electric Machines, Power Electronics and Drives–SDEMPED ‘03, Atlanta-GA, USA, Aug. 24-26,
2003, pp.177-181.
Liu, Z., Yin, X., Zhang, Z., Chen, D., Chen, W., 2004, ‘Online rotor mixed fault diagnosis way based
on spectrum analysis of instantaneous power in squirrel cage induction motors’, IEEE Transactions
on Energy Conversion, Vol. 19, Issue 3, pp. 485-490.
Malinowski, J., McCormick, J., 2002, ‘AC induction motor specifications-an update on currently
available procedures and options’, Conference Record of the 2002 Annual Pulp and Paper Industry
Technical Conference, Toronto-ON, Canada, June 17-21, 2002, pp. 181-189.
Maxwell, J. C., 1865, ‘A dynamical theory of the electromagnetic field’, Philosophical Transactions
of the Royal Society of London 155, 459-512 (1865). (This article accompanied a December 8, 1864
presentation by Maxwell to the Royal Society).
McCormick, A. C., Nandi, A. K., 1999, ‘Bispectral and trispectral features for machine condition
diagnosis’, IEEE Proceedings of Visual Image Signal Processing, Vol. 146, No. 5, pp. 229-234.
Melero, M.G., Cabanas, M.F., Rojas, C., Orcajo, G.A., Cano, J.M., Solares, J., 2003, ‘Study of an
induction motor working under stator winding inter-turn short circuit condition’, 4th IEEE
International Symposium on Diagnostics for Electric Machines, Power Electronics and Drives–
SDEMPED ‘03, Atlanta-GA, USA, Aug. 24-26, pp. 52-57.
Meireles, M.R.G., Almeida, P.E.M., Simoes, M.G., 2003, ‘A comprehensive review for industrial
applicability of artificial neural networks’, IEEE Transactions on Industrial Electronics, Vol. 50, Issue
3, pp. 585-601.
Muhlhaus, J., Ward, D. M., Lodge, I., 1985, ‘The detection of shorted turns in generator rotor
windings by measurement of circulating stator currents‘, Proceedings of the 2nd International
Conference on Electrical Machines Design and Applications–EMDA ‘85, IEE Publications, Vol. 254,
London, pp. 100-103.
Mäki-Ontto, P., 2006, ‘Modeling and reduction of shaft voltages in AC motors fed by frequency
converters’, Doctoral thesis, ISBN 951-22-8141-4, p. 140. Available on web at:
http://lib.tkk.fi/Diss/2006/isbn9512281414.
Nandi, S., Toliyat, H.A., Parlos, A.G., 1997, ’Performance analysis of a single phase induction motor
under eccentric conditions’, The 1997 IEEE Industry Applications Society Conference: The 32nd IAS
Annual Meeting, New Orleans-LA, USA, Vol. 1, Oct. 5-9, 1997, pp. 174-181.
Nandi, S., Toliyat, H.A., 1998, ’Detection of rotor slot and other eccentricity related harmonics in a
three phase induction motor with different rotor cages’, Conference Proceedings of International
Conference on Power Electronic Drives and Energy Systems for Industrial Growth–PEDES ’98,
Perth, Australia, Nov. 30-Dec. 3, 1998, Vol. 1, pp. 135-140.

126
Nandi, S., Toliyat, H.A., 1999, ‘Condition monitoring and fault diagnosis of electrical machines-a
review’, The 1999 IEEE Industry Applications Society Conference: The 34th IAS Annual Meeting,
Phoenix-AZ, USA, 3-7 Oct., 1999, Vol. 1, pp. 197-204.
Nandi, S., Toliyat, H.A., 2000, ‘Novel frequency domain based technique to detect incipient stator
inter-turn faults in induction machines’, The 2000 IEEE Industry Applications Society Conference:
The 35th IAS Annual Meeting, Rome, Italy, Vol. 1, Oct. 8-12, 2000, pp. 367- 374.
Nandi, S., Toliyat, H.A., 2002, ’Novel frequency-domain-based technique to detect stator inter-turn
faults in induction machines using stator-induced voltages after switch-off’, IEEE Transactions on
Industry Applications, Vol. 38, Issue 1, pp. 101-109.
Nandi, S. Bharadwaj, R.M. Toliyat, H.A., 2002, ‘Performance analysis of a three-phase induction
motor under mixed eccentricity condition’, IEEE Transactions on Energy Conversion, Vol. 17, Issue
3, pp. 392-399.
Nandi, S., Toliyat, H.A., Xiaodong Li, 2005, ‘Condition monitoring and fault diagnosis of electrical
motors-a review’, IEEE Transactions on Energy Conversion, Vol. 20, Issue 4, pp. 719-729.
Nattrass, D.A., 1993, ‘Partial discharge. XVII. The early history of partial discharge research’, IEEE
Electrical Insulation Magazine, Vol. 9, Issue 4, pp. 27-31.
Nau, S.L., Mello, H.G.G., 2000, ‘Acoustic noise in induction motors: causes and solutions’, Industry
Applications Society 47th Annual Conference on Petroleum and Chemical Industry, San Antonio-TX,
USA, Sept. 11-13, pp. 253-263.
Negrea, M., Arkkio, A., Jokinen, T., Hakuli, M., 2001, ‘Thermal analysis of a permanent magnet
synchronous motor’, 3rd IEEE International Symposium on Diagnostics for Electrical Machines,
Power Electronics and Drives–SDEMPED ‘03, Grado, Italy, Sept. 1-3, pp. 517-522.
Negrea, M., Pöyhönen, S., Jover, P., Arkkio, A., Hyötyniemi, H., 2002, ‘Modelling, simulation and
signal processing with application to electric machine fault diagnostics and condition monitoring’,
Proceedings of the 7th International Conference on Modeling and Simulation of Electric Machines,
Converters and Systems–ELECTRIMACS, Montreal, Canada, Aug. 18-21, 2002, on CD, 6 p.
Negrea, M., Jover, P., Arkkio, A., 2004, ‘A comparative investigation on the reliability and accuracy
of different diagnostic media when attempting to identify specific faults in an induction motor’,
Symposium on Power Electronics, Electrical Drives, Automation and Motion–SPEEDAM ‘04, Capri,
Italy, May 16-18, 2004, pp. 809-814.
Negrea, M., Jover, P., Arkkio, A., 2005 ‘Electromagnetic flux-based condition monitoring for
electrical machines’, 5th IEEE International Symposium on Diagnostics for Electric Machines, Power
Electronics and Drives–SDEMPED ‘05, Wien, Austria, Sept. 7-9, 2005, on CD, 6p.
Negrea, M., Jover, P., Arkkio, 2006, ‘A comparative investigation of electrical parameters for fault
detection and condition monitoring in induction motors–stator winding current and electromagnetic
flux’, 17th International Conference on Electrical Machines–ICEM ’06, Sept. 2–5, 2006, Chania, Crete
Island, Greece, on CD, 6 p.
Obaid, R.R., Habetler, T.G., Gritter, D.J., 2000, ‘A simplified technique for detecting mechanical
faults using stator current in small induction motors’, The 2000 IEEE Industry Applications Society
Conference: The 35th IAS Annual Meeting, Rome, Italy, Vol. 1, Oct. 8-12, 2000, pp. 479-483.
Obaid, R. R., Habetler, T. G., Stack, J. R., 2003, ‘Stator current analysis for bearing damage detection
in induction motors’, 4th IEEE International Symposium on Diagnostics for Electric Machines, Power
Electronics and Drives–SDEMPED ‘03, Atlanta-GA, USA, Aug. 24-26, 2003, pp.182-187.
Obaid, R. R., Habetler, T. G., 2003, ‘Effect of load on detecting mechanical faults in small induction
motors’, 4th IEEE International Symposium on Diagnostics for Electric Machines, Power Electronics
and Drives–SDEMPED ‘03, 24-26 Aug. 2003, pp. 307-311.

127
Ocak, H., Loparo, K. A., 2001, ‘A new bearing fault detection and diagnosis scheme based on hidden
Markov modeling of vibration signals’, Proceedings of IEEE International Conference on Acoustics,
Speech, and Signal Processing ICASSP ' 01, Salt Lake City-UT, USA, Vol. 5, May 7-11, 2001, pp.
3141-3144.
Ong, R., Dymond, J.H., Findlay, R.D., 1997, ‘Comparison of techniques for measurement of shaft
currents in rotating machines’, IEEE Transactions on Energy Conversion, Vol. 12 Issue 4, pp. 363-
367.
Ong, R., Dymond, J.H., Findlay, R.D., 2000, ‘Bearing damage analysis in a large oil-ring-lubricated
induction machine’, IEEE Transactions on Industrial Electronics, Vol. 47, Issue 5, pp. 1085-1091.
Parlos, A.G., Kim, K., 2001, ‘Development and experimental demonstration of a model-based fault
detection system for electromechanical systems’, Proceedings of the 2001 American Control
Conference, Arlington-VA, USA, June 25-27, 2001, Vol. 5, pp. 3866-3871.
Paterson, N., 1998, ‘The analysis and detection of faults in three-phase induction machines using finite
element techniques’, Doctoral Thesis, Robert Gordon University, Wetherby British Library, Aberdeen,
UK, 180 p .
Payne, B.S., Husband, S.M., Ball, A.D., 2002, ‘Development of condition monitoring techniques for a
transverse flux motor’, International Conference on Power Electronics, Machines and Drives–PEMD
‘02 (IEE Conference Publication No. 487), Bath, UK, June 4-7, 2002, pp. 139-144.
Penman, J., Dey, M.N., Tait, A.J., Bryan, W.E., 1986, ‘Condition monitoring of electrical drives’, IEE
Proceedings, Vol. 133, pt. B, No. 3, pp. 142-148.
Penman, J., Sedding, H. G., Lloyd, B. A., Fink, W. T., 1994. ‘Detection and location of interturn short
circuits in the stator windings of operating motors’, IEEE Transactions on Energy Conversion, Vol. 9,
No. 4, pp 652-658.
Penman, J., Jiang, H., 1996, ‘The detection of stator and rotor winding short circuits in synchronous
generators by analysing excitation current harmonics’, International Conference on Opportunities and
Advances in International Electric Power Generation, Durham, UK, Mar. 18-20, 1996, pp. 137-142.
Pöyhönen S., Negrea M., Arkkio A., Hyötyniemi H. and Koivo H., 2002a, ‘Support vector
classification for fault diagnostics of an electrical machine’, 6th International Conference on Signal
Processing–ICSP ' 02, Beijing, China, Aug. 26-30, 2002, Vol. 2, pp. 1719-1722.
Pöyhönen S., Negrea M., Arkkio A., Hyötyniemi H. and Koivo H., 2002b, ‘Fault diagnostics of an
electrical machine with multiple support vector classifiers’, 17th IEEE International Symposium on
Intelligent Control–ISIC '
02, Vancouver-BC, Canada, Oct. 27-30, 2002, Vol. 1, pp. 373-378.
Pöyhönen, S., Negrea, M., Jover, P., Arkkio, A., Hyötyniemi, H., 2003,’Numerical magnetic field
analysis and signal processing for fault diagnostics of electrical machines’, The International Journal
for Computation and Mathematics in Electrical and Electronic Engineering–COMPEL, Vol. 22, No.
4, pp. 969-981.
Pöyhönen, S., 2004, ‘Support Vector Machine Based Classification in Condition Monitoring of
Induction Motors’ Doctoral thesis, Helsinki University of Technology, ISBN: 951-22-7155-9.
Available on web at: http://lib.tkk.fi/Diss/2004/isbn9512271559/
Rankin, D.R. , Wilson, I.,1995, ‘The use of shaft voltage to detect air gap eccentricity and shorted
turns in salient pole alternators’, 7th International Conference on Electrical Machines and Drives,
1995, Durham, UK, Sept. 11–13, 1995, pp. 194-197.
Ramirez-Nino, J., Pascacio, A., 2001, ‘Detecting inter-turn short circuits in rotor windings’, IEEE
Computer Applications in Power, Vol. 14, Issue 4, pp. 39-42.
Rienstra, A., Hall, J., 2004, ‘Applying acoustic vibration monitoring to predictive maintenance’.
Available on web at: http://www.maintenanceworld.com/Articles/rienstraA/ultrasonic.pdf.

128
Rotating Electrical Machines Part 15, 1995, ‘Impulse voltage withstand levels of rotating a.c.
machines with form-wound stator coils’, IEC 34-15, 1995.
Salles, G., Filippetti, F., Tassoni, C., Crellet, G., Franceschini, G., 2000, ‘Monitoring of induction
motor load by neural network techniques’, IEEE Transactions on Power Electronics, Vol. 15, Issue
4, pp. 762-768.
Salon, S.J., 1995, ‘Finite elements analysis of electrical machines’, Kluwer Academic Publishers,
ISBN: 0792395948, 264 p.
Schoen, R.R., Lin, B.K., Habetler, T.G., Schlag, J.H., Farag, S., 1994, ‘An unsupervised, on-line
system for induction motor fault detection using stator current monitoring’, The 1994 IEEE Industry
Applications Society Conference: The 29th IAS Annual Meeting, Denver-CO, USA, Oct. 3-6, 1994,
Vol. 1, pp.103-109.
Schoen, R.R., Habetler, T.G., Kamran, F., Bartfield, R.G., 1995, ‘Motor bearing damage detection
using stator current monitoring’, IEEE Transactions on Industry Applications, Vol. 31, Issue 6, pp.
1274-1279.
Schoen, R. R., Habetler, T.G., 1995, ‘Effects of time-varying loads on rotor fault detection in
induction machines’, IEEE Transactions on Industry Applications, Vol. 31, Issue 6, pp. 900-906.
Schoen, R. R, Habetler, T. G., 1997, ‘Evaluation and implementation of a system to eliminate arbitrary
load effect in current-based monitoring of induction machines‘, IEEE Transactions on Industry
Applications,Vol. 33, No 6, pp. 1571-1577.
Silvester, P. P., Ferrari, R.L., 1996, ‘Finite Elements for Electrical Engineers’, Third edition,
Cambridge University Press, 1996, ISBN-13: 9780521449533, ISBN-10: 0521449537, 344 p.
Siau, J., Graff, A, Soong, W., Ertugrul, N., 2003, ‘Broken bar detection in induction motors using
current and flux spectral analysis‘, Australasian Universities Power Engineering Conference–
AUPEC2003,, Christchurch, New Zealand, Sept. 28-Oct. 1, 2003, ISBN 0-473-09867-9. Available on
web at: www.itee.uq.edu.au/~aupec/aupec03/papers/086%20Siau%20full%20paper.pdf.
Siddique, A., Yadava, G.S., Singh, B., 2003, ‘Applications of artificial intelligence techniques for
induction machine stator fault diagnostics: review’, 4th IEEE International Symposium on Diagnostics
for Electric Machines, Power Electronics and Drives–SDEMPED ‘03, Atlanta-GA, USA, Aug. 24-26,
2003, pp. 29-34.
Smith, A.C., Dorrell, D.G., 1996, ‘Calculation and measurement of unbalanced magnetic pull in cage
induction motors with eccentric rotors. I. Analytical model’, IEE Proceedings on Electric Power
Applications, Vol. 43, Issue 3, May 1996, pp. 193-201.
SpectraQuest, ‘Machinery Fault Simulator’, SpectraQuest Inc., Available on web at:
www.spectraquest.com.
Stack, J. R., Habetler, T. G., Harley, R. G., 2003, ‘Fault classification and fault signature production
for rolling element bearings in electric machines’, 4th IEEE International Symposium on Diagnostics
for Electric Machines, Power Electronics and Drives–SDEMPED ‘03, Atlanta-GA, USA, Aug. 24-26,
2003, pp. 172-176.
Stack, J.R. Habetler, T.G. Harley, R.G., 2004, ‘Bearing fault detection via autoregressive stator
current modeling’, IEEE Transactions on Industry Applications, Vol. 40, Issue 3, pp. 740-747.
Stavrou, A., Sedding, H.G., Penman, J., 2001, ‘Current monitoring for detecting inter-turn short
circuits in induction motors’, IEEE Transactions on Energy Conversion, Vol. 16, Issue 1, pp. 32-37.
Stone, G.C., Lloyd, B.A., Campbell, S.R., Sedding, H.G., 1997, ‘Development of automatic,
continuous partial discharge monitoring systems to detect motor and generator partial discharges’,
IEEE International Electric Machines and Drives Conference–IEMDC ‘97, Milwaukee-WI, USA,
May 18-21, 1997, pp. MA2/3.1-MA2/3.3.

129
Tallam, R.M., Habetler, T.G., Harley, R.G., Gritter, D.J., Burton, B.H., 2000, ‘Neural network based
on-line stator winding turn fault detection for induction motors’, The 2000 IEEE Industry Applications
Society Conference: The 35th IAS Annual Meeting, Rome, Italy, Vol. 1, Oct. 8-12, pp. 375-380.
Tallam, R.M., Lee, S.B., Stone, G., Kliman, G.B., Yoo, J., Habetler, T.G., Harley, R.G., 2003, ‘A
survey of methods for detection of stator related faults in induction machines’, 4th IEEE International
Symposium on Diagnostics for Electric Machines, Power Electronics and Drives–SDEMPED ‘03,
Atlanta-GA, USA, Aug. 24-26, 2003, pp. 35-46.
Tavner, P., Penman, J., 1987, ‘Condition Monitoring of Electrical Machines’, Research studies press
LTD, 1987. ISSN/ISBN: 0-86380-061-0, 302 p.
Tavner, P.J., Anderson, A.F., 2005, ‘Core faults in large generators’, IEE Proceedings on Electric
Power Applications, Vol. 152, Issue 6, pp. 1427-1439.
Tenhunen, A., Holopainen, T.P., Arkkio, A., 2003, ‘Impulse method to calculate the frequency
response of the electromagnetic forces on whirling cage rotors’, IEE Proceedings on Electric Power
Applications, Vol. 150, Issue 6, pp. 752-756.
Thailly, D. Romary, R. Brudny, J.-F., 2005, ‘Quantitative analysis of the external radial magnetic
field for detection of stator inter-turn short-circuit in induction machines’, European Conference on
Power Electronics and Applications EPE ’05, Dresden, Germany, Sept. 11-14, 2005, Available on
CD.
Thomas, V.V., Vasudevan, K., Kumar, V.J., 2003, ‘Online cage rotor fault detection using air-gap
torque spectra’, IEEE Transactions on Energy Conversion, Vol. 18, Issue 2, pp. 265-270.
Thomson, W. T., Deans, N. D., Leonard, R. A., Milne, A. J., 1983, ‘Monitoring strategy for
discriminating between different types of rotor cage faults’, Proceedings of the 18th Universities
Power Engineering Conference–UPEC, University of Surrey, Guildford, U.K., April 1983.
Thomson, W. T., 1992, ‘On-line current monitoring–The influence of mechanical loads or a unique
rotor design on the diagnosis of broken bars in induction motors‘, 10th International Conference on
Electrical Machines–ICEM ’92, Manchester, U.K., Sept. 15-17, 1992, pp. 1236-1240.
Thomson, W.T., 1994, ‘On-line current monitoring to detect electrical and mechanical faults in three-
phase induction motor drives’, International Conference on Life Management of Power Plants,
Edinburgh-UK, Dec. 12-14, pp. 66-73.
Thomson, W. T., A Barbour, F Filippetti and C. Tassoni, 1998, ‘An appraisal of the MMF-permeance
method and finite element models to study static air-gap cccentricity in induction machines‘, 13th
International Conference on Electrical Machines–ICEM ’98,Istanbul, Turkey, Sept. 2-4, 1998, pp.
218-227.
Thomson, W. T., 1999, ‘A review of on-line condition monitoring techniques for three-phase squirrel-
cage induction motors–past, present and future’, 2nd IEEE International Symposium on Diagnostics for
Electrical Machines, Power Electronics and Drives–SDEMPED ‘99, Gijon, Spain, Sept. 1-3, 1999,
pp. 3-17.
Thomson, W. T., 2001, ‘On-line MCSA to diagnose shorted turns in low voltage stator windings of 3-
phase induction motors prior to failure’, IEEE International Electric Machines and Drives Conference
– IEMDC ‘01, Cambridge-Massachusetts, USA, Jun. 17-20, pp. 891-898.
Thomson, W. T., Gilmore, R. J., 2003, ‘Motor current signature analysis to detect faults in induction
motor drives–fundamentals, data interpretation, and industrial case histories’ Proceedings of the 32nd
Turbomachinery Symposium, Houston, TX, USA, Sept. 8-11, pp. 145-156.
Thorsen, O. V., Dalva, M., 1999, ‘Failure identification and analysis for high voltage induction molars
in the petrochemical industry’, IEEE Transactions on Industry Applications, Vol. 35, No. 4, pp. 810-
818.

130
Toliyat, H.A., Lipo, T.A., 1995, ’Transient analysis of cage induction machines under stator, rotor bar
and end ring faults’, IEEE Transactions on Energy Conversion, Vol. 10, No. 2, pp. 241-247.
Trutt, F.C., Sottile, J., Kohler, J.L., 2002, ‘Condition monitoring of induction motor stator windings
using electrically excited vibrations’, The 2002 IEEE Industry Applications Society Conference: The
37th IAS Annual Meeting, Vol. 4, pp. 2301-2305.
Trzynadlowski, A.M., Ritchie, E., 2000, ‘Comparative investigation of diagnostic media for induction
motors: a case of rotor cage faults’, IEEE Transactions on Industrial Electronics, Vol. 47, Issue 5, pp.
1092-1099.
Tsoumas, I.; Mitronikas, E.; Safacas, A.; 2005, ’Induction motor mixed fault diagnosis based on
wavelet analysis of the current space vector’, 8th International Conference on Electrical Machines and
Systems-ICEMS 2005, Vol. 3, 27-29 Sept., 2005, pp. 2186-2191.
Vas, P., 1993, ‘Parameter estimation, condition monitoring, and diagnosis of electrical machines’,
Oxford, Clarendon, 1993, ISBN: 0198593759, 378 p.
Vas, P., 1999, ‘Artificial-Intelligence-based electrical machines and drives: Application of fuzzy,
fuzzy-neural, and genetic-algorithm-based techniques’, New York, Oxford Press, 1999, ISBN:
019859397X, 656 p.
Villada, F., Cadavid, D., Munoz, N., Valencia, D., Parra, D., 2003, ‘Fault diagnosis in induction
motors fed by PWM inverters’, 4th IEEE International Symposium on Diagnostics for Electric
Machines, Power Electronics and Drives–SDEMPED’ 03, Atlanta-GA, USA, Aug. 24-26, 2003, pp.
229-234.
Wang, C., Gao, R. X., 2000, ‘Sensor placement strategy for in-situ bearing defect detection’,
Proceedings of the 17th IEEE Instrumentation and Measurement Technology Conference–
IMTC, Baltimore-MD, USA, May 1-4, Vol. 3, pp. 1463-1467.
Wiedenbrug, E., Frey, G.; Wilson, J., 2003, ‘Impulse testing as a predictive maintenance tool’, 4th
IEEE International Symposium on Diagnostics for Electric Machines, Power Electronics and Drives–
SDEMPED ‘03, Atlanta-GA, USA, Aug. 24-26, pp. 13-19.
Wieser, R., Kral, C., Pirker, F., Schagginger, M., 1999, ‘On-line rotor cage monitoring of inverter-fed
induction machines by means of an improved method’, IEEE Transactions on Power Electronics, Vol.
14, Issue 5, pp. 858-865.
Williamson, S., Mirzoian, P., 1985, ‘Analysis of cage induction motor with stator winding faults’,
IEEE Transactions on Power Apparatus and Systems, July 1985, Vol. 104, No. 7, pp. 1838-1842.
Williamson, S., 2004, ‘The future of electrical machines’, Proceedings of International Conference on
Power Electronics Systems and Applications, Hong Kong, Nov. 9-11, 2004, pp. 11-12.
Yazidi, A., Thailly, D., Henao, H., Romary, R., Capolino, G.-A., Brudny, J.-F., 2004, ‘Detection of
stator short-circuit in induction machines using an external leakage flux sensor’, IEEE International
Conference on Industrial Technology–ICIT ' 04, Dec. 8-10, 2004, Hammamet, Tunisia, Vol. 1, pp 166-
169.
Yen, G. G., Lin, K. C., 2000, ‘Wavelet Packet Feature Extraction for Vibration Monitoring’, IEEE
Transactions on Industrial Electronics, Vol. 47, Issue 3, pp. 650-667.

131
132
Appendix I – Information about the measurement devices
Transient recorder W+W700
Mainframe: Includes measuring computer, 9-inch monochrome display screen, control unit,
3½ inch micro floppy drive for software load and as a mass storage medium for
measurement data. Space for 4 single or dual channels. Support for external colour monitor
for menu and signal presentation.
Retrofit to AT processor with external keyboard and 42- or 100-MB internal hard disk
available.
Expansion Frame: for further 8 single or dual channels. Up to three additional frames can
be connected for a total of 32 channels.
Single channels: They include pre-amplifiers with different inputs, 8 bits trigger, for level,
window, slew rate, and reference band.
Timeout with linkage possibilities.
Storage up to 256 Kwords, 1Mwords as option.
Dual time base.
Anti-aliasing filter.
Channel types:
50-MHz 8-bit
50-MHz 10-bit
100-MHz 12-bit
Dual channels
They include two pre-amplifiers with differential input,
8 bit trigger, level, window, slew rate.
Timeout with linkage possibilities.
Storage per channel 256 kWords, 1 Mword as option.
Dual time base.
Anti aliasing filter.
Channel types:
20-MHz 8-bit dual
1-MHz 12-bit dual
200-kHz 12-bit dual
-- In the main frame and expansion frame any combination of channels is possible.
Markers
As an option some channels can be equipped with marker inputs. Markers are signals which
can only assume the logical states 0 or 1. The number of markers available depends on the
bit count of the respective signal inputs:
8-bit 50-MHz 8 Markers
8-bit 20-MHz dual no Markers
12-bit 10-MHz 4 Markers
12-bit 1-MHz dual 2×4 Markers
12-bit 200-kHz dual 2×4 Markers
The pre-amplifiers are adapted to the maximum sampling rate and the resolution of the
channel. The differential outputs are switchable to asymmetric inputs. AC or DC signal
coupling can be selected.

133
The pre-amplifiers are equipped with an anti-aliasing filter of the four-pole vessel type
with five selectable frequencies.

Time base
Each channel is equipped with a dual time base with trigger delay selectable from -
100% to 400%.
The time elapsed since arming the channel assures the correct time relationship between
channels with different trigger timing.
Trigger
The trigger source of a channel can be defined as the main trigger for the whole
instrument or as a local trigger for the respective channel. In addition, the trigger
method listed below can be combined with the main trigger for the whole instrument:
OR/AND/OFF. Triggering from an external source and manual triggering (TRIGGER
KEY) are also possible.

For each channel an individual trigger with the following possibilities is available:
---level trigger for rising and falling edge.
---window trigger, in, out.
---slew rate.
---time-out trigger for the forgoing trigger method.
Trigger is activated if a trigger event does not occur within a specified time frame.
---reference band trigger for the comparison of two signals.

Channel memories
The channels are equipped with a memory of 25 kWords each. A 1-MWord option and,
for the 1-MHz and 200-kHz channels, a 64-kW memory option is available.
Signal presentation
On the screen up to 8 channels can be displayed simultaneously in correct relation to the
time axis. You can select, move, and zoom the signals with the help of the cursor and
control wheel.
The following forms of representation are available:
---Yt: up to 8 signals, amplitude in relation to time.
---DUAL: up to 4 signals with different presentation in two windows.
---Two separately: up to 8 signals in two independent windows with independent
manipulation.
---X-Y.
---X-Y-t: in preparation.
Operating
The combination of a pop-up menu and control wheel makes possible the simple and
fast handling of the W+W700.
User programming
The inherent versatility of the W+W700 can be further enhanced by user-originated
programs written in any desired program language.
Interactive control of the W+W700 by the user program enables the measurements
sequences to be automated and the processing of measuring data to be specific to the
application involved. For Turbo Pascal, version 5.0 and up, complete program units are
available, which makes programming easier.

134
Norma Wide Band Power analyser
Device configuration: Three configurations are preset by the manufacturer.
One configuration is stored on power-down.
5 configurations can be stored by the customer.

Zero suppression: RMS and Rectified Mean value <1% of range are displayed as 0.
Power values (P, S, Q) <0,01% are displayed as zero.
Integration values below these limits are ignored.

Angular error: between voltage and current:

Angular error influence by current range in Mode AC+DC without LP filter.

Valid for
Frequency range 0…100Hz 100…1kHz Add. Error
Range
50 mV 00.015º 0.020º 0.005 º/kHz
158 mV 0.005º 0.010º 0.005 º/kHz
500 mV 0.005º 0.005º 0.005 º/kHz
1580 mV 0.005º 0.005º 0.005 º/kHz

Channel separation: >5 dB at 100 kHz

Trigger input: TTL signal, overload max. 250 VRMS

Trigger distance> average time B+ 1 s,
Trigger delay: EXT.TRG<100 µs
IEEEBUS *TRG<600 µs

Synchronization: Internal: 10 Hz….10 kHz of CH1…..CH6 resp. CH12

External: 10 Hz…10kHz; 3V…..50 VRMS ;overload max. 250 VRMS ;
Common-mode voltage ≤ 5 Vp
Real Time Clock: 1 min. per month

Integrator: time error: ± 0.0025%+10 µs

The integration time always corresponds to integer multiples of averaging
Period A

Voltage channel
Nominal voltage: max. 1000 VRMS
Measuring voltage: max. 1000 VRMS or 2500 Vp , DC …..1MHz sinusoidal

Bandwidth: DC…2MHz
Sampling rate: 35…70 kHZ
Voltage frequency product: max. 109 VHz
Rise time: slew rate max. 10kV/ µs

135
Measuring accuracy: apply to a sinusoidal signal at ≤ 4% distortion factor and a level of
≥ ¨10% of range ( Vp ) in the nominal temperature range for 1 year after calibration at
23 ± 1K.
Limits of errors ± (% of m.v.+% of range)
Frequency range AC+DC AC
45Hz…65Hz ± (0.05+0.02) ± (0.05+0.005)
65Hz….1kHz ± (0.12+0.03) ± (0.12+0.005)

Current channel
Shunt factor: Is read automatically from EEPROM in shunt.

Measuring voltage: max. 1.1 VRMS or 1,58 VP . DC…..1MHz sinusoidal.

Bandwidth: DC…..>1.2 Mhz

Sampling rate: 35…70 kHz
Overload Protection up to: 250 VRMS ,350 VP in all range
Voltage frequency product: max. 107
Rise time: slew rate: 30 V/µs

Measuring accuracy: apply to a sinusoidal signal at ≤ 4% distortion factor and a level of

≥ ¨10% of range ( Vp ) in the nominal temperature range for 1 year after calibration at
23 ± 1K.

Limits of errors ± (% of m.v.+% of range)

Frequency range AC+DC AC
45Hz…65Hz ± (0.05+0.02) ± (0.05+0.005)
65Hz….1kHz ± (0.12+0.05) ± (0.12+0.005)

136
 Appendix II Stator winding configurations

“NB” no parallel branches

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
A+ A+ A+ A+ C- C- C- C- B+ B+ B+ B+ A- A- A A- C+ C+ C+ C+ B- B- B- B-

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48
A+ A+ A+ A+ C- C- C- C- B+ B+ B+ B+ A- A- A- A- C+ C+ C+ C+ B- B- B- B-

“2B” Two parallel branches

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
A1+ A1+ A1+ A1+ C1- C1- C1- C1- B1+ B1+ B1+ B1+ A1- A1- A1- A1- C1+ C1+ C1+ C1+ B1- B1- B1- B1-

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48
A2+ A2+ A2+ A2+ C2- C2- C2- C2- B2+ B2+ B2+ B2+ A2- A2- A2- A2- C2+ C2+ C2+ C2+ B2- B2- B2- B2-

“SB” Two branches; special

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
A1+ A1+ A2+ A2+ C1- C1- C2- C2- B1+ B1+ B2+ B2+ A1- A1- A2- A2- C1+ C1+ C2+ C2+ B1- B1- B2- B2-

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48
A1+ A1+ A2+ A2+ C1- C1- C2- C2- B1+ B1+ B2+ B2+ A1- A1- A2- A2- C1+ C1+ C2+ C2+ B1- B1- B2+ B2+

“4B” Four parallel branches

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
A1+ A1+ A1+ A1+ C1- C1- C1- C1- B1+ B1+ B1+ B1+ A2- A2- A2- A2- C2+ C2+ C2+ C2+ B2- B2- B2- B2-
A4+ A4+ A4+ A4+ C4- C4- C4- C4- B4+ B4+ B4+ B4+ A1- A1- A1- A1- C1+ C1+ C1+ C1+ B1- B1- B1- B1-

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48
A3+ A3+ A3+ A3+ C3- C3- C3- C3- B3+ B3+ B3+ B3+ A4- A4- A4- A4- C4+ C4+ C4+ C4+ B4- B4- B4- B4-
A2+ A2+ A2+ A2+ C2- C2- C2- C2- B2+ B2+ B2+ B2+ A3- A3- A3- A3- C3+ C3+ C3+ C3+ B3- B3- B3- B3-
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 21 22 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48
19 20 23 24

A
B
A C' B C C NB A' B'

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 45 46
24 43 44 47 48

A
B
2B
C
A B
C' A' C B'
 SB

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 45 46 48
24 43 44 47

A B' C
B C' B A'
A
C

139
 4B

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 30 31 32 33 34 35 37 38 40 41
24 29 36 39 42 43 44 45 46 47 48

A4+ A3 + A1 - A2+ A3 - A2- A4-

A1 +

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 45 46
43 44 47 48

B1 -
B1 B4+ C2- B3 C2+ B4- B3 -

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

C1 - C4- C2- C1+ C3 - C2+ C4- C3+

140
ISBN-13 978-951-22-8476-4
ISBN-10 951-22-8476-6
ISBN-13 978-951-22-8477-1 (PDF)
ISBN-10 951-22-8477-4 (PDF)
ISSN 1795-2239
ISSN 1795-4584 (PDF)