Testing of Stator Windings For Thermal Aging PDF

Testing of Stator Windings for Thermal
Aging
Technical Report
12065638
12065638
Testing of Stator Windings for
Thermal Aging
1000376
Final Report, August 2000
EPRI Project Manager

J. Stein
EPRI • 3412 Hillview Avenue, Palo Alto, California 94304 • PO Box 10412, Palo Alto, California 94303 • USA
800.313.3774 • 650.855.2121 • [email protected] • www.epri.com
12065638
DISCLAIMER OF WARRANTIES AND LIMITATION OF LIABILITIES
THIS DOCUMENT WAS PREPARED BY THE ORGANIZATION(S) NAMED BELOW AS AN
ACCOUNT OF WORK SPONSORED OR COSPONSORED BY THE ELECTRIC POWER RESEARCH
INSTITUTE, INC. (EPRI). NEITHER EPRI, ANY MEMBER OF EPRI, ANY COSPONSOR, THE
ORGANIZATION(S) BELOW, NOR ANY PERSON ACTING ON BEHALF OF ANY OF THEM:
(A) MAKES ANY WARRANTY OR REPRESENTATION WHATSOEVER, EXPRESS OR IMPLIED, (I)

WITH RESPECT TO THE USE OF ANY INFORMATION, APPARATUS, METHOD, PROCESS, OR
SIMILAR ITEM DISCLOSED IN THIS DOCUMENT, INCLUDING MERCHANTABILITY AND FITNESS
FOR A PARTICULAR PURPOSE, OR (II) THAT SUCH USE DOES NOT INFRINGE ON OR
INTERFERE WITH PRIVATELY OWNED RIGHTS, INCLUDING ANY PARTY’S INTELLECTUAL
PROPERTY, OR (III) THAT THIS DOCUMENT IS SUITABLE TO ANY PARTICULAR USER’S
CIRCUMSTANCE; OR
(B) ASSUMES RESPONSIBILITY FOR ANY DAMAGES OR OTHER LIABILITY WHATSOEVER

(INCLUDING ANY CONSEQUENTIAL DAMAGES, EVEN IF EPRI OR ANY EPRI REPRESENTATIVE
HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES) RESULTING FROM YOUR
SELECTION OR USE OF THIS DOCUMENT OR ANY INFORMATION, APPARATUS, METHOD,
PROCESS, OR SIMILAR ITEM DISCLOSED IN THIS DOCUMENT.
ORGANIZATION(S) THAT PREPARED THIS DOCUMENT
Ontario Power Technologies

Vanderbilt University
Vintek
ORDERING INFORMATION
Requests for copies of this report should be directed to the EPRI Distribution Center, 207 Coggins
Drive, P.O. Box 23205, Pleasant Hill, CA 94523, (800) 313-3774.
Electric Power Research Institute and EPRI are registered service marks of the Electric Power
Research Institute, Inc. EPRI. POWERING PROGRESS is a service mark of the Electric Power
Research Institute, Inc.
Copyright © 2000 Electric Power Research Institute, Inc. All rights reserved.
12065638
CITATIONS
This report was prepared by
Ontario Power Technologies

800 Kipling Avenue
Toronto, Ontario M8Z 6C4
Canada
Principal Investigators
J. Braun
H. Sedding
Vanderbilt University
Department of Physics and Astronomy
Nashville, TN 37240
Principal Investigators
Y. P. Ma
J. Wikswo, Jr.
Vintek
8311 W. 10th Avenue
Kennewick, WA 99366-9569
Principal Investigator
V. Neeley
This report describes research sponsored by EPRI.
The report is a corporate document that should be cited in the literature in the following manner:
Testing of Stator Windings for Thermal Aging, EPRI, Palo Alto, CA: 2000. 1000376.
iii
12065638
12065638
REPORT SUMMARY
The electrical insulation used in stator and rotor windings has a major impact on the reliability of
large motors and generators. Many tests and monitors used by utilities to determine the condition
of the insulation systems used in motors and generators provide little, if any, information on the
thermal aging of the insulation system. This report summarizes a series of experimental studies
employing the Superconducting Quantum Interference Device (SQUID) and other non-
destructive evaluation (NDE) methods to study thermal aging of stator windings.
Background
As stator windings age, their dielectric insulation may break down. Conventional non-destructive
diagnostic tests for thermal aging are limited to measurements of lumped electrical parameters.
The step voltage tests that are commonly used in the certification of generators may lead to
breakdown of the insulation and damage the system. An image of the spatial distribution of the
electrical properties of the windings may provide information about age-related failure. SQUID
imaging systems have been used for mapping magnetic fields due to currents through the
conductors and might supplement existing NDE techniques for stator windings.
Objectives
• To establish the feasibility of correlating thermal aging of stator coil/bar insulation to
dielectric changes measured at frequencies other than 60 Hz.
• To conduct experimental studies on the application of SQUID imaging systems to NDE for
thermal aging of stator windings in electric power systems
Approach
In the first part of the study, researchers conducted dielectric spectroscopy measurements at other
than 60 Hz on high voltage stator bars that had been removed from a hydraulic generator and
subjected to cyclic thermal aging for 2000 cycles. Researchers then conducted a series of
experimental studies on the magnetic imaging of the currents across dielectric insulation
materials, including studies on simulated parallel capacitors with a variety of flaw geometries
and sample segments from stator windings with different thermal aging.
Results
In the tests of high voltage stator bars that had been subjected to cyclic thermal aging for 2000
cycles, the most significant change in diagnostic parameters occurred within the first 250 thermal
cycles. The change in power frequency dissipation factor was more pronounced than either that
for partial discharge (PD) magnitude or discharge extinction voltage (DEV). Dielectric
spectroscopy measurements at other than 60 Hz provided little further information on the
condition of the insulation. The extent of aging of the bars was not determined.
v
12065638
Measurements on a parallel plate capacitor demonstrate that the distribution of charging currents
and the perturbations in this distribution that result from localized changes in dielectric constant
can be imaged magnetically. Measurements on stator bars demonstrate that it is possible to
image both resistive (through dielectric breakdown) or capacitive current flowing across the
insulation layer. The SQUID has adequate sensitivity to image the distribution of capacitive
charging currents even at 60 Hz. The pathway of the resistive current during dielectric
breakdown is clearly shown in the images. Measurements on the sample segments from stator
bars with different aging demonstrate that current distribution due to the displacement current
across the inner copper and the semicon layer can be imaged. Imaging of vector fields provides
additional information about the current distribution. The data also demonstrate that the
homogeneity of both the semicon tape and the copper conductors inside of the insulation may be
imaged.
The images show that the semicon layers of the samples are not homogeneous, especially for the
aged sample, and the inner copper conductors are twisted. The semicon tape becomes more
resistive and more inhomogeneous with thermal aging. The insulation layer between the copper
conductors degenerated with the thermal aging process. Due to the large irregularities of the
semicon and the copper, it is necessary to find out if the signal due to the variation of the
insulation material is large enough to overcome the signal due to the inhomogeneities. For a
fundamental understanding of how the dielectric constant of the insulation layer varies with
aging and how much the current on the semicon and the copper would be affected, the insulation
material needs to be characterized without the semicon or copper.
Keywords
Rotating machine
Stator
Thermal aging
Diagnostic
vi
12065638
Blank page
ABSTRACT
The electrical insulation used in stator and rotor windings, as well as core laminations, has a
major impact on the reliability of large motors and generators. Thermal aging of the insulation,
especially in conjunction with the differential expansion forces experienced by conductor and
insulation, can lead to a progressive embrittlement of the electrical insulation. The objective of
the present project is to establish the feasibility of correlating thermal aging of stator coil/bar
insulation to dielectric changes measured at frequencies other than 60 Hz. High voltage stator
bars, removed from a hydraulic generator were subjected to cyclic thermal aging for 2000 cycles
to accelerate the effects of thermal stress on the insulation system. The most significant change
in diagnostic parameters occurred within the first 250 thermal cycles. The change in power
frequency dissipation factor was more pronounced than either that for PD magnitude or DEV.
Dielectric spectroscopy measurements at other than 60 Hz provided little further information on
the condition of the insulation. The extent of aging of the bars was not determined.
Superconducting Quantum Interference Device magnetometry was performed on selected bars;
measurements showed that the thermally aged bars could be distinguished from the new bars.
vii
12065638
12065638
CONTENTS
1 INTRODUCTION ................................................................................................................. 1-1
2 SELECTION OF SUITABLE STATOR BARS ..................................................................... 2-1
3 INITIAL CHARACTERIZATION AND AGING OF STATOR BARS ..................................... 3-1
4 DIAGNOSTIC CHARACTERIZATION ................................................................................. 4-1

Test Results - Power Frequency Diagnostics ..................................................................... 4-2
Test Results - Dielectric Spectroscopy ............................................................................... 4-8
Test Results - UHF Dielectric Constant Mapping...............................................................4-10
Test Results - SQUID Mapping .........................................................................................4-17
5 SQUID MAPPING RESULTS .............................................................................................. 5-1

5.1 Introduction................................................................................................................... 5-1
5.1.1 Subject of the Work............................................................................................... 5-1
5.1.2 Previous Work ...................................................................................................... 5-2
5.1.3 Current Work ........................................................................................................ 5-2
5.2 Background .................................................................................................................. 5-3
5.2.1 Magnetic Imaging of Cable Currents ..................................................................... 5-3
5.2.2 Parallel-Plate Capacitor ........................................................................................ 5-4
5.2.3 Magnetic Imaging of the Current ........................................................................... 5-6
5.3 Measurements and Results .......................................................................................... 5-7
5.3.1 Stage 1 ................................................................................................................. 5-7
5.3.1.1 Measurements on a Roebel Bar .................................................................... 5-7
5.3.1.1.1 Bar Characteristics ................................................................................ 5-7
5.3.1.1.2 SQUID Measurements........................................................................... 5-8
5.3.1.2 Measurements on the Parallel-Plate Planar Capacitors................................5-14
5.3.1.2.1 Imaging of Charging Currents ...............................................................5-15
5.3.1.2.2 Minimization of Edge Cancellation ........................................................5-19
ix
12065638
5.3.1.2.3 Imaging Plates ......................................................................................5-21
5.3.2 Stage 2 ................................................................................................................5-25
5.3.2.1 Sample Preparation......................................................................................5-25
5.3.2.2 Measurement Procedure ..............................................................................5-26
5.3.2.3 Electrical Properties .....................................................................................5-28
5.3.2.4 Magnetic Imaging .........................................................................................5-32
5.4 Conclusions.................................................................................................................5-51
6 CONCLUSIONS .................................................................................................................. 6-1
7 REFERENCES .................................................................................................................... 7-1
x
12065638
LIST OF FIGURES
Figure 4-1 Discharge Extinction Voltage vs. No. of Thermal Cycles ....................................... 4-5
Figure 4-2 Negative PD Magnitude vs. No. of Thermal Cycles............................................... 4-6
Figure 4-3 Dissipation Factor Tip-Up vs. No. of Thermal Cycles ............................................ 4-6
Figure 4-4 Discharge Extinction Voltage vs. No. of Thermal Cycles ....................................... 4-7
Figure 4-5 Negative PD Magnitude vs. No. of Thermal Cycles............................................... 4-7
Figure 4-6 Dissipation Factor Tip-Up vs. No. of Thermal Cycles ............................................ 4-8
Figure 4-7 Tangent δ in the Range 0.1 Hz to 100 Hz, 61T 33, 1000 Thermal Cycles ............4-11
Figure 4-8 Tangent δ in the Range 0.1 Hz to 100 Hz, 61T 33, 2000 Thermal Cycles ............4-12
Figure 4-9 Tangent δ in the Range 0.1 Hz to 100 Hz, 61T 481, 0 Thermal Cycles ................4-12
Figure 4-10 Tangent δ in the Range 0.1 Hz to 100 Hz, 61T 481, 1000 Thermal Cycles ........4-13
Figure 4-11 Tangent δ in the Range 0.1 Hz to 100 Hz, 61T 215, 0 Thermal Cycles ..............4-13
Figure 4-12 Tangent δ in the Range 0.1 Hz to 100 Hz, 61T 420, 1000 Thermal Cycles ........4-14
Figure 4-15 Tangent δ in the Range 0.1 Hz to 100 Hz, 74B 768, 0 Thermal Cycles ..............4-15
Figure 4-16 Tangent δ in the Range 0.1 Hz to 100 Hz, 74B 774, 1000 Thermal Cycles ........4-16
Figure 4-17 Tangent δ in the Range 0.1 Hz to 100 Hz, 74B 774, 1000 Thermal Cycles ........4-16
Figure 5-1 The current distributions due to the leakage and capacitive currents between
the inner and outer conductor of a coaxial cable. (a) A cable with a circularly-
symmetric resistive fault. b) A cable with a non-symmetric resistive fault. c) The
charging current distribution when the inner and outer cables are energized at
opposite ends. d) The charging current distribution when the inner and outer cables
are energized at the same end. ....................................................................................... 5-4
Figure 5-2 Parallel-plate capacitor model. (a, c) Without flaw. (b, d) With flaws within the
dielectric medium. ........................................................................................................... 5-5
Figure 5-3 Three components of the magnetic fields produced by a current in (a) the x-y
plane, and (b) the x-z plane. The current in the x (or y) direction is silent in the
image of Bx (or By), while the two current segments in the z direction, which are
silent in the image Bz, produce the dipolar signals in the image of Bx............................... 5-6
Figure 5-4 Schematic drawings of the configurations for the magnetic imaging of the
Roebel bars. a) The unflawed test bar without dc leakage current. b) The test bar
with insulation failure and a dc leakage current. .............................................................. 5-8
xi
12065638
Figure 5-5 Magnetic field maps with 10 V at 200 Hz applied between the rear-end
copper lead #2 and the outer-layer lead #3. The in-phase signal is almost zero
because there is no resistive current between the copper and the outer layer, while
the quadrature component shows an amplitude of 50 pT as a result of a capacitive
current between the copper and the outer layer. The strongest current appears
between the two leads (bottom half of the lower-right figure). .........................................5-10
Figure 5-6 Magnetic field maps with 10 V at 200 Hz applied between the inner
conductor at front-end lead #1 and the outer layer lead #3. The most of the current
appears between the two leads along the upper half of the image..................................5-11
Figure 5-7 Magnetic images with 10V at 200Hz applied between the rear-end copper
lead #2 and the outer-layer lead #3. This demonstrates that a charging current
flows along the entire length of the bar, even far from the leads, as the capacitance
is charged over the entire dielectric. ...............................................................................5-12
Figure 5-8 Magnetic field maps with 10V at 200Hz applied between the front copper
lead #1 and the outer-layer lead #3. ...............................................................................5-13
Figure 5-9 (a) Magnetic field map for the bar with dielectric failure, with 5 V dc applied
between the rear copper lead #2 and the outer-layer lead #3. The dc current can
flow though the damaged insulation layer and back to the outer layer. The small
signal below the site of failure and lead #3 is due to the cancellation of the current
inside the copper and that in the outer layer. (b) Magnetic field with 5V dc between
the front copper lead #1 and the outer-layer lead #3. The dc current can flow though
the damaged insulation layer to the outer layer without cancellation...............................5-14
Figure 5-10 Planar capacitor formed by two 1/8 inch thick, 2 inch wide aluminum strips
and a layer of 1/8 inch G-10. (a) Different-end connection: one lead is connected to
one end of the top layer and another lead is connected to the other end of the
bottom layer. (b) Same-end connection: two leads are connected to the same end
of the top and bottom layers. In both (a) and (b), the arrows indicate the
displacement current that cross the insulation layer, and is then collected as an
Ohmic conduction current by the leads attached to the aluminum strips.........................5-16
Figure 5-11 Magnetic field maps (Bz). (a) Different end connection. The magnetic field is
the sum of the top and bottom layers. (b) Same end connection. The magnetic field
become smaller due to cancellation of the fields from the current in the two strips.
The left column shows the in-phase signals, and the right column shows the
quadrature signals. The hole in the G-10 layer is not detectable. ...................................5-17
Figure 5-12 The magnetic field maps for the aluminum capacitor for which the G-10
layer has a hole that filled with (a) a 3/4" diameter copper disk, and (b) a 3/8"
diameter disk. The dashed line in (a) indicates the mapping area. The connections
were on the same end. ...................................................................................................5-18
Figure 5-13 The unequal-plate capacitor formed by a copper-clad G-10 plate and an
aluminum strip. The copper is a 0.3 mm thick, 6 inch wide, and 12 inch long sheet.
An 1/8 inch thick, 2 inch wide 9 inch long aluminum bar is placed on the copper
sheet. An 1/8 inch thick G-10 layer that has been divided into several sections is
placed between them. The leads are copper wires attached to the copper sheet or
aluminum strip over the width of the conductor. The current densities in the two
conductors are different, so the field from the edges of the copper does not cancel
the field from the edges of the aluminum strip. ...............................................................5-19
xii
12065638
Figure 5-14 The data from the unequal-plate capacitor. The left column ((a) and (c)) is
for the different-end lead connection, and right column ((b) and (d)) is for the same-
end connection. (a) and (b) are the data from the unflawed capacitor, and (c) and
(d) are the data from the capacitor missing a 2 × 2 inch square of G-10 in the
middle. Each line is a scan of the magnetic field along a line parallel to the axis of
the aluminum strip. The fields become larger toward to the leads. The arrows
indicate the location of the missing portions; differences in the slope of the signals
can be seen between the flawed and unflawed cases. ...................................................5-20
Figure 5-15 Field map for the unequal-plate capacitor. The left column is for the
capacitor without flaws, and the right column is for the capacitor with the missing 2
by 2 inch portion at the middle. (a-d) Surface maps and the images of the fields. (e-
f) Images after derivative filtering. (g-h) Images after low pass filtering. The missing
portion of G-10 layer is evident in (h)..............................................................................5-22
Figure 5-16 Modified unequal-plate capacitor. The upper aluminum strip is slotted every
1/4 inch. One end is cut through, and another end is connected to the voltage
supply. The left column also shows the missing portions of the G-10 layer. The
middle column is the images after the derivative and the low pass filtering. The right
column is the images after applying a transverse derivative filter to the images in
the middle column. The location of the missing G-10 is clear. The smallest portion
at the middle of the capacitor is 1/2 × 1 inch, as shown in (e).........................................5-23
Figure 5-17 A modified unequal-plate capacitor that has flat-bottomed holes on the
upper surface of the aluminum strip. The images at the right column are obtained
by longitudinal derivative and low pass filtering. The missing portions that touch the
edge are evident; the interior flaws should become visible with appropriate digital
filtering. ..........................................................................................................................5-24
Figure 5-18 Schematic representation of the samples. Each bar consists of a bundle of
30 copper conductors that are separated by a thin layer of insulation. The copper
bundle is surrounded by a 5 mm thick insulation layer, wrapped by the electrically-
conducting semicon tape. The wider surfaces of the bar are covered with a silicon-
rubber coating that is not electrically conducting. The sides of the bar are usually
rubber-free and the semicon tape is exposed.................................................................5-25
Figure 5-19 Magnetic images for Bar 1-1. The ac voltage is applied to the semicon and
the one of the 30 copper conductors. (a) Raw data. (b) Filtered data. The one
copper conductor is oriented diagonally and twisted, as indicated by the arrow..............5-26
Figure 5-20 A schematic illustration of the connection of the copper conductors between
the two ends of the Bar 1-1 and Bar 2-1. The structure inside the bar (dashed lines)
is unknown. ....................................................................................................................5-28
Figure 5-21 The magnitude and the phase of the semicon current versus frequency
when a 10 V ac voltage is applied to the semicon electrodes. The semicon current
induces the out-of-phase current inside the copper. (a) Magnitude. (b) Phase. The
black lines are for the Bar 1-1, which are different from the others. ................................5-30
Figure 5-22 The magnitude and the phase of the capacitive current versus the
frequency when a 10 V ac voltage is applied to the copper and the semicon. The
current is due to the displacement current across the insulation layer. (a)
Magnitude. (b) Phase. The black lines are for the Bar 1-1, which are different from
others. ............................................................................................................................5-31
xiii
12065638
Figure 5-23 The images at phases of 20° through 160° for Bar 1-1, in which four copper
conductors are not connected to the electrodes. The color bars indicate the
amplitude of the signals, which vary with the phase angles. The phases at which
the signal is maximum are 10° for Bx, 50° for By and Bz. The Bz images show that the
copper conductors twist..................................................................................................5-34
Figure 5-24 The filtered images at phases of 20° through 160° for Bar 1-1, in which four
copper conductors are not connected to the electrodes. The images show where
the copper conductors twist, as indicated by arrows. ......................................................5-35
Figure 5-25 The images of the injected dc semicon current for Bar 1-1. (a) Raw data. (b)
Filtered data. ..................................................................................................................5-37
Figure 5-26 Magnetic images (raw data) for Bar 1-1 when 10 V 5.2 kHz is applied to the
two semicon electrodes. The current includes the semicon current and the induced
out-of-phase current in the copper. (a) At –40° phase (maximum amplitude). (b) At
phase 50° (minimum amplitude). ....................................................................................5-38
Figure 5-27 Magnetic images (filtered data) for Bar 1-1 when 10 V 5.2 kHz is applied to
the two semicon electrodes. The current includes the semicon current and the
induced out-of-phase current in the copper. (a) At -40° phase (maximum
amplitude). (b) At the phase 50° (minimum amplitude). ..................................................5-39
copper and the semicon layer. The capacitive current crosses the insulation layer
and is then collected by the leads attached to the copper and semicon. (a) At the
phase 30° (maximum amplitude). (b) At 120° phase (minimum amplitude).....................5-40
Figure 5-29 Magnetic images (filtered data) for Bar 1-1 when 10 V 5.2 kHz applied to
the copper and the semicon layer. The capacitive current crosses the insulation
layer and is then collected by the leads attached to the copper and semicon. (a) At
the phase 30° (maximum amplitude). (b) At 120° phase (minimum amplitude)...............5-41
Filtered data. ..................................................................................................................5-42
out-of-phase current in the copper. (a) At -40° phase (maximum amplitude). (b) At
50° phase (minimum amplitude). ....................................................................................5-43
amplitude). (b) At 50° phase (minimum amplitude). ........................................................5-44
and is then collected by the leads attached to the copper and semicon. (a) At 30°
phase (maximum amplitude). (b) At 120° phase (minimum amplitude). ..........................5-45
30° phase (maximum amplitude. (b) At 120° phase (minimum amplitude). .....................5-46
Filtered data. ..................................................................................................................5-47
xiv
12065638
out-of-phase current in the copper. (a) At -40° phase (maximum amplitude). (b) At
50° phase (minimum amplitude). ....................................................................................5-48
amplitude). (b) At 50° phase (minimum amplitude). ........................................................5-49
and is then collected by the leads attached to the copper and semicon. (a) At 30°
phase (maximum amplitude). (b) At 120° phase (minimum amplitude). ..........................5-50
30° phase (maximum amplitude. (b) At 120° phase (minimum amplitude). .....................5-51
xv
12065638
12065638
LIST OF TABLES
Table 4-1 Electrical Diagnostic Results on Stator Bars, as Received ..................................... 4-3
Table 4-2 Key Features of Figures 4-7 and 4-8, Bar 61T 33................................................... 4-9
Table 4-3 Key Features of Figures 4-9 and 4-10, Bar 61T 481..............................................4-10
Table 5-1 Bar Receipt Date, Age, and ID ..............................................................................5-25
Table 5-2 Electrical Properties ..............................................................................................5-29
Table 5-3 The Magnitude and the Phase of the Current at 5.2 kHz .......................................5-32
xvii
12065638
12065638
1
INTRODUCTION
The electrical insulation used in stator and rotor windings, as well as core laminations, has a
major impact on the reliability of large motors and generators. Failure of the insulation directly
or indirectly will result in failure of the machine, which in turn causes forced outages, reduced
reliability and increased maintenance and repair costs. Industry surveys have shown that
insulation problems are a predominant cause of motor and generator failure [1,2]. Since many
insulation failures originate with aging which occurs over many years, insulation failure rates
will increase in older equipment, and thus the electrical insulation will often determine the
remaining useful life of a machine. This latter aspect is important in asset management programs
to establish current condition.
Pressures on utilities to operate more efficiently, and in particular to minimize unnecessary

maintenance costs, together with the need to determine the remaining life of components, have
created a need to develop means for assessing insulation condition. Of special interest are tests or
monitors which can aid in the assessment of insulation condition while the machine is operating
normally. Manufacturers, consultants and utilities are all trying to improve their ability to assess
rotating machine insulation condition.
The machine environment creates several mechanisms which can age stator insulation. In
addition to electrical and mechanical stresses, heating of the insulation due to I2R losses in the
conductor is an important thermal aging factor, especially in conjunction with the differential
expansion forces experienced by conductor and insulation, leading to a progressive degradation
of the electrical insulation. Thermal aging will result in an initial hardening—which may even be
beneficial to the insulation performance—followed by gradual embrittlement of the insulation.
Diagnostic testing is intended to detect the progressive deterioration of the insulation system.
As well, insulation system deterioration by thermal cycling will result from the different
coefficients of expansion of copper and insulation. Also, since the primary heat source in a
winding is the losses in the copper, the copper reaches a higher temperature more rapidly than
the surrounding insulation. The result is that when a machine is quickly raised to full load, the
copper expands in the axial direction much more quickly than the insulation. Even if the
groundwall insulation were free to move with respect to the stator core, a mechanical shear stress
is created between the copper and the groundwall insulation, and within the groundwall
insulation. Depending on the length of the slot section and the cohesive strength of the
insulation, the bond to the copper may be broken and/or the insulation may delaminate.
Significant debonding or delamination creates air gaps in the insulation, leading to partial
discharges which erode the insulation. Alternatively, debonding, together with the 120 Hz
magnetic forces, gradually allow the strand conductors to move relative to one another, resulting
1-1
12065638
Introduction
in abrasion of the strand insulation and strand shorts. Over several years and hundreds of rapid
load cycles, the insulation eventually fails.
The physical properties of stator insulation are known to change with thermal aging. Typical
insulation characterization techniques include tensile strength & elongation, dynamic mechanical
analysis (DMA), differential scanning calorimetry (DSC), oxidation induction time (OIT),
oxidation induction temperature, Fourier transform infrared spectroscopy (FTIR), gel content,
swelling ratio, density and glass transition temperature. However, all the above techniques,
though very powerful, are destructive and cannot be used for in-situ tests of stator insulation.
Many in-situ tests and monitors are presently used by utilities to determine the condition of the
stator, rotor and coreplate insulation systems used in motors and generators. Most of these
methods are off-line tests which are done with the machine out of service. The most widely used
tests are a dc or ac high potential (hipot) test, together with an insulation resistance (IR) and
polarization index (PI, ratio of IR at ten and one minute) test. Some utilities augment this
information with partial discharge (PD), dissipation factor, dissipation factor tip-up, and/or
capacitance tests. However, these tests provide little, if any, information on thermal aging of the
insulation system.
Earlier studies on aged cable insulations underlined the value of performing tests at other than
power frequency and/or dc. DC insulation resistance and the related polarization index were
found to be insensitive to thermal aging for typical cable insulation (SBR, PVC, butyl, PE and
EPR), even in a very advanced state of deterioration. The dielectric characteristics studied over a
broad range of frequencies, and expressed for instance as tan delta, show variations of dielectric
properties, as a function of frequency, with the type of cable insulation.
The objective of the present project is to establish the feasibility of correlating thermal aging of
stator coil/bar insulation to dielectric changes measured at frequencies other than 60 Hz.
1-2
12065638
2
SELECTION OF SUITABLE STATOR BARS
In order to fulfil the objectives of the project, a sufficiently large number of stator bars were
required. Among the criteria considered when selecting suitable test objects were,
• Cost
• Materials
• Machine type
• Availability
Costs associated with manufacturing stator bars specifically for this project would have been
prohibitive, e.g., a single bar for a direct liquid-cooled winding can cost in the range of $80,000.
Consequently, a request for appropriate samples was sent to EPRI member utilities. This request
was for stator bars or coils from large generators, insulated with materials typical of those found
in present common use. Older, asphaltic-mica systems, although still encountered on operating
machines, were precluded from the study. Consequently, stator windings insulated with
thermoset materials, i.e., those in which the binder was an epoxy or polyester resin, were sought.
After a lengthy search period, stator bars insulated with a modern epoxy-mica insulation system
were obtained.
A total of 20 bars were received from the US Bureau of Reclamation’s Grand Coulee Dam
Generating Station. Eighteen of the bars were removed from unit 22 generator which was being
rewound, the remaining two bars were spares. The latter two test objects are required to provide
a reference with which to compare the results from the aged samples. The nominal rating of this
generator was 15 kV, 826 MVA and 86 r.p.m. At such a high rating, these bars were water-
cooled. Tests on the bars in unit 22 generator indicated that water was leaking from the brazes,
where hollow strands of the bars are connected to the headers. Approximately 60% of the bar
ends were deemed to have been affected by the presence of moisture. Consequently, this and
other units at Grand Coulee have been rewound.
The bars which were supplied from Grand Coulee were insulated with a resin-rich epoxy-mica
insulation system. Each bar has a total length of 3.10 m, of which the slot cell comprises 2.15 m.
The surface of the slot portion is coated, on one side only, with a conductive silicon rubber
which facilitates electrical contact between the bar surface and the stator core iron. Visual
inspection of the bars showed that they were in good condition considering the service history
and that some mechanical force is required to remove them from the stator slots.
2-1
12065638
12065638
3
INITIAL CHARACTERIZATION AND AGING OF STATOR
BARS
An accelerated aging test is intended to compare objectively the long term performance of
insulation systems. In an accelerated aging test, one or more stresses believed to contribute to
aging are applied to the insulation system at significantly higher levels than those experienced in
normal service. Under enhanced stresses the degradation of insulation occurs in hours, days or
weeks rather than years. Bars with different insulation systems are subjected to specified
enhanced stresses for a fixed period. The levels and duration of stresses applied in the test should
preferably be based on the functional specification of the insulation system. The insulation
system suffering less degradation in such a test is relatively the better insulation system.
Ontario Hydro has, for instance, been qualifying insulation systems, and hence sources of
supply, on the basis of a voltage endurance test. In this test, the capability of full size stator coils
or bars is assessed by energizing the coils at a prescribed voltage at elevated temperature. The
pass/fail criterion is based on whether the coil/bar survived the test for a predetermined time
period. As well, Ontario Hydro has applied a thermal cycling test to stator coils and bars from
machines subject to such stresses. This test is performed on full size stator bars and coils and
provides valuable data on the ability of the groundwall insulation to resist delamination and
remain well bonded to the copper stack.
The principal aging stress under consideration in this study is thermal, consequently,
consideration was given to various thermal aging methods. In principle, the simplest thermal
aging method is to place the bars in an oven and heat them at a fixed, elevated temperature for a
prescribed time. However, the size of the test objects, approximately 2.5 m in length, presents a
practical problem. Ovens of sufficient size are available, however, this method of heating the bar
is not satisfactory because in a relatively short time frame the temperature distribution in the bar
will be uniform. This situation is undesirable because it does not result in a temperature gradient
across the groundwall insulation, which exists in practice. Alternatively, heater plates can be
applied to the slot portion of the bar. This method is used in voltage endurance tests [1] of stator
coils or bars. Applying heat in this way produces a thermal gradient across the groundwall
insulation, albeit in the opposite sense from that encountered in service, i.e., the gradient is
driven from the outside in rather than inside out.
A third heating method involves thermally cycling the bars between two temperatures. This
technique has the advantage that a thermal gradient closely simulating that encountered in
operation is produced. Further, cyclic thermal aging has, in principle, a higher probability of
resulting in measurable aging in a shorter period of time because of the differential stresses
imposed by thermal cycling. This method of aging the stator bars was selected for this study.
3-1
12065638
Initial Characterization and Aging of Stator Bars
The test set up used is described in detail in IEEE standard 1310-1995 [2]. Essentially, the test
consists of heating and force cooling the test objects; stator coils or bars, and periodically
applying a range of electrical diagnostic tests to assess the aging effects induced by thermal
cycling. The lower and upper temperature limits chosen were 40 C and 140 C respectively. The
upper limit is below the thermal classification of the insulation (class F, 155 C) but higher than
the average operating temperature of the unit in service. Average heating and cooling rates of
2.5 C per minute were employed resulting in a cycle time of approximately 90 minutes.
Following initial (conventional) characterization the stator bars were subjected to accelerated
thermal aging, nominally to failure (as defined by the diagnostic tests), by thermal cycling. The
use of thermal cycling is critical as it introduces real thermomechanical stresses similar to those
experienced in service; bars/coils may survive for excessive, i.e., unrealistic, times at constant
oven aging temperatures. The stator bars were subjected to diagnostic characterization during the
aging phase, as discussed below. Bars were prepared for testing through removal of the
insulation from the bar.
3-2
12065638
4
DIAGNOSTIC CHARACTERIZATION
Diagnostic characterization of stator bars included conventional and novel testing through the
aging period. Novel techniques included other than 60 Hz frequency testing along with
exploratory techniques, primarily magnetic imaging (SQUID) and UHF (10-100 MHz) dielectric
constant measurements. Diagnostic sub-cycles were performed after 250 thermal cycles, or
375 hours.
The well established techniques include:

• dc leakage
• power factor tip-up
• partial discharge
• destructive hipot testing
The non-conventional techniques include:

• low frequency (0.01-1 Hz) tan δ
• UHF (10-100 MHz) dielectric constant mapping
• magnetic imaging (SQUID)
Partial discharge (PD) data provides information on the void size and distribution of a stator bar
or coil. The voids result either from improper impregnation and cure during manufacturing or
from delamination of the groundwall insulation in service. A typical PD test consists of
measuring the PD magnitude at a prescribed voltage, typically the nominal line-to-ground
voltage of the machine, as well as the discharge inception and extinction voltage, DIV and DEV
respectively. Proper analysis and interpretation of these parameters can provide insight into the
location, as well as size, of the voids.
Capacitance and dissipation factor tests also provide a measure of void content, however, this is
a spatially averaged value, i.e., the test cannot distinguish the contribution of one large defect to
that from a multiplicity of smaller voids. This test also provides a measure of the bulk properties
of the insulation and can be used to some extent to assess cure. Capacitance and dissipation
factor values are recorded at more than one voltage. Usually measurements are made at a voltage
below DIV and at the rated line-to-ground voltage of the stator winding. The difference between
the two values is known as the dissipation factor tip-up.
4-1
12065638
Diagnostic Characterization
Dielectric spectroscopy is a term generically applied to a number of techniques [3,4] in which

properties such as capacitance, permittivity, dielectric constant and tangent delta (loss factor) are
determined over a range of frequencies typically extending from mHz to kHz. Typically, these
methods have been applied to extruded polymers such as those used in cable insulation [5,6].
However, some applications to the investigation of aging in rotating machine insulation have
also been reported [7]. The basic properties measured are the same as described above, however,
they are recorded at multiple frequencies. Appropriate equipment to perform this measurement
was not available at the commencement of the project, consequently, data on this technique
contained in this report is limited.
Test Results - Power Frequency Diagnostics
With the exception of the two spare bars, the water boxes of the bars that had been recovered
from the generator were covered in an epoxy casting resin. This material had to be removed prior
to any electrical tests. Subsequent to removal of the epoxy casting, and to ensure that the bars
had not been damaged in the process of removal, a moderate ac hipot of 30 kV was performed on
each bar. All of the bars withstood the hipot.
Prior to commencement of the thermal aging program, all bars were subjected to standard power
frequency PD, capacitance and dissipation factor tests. The data from these measurements are
illustrated in Table 4-1. Partial discharge measurements were accomplished using a 375 pF
coupling capacitor, the output of which is connected, via a power frequency filter, to a 100 MHz
bandwidth analog oscilloscope. In addition to measurements of the maximum PD magnitude at
8.7 kV, the discharge inception (DIV) and discharge extinction voltage (DEV) were also
recorded. The PD magnitudes reported in the Table are recorded at 8.7 kV, the nominal
line-to-ground voltage of the winding. Examination of the Table shows that the lowest DEV
observed was 2.3 kV (bar 61T 256) and the highest DEV was 4.5 kV (bar 61T 458). Both values
of DEV relate to negative PD activity that results from discharges at, or close to, the copper
conductors. We consider negative PD data more representative of deterioration in this case,
because the surface condition of the bars is questionable due to the removal process. The average
negative DEV of the 20 bars is 3.3 kV. Typically, on new bars or coils of this voltage rating,
DEVs in the range of five to six kV are measured. Interestingly, the DEVs of the spare bars, 74B
768 and 74B 774, are below the average value recorded. Consequently, the DEVs of the bars
removed from the winding cannot be inferred as indicative of in-service aging. A similar analysis
of the negative PD magnitudes, recorded at 8.7 kV, shows that the average PD magnitude was
15 mV, a value that is well within the acceptable limits for new coils or bars. Again, there is no
difference between the results for the spare and in-service bars, indicating that no significant
delamination has resulted from operation.
4-2
12065638
Table 4-1
Electrical Diagnostic Results on Stator Bars, as Received
Bar ID DIV (kV)+ DIV (kV)- DEV (kV)+ DEV (kV)- PD (mV)+ PD (mV)-
61B 493 4.1 3.8 3.6 3.4 22 20

61T 215 3.3 3.1 2.8 3 20 17
61T 249 4.6 4.8 4.4 4.3 10 10
61T 256 2.8 2.8 2.3 2.3 15 15
61T 29 3.5 3.4 3.3 3.2 20 20
61T 313 4.3 4.3 3.7 3.7 13 10
61T 319 3.2 3.5 2.7 2.9 8 8
61T 33 2.9 3.1 2.6 2.8 15 15
61T 354 3.2 2.7 2.5 2.4 28 30
61T 388 3.8 3.6 3.6 3 50 15
61T 415 4.7 4.7 4.1 4.3 20 15
61T 420 3.6 3.6 2.9 3 10 10
61T 458 4 4.9 3.4 4.5 15 15
61T 46 3.5 3.7 3.2 3.2 15 15
61T 48 3.9 4.1 3.6 3.7 15 15
61T 481 4 4.1 3.7 3.7 15 15
61T 484 3.8 3.9 3.3 3.4 10 10
61T 492 3.7 3.7 3.4 3.3 20 20
74B 768 3.2 3.2 2.9 2.9 15 15
74B 774 3 3 2.6 2.6 15 13
Physical limitations in the thermal cycling rig, i.e., the capacity of the power frequency current
transformer, permitted only four bars to be aged at a time. One of these bars is used as the means
to control the temperature, which involves drilling into the groundwall insulation to make contact
with the copper conductors. This control bar cannot be used for any further high voltage
diagnostics. Hence, the true number of valid test objects being aged is three. The initial three
bars selected for aging were, 61T 33, 61T 420 and 74B 774. The latter bar is a spare that is used
to provide a reference with which to compare the aging characteristics of the service-aged
materials. Bars 61T 33 and 61T 420 were chosen as being representative of the service-aged
objects.
4-3
12065638
As mentioned above, diagnostic sub-cycles consisting of PD, capacitance and dissipation factor
were performed at 250 thermal cycle intervals. The results of these diagnostics are illustrated in
Figures 4-1 and 4-2 for negative DEV and negative PD magnitudes respectively as a function of
thermal cycles. Bar 61T 420 was removed from aging after 1000 thermal cycles for testing by
the two sub-contractors on the project. Examination of Figure 4-1 shows that apart from an
increase in the DEV of the spare bar after 1500 thermal cycles, there is little significant
difference between the DEVs of the aged and unaged bars.
Further, the DEV does not decrease as a function of thermal cycling, which might be expected,
i.e., in principle, if delamination is occurring, the DEV should decrease. The variation of PD
magnitude as a function of thermal cycling is illustrated in Figure 4-2. In this case, there appears
to be some correlation between an increase in PD magnitude and the number of thermal cycles.
Additionally, the PD magnitudes recorded from the spare bar are lower than those obtained from
the service-aged samples. Closer inspection of Figures 4-1 and 4-2 show that there is some
change in the PD behavior in the initial 250 thermal cycles which may be evidence of aging
occurring. However, subsequent cycling does not appear to produce much further change. After
1000 thermal cycles, because of the apparent lack of significant aging, evidenced by the small
changes in the diagnostic parameters, the maximum temperature was increased from 140 C to
155 C. Examination of Figure 4-1, shows that the change in maximum temperature has
apparently little immediate effect on the DEV of either bar 61T 33 or 74B 774. Bar 74B 774
demonstrates an increase in DEV at the higher temperature, whereas bar 61T33, after an initial
decrease in DEV, is little affected by the change in the upper limit of the thermal cycle.
Capacitance and dissipation factor were measured using a Multi-Amp CB605 bridge and a
1000 pF standard capacitor. These measurements were performed in broad accordance with
IEEE 286 and IEC 894. The stress grading material in the end regions of the bars was guarded
out. Figure 4-3 illustrates the variation of dissipation factor tip-up as a function of thermal
cycling over the same number of thermal cycles as Figures 4-1 and 4-2. Again, bar 61T 420 was
removed from test at 1000 thermal cycles, at which point the maximum temperature was
increased to 155 C. The dissipation factor tip-up data clearly show that the spare bar, 74B 774, is
unaffected by 2000 hours of thermal cycling, regardless of the upper temperature limit. However,
both service-aged bars demonstrate increases in tip-up of the order of 600 to 800% in the first
250 thermal cycles. After this initial increase, up to 1000 thermal cycles, the tip-up values tend to
stabilize at the higher values, indicating that the change is permanent. Observation of an increase
in tip-up during the initial stages of a thermal cycling test can, in some cases, be deemed
evidence of some post-curing of the thermoset resin binder. However, in such situations, the tip-
up values usually decrease in subsequent thermal cycles. Further, the stability of the tip-up
values exhibited by the spare bar indicate that the significant rise in dissipation factor cannot be
attributed to post-curing.
Figures 4-4, 4-5 and 4-6 include data obtained from bar 61T 481 that replaced bar 61T 420. All
1000 thermal cycles on this bar were up to the maximum temperature of 155 C. Inspection of
Figures 4-4, 4-5 and 4-6 demonstrates similar characteristics to the behavior of the other two
service-aged bars. Initially, there is a dramatic increase in tip-up, Figure 4-6, after which the
values stabilize. Similarly, an increase in the negative PD magnitude and concomitant decrease
in the DEV is observed after the first 250 thermal cycles. The change in the diagnostic
4-4
12065638
parameters for bar 61T 481, within this initial aging period, is greater than observed on either of
the two service-aged bars which were subject, initially, to the lower 4 maximum cycle
temperature of 140 C.
Overall, the power frequency diagnostic tests used in this study indicated that whatever aging did
occur, it happened predominantly within the first 250 thermal cycles. Dissipation factor tip-up
appears to be more sensitive to aging, e.g., the data displayed in Figure 4-3 for bar 61T 33, than
PD. The degree of aging in the initial thermal cycling period is influenced by the maximum
upper temperature limit.
6
6 1T 33
6 1 T 42 0
5 74 B 77 4
4
D E V (-) (k V )
0
0 50 0 1 000 1 500 2 000
N o . o f T h e r m a l C y c le s
Figure 4-1
Discharge Extinction Voltage vs. No. of Thermal Cycles
4-5
12065638
80
6 1T 3 3
70 6 1 T 42 0
P D M a g n it u d e ( -) ( m V )
74 B 77 4
60
50
40
30
20
10
0
0 50 0 1 00 0 1 50 0 2 00 0
Figure 4-2
Negative PD Magnitude vs. No. of Thermal Cycles
1 .6
6 1T 3 3
D i s s ip a t io n F a c t o r T ip - U p ( % )
1 .4 6 1 T 42 0
74 B 77 4
1 .2
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
0 50 0 1 00 0 1 50 0 200 0
N o . o f T h e rm a l C y c le s
Figure 4-3
Dissipation Factor Tip-Up vs. No. of Thermal Cycles
4-6
12065638
6
6 1T 33
61T 42 0
5 61T 48 1
74 B 77 4
4
D E V (-) (k V )
0
0 50 0 1 000 1 500 2 000
Figure 4-4
Discharge Extinction Voltage vs. No. of Thermal Cycles
90
61T 33
61T 42 0
80
61T 48 1
74 B 77 4
P D M a g n it u d e ( - ) ( m V )
70
60
50
40
30
20
10
0
0 50 0 1 00 0 1 50 0 2 00 0
Figure 4-5
Negative PD Magnitude vs. No. of Thermal Cycles
4-7
12065638
2 .0
6 1T 33
D is s ip a t io n F a c t o r T i p - U p ( % )
61T 42 0
61T 48 1
7 4B 77 4
1 .5
1 .0
0 .5
0 .0
0 50 0 1 000 1 500 2 0 00
Figure 4-6
Dissipation Factor Tip-Up vs. No. of Thermal Cycles
Test Results - Dielectric Spectroscopy
Capacitance and dissipation factor, or tangent δ, measurements over a range of frequencies from
0.1 to 100 Hz and voltages up to 8.7 kV were performed using a commercial instrument known
as the insulation diagnostic analyzer (IDA) manufactured by WaBTech AB. Dielectric loss
measurements at frequencies below power frequency had been investigated in oil-paper
insulation as well as for extruded polymers with this technique. This technique was investigated
to determine its potential use as a tool to quantify thermal aging of stator insulation systems. This
apparatus only became available some time after commencement of the project, therefore the
data reported here is somewhat limited.
Figure 4-7 illustrates tangent δ data obtained over the frequency range 0.1 Hz to 100 Hz at
various voltages for bar 61T 33 after 1000 thermal cycles. The absence of dielectric loss
information in the range 50 to 100 Hz at the higher voltages results from the variable frequency
power supply having insufficient capacity to cope with the higher capacitive load as the
frequency increases. This feature is common to all subsequent plots of dielectric loss as a
function of frequency and voltage, i.e., Figures 4-8 to 4-10. Figure 4-8 shows essentially the
same data as Figure 4-7, except that the measurements were made upon completion of
2000 thermal cycles. Figures 4-9 and 4-10 show similar data for bar 61T 481 at 0 and
1000 thermal cycles, respectively. Some key features of these data have been extracted and
compared with conventional power frequency data in Tables 4-2 and 4-3 for bars 61T 33 and
61T 481 respectively.
4-8
12065638
Table 4-2
Key Features of Figures 4-7 and 4-8, Bar 61T 33
Thermal Frequency Voltage Tangent δ 60 Hz Loss

Cycles (Hz) (kV) (%) (%)
1000 60 2 0.566 0.544

1000 0.1 2 0.797 -
1000 21.61 8.7 1.390 1.580
1000 0.1 8.7 1.640 -
2000 60 2 0.347 0.505
2000 0.1 2 0.777 -
2000 46.51 8.7 1.740 1.820
2000 0.1 8.7 1.970 -
4-9
12065638
Table 4-3
Key Features of Figures 4-9 and 4-10, Bar 61T 481
Thermal Frequency Voltage Tangent δ 60 Hz Loss

Cycles (Hz) (kV) (%) (%)
0 60 2 0.530 0.471
0 0.1 2 1.280 -
0 21.61 8.7 0.621 0.534
0 0.1 8.7 1.450 -
1000 60 2 0.232 0.417
1000 0.1 2 0.673 -
1000 46.51 8.7 1.950 2.072
1000 0.1 8.7 2.260 -
Inspection of Figures 4-7 to 4-10 shows that the dielectric loss generally increases inversely with
frequency. The losses at lower frequencies tend to be somewhat higher because they include the
contribution from ionic species that have lower mobilities. However, comparison of Figure 4-8
with Figure 4-7, i.e., measurements made after 2000 thermal cycles with respect to those made
after 1000 thermal cycles, demonstrates that there is no particular advantage, in this case, to be
gained by measuring the dielectric loss at frequencies other than 60 Hz. A similar comment can
be made with regard to the data displayed in Figures 4-9 and 4-10, as well as examination of the
key features of these Figures recorded in Tables 4-2 and 4- 3. Other bars were subject to
dielectric spectroscopy, however, no significant aging was performed on these test objects.
However, for completeness, the data obtained are illustrated in Figures 4-11 to 4-17.
Concluding that dielectric spectroscopy provides little, or no, additional information on thermal
aging in stator winding insulation would, at this stage, be premature. There is no firm evidence,
other than the observation that the power frequency dissipation factor increased, that any aging
of the test objects has been accomplished. Such a determination can only be made by dissection
of the bars which was precluded from this project.
Test Results - UHF Dielectric Constant Mapping
Vintek has developed a Stator Bar Wetness Detector (SBWD). Two bars were subjected to this
test; one bar had been failed at OHT and the failure location identified. Although the “blowout”
area was detected by Vintek in this test, no other significant anomalies were detected. The most
significant test data were obtained with the SBWD in a capacitance testing mode where the
probe is designed to sense at four separate depths into the insulation ranging from 0.050 inches
deep to the full insulation thickness. The SBWD software was modified to allow the test data to
be presented as a composite of the test data using the “good” bar sample as a standard compared
4-10
12065638
to be failed bar section. Testing was conducted at frequencies from 50 Hz to 100 kHz. Testing at
higher frequencies was precluded due to the semiconducting surface coating on the bars.
The probe used to take this data was 2.5 inches wide and was moved in 0.5 inch incremental
steps. The peak value of the data occurs at the position of the “blowout”. The peak value is
approximately two standard deviations higher than the average reading which is not very
significant; also, this higher reading implies a higher dielectric value in this area which seems
doubtful. Other factors such as slight geometry changes causing better coupling to the probe
cannot be discounted as possible sources as well.
Figure 4-7
Tangent δ in the Range 0.1 Hz to 100 Hz, 61T 33, 1000 Thermal Cycles
4-11
12065638
Figure 4-8
Figure 4-9
4-12
12065638
Figure 4-10
Figure 4-11
4-13
12065638
Figure 4-12
Figure 4-13
4-14
12065638
Figure 4-14
Figure 4-15
Tangent δ in the Range 0.1 Hz to 100 Hz, 74B 768, 0 Thermal Cycles
4-15
12065638
Figure 4-16
Figure 4-17
4-16
12065638
Test Results - SQUID Mapping
SQUID results are presented in Chapter 5.
4-17
12065638
12065638
5
SQUID MAPPING RESULTS
5.1 Introduction
5.1.1 Subject of the Work
At present, conventional non-destructive diagnostic tests for aging and breakdown of dielectric
insulation on high voltage cables and stator windings are limited to measurements, made
typically at the end of the cable or winding, of lumped electrical parameters such as dc insulation
resistance, dc and ac leakage current, polarization index, or power factor (tan delta), with the
later measurements being made over a range of frequencies. The situation is further compounded
by the fact that any tests that result in the breakdown of the insulation cannot be considered
non-destructive, and hence are better suited for laboratory studies than in situ testing of
operational systems; unfortunately, dc and ac step voltage tests which are commonly used in the
certification of generators may in fact lead to breakdown of the insulation and hence damage to
the system.
While lumped parameter analyses of cables are readily interpreted if the cables are spatially
homogeneous, it is highly unlikely that the electrical properties of the dielectric are uniform
along a complex stator winding; it may in fact be the case that a small number of localized
inhomogeneities are responsible for the age-related failure of a cable or winding that is otherwise
satisfactory. However, the study of the process of thermal aging of insulation in cables and
windings in the laboratory is hindered by the lack of knowledge of the voltage-, and
age-dependence of the spatial variation of the dielectric properties. Only now are tests being
developed that might provide a map of the spatial distribution of the electrical properties of the
cable. One such test utilizes microwaves to measure the dielectric constant and dielectric
absorption profile along a cable. Another approach, the subject of this report, would use a
SQUID magnetometer to map inhomogeneities of the current across the dielectric that
subsequently return through a conducting layer covering the outside of a cable or winding, and
then to determine discontinuities in the dielectric constant or the location of breakdown currents
along the cable.
While not the thrust of the present magnetic imaging effort at Vanderbilt, we recognize the
relevance of our studies to high-potential, or high-pot, testing, which is typically used as a means
to demonstrate the integrity of insulation in generator windings. A very large potential, much
larger than typical operating voltages, is placed across the insulating barrier and if a minimal
value leakage current flows, the insulation is deemed secure. The practice has recently come
under criticism, since some users worry that high-pot testing might actually damage the
insulation and that failure may be more likely after successful high-pot testing than before.
5-1
12065638
SQUID Mapping Results
5.1.2 Previous Work
As a first demonstration of the ability of a SQUID to detect a fault current in a coaxial cable, we
fabricated an air-gap coaxial cable with localized leakage current.1 The cable was constructed of
hollow brass tubes and a leakage current flowed when an artificial flaw was introduced between
the tubes. The flaw consisted of a bundle of cotton fibers which had been immersed in a saline
solution. The flaw was inserted close to one end of the coax, while the other end was shorted. As
a true coax is magnetically silent, the end effects and the perturbations caused by the flaw are all
that we expected to see. In this test, we showed that when the sample is scanned beneath the
SQUID, the contribution of the flaw can be seen against a nearly flat background. The
perturbations to that background were due to the asymmetry in current flow caused by the
asymmetric flaw. While this result was encouraging, it was an extremely simple phantom model
that needed many improvements before we could accurately determine the minimum leakage
current SQUID NDE can detect for partially and fully loaded cables.
5.1.3 Current Work
This work has been performed in two stages:
(1) Stage 1
The study in Stage 1 includes measurements on large-scale parallel-plate capacitors, and

measurements on two sections of stator bar. All the images were taken by MicroSQUID I, which
measures the z-component of the magnetic field only. The studies demonstrate the capability of
imaging both resistive and capacitive currents through the dielectric of a stator bar, and provide a
basis for understanding the generation of the magnetic fields.
(2) Stage 2
The study in Stage 2 is the continuing effort to examine both the role of inhomogeneities in the
imaging of resistive current distributions, and digital filtering approaches that would allow us to
better identify flaws in the dielectric. By measuring bars at different times during the aging
process, it may be possible to identify what role these inhomogeneities play in age-related
failures. All the recent images were obtained using our new vector SQUID (MicroSQUID III),
which measures all three components of the magnetic field. The images for the x- and
y-components provide us with additional information about the current distribution. The
uniformity of the charging current patterns in bars has been examined to ascertain the extent to
which these patterns reflect underlying inhomogeneities in either the conductors or the dielectric.
1
EPRI Report High Resolution Superconducting Magnetometry for Nondestructive Evaluation, W.G. Jenks, Y.P.
Ma, and J.P. Wikswo, EPRI TR-108649, August, 1997.
5-2
12065638
5.2 Background
5.2.1 Magnetic Imaging of Cable Currents
By Ampere’s law, there is no magnetic field outside of an infinite long axially-symmetric coaxial
cable carrying equal and opposite currents on the two conductors. This may provide a sensitive
measure for detecting departures from axial symmetry, or longitudinal variations in the resistive
or capacitive currents through the dielectric. Let us first consider the long cylindrical cable in
Figure 5-1a, to which a dc potential, V, has been applied in the absence of a load. In the steady
state, there will be no displacement currents. Since there is no load, the only current carried by
the cable will be that from leakage or fault currents. Let us furthermore suppose that the leakage
occurs only at a single, radially symmetric fault of resistance Rf . Since the fault is axially
symmetric, then the current will flow radially out from the center conductor to the outer
conductor with no angular dependence. As a result, this fault will be magnetically silent. If,
however, the fault is restricted to one side of the cable, as shown in Figure 5-1b, then the axial
symmetry of the radial fault current is broken, and a local magnetic field will be produced. There
can be a contribution to the magnetic field from the fault current and the change in the current
distribution along the inner and outer conductors. The breaking of the axial symmetry of the
current in each cable could lead to an imperfect cancellation of the field of one conductor by that
in the other. In either case, a SQUID magnetometer with a readily-achievable laboratory
sensitivity of 2 to 20 pT should be able to detect a 0.2 to 2 microamp fault current flowing
through 1 cm of dielectric.
If an AC voltage is applied to the cable, any displacement or conduction currents that are not
axially-symmetric would also be detectable. Figures 5-1c through 5-1d show the distribution of
capacitor charging currents that would flow along the cable if the two conductors are energized
at the opposite or same ends. The radial displacement current through the dielectric is manifest
along the cable as an axial Ohmic charging current. Note that the net magnetic field outside the
cable should be zero in Figure 5-1d because the two conductors carry the current in opposite
direction. Since there is a large magnetic field in Figure 5-1c, it may be difficult to detect the
small change of the magnetic field due to the fault. As we will see, the best geometry for
detecting insulation failures will be that in which the ideal cable is magnetically silent, i.e.,
Figure 5-1d, so that departures from ideality can be detected with the greatest sensitivity.
From Figure 5-1, we can see that in an ideal, cylindrical coaxial cable, any axially symmetric
perturbation in either the resistance or capacitance of the dielectric would not produce an
external magnetic field. However, typical rectangular stator windings are not axially symmetric.
It will be necessary to determine for each geometry the distribution of the magnetic field around
the cable due to both the axial currents along the cable, and the radial fault currents. If the axial
currents produce a magnetic field, their magnetic field will have a differing orientation and
spatial pattern than that from a radial fault; mapping of the vector magnetic field should allow
the separation of the two types of current. In addition, the presence of a fault will result in a
greater axial current on one side of the fault than on the other. The wide dynamic range of
SQUID magnetometers may make it possible to localize changes in the magnetic field that are a
small fraction of the total field. In the case of stator windings fabricated from multiple wires,
5-3
12065638
careful measurements of the distribution of the magnetic around the winding may allow
determination of whether the axial current in the cable is carried uniformly by all wires.
Figure 5-1
The current distributions due to the leakage and capacitive currents between the inner and
outer conductor of a coaxial cable. (a) A cable with a circularly-symmetric resistive fault.
b) A cable with a non-symmetric resistive fault. c) The charging current distribution when
the inner and outer cables are energized at opposite ends. d) The charging current
distribution when the inner and outer cables are energized at the same end.
5.2.2 Parallel-Plate Capacitor
The cable bar can be idealized as a rectangular capacitor. For better understanding how the
capacitive currents flow through the bar, we may use a simple parallel-plate capacitor model.
Figure 5-2a shows a simple parallel-plate capacitor with two extended conducting plates
insulated by a layer of dielectric material. The voltage source is connected to the two conducting
plates. A dc voltage applied across the two electrodes does not produce a current unless there is
breakdown of the dielectric layer, whereas an ac voltage across the two conducting plates
produces a displacement, or capacitive, current across the intact dielectric between the two
conducting plates.
5-4
12065638
Figure 5-2
Parallel-plate capacitor model. (a, c) Without flaw. (b, d) With flaws within the dielectric
medium.
The spatially-extended capacitor may be visualized as a number of small capacitors, whose

displacement currents across the dielectric layer are collected by the two electrodes attached to
the two conducting plates, as shown in Figure 5-2c. The current flowing on each conducting
plate becomes resistive, and travels in opposite directions. At the location near the electrode, the
current becomes larger because it is the sum of the displacement currents across all of the
capacitors. If the electrodes are connected to the opposite edges of the conductors, the current is
also larger at the location near the electrode for each conductor, however, the currents in the two
conductors flow in the same direction.
If both the conductor and the dielectric layer are homogeneous, all of the small capacitive
elements that represent the distributed system are identical, as shown in Figure 5-2c. The current
flowing in the conducting plate is longitudinal, and increases gradually towards the leads. The
current density is only a function of the distance from the electrode.
Any imperfections of the insulator cause alter the value of the capacitive element associated with
that region, and then the current density on the conducting plate at that location varies, as shown
in Figure 5-2b and 5-2d. As a consequence, this can even produce a transverse current. Any
imperfections in either conductor could affect both the electrical potential and current
distributions on the conducting plate.
5-5
12065638
5.2.3 Magnetic Imaging of the Current
The current flowing in the conductor has associated with it a magnetic field, which can be
measured by a SQUID magnetometer. By the Law of Biot and Savart, the local magnetic field is
perpendicular to the local current. Hence a current in the x direction will produce a magnetic
field in the y-z plane. This is demonstrated in Figure 5-3, which shows the images of the three
components of the magnetic field produced by a 0.5 mA current segments in x-y plane (5-3a),
and in x-z plane (5-3b). (Obviously, the required continuity of current does not allow isolated
current segments, but this approach is useful for understanding which currents contribute to a
particular field component.) The pick-up coils of the vector SQUID gradiometer are located
6 mm above the current segments.
a) C u rre n t in th e x-y p lan e
BX BY BZ
y
b ) C u rren t in th e x-z p lan e BX BY BZ
Figure 5-3
Three components of the magnetic fields produced by a current in (a) the x-y plane, and
(b) the x-z plane. The current in the x (or y) direction is silent in the image of Bx (or By),
while the two current segments in the z direction, which are silent in the image Bz, produce
the dipolar signals in the image of Bx.
Figure 5-3a shows that the current in the x direction is silent in the image for Bx, while the
current in the y direction is silent for By. All currents in the x-y plane contribute to the
z-component images. Figure 5-3b shows that the two current segments in the z direction are
silent in Bz. It is clear that the image for Bx shows the two current segments in the z direction.
Those two segments do appear in the image for By, however, it is difficult to distinguish them
from the By field for the current segment in the x-direction.
Given this background, we can proceed to examine the magnetic fields from several test samples.
5-6
12065638
5.3 Measurements and Results
5.3.1 Stage 1
5.3.1.1 Measurements on a Roebel Bar
5.3.1.1.1 Bar Characteristics
We have examined two sections of a Roebel bar, also known as a half-turn bar.2 The nominal
voltage rating of the bar is 13.8 kV. The ground-wall, or main, insulation is built up from many
half-lapped layers of a mica paper tape on a glass fibre backing pre-impregnated with an epoxy
resin. The resin is in the so-called B stage, i.e., not fully cured, when the tape is applied to the
copper conductors. Because there is more resin present than required, this is known as a
resin-rich tape. After the tape is applied, by a six-axis taping robot, a further layer of sacrificial
tape, probably made from Mylar or the equivalent, is applied. The taped bars are then placed in
an autoclave and curing and impregnation take place in an environment of asphalt.
After curing is complete, and the sacrificial layer removed, a robust semicon tape consisting of a
dispersion of carbon particles in an epoxy matrix is applied to the bar surface. The conductivity
of this tape ensures that charge cannot accumulate on the outside of the insulation and discharge
to the metal slot that contains the stator, which could lead to insulation failure. In terms of
physical parameters, the thickness of the semicon layer is around 0.15 mm; the resistivity is
functionally dependent on densification of the host matrix, which in turn depends upon
processing parameters such as how tightly the tape was wound on, properties of the shrink tape
applied on top of the semicon tape, and cure temperature. However, most manufacturers aim to
have a resistivity in the range of 5 kΣ/square to 50 kΣ/square. The figure is a trade off between
ensuring good electrical contact between bar surface and core iron, i.e., minimize resistivity, and
preventing the flow of interlaminar currents in the core iron, i.e., maximize resistivity. The
resistivity value should be fairly uniform over the length of the bar. The bar is then coated with a
silicon rubber coating to provide a conformable coating so that the bar will remain tight within
the slot. Since at least some fraction of the silicon rubber also has to be conductive, portions of it
are carbon loaded.
Prior to the SQUID tests, the insulation systems of one of the bars was subjected to a voltage
endurance test, 35 kV, 110 C to failure or at least 250 hours. As a result, the section contains a
puncture site due to electrical breakdown.
2
Information for this section was provided by Howard Sedding of Ontario Hydro.
5-7
12065638
5.3.1.1.2 SQUID Measurements
Figure 5-4 is a schematic of the measurement geometry for the two Roebel samples. Each bar
has an inner bundle of copper conductors, an insulation layer and an outer conducting layer. The
samples are placed on a motorized stage that allows the sample to be scanned beneath a SQUID
magnetometer.3 For these tests, A ‘front’ refers to the end of the bar. At the front end of the bars,
the insulation layers were bared by removing the outer conducting layer. We attached three leads
to each sample: two leads connected to each end of the inner copper (front lead #1 and rear lead
#2), and lead #3 wrapped as a single turn around the outer semicon layer near the rear end of the
bar. The dashed area indicates the mapping areas A and B, which are 240 mm by 120 mm. A
low-pass filter comprised of a 1 kΣ resistor and a 1 µF capacitor is connected in series with the
voltage source and the bar to block whatever high frequency noise emanates from the voltage
source.
(a) U n flaw ed B ar
M ap p in g area B M ap p in g area A
F ro n t R ear
lead # 1 lead # 2
co p p er in su latio n o u ter lay er lead # 3
(b ) F law ed b ar
D am ag ed
lead # 1 lead # 2
M ap p in g area B lead # 3
Figure 5-4
Schematic drawings of the configurations for the magnetic imaging of the Roebel bars.
a) The unflawed test bar without dc leakage current. b) The test bar with insulation failure
and a dc leakage current.
The bars were imaged in three ways: 1) without an applied voltage to check for magnetic
contamination; 2) with a dc voltage applied across two of the three leads to detect resistive faults
in the insulation; and 3) with an ac voltage applied across two of the three leads to detect
resistive (in-phase) and capacitive (quadrature) signals. For the ac measurements, the SQUID
output was connected to a lock-in amplifier and both the in-phase and quadrature components
were recorded. At frequencies higher than 2 kHz, the fast mode of the SQUID electronics was
3
EPRI TR-108649, Op. Cit.
5-8
12065638
used. Figures 5-5 through 5-9 present the data from the Roebel bars. Each figure includes surface
maps and image maps for both in-phase (left column) and quadrature (right column) signals,
except the figures for dc measurements.
Results on Unflawed Bar
We scanned the unflawed bar above the two areas shown by the dashed lines in Figure 5-4a. The
frequencies of the voltage applied across the two of three leads ranged from 60 Hz to 5 kHz. As
expected, larger displacement currents are seen at higher frequencies.
(1) Mapping area A above the lead #3 wrapped on the outer layer
First we measured the magnetic field above lead #3, which is wrapped on the outer semicon
layer. When a dc voltage was applied between the inner copper and the outer layer, there was no
magnetic field except the field due to minor magnetic contamination within or on the sample. To
image the charging current in the conductor, we applied 10 V ac between the inner copper and
outer semicon layer. Figure 5-5 shows the magnetic field maps when 10 V at 200 Hz was applied
to leads #2 and #3. Figure 5-6 is the field map when the same voltage was applied across leads
#1 and #3. In both Figures 5-5 and 5-6, the in-phase signal is very small and the quadrature
signal is much larger, as expected because the capacitive reactance of the undamaged insulation
is less than the resistance. The current flowed primarily between the two leads connected to the
voltage source B from the center of the image downward in Figure 5-5 and from the center
upward in Figure 5-6. The irregular distribution of the currents in the center of the image
probably indicates a non-uniform connection between the wire lead and the semicon, possibly
because of spatial variations in the conductivity of the silicon rubber coating. Most importantly,
the charging current along the bar is readily evident.
5-9
12065638
#3
#2
U n flaw ed b ar
1 0 V 2 0 0 H z v o ltag e
Figure 5-5
Magnetic field maps with 10 V at 200 Hz applied between the rear-end copper lead #2 and
the outer-layer lead #3. The in-phase signal is almost zero because there is no resistive
current between the copper and the outer layer, while the quadrature component shows an
amplitude of 50 pT as a result of a capacitive current between the copper and the outer
layer. The strongest current appears between the two leads (bottom half of the lower-right
figure).
5-10
12065638
#1
#3
U n flaw ed b ar
1 0V 2 00 H z v oltag e
Figure 5-6
Magnetic field maps with 10 V at 200 Hz applied between the inner conductor at front-end
lead #1 and the outer layer lead #3. The most of the current appears between the two leads
along the upper half of the image.
5-11
12065638
(2) Mapping Area B above the Front End
We then measured the magnetic field above area B shown in Figure 5-4b. Figure 5-7 is the
magnetic field when 10 V at 200 Hz was applied across leads #2 and #3, so that the mapping
area is not between the two leads connected to the voltage source. The charging current for the
capacitance at the upper half of the bar was readily detected. Figure 5-8 is the field map when the
same voltage was applied across leads #1 and #3, so that the mapping area is within the area
between the two leads connected to the voltage source. The quadrature current was much larger
than that in Figure 5-7 because the conductor beneath the region imaged was carrying the
charging current for all of the bar, instead of just that flowing into the far end as in Figure 5-7.
#3
#2
U n fla w e d b ar
1 0 V 2 0 0 H z v o ltag e
Figure 5-7
Magnetic images with 10V at 200Hz applied between the rear-end copper lead #2 and the
outer-layer lead #3. This demonstrates that a charging current flows along the entire
length of the bar, even far from the leads, as the capacitance is charged over the entire
dielectric.
5-12
12065638
#1
#3
U nflaw ed bar
10 V 20 0 H z v oltage
Figure 5-8
Magnetic field maps with 10V at 200Hz applied between the front copper lead #1 and the
outer-layer lead #3.
When the frequency of 2 kHz was applied, the amplitude of the magnetic signal increased by a
factor of five. In contrast to magnetic signals from the in-phase component, at 60 Hz the
quadrature signal was smaller but still detectable.
Results on Flawed Bar
We have measured magnetic field for the sample with dielectric failure (see Figure 5-4b) above
the area B (front) only. Both dc and ac current can flow through the damaged insulation layer
when the voltage is applied between the copper and the outer layer, in contrast to the unflawed
bar. Figures 5-9a and 5-9b are the magnetic maps when 5 V dc was applied between the rear
leads #2 and #3, and between the front leads #1 and #3, respectively. The data in both
Figures 5-9a and 5-9b show that the failure point divides the images into two parts: between the
breakage and the rear lead #3 and between the breakage and the front lead #1. The current
flowing between the rear lead #2 and the outer-layer lead #3 is smaller because the current flow
in the copper and the current flow in the outer layer are in the opposite direction and hence
partially cancel.
5-13
12065638
#1
#3
#3
#2
F law ed b ar w ith leak ag e
5 V d c v o ltag e
Figure 5-9
(a) Magnetic field map for the bar with dielectric failure, with 5 V dc applied between the
rear copper lead #2 and the outer-layer lead #3. The dc current can flow though the
damaged insulation layer and back to the outer layer. The small signal below the site of
failure and lead #3 is due to the cancellation of the current inside the copper and that in
the outer layer. (b) Magnetic field with 5V dc between the front copper lead #1 and the
outer-layer lead #3. The dc current can flow though the damaged insulation layer to the
outer layer without cancellation.
When an ac voltage is applied across between either #1 and #3, or #2 and #3, the field maps are
similar to the map with applied dc voltage because the leakage current is much larger than the
displacement current.
5.3.1.2 Measurements on the Parallel-Plate Planar Capacitors
In order to explore the sensitivity and spatial resolution of SQUID imaging of capacitive and
resistive fault currents flowing in an insulated stator bar, we fabricated a sample with a simple,
controlled geometry. Two aluminum strips, each 2 inch wide and 21 inch long, were placed on
either side of a 1/8 inch thick G-10 layer. Holes of various diameters in the G-10 layer were used
to simulate a reduction in the dielectric constant; the holes could be filled with insulated disks of
copper to provide regions with localized increases in capacitance. Sinusoidal voltages of 20 V at
frequencies of 6 kHz, 4 kHz, 3 kHz and 2 kHz were applied across the aluminum strips. A
3.5 kHz low-pass filter was connected in series with the voltage source. The SQUID output was
5-14
12065638
connected to a two-phase lock-in amplifier, and both the in-phase and quadrature components
were recorded.
5.3.1.2.1 Imaging of Charging Currents
The lead connections were made to either different ends of the strip (see Figure 5-10a) or the
same end (see Figure 5-10b). The displacement current flows across the G-10 layer between the
two aluminum strips such that the associated Ohmic charging current flows along the strips and
is collected by the leads attached to the end of each strip. The Ohmic current density is uniform
in the transverse direction except near the end. As can be seen in both Figures 5-10a and 5-10b,
for either strip the current is near zero at the end without the lead, and increases linearly to the
maximum at the end with the lead.
The displacement current across the G-10 layer is in the vertical direction, which does not
produce a magnetic field in the vertical direction and hence is silent to the Bz pick up coil for the
SQUID magnetometer used in these measurements. The uniform resistive current flowing
longitudinally will not produce the magnetic field in the vertical direction except near the edges
of the strip. If the leads are placed at different ends, as shown as in Figure 5-10a, the current in
the top strip and bottom strip are in the same direction. The fields due to the edges of the two
strips are added and the resultant field has the same amplitude over the entire length of the strip
length because the gradient of the current in the top strip is compensated by that in the bottom
strip. If the leads are placed at the same end of the two strips as shown as in Figure 5-10b, the
current flows in the opposite direction in the two strips and then the fields due to the edges of
two strips are almost cancel each other. The resultant field is small and has a slope with a
gradient towards to the end with the leads attached to the top strip.
5-15
12065638
a) D ifferen t en d le ad s c o n n ec tio n
b ) S am e en d le ad s c o n n ec tio n
Figure 5-10
Planar capacitor formed by two 1/8 inch thick, 2 inch wide aluminum strips and a layer of
1/8 inch G-10. (a) Different-end connection: one lead is connected to one end of the top
layer and another lead is connected to the other end of the bottom layer. (b) Same-end
connection: two leads are connected to the same end of the top and bottom layers. In both
(a) and (b), the arrows indicate the displacement current that cross the insulation layer,
and is then collected as an Ohmic conduction current by the leads attached to the
aluminum strips.
Figure 5-11 shows the magnetic field map for the two leads at the different ends (a), and at the
same end (b), respectively. The left and right columns are the in-phase and quadrature
components. The frequency is 2 kHz. Flaws of the capacitor are simulated by 3/8, and 3/16 inch
diameter holes at the center of the G-10 layer. The measured magnetic map does not reveal the
presence of the holes. When the holes were filed by a copper disk with the same diameter, we
could identify the location of the copper disk, as shown in Figure 5-12. Although the copper disk
is insulated from the two aluminum strips, it increases the capacitance between the strips at that
point because of the reduced air gap. This should provide a localized increase in the
displacement current, and an associated distortion of the conduction currents flowing in the plate.
However, the cancellation between the top and bottom strip makes detection of the flaw difficult.
We have not determined the relative contribution of the flaw signal from displacement currents
and from eddy currents induced by the current flowing in the aluminum strip.
5-16
12065638
a) D iffere nt en d lead s co n ne ction
b ) S a m e e n d lea d s co n n ec tio n
Figure 5-11
Magnetic field maps (Bz). (a) Different end connection. The magnetic field is the sum of the
top and bottom layers. (b) Same end connection. The magnetic field become smaller due
to cancellation of the fields from the current in the two strips. The left column shows the
in-phase signals, and the right column shows the quadrature signals. The hole in the
G-10 layer is not detectable.
5-17
12065638
(a) G -1 0 lay er h as a co p p er d isk
(b ) Im ag e o f a 3 /4 in ch d iam eter co p p er d isk
(c) Im ag e o f a 3 /8 in ch d iam eter co p p er d isk
Figure 5-12
The magnetic field maps for the aluminum capacitor for which the G-10 layer has a hole
that filled with (a) a 3/4" diameter copper disk, and (b) a 3/8" diameter disk. The dashed
line in (a) indicates the mapping area. The connections were on the same end.
5-18
12065638
5.3.1.2.2 Minimization of Edge Cancellation
Recognizing that cancellation of the currents in the two plates during same-end excitation would
reduce our ability to image the charging currents, we fabricated a capacitor which consists a 1/8"
thick, 2" wide and 9" long aluminum strip and a 0.3 mm thick, 6" wide and 12 long copper sheet.
A 1/8" thick G-10 layer with the same size of the aluminum strip was placed between the
aluminum strip and copper sheet, as shown in Figure 5-13. The currents in the lower plate spread
over a much larger area and hence are not effective in canceling the currents in the upper plate
(Note that in the stator bars, the rectangular geometry and other asymmetries prevent perfect
cancellation of the two currents). The G-10 layer has been divided into small portions, so that
flaws can be simulated by removing sections of the G-10. Figure 5-14 is the data for two
different leads connections: (a) and (c) the different-end leads, while (b) and (d) are for the
same-end leads. Figures 5-14a and 5-14b are for the unflawed capacitor. Figures 5-14c and 5-14d
are for a capacitor that is missing a 2 × 2 inch square of G-10 at the middle of the strip. The
arrows in Figures 5-14c and 5-14d indicate the location of the missing portion. The small
difference between the flawed capacitor and the unflawed is visible in the traces recorded at the
edges.
C o p p er sh eet A lum in um b ar
G -1 0
Figure 5-13
The unequal-plate capacitor formed by a copper-clad G-10 plate and an aluminum strip.
The copper is a 0.3 mm thick, 6 inch wide, and 12 inch long sheet. An 1/8 inch thick, 2 inch
wide 9 inch long aluminum bar is placed on the copper sheet. An 1/8 inch thick G-10 layer
that has been divided into several sections is placed between them. The leads are copper
wires attached to the copper sheet or aluminum strip over the width of the conductor. The
current densities in the two conductors are different, so the field from the edges of the
copper does not cancel the field from the edges of the aluminum strip.
5-19
12065638
a) b)
A B
B
BZ BZ
X (m m ) X (m m )
c) d)
C D E F
BZ BZ
C D E F
X (m m ) X (m m )
Figure 5-14
The data from the unequal-plate capacitor. The left column ((a) and (c)) is for the
different-end lead connection, and right column ((b) and (d)) is for the same-end
connection. (a) and (b) are the data from the unflawed capacitor, and (c) and (d) are the
data from the capacitor missing a 2 × 2 inch square of G-10 in the middle. Each line is a
scan of the magnetic field along a line parallel to the axis of the aluminum strip. The fields
become larger toward to the leads. The arrows indicate the location of the missing
portions; differences in the slope of the signals can be seen between the flawed and
unflawed cases.
5-20
12065638
To clarify the difference due to the missing portion, we processed the data by applying suitable
spatial filters. Figure 5-15 shows the imaging processing steps, where the left column is for the
unflawed capacitor, and the right column is for the capacitor with 2 × 2 inch portion missing.
The raw data has been averaged to reduce the noise (see (a) - (d)). A spatial filter is applied to
the data to obtain the gradient along the longitudinal direction (e) and (f). Then the gradient is
filtered by a spatial low pass filter (g) and (h), so that the discontinuity of the gradient shows the
location of the missing portion.
This arrangement can readily detect a one-inch segment of G-10 missing at the edge of the
capacitor, but does not show as clearly a missing segment that is far from the edge, since the
missing portion away from the edge does not affect the field near the edges very much.
5.3.1.2.3 Imaging Plates
One interpretation of the preceding measurements is that the Ohmic currents flowing in the bulk
of either plate are magnetically silent, and that the magnetic signals arise solely from the
discontinuity in the tangential current at the edge of the plate. The use of the unequal plate
capacitor separated the edges of the upper and lower plates so that the cancellation of their edge
contributions would not be as complete, and the resulting signals were larger, as expected. To
enhance further the sensitivity of the measurement to flaws near the center of the strip, we
modified the narrow aluminum strip by adding more edges at the center in the form of thin,
longitudinal slots every 1/4 inch, as shown in Figure 5-16. The slots are through the one end of
the strip, like a fork, so that the current can not flow across the slots. A copper wire is attached to
another end of the aluminum strip. The current is restricted to the longitudinal direction except
near the leads. Because the slots are narrow, they do not significantly reduce the capacitance
between the two plates. In Figure 5-16, the left column shows the location of the missing G-10.
The middle column has the images processed by a longitudinal derivative filter and a low pass
filter. The missing potion can be identified. The right column is the image after a transverse
gradient filter is applied to the images in the middle column. The missing 2 by 1 inch segment of
G-10 at the center of the sample is readily detected.
As an alternative to the slotted plate, an aluminum strip was fabricated with 30 flat bottom holes,
each of which is 1/4 inch diameter and 1/16 inch deep in one surface. Figure 5-17 shows the
images for the capacitors with different missing portions. The signal is noisier than the signal
from the slotted aluminum strip, possibly because of the more complicated distortion of the
current distribution by the holes, or because the holes go only one-half the way through the plate.
The right column shows the flaw through the edges. The transverse derivative filter is not as
effective as for the slotted plate; an appropriate filter needs to be identified.
5-21
12065638
U n fla w e d G -1 0 F la w ed G -1 0
a ) R aw d ata b ) R aw d ata
c ) R aw d ata d ) R a w d a ta
e ) D eriv a tiv e filte rin g f) D e riv ativ e filterin g
g ) L o w p ass filte rin g h ) L o w p ass filte rin g
Figure 5-15
Field map for the unequal-plate capacitor. The left column is for the capacitor without
flaws, and the right column is for the capacitor with the missing 2 by 2 inch portion at the
middle. (a-d) Surface maps and the images of the fields. (e-f) Images after derivative
filtering. (g-h) Images after low pass filtering. The missing portion of G-10 layer is evident
in (h).
5-22
12065638
a) L o catio n o f flaw D o u b le filterin g Trip le filterin g
b)
c)
d)
e)
Figure 5-16
Modified unequal-plate capacitor. The upper aluminum strip is slotted every 1/4 inch. One
end is cut through, and another end is connected to the voltage supply. The left column
also shows the missing portions of the G-10 layer. The middle column is the images after
the derivative and the low pass filtering. The right column is the images after applying a
transverse derivative filter to the images in the middle column. The location of the missing
G-10 is clear. The smallest portion at the middle of the capacitor is 1/2 × 1 inch, as shown
in (e).
5-23
12065638
L o catio n o f th e flaw D o u b le filterin g
Figure 5-17
A modified unequal-plate capacitor that has flat-bottomed holes on the upper surface of
the aluminum strip. The images at the right column are obtained by longitudinal derivative
and low pass filtering. The missing portions that touch the edge are evident; the interior
flaws should become visible with appropriate digital filtering.
5-24
12065638
5.3.2 Stage 2
5.3.2.1 Sample Preparation
After the preliminary studies described above, we turned to the final tests on Roebel bars. We
received six test samples in three shipments from Ontario Hydro, as shown in Table 5-1. Each
shipment included two test segments cut from the same stator winding, one of which had a
painted end, as shown in Figure 5-18.
Table 5-1
Bar Receipt Date, Age, and ID
2/6/98 (61T29) no information about age Bar 1-1, Bar 1-2
11/20/98 (61T420) 1000 cyc., 40°C-140°C-40°C --- 1.5h Bar 2-1, Bar 2-2
4/27/99 (61T33) 100 cyc., 40°C-140°C-40°C --- 1.5h, Bar 3-1, Bar 3-2
1000 cyc., 40°C-155°C-40°C --- 1.5h
a) M app in g area: 360 x 1 20 m m
A ll c o p p e r strip s are
co n n ecte d
b)
G ray p ain t
Figure 5-18
Schematic representation of the samples.
Each bar consists of a bundle of 30 copper conductors that are separated by a thin layer
of insulation. The copper bundle is surrounded by a 5 mm thick insulation layer, wrapped
by the electrically-conducting semicon tape. The wider surfaces of the bar are covered
with a silicon-rubber coating that is not electrically conducting. The sides of the bar are
usually rubber-free and the semicon tape is exposed.
5-25
12065638
We connected three copper wire electrodes to each bar: one was wrapped and soldered to one
end of all the individual conductors, and one to the black semicon tape at each of the two ends of
the bar, as shown in Figure 5-18.
5.3.2.2 Measurement Procedure
The first measurement of Bar 1-1 shows the internal twisting of the copper conductors, which
was not evident from a casual visual inspection of the bar. To confirm the orientation of the
copper conductors, we imaged one copper conductor by connecting the first electrode to a single
conductor rather than the entire bundle. Figure 5-19 shows the magnetic images of the capacitive
current for one copper conductor. The dashed lines indicate the position of the bar relative to the
image. It is clear that the copper conductor does twist.
a) R aw d ata b ) F iltere d d ata

Bx F iltere d B x
By F iltere d B y
Bz F iltere d B z
Figure 5-19
Magnetic images for Bar 1-1. The ac voltage is applied to the semicon and the one of the
30 copper conductors. (a) Raw data. (b) Filtered data. The one copper conductor is
oriented diagonally and twisted, as indicated by the arrow.
Because any inhomogeneities of the copper would complicate the magnetic image, we then
tested the electrical properties of the bars before magnetic imaging using the following steps:
1. Check the copper conductors by ohm meter to find how the bar twists.
2. Measure the semicon resistance.
3. Solder all copper conductors to a common wire electrode at one end.
4. Apply a dc voltage through the semicon electrodes and measure the current.
5-26
12065638
5. Apply an ac voltage to the two semicon electrodes and measure the current through the
semicon tape as a function of frequency.
6. Apply an ac voltage between the copper and the semicon, and measure the current through
the insulation layer as a function of frequency
We then acquired the magnetic images as follows:
1. Demagnetize each sample inside the magnetic shield.
2. Scan the background magnetic field profile of the sample without applying a current.
3. Inject a dc current into the semicon layer to characterize the homogeneity of the semicon
tape.
4. Apply 10 volts at 5.2 kHz between the two ends of the semicon tape to induce out-of-phase
currents inside the copper conductors to characterize any homogeneities in the copper.
Record both the in-phase and quadrature components.
5. Apply 10 V at 5.2 kHz ac between the copper bundle and the semicon tape to image the
capacitive current. Record both the in-phase and the quadrature components.
5-27
12065638
5.3.2.3 Electrical Properties
We have examined the twist in two samples, Bar 1-1 and Bar 2-1, which have similar
configuration. Figure 5-20 shows the direction of each copper conductor; however, the internal
structure could not be determined electrically. Because the insulation layers between the copper
conductors of the Bar 3-1 and Bar 3-2 are so bad, almost all the copper conductors were
connected each other, and it was impossible to determine the bar twists for Bar 3.
B ar1 -1
L 4 R
B ar2 -1
L R
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
22 21 20 19 18 17 16 1 2 3 4 5 6 7 8
Figure 5-20
A schematic illustration of the connection of the copper conductors between the two ends
of the Bar 1-1 and Bar 2-1. The structure inside the bar (dashed lines) is unknown.
The electric properties of the six bars are shown in Table 5-2. The insulation layer between the
copper conductors of the Bar 3 is so thin, that almost all the copper conductors are connected
each other electrically through the adjacent conductors.
5-28
12065638
Table 5-2
Electrical Properties
Sample Insulation Between Conductors Resistance of Semicon

Tape (Ohm/mm)
Bar 1-1 All the conductors were insulated 34
Bar 1-2 All the conductors were insulated 60
Bar 2-1 A few conductors were connected 160
Bar 2-2 A few conductors were connected 234
Bar 3-1 Most of conductors were connected 235
Bar 3-2 Most of conductors were connected 411
The resistance of the semicon tape for each sample is listed in Table 5-2. The Bar 3 has the worst
insulation between the copper conductors and the largest semicon resistance, both of which may
be due to the aging.
When a 10 V ac voltage was applied between the two electrodes on the surface of the semicon
layer, the current in the semicon induced an out-of-phase current inside the copper conductors.
The associated reactance altered the magnitude and phase of the current in the semicon. A
100 ohm resistor, which is much smaller than the resistance of the semicon tape, was serially
connected to the circuit and the magnitude and the phase of the voltage across the resister was
recorded by a lock-in amplifier. Then the voltage across the resistor would reflect the total
current through the electrode.
Figure 5-21 shows the magnitude and the phase of the current versus frequency for Bar 1-1,
Bar 2-1 and Bar 3-1. At very low frequencies (f ~ 0), the current is almost same as that due to a
dc voltage. At such low frequencies, the inductance is small and the resistive component
dominates the current. Because the semicon resistance of Bar 1 is smaller than that of Bar 2 or
Bar 3, the current for Bar 1 is larger at the lower frequencies, as shown in Figure 5-21a, and the
phase of the current for the Bar 1 is smaller as shown in Figure 5-21b. At high frequencies the
inductance and the capacitance would dominate. Figure 5-21a shows that the current increases
linearly with the frequency, probably indicating that the dielectric capacitance is connecting the
low-resistance copper in parallel with the higher resistance semicon.
5-29
12065638
a) M ag n itu d e
3 .0
S u rface cu rrent (m A )
2 .0
V ~= 1 0 v
1 .0 # 1 -1
# 2 -1
# 3 -1
0
0 20 40 60 80 100
F req u en cy ( k H z)
b ) P h ase
100
P h ase of cu rren t (d eg ree)
80
60
40
# 1 -1
# 2 -1
20 # 3 -1
0
0 20 40 60 80 100
F req u en cy ( k H z)
Figure 5-21
The magnitude and the phase of the semicon current versus frequency when a 10 V ac
voltage is applied to the semicon electrodes. The semicon current induces the
out-of-phase current inside the copper. (a) Magnitude. (b) Phase. The black lines are for
the Bar 1-1, which are different from the others.
Figure 5-22 is the result of the similar measurement for the displacement current when a 10 V ac
voltage is applied across the copper and the semicon (both electrodes are at the same end of the
sample). At very low frequencies (f ~ 0), the current is very small, as shown in Figure 5-22a, and
the phase of the total current is almost 90°, as shown in Figure 5-22b. At high frequencies, the
current becomes linear with frequency, similar to the semicon current in Figure 5-21a. We
wonder whether the differences between the three bars in Figure 5-22b may be the result of
aging.
5-30
12065638
a) M ag n itu d e
3 .0
C ap acitiv e cu rren t (m A )
2 .0
1 .0 # 1 -1
# 2 -1
# 3 -1
V ~= 1 0 v
0
0 20 40 60 80 100
F req u ency (k H z)
b ) P h ase
100
P h ase of cu rren t (d eg ree)
90
80
# 1 -1
# 2 -1
# 3 -1
70
0 20 40 60 80 100
F req u ency (k H z)
Figure 5-22
The magnitude and the phase of the capacitive current versus the frequency when a
10 V ac voltage is applied to the copper and the semicon. The current is due to the
displacement current across the insulation layer. (a) Magnitude. (b) Phase. The black lines
are for the Bar 1-1, which are different from others.
At 5.2 kHz, the frequency used for magnetic imaging, the magnitude and the phase of the current
through the electrodes are listed in Table 5-3. Both the semicon current and the current across the
semicon and the copper have a phase difference with respect to the applied voltage, which is
expected because of the non-zero capacitance and inductance.
5-31
12065638
Table 5-3
The Magnitude and the Phase of the Current at 5.2 kHz
Sample Semicon Current (Voltage Applied Capacitive Current (Voltage Applied

on the Surface) Across the Semicon and the Copper)
Magnitude (mA) Phase (degree) Magnitude (mA) Phase (degree)
Bar 1-1 0.24 35° 0.23 86°
Bar 2-1 0.14 73° 0.19 79°
Bar 3-1 0.14 79° 0.17 78°
In Figures 5-21, 5-22 and Table 5-3, the data for Bar 1 may be distinguished from the Bar 2 and
Bar 3. We believe that Bar 1 was the least aged as compared with the data from Bar 2 and Bar 3
that were thermally aged. For the detailed analysis, the frequency and age dependence of each
parameter, such as the resistance of the semicon and the dielectric constant of the insulation
layer, need to be studied separately.
5.3.2.4 Magnetic Imaging
We have imaged all six samples that we received. Because the thicker insulation layer and
irregular geometry of the section with the gray paint (Bar 1-2, Bar 2-2 and Bar 3-2) causes a
complicated current distribution, we present the images for the sections without the paint
(Bar 1-1, Bar 2-1 and Bar 3-1) only for comparison between differences in aging.
For better signal-to-noise ratio, we used 10 V at 5.2 kHz for the ac measurement, which is the
upper frequency limit of the SQUID gradiometer.
The vector SQUID gradiometer can measure three components of the magnetic field. First of all,
we define the x direction as the longitudinal direction of the bar for all the images. Because the
electrodes are attached at the end of the bar, the currents flowed mainly in the x direction, which
produces a large signal in the By and Bz images. If the currents are perfectly in the x direction
(along the bar), there is no signal in the image of Bx. In general, a non-zero signal in Bx may be
due to the imperfect alignment of the sample (or the current) with the x coil of the SQUID, or the
presence of a non-zero transverse or vertical current. The non-alignment with the x-coil produces
a small signal which should be similar to that of By. Otherwise the signal of Bx should be
different from By and Bz. Thus for our measurements, the Bx image is valuable for monitoring
any transverse current due to an inhomogeneity in the sample because the large longitudinal
current is magnetically silent in Bx.
The magnetic field is produced not only by the current flowing in the semicon, but also by the
current in copper conductors as well as through the insulation layer. The phase of the magnetic
signal is related to the phase of the corresponding current distribution. We have recorded both
the in-phase and quadrature components of the signals (relative to the applied voltage). We are
5-32
12065638
able to obtain the magnetic images corresponding to the current distribution at any desired phase
from the recorded in-phase and quadrature components.
As an example, we present the phase-related images in Figures 5-23 and 5-24. The images are
the capacitive current for Bar 1 with four copper conductors disconnected from the electrodes.
Figure 5-23 is the raw data and Figure 5-24 is the filtered data. Figure 5-23 shows that the
amplitude of the signal, indicated by the relative color bar, varies with the phase angle. The
phase at which the signal has the maximum amplitude is defined as the phase of the image,
which is different from the phase for the semicon current shown in Table 3.
In Figure 5-23, the phase of Bx is 40° and the phases for By and Bz are both 10°. The phase of the
transverse or the vertical current is different from that for the longitudinal current.
In Figure 5-24, the filtered version of the images in Figure 5-23, the Bx images clearly show the
locations where the copper conductors have their greatest twist, as indicated by the black arrows.
This is consistent with the By and Bz images.
5-33
12065638
Bx o o o o o o o o
20 40 60 80 10 0 12 0 14 0 16 0
1 0.5 0.5 1
1 1 0.5 0.5
0 0 0
0 0 0 0
0
-0.5
-0.5 -1
-1 -1
-1 -0.5 -0.5
By
3
4 4 1
0.5
2 1
2
0 0
0 0 0 0 0 0
-1
-2 -2 -1
-2 -2 -1 -1
Bz
5 5 4 2 2
4
2 1
0
0 0 0 0 0 0 0
-2 -1 -2
-4 -4 -1
-5 -5
Figure 5-23
The images at phases of 20° through 160° for Bar 1-1, in which four copper conductors are
not connected to the electrodes. The color bars indicate the amplitude of the signals,
which vary with the phase angles. The phases at which the signal is maximum are 10° for
Bx, 50° for By and Bz. The Bz images show that the copper conductors twist.
5-34
12065638
20 40 60 80 100 120 140 160
Bx
By
Bz
Figure 5-24
The filtered images at phases of 20° through 160° for Bar 1-1, in which four copper
conductors are not connected to the electrodes. The images show where the copper
conductors twist, as indicated by arrows.
The remaining images (Figures 5-25 through 5-39) are for three samples, Bar 1-1, Bar 2-1 and
Bar 3-1. For each sample, there are three sets of images: 1) the images for dc current applied to
the semicon layer to characterize the semicon layer; 2) the images for a 10 V, 5.2 kHz voltage
applied to the two electrodes on the semicon layer, and 3) the images for a 10 V, 5.2 kHz voltage
applied between the electrode on the copper and one on the semicon, i.e., the capacitive current.
Each set of images includes the three components of the magnetic field at the two phases for
which the amplitude of the signal is either maximum or minimum. The filtered images are also
presented, which show more details about the field distribution.
Figures 5-25, 5-30 and 5-35 are the images when a dc current is injected to the semicon layer of
Bars 1-1, 2-1 and 3-1, respectively. The background due to the magnetic contamination has been
subtracted for each image. The left columns are the raw data for the three components of the
magnetic field. The spatially filtered images are shown in the right columns. Evidently, the
5-35
12065638
semicon layers are not homogeneous for all three samples, especially evident in the images of
Bx. It is clear that the semicon layers of Bar 2-1 and Bar 3-1 are much less homogeneous than
that of Bar 1-1. The irregular By and Bz images for Bar 2-1 (Figure 5-30) indicate that the
semicon tape is corrupted, possibly due to aging.
Figures 5-26, 5-31 and 5-36 are the raw images when an ac voltage is applied to the semicon
layer for samples Bar 1-1, Bar 2-1 and Bar 3-1, respectively. The filtered images are in
Figures 5-27, 5-32 and 5-37. In all the figures, the images in the left and right columns are
related to the phase at which the signal is maximum (phase –40°) and minimum (phase 50°),
respectively. The semicon current induces the out-of-phase current inside the copper conductors,
with a phase that is different from the phase of the semicon current. For Bar 1-1 and Bar 1-3, the
images at the left column in Figures 5-26 – 5-27 and 5-36 – 5-37 are similar to the dc images in
Figures 5-25 and 5-35, respectively. For Bar 2-1, the current in the images at the left column in
Figure 5-31 – 5-32 are smoother than the dc images in Figure 5-30, which may be the result of
the out-of-phase current either smoothing the semicon current or dominating it. Because of the
phase, the right column images for all three samples is determined by the current distribution in
the copper, but it does not show as significant structural details as might have been expected.
Figures 5-28, 5-33 and 5-38 are the raw images when an ac voltage is applied across the copper
and the semicon layer for the samples Bar 1-1, Bar 2-1 and Bar 3-1, respectively. The filtered
images are in Figures 5-29, 5-34 and 5-39. The displacement currents through the insulation
layer are collected by the electrodes. The images at the left column are more likely the same as
the images of the semicon current, while the images at the right column show the structure of the
copper conductors.
Because the highly irregular semicon layer of Bar 2-1, it is difficult to compare its images with
others. Now we compare the images of Bar 1-1 (Figure 5-29) with the images of Bar 3-1
(Figure 5-39). In both figures the right columns show that the copper conductors have similar
structures. However, the Bar 3-1 (Figure 5-39) shows more irregularity than that for Bar 1-1.
5-36
12065638
a) R a w d ata b ) F iltered d ata

Bx Bx
0 0 .1 0 .2 0 .3 0 .4 nT
By By
0 1 2 nT
Bz Bz
0 1 2 3 4 nT
B ar1 -1
Figure 5-25
The images of the injected dc semicon current for Bar 1-1. (a) Raw data. (b) Filtered data.
5-37
12065638
a) P h ase -4 0 b ) P h ase 5 0
Bx Bx
-2 0 -1 0 0 pT 0.5 0 1 .5 p T
By By
-1 0 0 0 600 pT -4 0 0 10 pT
Bz Bz
-6 0 0 0 400 pT 0 40 pT
Figure 5-26
Magnetic images (raw data) for Bar 1-1 when 10 V 5.2 kHz is applied to the two semicon
electrodes. The current includes the semicon current and the induced out-of-phase
current in the copper. (a) At –40° phase (maximum amplitude). (b) At phase 50° (minimum
amplitude).
5-38
12065638
a) P h a se -4 0 b ) P h ase 5 0
Bx Bx
By By
Bz Bz
Figure 5-27
Magnetic images (filtered data) for Bar 1-1 when 10 V 5.2 kHz is applied to the two semicon
current in the copper. (a) At -40° phase (maximum amplitude). (b) At the phase 50°
(minimum amplitude).
5-39
12065638
a) P h a se 3 0 b ) P h a se 1 2 0
Bx Bx
-5 0 0 pT -5 0 2 pT
By By
-5 0 0 300 pT -5 0 35 pT
Bz
Bz
-3 0 0 0 200 pT -2 0 0 30 pT
B ar1-1
V~
Figure 5-28
Magnetic images (raw data) for Bar 1-1 when 10 V 5.2 kHz is applied to the copper and the
semicon layer. The capacitive current crosses the insulation layer and is then collected by
the leads attached to the copper and semicon. (a) At the phase 30° (maximum amplitude).
(b) At 120° phase (minimum amplitude).
5-40
12065638
a) P h a se 3 0 b ) P h ase 1 2 0
Bx Bx
By By
Bz Bz
B a r1 -1
V~
Figure 5-29
Magnetic images (filtered data) for Bar 1-1 when 10 V 5.2 kHz applied to the copper and the
the leads attached to the copper and semicon. (a) At the phase 30° (maximum amplitude).
5-41
12065638
a) R aw d ata b ) F iltered d ata

Bx Bx
0 0.2 n T
By By
0 0.45 nT
Bz Bz
0 0.7 n T
B ar2 -1
Figure 5-30
5-42
12065638
a ) P h ase -4 0 b ) P h ase 5 0
Bx Bx
--2.5 0 2 pT -0.8 0 0.6 p T
By By
-20 0 12 0 p T -5 0 20 pT
Bz Bz
-10 0 0 10 0 p T 0
Figure 5-31
current in the copper. (a) At -40° phase (maximum amplitude). (b) At 50° phase (minimum
amplitude).
5-43
12065638
a) P hase -4 0 b ) P h ase 5 0
Bx Bx
By By
Bz Bz
Figure 5-32
amplitude).
5-44
12065638
a) P h ase 0 b ) P h ase 90
Bx Bx
-4 0 8 pT -1 0 2 pT
By By
-3 0 0 50 pT -6 0 6 pT
Bz Bz
-4 0 0 60 pT -4 0 14 pT
B ar2 -1
V~
Figure 5-33
the leads attached to the copper and semicon. (a) At 30° phase (maximum amplitude).
5-45
12065638
a) P h ase 0 b ) P h ase 9 0
Bx Bx
By By
Bz Bz
B ar2 -1
V~
Figure 5-34
the leads attached to the copper and semicon. (a) At 30° phase (maximum amplitude.
5-46
12065638
a) R aw data b ) F iltered d ata

Bx Bx
0 0 .1 2 n T
By By
0 0 .3 5 n T
Bz Bz
0 0 .6 n T
B ar3 -1
Figure 5-35
5-47
12065638
a) P h a se -4 0 b ) P h ase 5 0
Bx Bx
-2 0 1 .5 p T -0 .6 0 0 .4 n T
By By
-1 0 0 60 pT 15 pT
Bz Bz
-8 0 0 80 pT 0 15 pT
Figure 5-36
amplitude).
5-48
12065638
a) P h ase -4 0 b ) P h ase 5 0
Bx Bx
By By
Bz Bz
Figure 5-37
amplitude).
5-49
12065638
Bx Bx
-20 0 10 p T -4 0 4 pT
By By
0 15 0 pT 0 20 p T
Bz Bz
-150 0 15 0 pT -10 0 20 p T
B ar3 -1
V~
Figure 5-38
the leads attached to the copper and semicon. (a) At 30° phase (maximum amplitude).
5-50
12065638
Bx Bx
By By
Bz Bz
B ar3-1
V~
Figure 5-39
the leads attached to the copper and semicon. (a) At 30° phase (maximum amplitude.
5.4 Conclusions
1. The measurements on the stator bars in Stage 1 demonstrate that it is possible to image the
currents flowing in a stator bar that result from either resistive (through dielectric
breakdown) or capacitive current flowing across the insulation layer. The SQUID has
adequate sensitivity to image the distribution of capacitive charging currents even at 60 Hz,
although the signal-to-noise ratio can be improved by working at higher frequencies. The
pathway of the resistive current during the dielectric breakdown is clearly shown in the
images. We saw evidence of geometric cancellation arising from the coaxial nature of the
conductor and its surrounding semicon layer, but the cancellation was far from perfect as
would be expected for a homogeneous, axially symmetric cylindrical cable.
2. The measurements on the parallel plate capacitor demonstrate that we can image
magnetically the distribution of charging currents and the perturbations in this distribution
that result from localized changes in dielectric constant. The measurements also
demonstrated that conductor symmetries can reduce the size of the observed signals, but this
5-51
12065638
can be corrected in part by using a capacitor plate that is either slotted or contains an array of
blind holes.
3. The measurements on the sample segments from the three stator bars with different aging in
stage 2 demonstrate that we can image the current distribution due to the displacement
current across the inner copper and the semicon layer. The imaging of the three components
of the magnetic field provides us more information about the current distribution. The x
component of the field, Bx, is particularly useful because the longitudinal current (x-
direction) is silent, whereas the signal due to the transverse or vertical current that is the
result of the inhomogeneities in the sample is enhanced.
Our data also demonstrates that we can image the homogeneity of the semicon tape as well as
the copper conductors inside of the insulation. The results show that the semicon layers of the
samples are not homogeneous, especially for the aged Bar 2 and Bar 3. The images also show
that the inner copper conductors are twisted.
Combining the electric and the magnetic measurements, we conclude that Bar 1 is least aged.
Bars 2 is more aged and bar 3 is most aged, which are in good agreement with our
information about the aging process. The semicon tape becomes more resistive and more
inhomogeneous with the thermal aging. The insulation layer between the copper conductors
degenerated with the thermal aging process.
4. Unfortunately, the results show that the field variation due to the irregularities of the semicon
layer and the twisted inner copper conductors are superimposed to the field due to the
variation of the displacement current. In general, we have shown that the use of a two-phase
lock-in amplifier is useful for extracting the phase information, in that by choosing an
appropriate phase angle, we can separate the signal due to the semicon from the signal due to
the copper. However, we could not obtain isolated samples of the material used for the
insulation layer for this study, and we lack the knowledge of how the dielectric constant of
the insulation layer changes with frequencies and aging. It is impossible for us to estimate
quantitatively how much the displacement current would change relative to the semicon
variation or the twisted copper. For these reasons, it is difficult to extract the signal that
would be due to any inhomogeneities in the insulation layer.
While we had hoped to compare the images taken at Vanderbilt with those obtained at
Ontario Hydro and Vintek, so as to ascertain whether magnetic imaging can identify changes
in the properties of a bar that are associated with dielectric aging, we were never provided
with these data.
5. Due to the large irregularities of the semicon and the copper, it is necessary to find out if the
signal due to the variation of the insulation material is large enough to overcome the signal
due to the inhomigeneities. For the fundamental understanding how the dielectric constant of
the insulation layer vary with the aging and how much the current on the semicon and the
copper would, the insulation material needs to be characterized without the semicon or
copper.
5-52
12065638
6
CONCLUSIONS
A number of high voltage stator bars, removed from a hydraulic generator were subject to cyclic
thermal aging as part of an effort to develop effective tools to detect the effects of thermal stress
on the insulation system. The main findings of this component of the study were,
• The most significant change in diagnostic parameters occurred within the first 250 thermal
cycles for the service-aged bars, after this point little change was observed up to
2000 thermal cycles.
• The change in power frequency dissipation factor was more pronounced than either that for
PD magnitude or DEV.
• Very little change in the diagnostic parameters was observed for the spare bar as a function
of thermal cycling.
• Despite the changes observed in, e.g., dissipation factor, how much, if any aging of the bars
was effected by thermal cycling cannot, at this time, be determined.
• Dielectric spectroscopy measurements on the bars provided little further information on the
condition of the insulation.
• SQUID measurements demonstrate that localized changes in dielectric constant can be
detected. The results rank the aged bars correctly but inherent inhomogeneities of the
materials (semicon and conductors) limit sensitivity of the technique. More work on the
dielectric materials excluding semicon and conductors is required to further enhance the
technique.
6-1
12065638
12065638
7
REFERENCES
1. D.L. Evans, “IEEE working group report of problems with hydrogenerator thermoset stator
windings, part I”, IEEE Trans. Power App. & Systems, PAS-100, 1981, pp. 3284 – 91.
2. P. Mighdoll, R.P. Bloss and F. Hayashi, “Improved Motors for Utility Applications, Vol. 2:
Industry Assessment Study”, EPRI Report EL-2678, 1982.
3. G.C. Stone, B.K. Gupta, J.F. Lyles and H.G. Sedding, “Experience with accelerated aging
tests on stator bars and coils”, IEEE International Symposium on Electrical Insulation,
Toronto, Canada, 1990, pp. 356 – 60.
4. IEEE Standard 1310-1995, “Recommended Practice for Thermal Cycle Testing of Form
Wound Stator Bars and Coils for Large Generators”, IEEE Standards Dept., Piscataway, NJ,
1995.
5. P.J. Hyde, “A wide frequency range dielectric spectrometer”, Proc. IEE, vol. 117, 1970,
pp. 1891 – 1901.
6. F.I. Mopsik, “A precision time domain dielectric spectrometer”, Rev. Sci. Instrum., vol. 55,
1984, pp. 79 – 87.
7. J.M. Braun, R.M. Morra and H.G. Sedding, “Assessment of cable insulation by low-
frequency dielectric characterization”, Conference on Electrical Insulation and Dielectric
Phenomena, Knoxville, TN,1991, pp. 408 – 13.
8. J.P. Steiner, J.P. Quinlan, J.M. Braun, R.M. Morra and H.G. Sedding, “A new low frequency
high voltage insulation analyzer using transient methods”, Conference on Electrical
Insulation and Dielectric Phenomena, Victoria, Canada, 1992, pp. 328 – 33.
9. A. Helgeson and U. Gafvert, “Dielectric response during curing of a resin-rich insulation

system for rotating”, Conference on Electrical Insulation and Dielectric Phenomena, Austin,
TX, 1999, pp. 289 – 92.
7-1
12065638
12065638
12065638
Target: About EPRI
Steam Turbines, Generators, and Balance of Plant EPRI creates science and technology solutions for
the global energy and energy services industry. U.S.
electric utilities established the Electric Power
Research Institute in 1973 as a nonprofit research
consortium for the benefit of utility members, their
customers, and society. Now known simply as EPRI,
the company provides a wide range of innovative
products and services to more than 1000 energy-
related organizations in 40 countries. EPRI’s
multidisciplinary team of scientists and engineers
draws on a worldwide network of technical and
business expertise to help solve today’s toughest
energy and environmental problems.
EPRI. Powering Progress
© 2000 Electric Power Research Institute (EPRI), Inc. All rights

reserved. Electric Power Research Institute and EPRI are registered
service marks of the Electric Power Research Institute, Inc.
EPRI. POWERING PROGRESS is a service mark of the Electric
Power Research Institute, Inc.
Printed on recycled paper in the United States of America
1000376
EPRI • 3412 Hillview Avenue, Palo Alto, California 94304 • PO Box 10412, Palo Alto, California 94303 • USA
800.313.3774 • 650.855.2121 • [email protected] • www.epri.com
12065638

Testing of Stator Windings For Thermal Aging PDF

Uploaded by

Copyright:

Available Formats

Testing of Stator Windings For Thermal Aging PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Testing of Stator Windings For Thermal Aging PDF

Uploaded by

Copyright:

Available Formats

Testing of Stator Windings for Thermal

Final Report, August 2000

EPRI Project Manager

(A) MAKES ANY WARRANTY OR REPRESENTATION WHATSOEVER, EXPRESS OR IMPLIED, (I)

(B) ASSUMES RESPONSIBILITY FOR ANY DAMAGES OR OTHER LIABILITY WHATSOEVER

ORGANIZATION(S) THAT PREPARED THIS DOCUMENT

Ontario Power Technologies

This report was prepared by

Ontario Power Technologies

This report describes research sponsored by EPRI.

1 INTRODUCTION ................................................................................................................. 1-1

2 SELECTION OF SUITABLE STATOR BARS ..................................................................... 2-1

3 INITIAL CHARACTERIZATION AND AGING OF STATOR BARS ..................................... 3-1

4 DIAGNOSTIC CHARACTERIZATION ................................................................................. 4-1

5 SQUID MAPPING RESULTS .............................................................................................. 5-1

6 CONCLUSIONS .................................................................................................................. 6-1

7 REFERENCES .................................................................................................................... 7-1

Pressures on utilities to operate more efficiently, and in particular to minimize unnecessary

The well established techniques include:

The non-conventional techniques include:

Dielectric spectroscopy is a term generically applied to a number of techniques [3,4] in which

Test Results - Power Frequency Diagnostics

61B 493 4.1 3.8 3.6 3.4 22 20

Test Results - Dielectric Spectroscopy

Thermal Frequency Voltage Tangent δ 60 Hz Loss

1000 60 2 0.566 0.544

Thermal Frequency Voltage Tangent δ 60 Hz Loss

Test Results - UHF Dielectric Constant Mapping

Test Results - SQUID Mapping

SQUID results are presented in Chapter 5.

5.1.1 Subject of the Work

5.1.2 Previous Work

5.1.3 Current Work

This work has been performed in two stages:

The study in Stage 1 includes measurements on large-scale parallel-plate capacitors, and

5.2.1 Magnetic Imaging of Cable Currents

5.2.2 Parallel-Plate Capacitor

The spatially-extended capacitor may be visualized as a number of small capacitors, whose

5.2.3 Magnetic Imaging of the Current

b ) C u rren t in th e x-z p lan e BX BY BZ

5.3 Measurements and Results

5.3.1.1 Measurements on a Roebel Bar

5.3.1.1.1 Bar Characteristics

5.3.1.1.2 SQUID Measurements

co p p er in su latio n o u ter lay er lead # 3

Results on Unflawed Bar

(2) Mapping Area B above the Front End

Results on Flawed Bar

5.3.1.2 Measurements on the Parallel-Plate Planar Capacitors

5.3.1.2.1 Imaging of Charging Currents

a) D iffere nt en d lead s co n ne ction

(a) G -1 0 lay er h as a co p p er d isk

(b ) Im ag e o f a 3 /4 in ch d iam eter co p p er d isk

(c) Im ag e o f a 3 /8 in ch d iam eter co p p er d isk

5.3.1.2.2 Minimization of Edge Cancellation

5.3.1.2.3 Imaging Plates

e ) D eriv a tiv e filte rin g f) D e riv ativ e filterin g

g ) L o w p ass filte rin g h ) L o w p ass filte rin g

a) L o catio n o f flaw D o u b le filterin g Trip le filterin g