21, rue d’Artois, F-75008 PARIS A1-203 CIGRE 2016

http : //www.cigre.org

Assessment of manufacture quality of stator windings insulation of rotating

machines by means of macrographic analysis

R. BAENA, S. RODRÍGUEZ, A. VILLARRUBIA

B.REMARTINEZ, O.MARTINEZ
Iberdrola Generación
Spain

SUMMARY

Beyond the assessment of the manufacturing quality of the insulation to ground of the stator
windings of electrical rotating machines, based on carrying out quality control electric tests
such as dielectric strength, capacitance and tan delta, partial discharges analysis, or endurance
tests; occasionally the preparation of samples for their analysis by means of a microscope is
requested.
The preparation consists essentially in cutting a cross section and finely grinding it, which
must be carried out according to usual metallographic preparation criteria. This implies
abrasive cutting of the section to be analyzed and several grinding steps with increasing grit
number.
The aim of the preparation is to obtain a flat and even surface apt for microscope observation,
while introducing the minimum possible damage and/or dimensional distortions that may
affect the analysis.

Thus far, these section analyses have allowed examining visually and with a certain degree of
subjectivity the finishing of an insulation system: structure, homogeneity, impregnation
degree, presence of voids or folds, etc. While this subjectivity may be good enough for the
former features, it is insufficient for evaluating more quantifiable factors.

However, nowadays, with the introduction of digital imaging systems, it is possible to

perform these analyses in a more objective way, being able to make real measurements of size
and angle of the different geometric features, heterogeneities or imperfections of these
insulation systems, and even, after an in depth analysis of these figures for each
manufacturing system, their correlation with other electric parameters and their evolution
throughout time. At the moment, there are no International Standards to cover stator bars
dissection process and analysis and it is based on the expertise of the operator. However,
these analyses give valuable information to ascertain the cause of failure during any quality
control electrical test and/or the amount of degradation in the case of endurance type tests,
giving somehow an idea of the insulation life expectancy. In the case of final users, this
information can also support someway the manufacturer selection in future projects.

[email protected]
Iberdrola Generación has been carrying out macrographic analyses of the insulation systems
of electric generators for over two decades. However, although large amount of information
has been available, most of it is not in digital format and it is not possible to get numerical
data. During the last years, with the implementation of digital imaging systems in these
analyses, a systematic analysis procedure, which is described in this paper, has been
developed and is intended to be applied routinely during the procurement and reception of
new equipment.

This procedure includes the definition of the zones of the bar/coil to analyze, the general
sample preparation criteria and the different parameters that are going to be analyzed and
measured, such as the strand radius, apex angle, rugosity radius, dimensions of
delaminations, voids, lack of resin penetration/impregnation, etc.

This paper includes the application of this procedure to a variety of samples of different
manufacturing systems and OEMs, as well as a first approach of correlation between the
results of these macrographic analyses and the values obtained at the electric tests during the
acceptance process, like power factor, partial discharge and breakdown voltage.

KEYWORDS

Stator Bar, macrographic analysis, internal corona protection, quality control.

1
 1. Introduction

The main manufacturing quality assessment of the insulation to ground of the stator windings
of electrical rotating machines is based on a series of quality control and acceptance electrical
tests such as dielectric strength, capacitance and tan delta, or partial discharge analysis, and in
some cases endurance tests.
In case of failure during these quality control electrical tests and/or endurance type tests or,
even when no failure exists, in order to recover information about the amount of degradation
accumulated during these accelerated ageing tests, it is usual in the industry to request the
preparation of sample sections for their micro/macrographic observation and analysis. Thus
far, these section analyses have allowed examining visually the finishing of an insulation
system: structure, homogeneity, impregnation degree, presence of voids or wrinkles, etc.
However, it implies a certain degree of subjectivity, which is clearly insufficient when
pretending to perform geometric measurements and correlate them with the theoretical
electric parameters.

Nowadays, digital imaging has become widespread in laboratory microscopy and

macroscopy. This, in some market sectors, has led to a trend towards integrating the digital
image acquisition hardware and image treatment software with the micro/macroscope itself.
The increased quality of digital image acquisition together with the capabilities of image
treatment software allow to make more accurate geometric measurements, which should
reduce the subjectivity of the analyses and provide appropriate data in order to relate the
geometric parameters and possible defects to the measured and expected electric performance.
Obtaining and storing quantifiable data in a systematic way will also allow its use in the
future, after the analysed elements have ended their service life.
Therefore, a procedure has been developed in Iberdrola Generación in order to carry out these
analyses in a systematic and repetitive way, with the intention of performing them routinely
during new equipment procurement and reception processes.

2. Sample selection

Those extra bars manufactured for quality control destructive tests, which fail during a
breakdown test should be subjected to dissection analysis for further investigation about the
failure itself and if the defect was localised or widespread so that, the point of surface
puncture and the carbonised path should be analysed, as well as the middle slot area and both
slot ends.

In the same manner, those extra bars manufactured for endurance tests which fail or pass
those accelerated ageing tests, should also be analysed for defect investigation, in case of
failure, or for better knowledge about the amount of degradation in case of success.

Bars that do not comply with the acceptance limits of tan delta or partial discharge tests,
should also be analysed. Middle slot area as well as each slot end should be selected for
dissection.

The general criteria to take into account for selecting the dissection sites can be those stated in
8.3 of IEEE 1310.

Although all previous statements are mandatory, it has to be highlighted that, all bar sets of
the sample analysed belong to projects with no defects during the manufacturing quality

2
assurance control (including voltage endurance tests) and so that, the main goal of the
analysis was not focused in locating any internal defect.

Picture 1: OEM 1. Nominal voltage 14,5KV 212

MVA.

Picture 2: OEM 2. Nominal voltage 8,5KV 40,5

MVA.

Picture 3: OEM 1. Nominal voltage 15KV 25,8MVA.

Picture 4: OEM 3. Nominal voltage 15KV 24 MVA.

3. Sample preparation procedure

The sample preparation is crucial for the analysis, as the quality of the former will limit the
accuracy and the obtainable data of the latter. The main drawback arises due to the human
manual work required, assuming an appropriate and specific specimen holder is not available.
The aim of the preparation is to obtain a cross section (perpendicular to the local strand axis
direction), the surface of which will be finely ground in order to obtain a flat and even
surface.
Finally, the sample must be able to stand with the prepared cross section perpendicular to the
optical axis of the micro/macroscope. Although this last requirement may seem not so trivial
for curved sections, it can be easily accomplished for straight sections.
From now on, the preparation guidelines on this paper will refer to stator winding straight
sections, although its content is applicable to curved sections taking into account their
particular features.

3.1. Cutting

The first stage consists in cutting out a sample, from which one of the cross surfaces is going
to be analysed, preferably by performing two parallel cuts as this will allow the straight
sections to stand suitably. The size of the sample must be suitable both for adequate handling
during preparation and for its observation with the analysis hardware.
The cutting process can be very aggressive for the insulation and copper strands, leading to
undesired damage and geometric distress. To prevent this from happening, a good method is
to use an abrasive metallographic cutter at a low cutting force and speed, and enough
refrigeration. During the last years, the samples analysed by Iberdrola Generación have been

3
cut with a speed not higher than 0.5 mm/sec, which results in an acceptable lead time taking
into account that the amount of samples available has been small.

3.2. Grinding

The next stage consists in grinding the cross section to be analysed following a sequence of
decreasing grain size, in accordance with metallographic preparation good practices, until the
achievement of a flat and even surface where no significant macroscopic scratches can be
observed. Typically, the grinding will start at 200/220 grit, followed by 500 and 1000 grit. If
the result isn’t good enough, finer grain sizes can be used.
The use of water refrigeration during the grinding process is advisable, as this will remove the
particles produced avoiding their embedding in the resin.
After each grinding step, the surface must be cleaned in order to remove the particles present,
which could produce scratches during the next grinding step. Special care must be taken in
those cases in which there are large voids, delaminations and lack of impregnation, as water
and debris can end up trapped in them. After the last grinding step a thorough drying of the
sample is required, as water may remain trapped by capillarity and reveal further on
complicating certain measurements or staining the strands.
The parameters of the grinding process will depend on the state of the surface and the
technician’s experience and ability. However, our experience shows that good results have
been achieved at 300 rpm grinding speed and applying low force during the first steps and
low to moderate force during the final steps.
This stage is not only the most critical of the sample preparation; it is also the stage where
most of the preparation defects will be introduced.

4. Preparation defects

The three main sources of defects during preparation that will affect the sample observation
are the sample cutting, the grinding process, and the standing of the sample for its
observation.

As it has been mentioned previously, the cutting process can be very aggressive due to the
damage required to produce the cut and the heat generated during the process.
Higher cutting speeds and cutting force are more likely to cause deformations and slight
displacements of the strands. At the same time, more aggressive cutting parameters increase
the heat generation, which may lead to the deterioration of the insulation. However, this
should not be a concern when using refrigerated and automatic cutting machines, with which
the parameters can be set.

The grinding process, when the sample is held by hand, depends on the technician’s ability
and will almost certainly introduce defects regarding the ideal result, therefore it is
recommended to be done by a qualified technician with experience in metallographic
preparation. These defects typically come in the form of:
• Loss of flatness: consists in a certain degree of convexity on the whole analysis
surface or only on part of it, and occurs as the result of an uneven grinding (different
speed/wear along the sand paper disc on rotating grinding machines, uneven pressure
applied to the sample,...).
• Surface inclination: occurs when the analysis surface isn’t perpendicular to the local
strand axis, and it is usually caused by an uneven pressure or the sample being
incorrectly held during grinding.

4
 • Grinding marks: these are not expected to be evident after the grinding process.
However, as a consequence of the previous two defects, some grinding marks from
previous grinding steps may be present at the borders, which are not always easily
detectable by naked eye. These marks are evident on the microscope and sometimes
make the analysis impossible. Correcting this defect is usually quite hard as it tends to
increase the convexity or the inclination of the surface.
Out of these three kinds of defects, the grinding marks are the worst as they can make
impossible to carry out certain measurements.
In the case of a loss of flatness or surface inclination, when they are not too severe, if the
sample is placed with the strands parallel to the optical axis the distortion of the image and
measurement will be very small. This is due to the vertical perspective of the
micro/macroscope and the multifocusing capabilities of the image treatment software.

Finally, although it is not strictly a preparation defect, an incorrect placement of the sample
under the micro/macroscope may result in tilting the surface, which will distort the image and
most of the measurements.

5. Sample analysis

The final stage of the procedure is the sample analysis under the micro/macroscope, which
will consist of a series measurements and the assessment of a series of aspects not so easily
quantifiable; and the appropriate treatment and storage of the data obtained.

5.1. Data acquisition

Although the measurements are expected to give objective results, these can be affected by
the technician’s operation, as its subjectivity may give different results for a variety of
reasons:
• Measurement method selected: image treatment software usually have several
measurement tools which properly used can reduce significantly the measure error.
For example, an effective method to measure a radius is using a 3-point circle, and to
measure a thickness the use of a baseline and a perpendicular line can be quite helpful.
• Non-uniform geometry: despite the quality of the manufacturing, usually neither the
winding insulation radius at corners nor its thickness are constant at any given sector.
In these cases, the reference points chosen by the technician in order to carry out the
measurement can significantly affect its repeatability.
• Technician’s operation: the accuracy of the person carrying out the measurement will
usually imply a shift of a few pixels, which tends to result in a very low error
compared to other factors.

The procedure aims for a systematised method of analysis and measurement, specifying the
parameters and features to be considered.
The quantitative measurements to be carried out are the following:
• Presence and configuration of the Internal Corona Protection (ICP).
• Inner (in case of ICP presence) and outer radius of the groundwall insulation at the
corners, preferably by the 3-point circle method.
• Thickness of the groundwall insulation at corners and sides.
• Layer thickness of the winding (at a random area on one of the sides).
• Tape overlapping (Apex angle).
• Tape bend radius in case of any tape wrinkle detected.
5
• Radius at corners of strands (if no ICP is present), preferably by the 3-point circle
method.
• Void size perpendicular to ground plane, if detected.

Picture 5: Measurement of inner and outer radius of

the groundwall insulation at a corner, which is
obtained as the radius of a circle that includes 3-points
set along the curve.
Care has to be taken on those samples where the
radius of the curve may vary significantly.
(Measurements expressed in mm)

Picture 6: Measurement of the thickness of the

groundwall.
At the corners the thickness has to be measured, not
calculated as the substraction of the inner radius from
the outer radius. This is because, as it can be observed
in picture 5, generally the centers of curvature do not
coincide.
(Measurements expressed in mm)

Picture 7: Measurement of the layer thickness at

different positions.
(Measurements expressed in µm)

6
 Picture 8: Measurement of the tape bend radius at
wrinkles, which is obtained as the radius of a circle
that includes 3-points set along the curve.
These measurements require a higher magnification.
(Measurements expressed in µm)

Picture 9: Measurement of the radius at corners of

strands, which is obtained as the radius of a circle that
includes 3-points set along the curve.
(Measurements expressed in µm)

The other well defined qualitative aspects that must be reported are the following:
• Manufacturing characteristics: materials, manufacturing process, number of layers,
ICP, bar length ...
• Nominal operation parameters: voltage, power.
• General appearance (homogeneity, resin distribution,...).

5.2. Defect assessment

The defect assessment is not so easily quantifiable and it requires not only the identification
of the defect, but also a description of its features and if it is an isolated case or if it is more
widespread. In some cases it may be convenient to attach photographs.
Some of the defects that might appear and should be registered are the following:
• Lack of homogeneity: only if it is significant.
• Delaminations: presence, quantity, distribution and length.
• Tape wrinkles: presence, quantity, distribution and radius.
• Lack of resin: isolated or more widespread.
• Strand deformation or displacement.
• Location and distribution of voids in the groundwall or around the strands.

7
 5.3. Results treatment

All the data obtained previously (including pictures) will be systematically recorded and
stored suitably, so it is easily accessible in the future. A good option is to attach the generated
results and documents to the acceptance tests’ documentation.

5.4. Electrical Parameters obtained during bars acceptance electrical testing

Electrical parameters obtained during quality control process of the stator bars manufacturing
are the main tool for accepting or rejecting a bars set. The establishment of pass-fail criteria in
technical specifications of new generator projects is globally accepted in the industry, as well
as most of them are specified in international standards, which is not at the moment extended
to micrographic measurements.

The main parameters taken into account for stator bars insulation manufacturing quality
assessment are:
• Power factor (Tangent Delta) IEC60894:
o Low voltage power factor at 0,2 times nominal voltage.
o Power factor increase per voltage step.
o Power factor TIP-UP, being (PF 0,6Un – PF 0,2Un)/2.
• Partial Discharges:
o Qmax at nominal voltage according IEC60270.

5.5. Calculated Electrical Parameters.

With the dimensional parameters obtained through the micro/macroscopic analysis of the
samples, some electrical parameters could be calculated for further analysis and insulation
characteristics knowledge. These parameters can be useful to determine the strong / weak
aspects of the insulation system, as well as to be a complement for future analysis of the
insulation ageing assessment.

From this set of electrical parameters the following could be highlighted:

• Average Electrical Stress (kV/mm).

• Electrical Stress at corners (kV/mm).
• Stress Concentration Coefficient (Max Stress / Average Stress).
• KV per layer.

6. IBERDROLA Sample Analysis.

6.1. Sample Description

As it was introduced in this paper’s summary, Iberdrola Generación has been carrying out
macrographic analyses of the insulation systems of new or rewound electric generators for
over two decades. However, although large amount of information has been available, most of
it is not in digital format and it is not possible to get numerical data.

This paper only includes the application of the described procedure during the last years after
the implementation of digital imaging systems in these analyses. Therefore, although a variety
of samples of different manufacturing systems and OEMs, ICP configuration, operation

8
voltage and power have been analysed, the current number of samples needs to be increased
to reach better conclusions, mainly regarding correlation with electrical parameters obtained
through bars testing and ageing profiles in service.

All samples included in this paper are from hydro-generators from different OEM’s, with
several variations of VPI based insulation systems and operation voltage within 8,5 and 15
kV.

7. Preliminary Conclusions.

The survey has been done over bars with different ICP (internal corona protection)
configuration:
- Fully shied with bare copper wire in the narrow side.
- Partially shield with U type semiconductive mastic /varnish.

1. Those bars with fully shield ICP presented lower 0,2Un Tan Delta (from 4,5 0/00 to 8,0
0/00 ) than those with partially shielded (from 9,5 0/00 to 11,00 0/00). As expected per [2].
2. Calculated Tip-Up as per IEC 60894 standard is higher in those bars with fully shield
ICP configuration (in the range of 0,20 0/00) while those with partially shield ICP
showed none or slightly negative Tip-Up (-0,01 0/00 to -0,23 0/00). As expected per [2].
3. Calculated maximum electrical stress at the corner in all the samples are in the range
of 4,36 kV/mm to 5,23 kV/mm.
4. The ratio of maximum electrical stress at the corner to the average electrical stress
(Stress Concentration Coefficient) are in the range of 1,49 to 1,90.
5. Those bars with fully shield ICP tend to have higher ICP radius (1,85 mm to 2,0 mm)
than those with partially shield ICP (1,3mm to 1,5 mm). As expected per [2].
6. All bar samples are in the range of 0,36 to 0,38 kV per insulation layer.
7. No significant voids have been found in the groundwall insulation and, those voids
found around the conductor stack perpendicular to ground plane in partially shielded
bars (in the area not shielded) were in the range of 200 µm to 500 µm .

152

244

Picture 10: Measurement of void size around Picture 11: Measuremente of void size in groundwall
conductors (Measurements expressed in µm) insulation (Measurements expressed in µm)

8. Tape wrinkles found in the groundwall insulation had a tape bending radius from 36
µm to 1.091 µm.

9
Picture 12: Measurement of tape wrinkles’ radius Picture 13: Measurement of tape wrinkles’ radius
(Measurements expressed in µm) (Measurements expressed in µm)

9. Ratio of corner bar thickness to lateral one because of tape tension was between 0,96
to 1,03 in all samples. Samples of an specific OEM were found between 0,73 to 0,83.

Picture 14: Thickness difference between corner

and side

8. Future work

The implementation of this procedure in Iberdrola Generación, regarding new equipment, will
grant the possibility of acquiring accurate geometric data and the effective storage of the
individual measurements. This may provide, in the years to come, useful relations between
the initial features of the insulation system and its subsequent performance during its service
life, and could provide information about expected results with different OEM manufacturing
processes, helping to determine specification criteria.

On the other hand, regarding the preparation, in order to reduce the amount of defects and
dedicated man-hours, improvements can be made, such as introducing the use of a suitable
specimen holder during the grinding stage.

10
BIBLIOGRAPHY

[1] “Calculation of Electric Field and Structure Design of Semi-Conductor Layer for Stator Bar of
1000 MW Turbo Generator” Harbin Research Institute of Large Electrical Machines and Harbin
University of Science and Technology. CIGRE SCA1 Meeting – Beijing, September 2011.
[2] “Study of the Effect of Different Internal Corona Protection on Electrical Properties of
Generator Stator Bars” Harbin Research Institute of Large Electrical Machines and Harbin
University of Science and Technology. 9th International Conference on Properties and
Applications of Dielectric Materials – Harbin, July 2009.

11