Studies into the effectiveness of
the electromagnetic stator core test
D R Bertenshaw*, ENELEC LTD, UK
A C Smith, University of Manchester, UK
Abstract
Stator interlamination insulation in large electrical
machines relies on routine testing for integrity. Thermal
testing originally using a high flux test (HFT) has to a large
extent now been replaced by a low flux electromagnetic
test (EMT) where the fault current in any interlamination
insulation failure is detected. A number of alternate
approaches have been developed with differing metrics
and acceptance criteria. This research analyses how
the differing EMT techniques correlate to each other
and to the HFT, plus how lamination segmentation can
affect HFT test levels. It further draws together both
recently published work and new insights from research
into EMT effectiveness (particularly of the dominant
EL CID system) to improve the reliability of its result
interpretation. The impact of core loss, fault inductance,
artefacts from non-uniform loss and permeability, plus
multiple aligned faults are considered.
Figure 1: Eddy currents induced in a damaged core
Testing for interlamination insulation damage became
routine since c1950 [4, 5], beginning with the traditional
HFT method (aka ‘Loop Test’ and ‘Ring Flux Test’) for
which five alternate public standards exist [6-10]. The
most current standard IEEE Std 62.2 [10] requires a high
induced toroidal flux at or near 100% of the operating
flux level, with infra-red (IR) thermography used to
detect those areas where some insulation breakdown has
occurred. This test remains the most trusted and thus
reference test [11].
1. Introduction
In large electrical machines such as utility generators,
if the insulation on a number of laminations becomes
shorted together, the consequent induced (eddy) currents
shown in Figure 1 cause local ‘hotspots’ to occur at the
damage. Such local heating, even if modest, can affect the
life expectancy of nearby winding insulation [1], and if
left unattended, can propagate and lead in extreme cases
to the catastrophic failure of the generator [2]. However
stator interlamination insulation cannot be reliably
monitored online [3]. In consequence, with stator testing
only possible at major service intervals, users must rely on
the ability of tests to detect nascent developing faults such
that they can be corrected before becoming dangerous.
Post WW2, turbo-generator power ratings grew rapidly,
at a rate of 32% pa between 1950 to 1975 for hydrogen
and water-cooled units [12], leading to a spate of stator
core failure rate in the UK [13] and increased testing
need. The high energy, safety hazards and lengthy test
time of the HFT motivated the development of the low
flux electromagnetic stator core test in 1979 [14]. The
*drb@bertenshaw.org.uk
KEYWORDS
Stator core, interlaminar insulation, electromagnetic test, EL CID, high flux test.
Cigre Science & Engineering • N°4 February 2016
37
resultant EL CID (ELectromagnetic Core Imperfection
Detector) system, operating at typically 4% of service
flux, is now in use worldwide [15].
have already been published and are summarized here
for reference, together with several new phenomena that
have been researched and are more fully detailed. Due to
its dominance the EL CID system is used in the majority
of the studies, but most of the discussed issues can be
expected to impact the other EMT systems.
In this test, fault currents induced in damaged areas
are measured by sensing their ac magnetic potential
difference (mpd) across slot teeth edges using an narrow
air-cored coil known as a Chattock potentiometer [16],
as shown in Figure 2. Any fault current is detected as
that mpd in quadrature to the excitation flux at test flux
level, indicated as a ‘Quad’ current [17]. The normal
recommendation [18] is that signals above 100 mA are ‘a
level at which faults should be investigated further.’
2. The growth of EMT systems
2.1. EMT systems in use
EL CID was the first and is now the dominant EMT.
While it has undergone two revisions to improve ease of
use [15], the basic principle of operation has remained
unchanged. However it has not remained the only EMT
system, the ease and speed of the test compared to the
HFT stimulated the development of several competing
low flux EMT products for in-house use or sale [12].
The principal alternate system in use is the Alstom/
ABB DIRIS (Diagnostic Investigation with Rotor in Situ)
system, patented in 1991 and 2004 [22, 23]. It uses an
air-cored coil to detect the fault current by mpd phase
angle change, computing the total fault power at service
flux level. The threshold of damage is set at 15 W in a
4–10 mm fault for turbo-generators and 10–20 mm in
hydro-generators to avoid the risk of ‘core melting’ [24].
Siemens in the USA also developed an EMT for inhouse use, claimed to ‘provide standard ELCID core
data’[25] with additional features. This was patented in
2004 [26] and launched in 2007 [27] as SMCAS (Siemens
Multi-frequency Core Analysis System). The system
can energise the core at low flux levels simultaneously
at 50/60Hz and higher frequencies (500–1000 Hz) to
improve diagnostic capability. The system detects core
faults with a normal Chattock and records Phase and
Quad values equivalent to EL CID, with the same fault
thresholds.
Figure 2. Chattock positioned across stator fault
Both tests have difficulties. While the HFT directly
detects the heating phenomena of concern, the required
power level can be very high and hazardous (>1 MVA),
and the test take several days to conduct. In addition
stator windings obstruct and thus attenuate slot and core
yoke thermal defect signals. The EMT is not attenuated
by stator windings, but is more sensitive to fault length.
Further, since the EMT is sensing the magnetic field
from a fault current that is within a similarly magnetised
stator core, misleading results can occur in some
circumstances, giving both false positives and negatives.
Two Russian EMT systems (EMK ЭлектроМагнитный
Контроль) were devised. The first in 1995–2000 [28, 29]
from the Russian Electric Power Research Institute used
eventually a 10 mm dia. air-cored coil. The detected mpd
phase change at a fault is used to compute an apparent
full flux fault power, assessed in Table 2 with the Russian
Turbo-generator Life-expired Guidelines [30]. From
2003 Gorodov et al. [31, 32] researched iron-cored
sensors, establishing a relative loss metric Ka from the
fault phase angle. The product was commercialised in
The EMT has achieved a wide degree of confidence from
published studies [11, 19, 20] and many test reports
(some listed in [21]). This research study is not pejorative
to that confidence, but seeks to better ensure that users
appreciate the issues that can affect the test results, and
how to accommodate them. A number of these topics
Cigre Science & Engineering • N°4 February 2016
38
2003 as the Introskan-IS200 core tester [33]. Gorodov et
al.’s recommendations are also found to be comparable
to typical EL CID practice for a one packet fault (~50mm
long) [34], also shown in Table 2.
use of severe sensitivity to varying tooth/yoke contact,
and research appears to have ceased. Replacing it, GE
described the RACER (Rapid Analysis Core Evaluation
Report) test system between 2003–8 in a series of papers
[42-45] and patents [46-49]. This uses an in-slot, ironcored probe shown in Figure 4 to notionally reduce noise
and improve reliability in fault detection.
Other systems in use are all similar to the EL CID.
The CDA system [35] from KES International Ltd in
Canada uses a higher induction of 5–15% of full flux,
with fault currents scaled to equivalent full flux current
for recording. A system called PROFIM was developed
by D. and R. Zlatanovici [36] for use by the Romanian
ICEMENERG Institute. It operates at a low ~4% flux and
any fault current is detected by analysing the Chattock
sensor signal’s phase angle. Finally another product
called EL CID used by INDUCOR in Argentina was
disclosed in 2013 [37], also operating at 4% flux and
providing Phase and Quad signals.
2.2. EMT systems not commercialised
Figure 4. GE iron-cored probe (© IEEE 2003 [42])
Whilst this elegantly solves the problem of obtaining
constant net magnetic air gap, the sensitivity to slot width
means the system still needs calibration to the machine.
Other problems also result, particularly of very poor
sensitivity to tooth tip faults and the problem of sensing
flush-wedged machines. Service using the product was
advertised up to 2011 [50] but does not now appear to
be in use.
Several researchers have investigated stator core testing
using local injection of excitation flux rather than bulk
toroidal induction. The benefits seen are reduced test
power, absence of excitation windings, and the ability to
test cores which might not be complete. A 1999 invention
by Bourgeois and Lalonde from Hydro Quebec [38]
induced a local flux from a C-shaped probe across 2
teeth, shown in Figure 3. The drive field and flux density
are measured to determine the core loss locally, with any
local loss increase ascribed to a core fault.
Some researchers have investigated resonant systems,
where a pulse of flux is locally induced in the core through
a wound, iron-cored probe across 2 teeth (comparable to
Figure 3), with the severity of any local fault determined
by the damping of the response. ELIN first proposed this
in 1984 [51], after which Ramırez-Nino and Pascacio in
2003 [52] described their similar MLM system. However
all such systems are very sensitive to probe-core spacing,
and require custom probes for each machine geometry.
They have not been adopted.
3. Correlation between test
methods
3.1. Correlation between EL CID and HFT
Figure 3. Hydro Quebec test flux injection probe
Later work by Kliman et al. for GE [39-41] similarly
used local flux injection with an improved differential
probe to attempt to overcome poor fault sensitivity.
This system still had the expected major drawback in
The main threat of a stator core fault is the local heating
effect, and the EMT is undertaken on the basis that the
results reflect the expected service overheating. While
there is considerable confidence in this assumption,
Cigre Science & Engineering • N°4 February 2016
39
Figure 5. Corrected correlation – all results
Figure 6. Test data and normal distributions
The identified correction factors were applied to the test
results and plotted as shown in Figure 5. The Mean Least
Squares (MLS) trend line slope mean was chosen over
the statistical mean, since it allows enforcing a logical
0 °C/0 mA origin on the trend. The corrected data,
binned at 2 °C/100 mA increments, is plotted in Figure
6 with Normal distributions overlaid for comparison. It
can be seen that the test correlation distribution has a
clear Normal tendency, with a modest skew to the right
tail. The red solid line shows the Normal distribution
for all values. The green dashed line shows the Normal
distribution if only test points of <20 °C/100 mA are
considered, and displays a close match. The effect of
enabling each of the correction factors described above
was tested and all except the core size correction caused
some reduction in the variance of the data, albeit small.
there is no known proof of this. All stator core faults are
considered problematic and a risk to winding insulation
life well before they pose any threat to lamination
insulation. Consequently both tests independently
set thresholds sufficiently above normal background
variation that a fault is evident, which could benefit
from attention. Subsequently the two test threshold
values became synonymous, giving the commonplace
expectation that the Quad signal detected by EL CID is
correlated to the HFT, at the rate of 5–10 °C/100mA [18].
Occasional evidence has been supportive but
inconclusive, with a 2004 EPRI study [53] failing to
demonstrate a correlation. To obtain more reliable
results, the published data on 106 reported core fault test
results from the field on turbo- and hydro-generators was
collated, where both tests were conducted on the same
fault. This analysed the correlation actually occurring on
electrical machines in service in the field, and also the
test variables that can affect them.
The data was corrected where possible for a number
of issues. HFT tests were thermally normalized to an
assumed average field test flux level of 90%, scaling
by flux squared, with a further small correction also
for particularly short test times. The EL CID Chattock
signal was also compensated for particularly short
faults <15 mm due to its decreasing sensitivity as shown
later in Figure 8. In some tests (esp. the EPRI study)
inaccuracies had occurred due to fault interactions
(Quad Recovery as described in section 6) between
severe faults [54]. Discrepancies found in recording HFT
temperature differentials in the EPRI study were also reestimated. It is known that the signal from buried faults
(under windings and in core yoke) is more severe for the
HFT than EMT, and compensation was made for these
where declared. It was further expected that the stator
core size and resulting excitation field would affect EMT
results, and smaller cores tested at <3 V/m were scaled
up. However, contrary to expectation, the correlation
was generally already higher in smaller cores than
larger before scaling, indicating that this correction was
probably unsound.
(* °C/100 mA)
All Tests
Turbo’s
Hydro’s
Qty. faults/tests
106
73
33
95% confidence mean*
8.5–10.7
8.0–10.9
9.8–12.9
Correlation coefficient (R2)
0.84
0.85
0.79
Lower-upper quartiles*
6.4–12.2
6.7–12.5
6.1–12.9
95% confidence upper limit*
18.1
18.6
18.1
Table 1. HFT-EL CID correlation analysis results
The analysis in Table 1 shows that the mean field
population correlation between the two tests is between
8.5–10.7 °C/100 mA to a 95% confidence level. It was
found that 64% of the sample tests lie between 5–11
°C/100 mA, supporting the industry expectation of
5–10 °C/100 mA. There is an indication that hydrogenerators have a 19% higher mean correlation than
turbo-generators. However their population variances
are sufficiently large that one would not perceive the
difference from individual tests. Full details of the
correlation analysis is given in [21].
3.2. Correlation between EL CID and DIRIS
Of the main EMT systems in use, EL CID and SMCAS
Cigre Science & Engineering • N°4 February 2016
40
Figure 7. DIRIS measurement coil on stator core (© IEEE 2001 [24])
Figure 8. Modelled and experimental Chattock sensitivity with length
for surface faults (Fault_1 = tooth tip centre,
Fault_2 = tooth tip edge, Fault_3 = slot base)
at 15 mm fault length (shown in section 3.1), the fault
temperature variation with length is shown in Figure
9. For the same nominal 10 mm dia. fault region, fault
temperatures from 5 °C to >50 °C may occur for both
tests at 100 mA and 15 W, as fault length reduces. This
also demonstrates the reason for the occasional user
complaint that an EMT appears to not detect very small
faults that still appear as a distinct pinprick of heat in the
IR images from an HFT.
(at 50/60 Hz) produce essentially identical signals, so
result interpretation remains the same. However DIRIS
reports a full flux, total fault power signal, contrasted to
EL CID’s test flux, fault current signal. While it appears
there is no correlation between the two measures, it can
be shown that this is not the case.
DIRIS uses a wide air-cored coil shown in Figure 7,
whose sensitivity to surface fault currents is calibrated
as a response per A-mm, and is almost linear with fault
length for shorter faults (<50 mm). Since the fault length
is unknown, the fault signal is perforce computed as the
total fault power in watts scaled to full service flux [24].
This power appears to be both a very different measure
and the 15 W threshold a substantially higher threshold
than EL CID’s 100 mA Quad current, which yields an
apparent ~3 W power in a 10 mm fault at rated flux
in a typical larger turbo-generator (tested at 5 V/m).
However when allowance is made for the rapidly
reducing Chattock sensitivity to short faults as analysed
by Ho et al. [55] (shown in Figure 8), the power rises
substantially. Figure 9 shows the total fault service power
for an EL CID Quad signal of 100 mA remains within
±20% of the DIRIS’s 15 W for a 2–35 mm fault length.
Thus the DIRIS and EL CID thresholds for fault warning
are very comparable for modest fault lengths.
Figure 9: Total fault powers and temperature rise for EL CID and DIRIS
detected threshold faults
The above analyses allow a combined table to be
completed in Table 2 approximately comparing the
metrics and thresholds for the various stator core test
methods.
The variation in internal fault temperatures at the test
thresholds can also be estimated for a nominal 10 mm
dia. fault region where the laminated axial thermal
conductivity is assumed to be 10% of the radial [56].
To avoid the need to develop an analytic full 3D model,
it is assumed the net radial thermal conductivity is
proportional to fault length, the axial net thermal
conductivity constant, and their effective dissipative
capacities are in proportion to their axial/radial surface
areas and thermal conductivity.
4. Problems from lamination
joint reluctance
Stator cores of large electrical machines most commonly
use half-overlap lamination joints. The flanking
laminations thus locally carry double normal flux
density and saturate at service flux, developing a large
local mmf. While this phenomenon is well known, its
quantification is rarely advised. By flux levels of 1.2 T the
Since the reported power is that in the total fault, the heat
dissipation capacity of the fault will differ substantially
with length. If the total effective thermal conductivity is
set such that the EL CID 100 mA results in 10 °C rise
Cigre Science & Engineering • N°4 February 2016
41
HFT
EL CID et al.
DIRIS
Gorodov / Introscan & EMK
Hotspot
<10 °C
<100 mA
Normal
Power <15 W
Ka <1.3 (< 3–4°)
Power <20 W, No fault
100–260 mA
Investigation region
Power 15–40 W
Operational safety endangered
Ka = 1.3–1.8 (4–10°)
Power 20–40 W
Exceeds the HFT acceptance (>15 °C)
>260 mA
Power >40 W
Ka >1.8 (>10°)
Power >40 W
May exceed the generator capability
Hotspot
10–25 °C
Hotspot
>25 °C
Power >120 W
Risk of ‘iron fire’
Table 2. Correlation between HFT, EL CID, DIRIS, Ka , and EMK fault power
of the test core joints, using a new approach to
extrapolate the susceptibility of electrical steel into
deep saturation. Counterbalancing this, notionally
isotropic (non-oriented) electrical steel still retains
considerable magnetic anisotropy, such that the rolling
direction permeability can exceed 200% of the transverse
direction. Since toroidal flux in laminations mostly
flows in the rolling direction, this can reduce effective
toroidal reluctance by 18–30% from published data. The
FE results show the impact of the joints and directional
permeability by the close match of the model to tests in
Figure 12.
toroidal reluctance of such a core can increase by 200%
over that expected without core joints and can make a
large difference to the stator HFT excitation requirement.
The impact is shown in Figure 10 for both a FE model
and test results of an actual 100 MVA test stator, with 60°
lamination segments.
The problem has in fact been long researched with Bohle
[57] identifying the issue as early as 1908, but remains
ignored in modern design texts. Further IEEE Std 62.2
[10] just instructs ‘use the manufacturer’s B/H curves’ to
compute the HFT excitation needed, as does published
test advice [58]. Additional disparity occurs where both
above sources also give a ‘typical’ B/H curve that the
excitation may follow. The flux density B field is peak
but the magnetising H field is clearly assumed rms and
must be converted to peak using an appropriate crest
factor. These are compared to published data for Cogent
M310-50A steel [59] (B and H values are conventionally
specified as ac peak) and the measured B/H curve of a test
core built with M310-50A steel. They are all remarkably
different as shown in Figure 11, though the two ‘typical’
curves converge at the same level as the test result for a
typical HFT 100% flux level of 1.4 T.
A final problem emerges from the changing crest factor
of the excitation current. For a sinusoidal flux excitation,
the saturating steel causes the crest factors to rise from
1.4 to >2.1 at flux levels from 0.6 T to 1.4 T as shown in
Figure 13. This tends to reduce the rms current required
compared to sinusoidal assumptions.
The use of slightly under full flux density in the HFT is
common. Thus when estimating the required excitation
curent from electrical steel B/H data (and correctly
interpreting H as peak magnetic field strength), the
rms excitation at worst case of ~1.2 T may be ~200%
greater than expected due to joint reluctance, but over
To study the full situation, an FE model was completed
Figure 10. Core mmf increase due to half-overlap lamination joints.
Figure 11. Typical and actual B/H (ac peak) curves for HFT.
Cigre Science & Engineering • N°4 February 2016
42
Figure 12. Modelled and actual test B/H results for stator
Figure 13. Flux and excitation current at ~1.2 T flux density with halfoverlap lamination joints
estimated by typically 25% in the unsaturated core due
to anisotropy, and by 1.4 due to a current crest factor of 2
vs 1.4. Thus the total rms excitation percentage increase
compared to assuming basic B/H values may be typically
computed from the net impact on the core’s reluctance
giving here a 93% increase.
15–20° is typical from tests on assembled cores of two
sources of M310-50A electrical steel, at the usual EMT
flux level (4% of full flux density)
In EL CID, the Reference is normally aligned to the
rms excitation current Ie of rotational frequency ω, and
adjusted such that a zero Quad signal (mpd across the
fault region) is developed without a fault, thus the Quad
signal is solely the induced fault current resolved to the
quadrature axis [17].
Without allowance for lamination joint saturation,
permeability anisotropy, and magnetic field crest factors,
inadequate HFT stator core excitation may be provided,
leading to under-testing if thermal scaling for reduced
flux density is not applied. Further, since the concern is
induced voltage heating a resistive fault, to compensate
for any flux wave distortion the flux sense coil signal
should always be measured with an rms-indicating
instrument. Full details of the study are given in [60].
In the fault modelled in Figure 15, peak flux density B is
induced by excitation current Ie around mean magnetic
path radius r. For a fault of length F, conducting area
A in steel resistivity ρ, the fault’s resistance Rf and selfinductance Lf are computed conventionally, assuming
the flux induced by the fault current remains planar in
the core circumference.
5. The impact of core loss on
severe faults
Assuming the lamination and rear keybar resistance are
negligible compared to the fault, allows the fault power
Pf to be computed as
In a stator core, the flux significantly lags the excitation
mmf due to hysteresis and eddy current losses in the
core. This varies with type of the steel and especially
the excitation level. Figure 14 shows a loss angle (θ) of
Figure 14. Core loss angle variation with excitation
Figure 15. Stator core excitation and fault circuit abcda
Cigre Science & Engineering • N°4 February 2016
43
Figure 16. Normalised quad signal and service fault power
Figure 17. Excitation and Chattock relationship
shows that for the ideal loss free core, the Quad signal
reaches a peak at the maximum test flux density power
point of Rf = ωLf, though at service flux density fault
power peak is at Rf = 0.3ωLf. .
If the resistance dominates (Rf >>ωLf) as is normal for a
modest fault,
This shows the presence of core loss substantially affects
the maximum detected Quad signal at high fault powers,
both due to the core loss and the increased reactance
at test flux levels. During early fault development with
Rf >10ωLf, the error caused by a typical 20° core loss
remains <10%. However for more severe faults, the errors
become substantial, such that the Quad signal can even
suffer a polarity inversion as Rf<ωLf, shown in Figure 16.
while if the inductance dominates as may occur for a
very severe fault (Rf>>ωLf),
Equations (3) and (4) show that the power in a fault
dominated by fault resistance is proportional to the fault
length and area, and as expected, to the square of the
factors that control the fault voltage. However in a fault
dominated by inductance, the power is still proportional
to the fault length but now inversely proportional to
area, a key reason why molten core faults tend to reach
a limiting diameter but not length. It also interestingly
reduces to the equivalent of a 1:1 current transformer
from excitation current into fault resistance.
While the assumption that the lamination and keybar
resistances are insignificant may fail in severe faults, fault
inductance has been shown to limit fault currents and
powers for quite small area faults [61]. The impact of this
effect is still minimal at clearly dangerous fault levels of
<1 A Quad signal. However in the case of very severe
faults, such as those caused by core melts, the effect
of core loss and fault inductance can act to reduce the
apparent fault intensity. This phenomenon particularly
needs to be considered when interpreting signals from
severe, perhaps melted, faults buried in the core yoke.
The flux lags the excitation current by an angle θ due to
the core loss, however the Quad signal IQ is detected as
that current in quadrature to the excitation current and,
assuming 100% Chattock detection sensitivity, is given
by
6. The interaction of multiple
faults
In the majority of cases of electromagnetic stator core
testing, the stator has either no significant faults, a few
of modest amplitude (i.e. 100–300 mA Quad), or one
serious fault. In these situations interpretation of the fault
signals is conventional. However if multiple faults occur
that are circumferentially aligned, and especially if any
are severe faults, the fault fluxes can interact and cause
reduction of their detected Quad values. This effect is
termed ‘Quad Recovery’ and has led to under-recording
of the severity of faults in field examples.
The detected Quad signal in (5) is explored for a range of
core loss angles as fault resistance reduces (i.e. intensifies)
as a proportion of the test flux reactance (Rf /ωLf). This
assumes a typical relative permeability of 3000 at test flux
density and 1000 at service flux density levels. Figure 16
Cigre Science & Engineering • N°4 February 2016
44
The cause of the effect is due to the method of measurement
of the magnetic potentials. This is shown in Figure 17
for a homogeneous core with long faults, where the mpd
measured by the Chattock has contributions from the
excitation current phasor Ie. and fault current phasor If.
The flux from the fault induced in the core will establish
a magnetic field around the core, and this field will be
detected by the Chattock away from the fault. Since the
majority of the magnetic field strength from the fault
mmf occurs in the air, the Chattock signal will be the
sum of the fault mmf and the section proportions of the
fault and excitation magnetic field strengths developed
within the core [17].
48 slots and a fault of 800mA imposed on a tooth tip.
It can be seen that the fault gives a small 17 mA Quad
Recovery signal on all other slots.
Thus the mpd detected by the Chattock at the span angle
α is
Figure 18. Plot of Quad Recovery mpd obtained from 2D FE
Since  for single slot Chattock spans, this shows
that the Chattock detects nearly all the fault current mmf
and the inverse of the circumferential proportion of the
excitation current mmf.
This 2D analysis only valid for long faults, a rare situation.
To investigate the effect on shorter faults, a 2D FE turbogenerator model with 40 mm packets was developed
considering the laminations viewed on their edges (red
arrow) as shown in Figure 19. The core is modelled as
essentially ‘unrolled’ along the yoke magnetic mean
line, with laminations having low axial (z direction)
permeability to simulate the core stacking factor (0.97).
Fault currents are set in the y direction with the Chattock
signal derived from integration of H.dx between its endpoints on the x/z plane.
The sum total of all the mpds detected by the Chattock
across each slot around the bore in the air should sum
in phase and amplitude to the enclosed current from
Ampère’s law. Since the fault current is outside this
summation this can only be the excitation current. The
EL CID is set up normally so that its Phase resolution is
in phase with the excitation current Ie. thus
The measurement interpretation of (7) is that the
Quad values around the core surface must sum to zero,
regardless of the presence or absence of faults. From (6)
the detected mpd away from the fault will be the sum of
the excitation current and fault current, proportioned by
the slot and Chattock span angle.
Figure 19. Model orientation and axes
The initial results showed that the fault flux Quad
Recovery was very strongly biased towards the fault
region if eddy currents were not permitted to flow in the
laminations. This is caused by the fault flux circulating
back to the fault region over a considerable distance,
despite the low axial permeability. However if eddy
fault signal element
The Quad component of the
is the Quad Recovery signal and is the opposite polarity
to the usual detected fault current signal. It can act to
reduce another local fault signal. The effect can be simply
modelled in 2D FE with the result shown in Figure 18 for
Cigre Science & Engineering • N°4 February 2016
45
Figure 20. Flux density flow down laminations from 30 mm fault
Secondly in the earlier EPRI study [53], multiple severe
faults were experimentally applied to a stator and the
EL CID results found to be difficult to correlate to the
thermal results. When the Quad Recovery effect is
accomodated the results correlate much closer to the
expected values [21].
currents are permitted by setting a rational lamination
resistivity, then these strongly oppose the axial
divergence of fault flux, and fault flux density around the
core becomes very uniform as shown in Figure 20.
The results plotted in Figure 21 for four lengths of fault,
show that for a very short fault length of 10 mm the Quad
recovery is still modestly biased towards the source,
however once the fault reaches 20 mm the recovery is
virtually uniform.
In a third situation, a stator bore had been bead blasted and
the surface damaged causing substantial areas of stator
core surface to become faulty, such that the cummulative
Quad Recovery was exceeding local fault signals. This
was resolved by development of a global compensation
algorithm for the results. The Quad values around the
core surface must sum to zero, and the Phase values sum
to the excitation current (as usually expected). In reality
since the Chattock is normally set to span the opposite
sides of the teeth across a slot, there is double-counting
of the potentials across each tooth surface. For the Quad
Recovery potential deriving from the circulating flux
this is a small potential referred to the tooth surface and
may be ignored (or a notional -10% allowed). However
any fault on a tooth tip may be counted twice, and must
be allowed for in the summation.
If Qin and Qn are the indicated Quad signal and actual
fault Quad value for slot n at each core axial position/
packet, and ε a global error value from any error in the EL
CID phase reference, each slot signal at an axial position/
packet on a core of N slots will thus be
Figure 21. Quad Recovery 100 mm mpd around x-axis
The attenuation of fault signals due to Quad Recovery is
fortunately rare since it requires a very substantial fault to
develop a Recovery signal significant enough to impact
measurements of other faults. However several instances have
occurred where accommodation of its effect was required to
understand test results. In one case a major fault caused by a
local core melt developed enough Quad Recovery to mask
another fault in the same axial region [54]. This is illustrated
in Figure 22 where a significant -180 mA Quad signal had
been depressed to just -80 mA and ignored.
Summing round the bore gives
Hence around each packet the Quad values should only
accumulate the standing error values. However in order
to determine the actual fault values, it must be assumed
or known that at least one or more slots at each axial
position/packet are unfaulted, thus their positive value
(assuming negative Quad signals are indicative of a fault,
as is conventional) represents only the Quad Recovery
potential. Further, in a normal stator core, natural core
loss variations lead to natural Quad fluctuations around
Figure 22. Before and after correction for Quad Recovery
Cigre Science & Engineering • N°4 February 2016
46
the bore, not representative of fault or measurement
error. A further variance value Cq of up to -50 mA may
be added to ε to reflect the age and dilapidation of the
machine to avoid overstating any fault.
variations are not uniform in the core this conclusion
is not true. The study recognises that due to low axial
permeability from the stacking plus eddy currents, the
flux around the core in each lamination plane will remain
mostly constrained to that plane, and thus approximately
circumferentially constant in amplitude and phase angle
to the excitation current.
Unfaulted slots will carry a positive Quad signal averaged
as Q+max which, assuming uniform Quad Recovery
around the core, is given from (9) as
By integrating over an ac cycle it can be shown that
for a sinusoidal flux density phasor B of rotational
frequency ω, with lagging phase angle θ, induced by coaxial magnetic field strength phasor H, the net absorbed
power density PL (core loss spatial density) is given by
(13). Here B and H represent the fundamental rms values
of B and H.
Thus for each slot in the packet, the true Quad signal can
be estimated from (10) and (11) as
A Quad signal is developed when the magnetic potential
across a slot is not in phase with the excitation mmf
Ie. To determine how both a variation in core loss and
permeability can develop Quad signals, the locally
developed rms magnetic field strength phasor H is
resolved into components Hinductive in phase and Hreal­ in
quadrature to the flux density phasor B. The resultant
phasor diagram around a simple 4 slot core is shown in
Figure 23. Since for slot n Hreal=Hnsinθn, it can be seen
that the core loss power is wholly transferred in the Hreal
component, the component Hinductive in phase with the
flux density carries no net power.
This is amenable to a spreadsheet analysis. A set of the
maximum Quad +ve values for each packet or axial
region are ranked to check that there are no odd-looking
values which might derive from human input error,
and that the maximum +ve values seem to be trending
to a stable value. This indicates that the hypothesis that
at least one or two slots in each packet are unfaulted is
true. The mean of these maximum +ve values is then
computed and true fault values determined from (12).
The values of Cq and ε are seen to be automatically
accomodated by this approach. The phenomena of Quad
Recovery is discussed further in [54].
7. The impact of local variation
in core loss and permeability
The standard model for the EMT assumes that the
stator core is composed of electrical steel with uniform
magnetic properties. However variances in core loss and
permeability caused by non-uniform steel have been
shown to occasionally give rise to substantial fault signal
artefacts not due to actual interlamination insulation
defects, in particular due to the use of circumferentially
differing steels in robotic assembly.
Figure 23. Varying core loss phasor diagram around core
The local and aggregate core loss angles θn and θaggregate
are given for slot n with N slots
In the EL CID test, overall variations in core loss or
permeability will only affect the magnetic field strength
(hence excitation current) required to induce the flux,
thus only vary the Phase signal. However when these
Cigre Science & Engineering • N°4 February 2016
47
Figure 26. False colour map of Quad signals
The total magnetic field strength equating to the
excitation mmf I­e will be
Thus from (16) the Quad magnetic field strength Q
(A/m) in the variance region is given by
Thus from (14) and (15) the local Quad field Qn developed
for slot n shown in is given by
Expanding (18) only produces a complex expression
which does not provide any better insight into the
effects of the variance. A numeric example provides a
case specific illustration of the scale of the effect, using
a 110 mm slot pitch with M310/50A electrical steel
(relative permeability 3800 and core loss 0.006 W/kg at
0.056 T, 50 Hz). The curves in Figure 25 illustrate the
Quad signals that are developed in the variance region a
of the core, modelled at 10% and 50% of the core, for up
to +/-50% variation in core loss and permeability.
To study the practical scale of these phenomena, a model
in Figure 24 is considered of a stator core yoke of density
ρ and unit depth illustrated in Figure 24 with a material
variance region a. In the two regions the specific mass
core loss (W/kg) are W and Wa and permeability are µ
and µa. The constant induced rms flux density B develops
rms magnetic field strengths H and Ha. The phase angle
(core loss angle) between flux density and magnetic field
strength is θ and θa. Factors XW and Xμ proportion the
variances such that
Figure 25. Quad signal from local 10 or 50% circumferential variation in
core loss or permeability
This shows that Quad signal artefacts above the -100 mA
warning threshold can be developed in stator cores with
significant magnetic non-uniformity. The Quad signals
from the two parameter variances may accumulate or
offset each other, but the combination is not linear.
These artefacts have been seen to occur on a robotically
stacked core constructed with opposing 180° segments of
Figure 24. Stator core loss and permeability in normal and variance
region a
Cigre Science & Engineering • N°4 February 2016
48
Figure 27. Spread of core loss and permeability at 0.05 T
Figure 28. Modelled and measured M310-50A Rayleigh hysteresis loops.
(maximum ) and induced flux density (maximum
low flux levels, given by
differing lamination materials [62]. Despite satisfactory
high flux tests, when an EL CID test was conducted on
the core, high Quad signal readings occurred as shown
in Figure 26, which significantly exceeded -100 mA and
thus a serious concern. When measured, the M270-50A
lamination steel properties at low flux density were found
to have large variations of relative permeability and core
loss at the test flux level of 0.05 T, with little correlation
between them as shown in Figure 27
) at
Assuming no residual magnetisation, where µi is the
residual permeability at zero flux and γ the “Rayleigh
constant” for each material, where K = +1/-1 for
descending/ascending H with sinusoidal flux, and where
constants D1 – D3 are expressions of µi, γ, and , gives
a periodic function for H of
Using the analytical approach above, a numerical model
showed that the maximum local variance in core loss
in one 180° segment could cause up to 368 mA Quad
signal artefacts to appear, with up to 111 mA from the
same maximum spread in permeability. These were quite
sufficient to produce the discovered worst case values of
~200 mA. In conjunction with other analysis and tests, it
was concluded that there was no actual interlamination
insulation damage.
A novel regression technique was developed to
interpolate µi and γ from published B/H specification
0 for a variety of electrical steels (M270–
data as B
M600), showing relative µi varying from 452–1057 in
line with measurements. Since γ governs the hysteresis,
hysteresis loss proportion was found to vary from 51–
82%, allowing D1 – D3 to be evaluated.
Study of the EMT vector diagram shows that any actual
fault Quad signals are linearly superimposed on the
non-fault Quad signal artefacts. This still allows the
measurement of any genuine Quad fault level by simple
variance from a no-fault template test level rather than
zero. An earlier analysis without the above analytic
extension is given in [62].
Fourier analysis of equation (20) however yields insoluble
elliptic integrals, consequently numeric integration was
used to compute the resultant harmonics, giving up to
6% 3rd harmonic. The impact of these on the EL CID test
was analysed and shown to be dependent on the method
of phase analysis (demodulation) of the fault signal.
However even with the worst case method, the impact
on the EMT was found to be very minor even for major
variations in the core, with typical Quad signal change
<12 mA. The full analysis is given in [63].
8. Impact of Rayleigh hysteresis
A further phenomenon that could affect EMT readings
is the non-linear permeability and hysteresis of electrical
steel, illustrated by the hysteresis loop at 0.05 T in Figure
28. For sinusoidal flux, this causes the development
of harmonics in the magnetic field. Depending on the
sensitivity of the EMT to harmonics, this can affect fault
interpretation if the hysteresis varies around the stator
core in the same manner as in Figure 24.
9. Conclusions
The electromagnetic stator core test has reached a
substantial maturity in large stator core testing, with its
ease and safety of application assisting its acceptance.
While many designs have been tried, the principal
systems in use have settled on just two metrics, apparent
At low EMT flux levels, induction follows the Rayleigh
parabolic hysteretic relationship between magnetising field
Cigre Science & Engineering • N°4 February 2016
49
test fault current or apparent service fault power. Despite
their difference, it is found that they are closely correlated
for normal lengths of developing fault.
give this effect. Fortunately these artefacts have a distinct
pattern and are linearly superimposed on any actual core
fault signals, allowing simple accommodation of their
presence.
It is also shown that the EMT has a clear correlation
to the thermal HFT and can thus provide the same
indication of threat from a developing stator core fault.
However the apparent simplicity of the reference HFT
belies a number of problems relating to its practice,
with little consideration paid to the square-law effect
that reducing test flux density has on recorded fault
temperature. This is compounded by conflicting
advice on core excitation requirements and the littlerecognised impact of lamination segment joints at
high flux levels. Consequently, even though offset by
anisotropic permeability and high current crest factors,
much greater excitation may be required compared
to that assumed from catalogue steel magnetic data,
risking undertesting.
The magnetising harmonics from low flux Rayleigh
hysteresis raise a similar possibility that non-uniformity
of hysteresis loss could also cause signal artefacts. This is
analysed and shown not to be a concern.
This paper describes and resolves a number of issues that
should be recognised in stator core testing. It is hoped that
with these further understandings the electromagnetic
test can continue to be used with improved confidence
to monitor the stator core condition of large electrical
machines.
10. Acknowledgements
This work was completed as part of a PhD research
programme at the University of Manchester, UK,
supported by ENELEC LTD, UK.
The EL CID test (and similar EMTs) derive their fault
current signal from that mmf in quadrature to the
excitation field, rather than flux. In consequence the
core loss and difference in permeability of the core
between test and operation, coupled with the fault’s selfinductance, cause the detected Quad signal to limit and
even invert for very severe faults. While not an issue for
normal faults, it may impact the interpretation of very
severe buried faults, such as those caused by core melts.
11. References
[1] G. C. Stone, I. Culbert, E. A. Boulter, H. Dhirani, Electrical
insulation for rotating machines: design, evaluation, aging, testing
and repair, 2 ed: Wiley, 2014, p. 672.
[2] J. S. Edmonds, A. Daneshpooy, S. J. Murray, R. A. Sire,
"Turbogenerator stator core study," presented at: SDEMPED
2007, IEEE International Symposium on Diagnostics for Electric
Machines, Power Electronics and Drives, September, Cracow,
Poland, 2007.
The interpretation of EMT results can also be affected
by several uncommon but not unknown effects. Due
to the EMT measurement method, the sum of all Quad
signals must circumferentially sum to zero. This leads to
an effect where a severe fault can depress the signal from
another axially aligned fault due to Quad Recovery. This
needs to be recognised in cases of severe core damage,
and an algorithm is presented that allows correction of
results.
[3] P. Tavner, L. Ran, J. Penman, H. Sedding, Condition monitoring
of rotating electrical machines, 2 ed: Institution of Engineering
Technology, 2008, p. 250.
[4] H. R. Tomlinson. "Interlaminar insulation test for synchronous
machine stators," Power Apparatus and Systems, Part III
Transactions of the American Institute of Electrical Engineers, vol.
71(1), pp. 676-7, 1952.
Similarly EMT signal artefacts not reflecting core damage
can occur in situations when stator cores are stacked
with laminations having differing magnetic properties
(e.g. not from the same steel batch) and the laminations
are not mixed in stacking. In this case circumferential
variations in core loss and permeability can produce
unexpected signal results in the EMT. A new technique
analyzing the vectors transferring the core loss energy
demonstrates how these artefacts can come about, and
how variations of both permeability and core loss can
[5] C. Maughan, "Testing of turbine-generator stator core lamination
insulation," presented at: EPRI 2004 Winter Workshop and Turbine
& Generator Users Group (TGUG) Meeting, January, Albuquerque,
USA, 2004.
[6] Guide for insulation maintenance of large alternating-current
rotating machinery (10 000 kVA and Larger), IEEE, IEEE Std 561977, 1977 (reaffirmed 1991).
[7] Guide for insulation maintenance for rotating electric machinery (5
hp to less than 10 000 hp), IEEE, IEEE Std 432-1992, 1992.
[8]NORMATIV
Cigre Science & Engineering • N°4 February 2016
50
DE
ÎNCERCĂRI
ŞI
MĂSURĂTORI
LA
ECHIPAMENTE ŞI INSTALAŢII ELECTRICE (Testing and
measurements to regulatory and electrical equipment), RENEL
Romania, PE116/1994, 1994.
[26] M. W. Fischer, J. W. Shelton, M. J. Metala, "Electromagnetic stator
insulation flaw detector," Patent US 6,791,351, 2004.
[27] M. W. Fischer, M. J. Metala, J. W. Shelton, "Siemens electromagnetic
Core Analysis System (SMCAS) for generator inspections,"
presented at: EPRI Turbine Generator User Group Meeting, July,
Phoenix, USA, 2007.
[9] Volumes and norms for testing electrical equipment (in Russian),
RAO "UES of Russia", RD 34.45-51.300-97, 1997.
[10] Guide for diagnostic field testing of electric power apparatus, IEEE,
IEEE Std 62.2-2004, 2005.
[28] F. A. Polyakov. "The steady temperature field in the region of a
local short circuit of the iron laminations of the stator core of a
turbogenerator," Electrical Technology Russia, vol. 4, pp. 49-57,
2000.
[11] "EL CID (Electromagnetic - Core Imperfection Detector) testing of
large steam-turbine-driven generators," CIGRE, Paris, Report No.
TB 257, 2004.
[29] V. B. Berezhansky, L. I. Chubraeva, G. V. Rostik, "Experience
with modified iron fault control technique for the stator cores of
electrical machines," presented at: Proc IEEE Stockholm Power Tech
Conference Int Symp Electr Power Eng, 18-22 June, Stockholm,
Sweden, 1995.
[12] D. R. Bertenshaw, "A history of stator core testing," presented at:
EPRI 2013 Winter Turbine Generator User Group Workshop,
January, Savannah, USA, 2013.
[13] D. C. Renew. "Core faults on 500MW generators. Part 1 historical
review," CEGB, Report No. RD/L/N152/76, 1976.
[30] Guidelines - guidance on assessing the technical condition of turbine
generators, exhaust stipulated normative life (in Russian), RAO
UES of Russia under Rule No 200, 2008.
[14] J. Sutton, "Method of and apparatus for testing laminated magnetic
cores," Patent GB 2044936A, 1979.
[31] V. V. Gorodov, A. G. Vdovikov, V. M. Gandshu, A. P. Lashitsky.
"The criterion of the defectiveness when using an electromagnetic
method of detecting short-circuits between laminations of electrical
machines (in Russian)," ЭЛЕКТРОТЕХНИКА (Elektrotekhnika),
vol. (4), pp. 35-41, 2003.
[15] J. Sutton, M. Sasic, D. R. Bertenshaw, "30 years experience with
EL CID stator core testing," presented at: Iris Rotating Machine
Conference, 17 June, Long Beach, USA, 2008.
[16] A. P. Chattock. "On a magnetic potentiometer," Proc Phys Soc
London, vol. 9, pp. 23-6, 1887.
[32] V. V. Gorodov, A. G. Vdovikov, Y. U. N. Samorodov, V. N. V.
Grechenkov, "Facility testing fault of sheets of active steel in cores of
stators of ac electric machines," Patent RU2195681-C1, 2003.
[17] D. R. Bertenshaw, "Analysis of stator core faults - a fresh look at
the EL CID vector diagram," presented at: Hydro 2006, September,
Porto Carras, Greece, 2006.
[18] ELCID Evolution User Manual, V6 ed, Toronto, Canada: Qualitrol
- Iris Power, 2014, p. 214.
[33] "Introskan-IS200 system for automated control circuits active sheet
steel cores of electric machines," Elektrotekhnicheskie Sistemy 1
(ES1), 2010, http://ets1.spb.ru/?page_id=95.
[19]J. W. Shelton, B. M. Reichman, "A comparative analysis of
turbogenerator core inspection techniques," presented at:
Proceedings of the American Power Conference, 22-24 April,
Chicago, USA, 1985.
[34] V. M. Gandshu, V. V. Gorodov, S. Dubitsky, V. Podnos, "Theoretical
analysis of the ELCID method usıng QuıckField FEA package,"
presented at: ICEM2006 XVIIth International Conference on
Electrical Machines, 2-5 September, Chania, Crete, 2006.
[20] "Repair and testing guide for generator laminated cores grounded
at the core outside diameter," EPRI, Palo Alto, USA, Report No.
1007441, 2002.
[35] M. Gavric, M. Senjamin, M. Kotlica, "Comparative analysis of
stator iron core damage using nominal and reduced flux," presented
at: Electrical Insulation Conference and Electrical Manufacturing
& Coil Winding Conference, October 26-28, Cincinnati, USA,
1999.
[21]D. R. Bertenshaw, A. C. Smith, "Field correlation between
electromagnetic and high flux stator core tests," presented at: PEMD
2012, The 6th IET International Conference on Power Electronics,
Machines and Drives, March, Bristol, UK, 2012.
[36] D. Zlatanovici, R. Zlatanovici. "Diagnosis of the electric generator
stator core by means of the low induction iron test, PROFIM,"
Journal of sustainable energy - Faculty of Power Engineering,
Oradea, Romania, vol. 2(2), pp. 24-9, 2011.
[22] M. Hobelsberger, I. Kirchhoff, E. Nodwell, Z. Posedel, "Method and
device for inspecting laminated iron cores of electrical machines for
interlamination shorts," Patent US 6,815,957, 2004.
[37] "Pruebas de baja induccion en estatores de grandes generadores
y motores core imperfection detection – ELCID test," www.
inducor.com.ar/articulostecnicos/pruebas_de_baja_induccion_
en_estatores_de_grandes%20_generadores.pdf, Buenos Aires,
Argentina,Inducor Ingenieria S.A., 2013.
[23] Z. Posedel, "Method and apparatus for inspecting laminated iron
core stacks for interlamination shorts," Patent US 4,996,486,
1991.
[24] Z. Posedel. "Inspection of stator cores in large machines with
a low yoke induction method—measurement and analysis of
interlamination short-circuits," IEEE Transactions on Energy
Conversion, vol. 16(1), pp. 81-6, 2001.
[38] J.-M. Bourgeois, F. Lalonde, "Apparatus and method for evaluation
a condition of a magnetic circuit of an electric machine," Patent US
5,990,688, 1999.
[39] G. B. Kliman, S. B. Lee, M. R. Shah, R. M. Lusted, N. K. Nair.
"A new method for synchronous generator core quality evaluation,"
Proceedings of IEEE Power Engineering Society General Meeting,
vol. 3(3), pp. 1888-93, 2003.
[25] M. Fischer, M. Metala, W. Abbasi, J. Shelton, "Siemens MultiFrequency Core Analysis System (SMCAS) - progress update"
presented at: EPRI Turbine Generator Users Group Meeting 18-20
August, St Louis, USA, 2010.
Cigre Science & Engineering • N°4 February 2016
51
[40] G. B. Kliman, J. A. Mallick, M. R. Shah, "Method and system for
detecting core faults," Patent US 6,469,504, 2002.
of Mechanical Engineers, Conference Proceedings, vol. 184(9), pp.
84-92, 1969.
[41] G. B. Kliman, M. R. Shah, S. B. Lee, "Differential sensor apparatus
and method for laminated core fault detection," Patent US
6,873,152, 2005.
[57] H. Bohle. "Magnetic reluctance of joints in transformer iron," IEE
Journal, vol. 41(191), pp. 527 - 32, 1908.
[58] G. Klempner, I. Kerszenbaum, Handbook of large turbo-generator
operation and maintenance, 2 ed. El-Hawary ME, editor: IEEE
Press / Wiley, 2008,
[42] S. B. Lee, M. R. Shah, G. B. Kliman, T. W. Mall, N. K. Nair, R.
M. Lusted, "An advanced technique for detecting inter-laminar
stator core faults in large electric machines," presented at: Industry
Applications Conference, Conference Record of the 38th IAS Annual
Meeting 12-16 October 2003.
[59] "Cogent non oriented electrical steel. Typical data," Surahammars
Bruks AB, 2013.
[60] D. R. Bertenshaw, A. C. Smith, "Lamination joint reluctance in
stator cores - a forgotten problem?," presented at: PEMD 2014, The
7th IET International Conference on Power Electronics, Machines
and Drives, 8-10 April, Manchester, UK, 2014.
[43] S. B. Lee, G. B. Kliman, M. R. Shah, T. W. Mall, N. K. Nair, R.
M. Lusted. "An advanced technique for detecting inter-laminar
stator core faults in large electric machines," IEEE Transactions on
Industry Applications, vol. 41(5), pp. 1185- 93, 2005.
[61] D. R. Bertenshaw, A. C. Smith, C. W. Ho, T. Chan, M. Sasic.
"Detection of stator core faults in large electrical machines," IET
Electric Power Applications, vol. 6(6), pp. 295-301, 2012.
[44] S. B. Lee, G. B. Kliman, M. R. Shah, N. K. Nair, R. M. Lusted. "An
iron core probe based inter-laminar core fault detection technique
for generator stator cores," IEEE Transaction on Energy Conversion,
vol. 20(2), pp. 2005.
[62] D. R. Bertenshaw, J. F. Lau, D. J. Conley, "Evaluation of EL CID
indications not associated with stator core interlaminar insulation
faults," presented at: Electrical Insulation Conference, June,
Annapolis, USA, 2011.
[45] S. B. Lee, G. B. Kliman, M. R. Shah, D. Kim, T. W. Mall, N. K.
Nair, R. M. Lusted. "Experimental study of inter-laminar core
fault detection techniques based on low flux core excitation," IEEE
Transactions on Energy Conversion, vol. 21(1), pp. 85-94, 2006.
[63]D. R. Bertenshaw, A. C. Smith, "The impact of non-linear
permeability on electromagnetic stator core tests," presented at:
PEMD 2014, The 7th IET International Conference on Power
Electronics, Machines and Drives, 8-10 April, Manchester, UK,
2014.
[46] S. B. Lee, G. B. Kliman, T. W. Mall, "Test probe for electrical devices
having low or no wedge depression" Patent US 6,927,598, 2005.
[47] S. B. Lee, G. B. Kliman, M. R. Shah, T. G. Richter, "Test probe,"
Patent US 6,847,224, 2005.
[48] S. B. Lee, D. Kim, T. W. Mall, "Manual probe carriage system and
method of using the same" Patent US 7,208,971, 2007.
12. Biographies
[49] G. B. Kliman, R. J. Carl, M. R. Shah, S. B. Lee, "Universal sensor
probe with adjustable members configured to fit between a plurality
of slot openings of varying widths" Patent US 7,466,126, 2008.
David Bertenshaw obtained a BSc(Eng) degree at
Imperial College London in 1970, is a Fellow of the IET
(formerly IEE) and was Director of R&D for lighting and
control system manufacturers for many years. He moved
to large electrical machine test systems in 1998 with
ADWEL International, now part of Iris Power, managing
their EMEA business and development of stator core
test systems. Having completed a PhD at University of
Manchester in 2015 researching electromagnetic stator
core testing, he now provides consultancy services
through ENELEC LTD.
[50] "Generator test and inspection fact sheet," GEA13004B, GE Energy,
2011.
[51] ELIN-UNION, "Arrangement for measuring magnetic characteristics," Patent AT390522B, 1984.
[52]J. Ramirez-Nino, S. A. Pascacio. "Novel method for detecting
interlamination short-circuits in a stator core based on
measurement of eddy-current losses in power generators," Institute
of Physics Publishing Measurement Science and Technology, vol. 14,
pp. 2104-10, 2003.
[53] "Generator core overheating risk assessment - core model studies,"
EPRI, USA., Report No. 1009855, 2004.
Sandy Smith was previously at Imperial College,
Cambridge University and Brook Crompton as Head
of Research. He joined UMIST (Manchester) in 2000
as a Senior Lecturer, becoming Professor of Electrical
Machines in 2007 and also Director of the Roll-Royce
University Technology Centre on Power Conversion
Systems. He is a Fellow of the IET (formerly IEE),
SMIEEE, and currently Editor of the IET Journal –
Electrical Systems in Transportation. Research interests
are in design and modelling of motors, generators and
drives.
[54] D. R. Bertenshaw, B. Chapman, "EL CID results interpretation in
the presence of substantial core faults," presented at: EPRI European
Turbine Generator User Group Workshop, June Madrid, Spain,
2009.
[55] C. W. Ho, D. R. Bertenshaw, A. C. Smith, T. Chan, M. Sasic. "Threedimensional finite element analysis of large electrical machine
stator core faults," Electric Power Applications, IET, vol. 8(2), pp.
60-7, 2014.
[56] T. J. Roberts. "Determination of the thermal constants of the heat
flow equations of electrical machines," Proceedings of the Institution
Cigre Science & Engineering • N°4 February 2016
52