The Impact of Humidity on PD Inception Voltage as a Function of Rise-Time in
Random Wound Motors of Different Designs
M. Fenger, G. C. Stone and B. A. Lloyd
Iris Power Engineering Inc.
1 Westside Drive, Unit #2
Toronto, Ontario – M9C 1B2
CANADA
Abstract: Random wound stator windings in motors,
operating in utility and industrial plants, have been
reported to suffer from premature failures. Although
several failure mechanisms exist, dissection of failed
stators show some of these failures have been caused by
exposure to fast rise-time voltage surges coming from
inverter drives. Partial discharges may occur between
turns during fast rise-time voltage surges. Previously,
the relationship between Discharge Inception Voltage
(DIV) and rise-time has been characterized for magnet
wire insulation samples and stators of various designs
under laboratory conditions. As well, DIV and surge
environment measurements performed in the field on
motors suspected to exhibit stator winding problems
related to surge voltages have been carried out.
In this paper, the influence of ambient humidity on DIV
for a magnet wire insulation sample is studied. The
results are discussed in relation to DIV curves obtained
on stators with confirmed problems in regards to
withstanding the surge voltage environment applied
during normal operation. The paper presents and
discusses the results of the experiments and does not
attempt to put forward a theoretical model for discharge
behaviour under surge conditions in random wound
stators.
modulated (PWM) type that use insulated gate bipolar
junction transistors (IGBTs) can create tens of
thousands of fast rise-time voltage surges per second.
Previously, anecdotal evidence suggested that voltage
surges from IFDs can lead to gradual deterioration and
eventual failure of the insulation [4-6]. In this
publication, data, which directly links the presence of
PD to the applied voltage surge environment, is
presented.
Previously, it was experimentally shown that as much
as 75% of the surge voltage applied to the terminals can
be distributed throughout the first coil [8]. It was also
argued that, in order to assess the severity of the
electrical surge environment in which a motor operates,
a measurement fully characterizing the voltage surge
distribution,
i.e.
quantitatively
characterizing
relationship between rise-time, surge magnitude and
pulse repetition rate, must be performed.
An example of a fully characterized surge environment
is given in Figure 1. As can be seen, the plot shows the
relationship between surge rise-time and voltage surge
"Calibrated" Surge Plot
Introduction
3.65
3.30
2.55
2.20
1.86
1.45
1.10
Magnitude (pu)
2.89
DIV Curve
0.76
0.36
0
100
200
300
400
500
600
700
800
900
1000
1100
1200
1300
1400
1500
Researchers have understood for over 70 years that fast
rise-time voltage surges from a circuit breaker closing
can lead to an electrical breakdown of the turn
insulation in motor stator windings [1]. If the turn
insulation is of an insufficient thickness, or has aged in
service, the insulation will puncture when a short risetime voltage surge occurs. Punctured turn insulation
allows for a very high circulating current to flow into
the affected copper turn, rapidly melting the copper
conductors, which, in turn, results in burning/melting of
the slot liner insulation, leading to a stator winding
ground fault [2,3].
0.00
Rise-Time tr (ns)
Rapid advances in power electronic components in the
past decade have lead to a new source of voltage
surges. Inverter-fed drives (IFDs) of the pulse width
Figure 1: Example of a Calibrated Surge Plot [12].
Figure 3: Picture of Motor 1 After Failure
magnitude.
A color-coding scheme provides
information of the pulse repetition rate. Voltage surges
above the DIV curve are likely to give rise to partial
discharges while voltage surges below the DIV curve
will not.
Background
The following case study originates from a power
station. The net electrical output is 390 MW. The plant
was put into operation in July 1997. A large number of
variable speed drives have been installed for the
operation of pumps to reduce the unit’s house load.
Motors rated 90 kW (120 H.P.) and below are mainly
supplied from 400 V supplies whereas motors rated
above 90 kW (120 H.P.) are supplied from the 690 V
supplies. A total of nine motors, of which seven
participate in a surge environment survey, are supplied
from the 690 V bus bars. The sizes of these motors
vary from 130 kW (175 H.P.) up to 1,890 kW (2,520
H.P.). The motors and the associated inverters are from
the same supplier.
Since the commissioning two of the seven motors have
been subjected to stator winding failures. As can be see
from Table 1, one motor (850 kW) failed three times
within the first 36 months of service and another motor
(680 kW) failed once after 36 months. For Motor 1, the
Motor
Application
Power Rating
[kW]
1
Main cooling pump
850
2
Condensation pump
680
Table 1: Failure Times for Motors 1 and 2.
Figure 2: Enhanced Close Up of Motor 1 Failure. Local
Melting of the Stator Winding Is Clearly Visible.
third failure occurred after less than 6 days of service
after the 2nd failure – see Figure 3 and Figure 2. After
the first failure on Motor 1, the motor manufacturer was
contacted and it was suggested that fast rise-time
voltage surges originating from the IGBT inverter may
have lead to the premature winding failure. The
manufacturer, who concluded that the failure was
accidental and that it was quite unlikely that a similar
failure would occur, rejected this theory.
Following the second failure on the same motor, the
theory of fast rise-time voltage surges being the cause
of failure was brought up again. Once more, the
manufacturer rejected this theory. The conclusion was,
as before, that the failure was accidental.
The
manufacturer accepted however to perform on site
voltage measurements at the motor terminals in order to
assure the customer that the failure was accidental and
not caused by voltage surges. These measurements
were performed only two days before the third failure
on the same motor.
One month later, a second motor failed. Following this,
Date of fault
10-06-1998
10-02-2000
14-09-2000
13-10-2000
Total operating
hours
2140
13044
13177
9440
Number of
starts
4
119
152
753
a decision to perform an independent measurement of
the electrical surge environment on all 690 V motors,
using on-line surge monitoring equipment was made
[12].
"Calibrated" Surge Plot For Motor 1 3.65
3.30
2.89
DIV Curve
2.20
1.86
1.45
Instrumentation: Voltage Surge Test Setup
Surges resulting in PD
Magnitude (pu)
2.55
Although both motors that failed were still covered by
the manufacturer’s warranty, the unforeseen failures
have lead to considerable expense to the power station.
1.10
TM
The instrumentation used to measure Partial Discharge
Inception Voltage (DIV) is an off-line test and is
thoroughly described in [13]. In short, via a Baker
Surge Tester a 50 ns rise-time surge is applied to an
insulation sample or a stator winding. If of sufficient
magnitude, the surge voltage will give rise to a partial
discharge. Using specialized equipment, PDAlertTM,
the induced current signal on the insulation sample
conductor from the partial discharge may be extracted
from the surge signal it self. The net output from the
instrument is thus a voltage signal originating from the
partial discharge current pulse.
Instrumentation: DIV Test Setup
The test procedure is described thoroughly in [12] but
repeated in short here: Having connected a stator to the
surge source, the surge magnitude was increased at
approximately 200 volts per second from zero volts
until a partial discharge was observed. The surge
magnitude was then quickly decreased to zero volts.
The procedure was repeated seven times for each risetime. Based on this, the mean (average) DIV was
calculated for each rise-time. In addition, the ambient
temperature and humidity was logged.
Surges not resulting in PD
0.76
0.36
1500
1400
1300
1200
1100
900
1000
800
700
600
500
400
300
200
100
0
0.00
Rise-Time tr (ns)
Figure 4: Calibrated Surge Plot for Motor 1
shows the opposite relationship, namely an increase in
DIV with increasing rise-time as explained by the
distribution of voltage across the first turn as a function
of surge rise-time [12]. One possible explanation is that
the discharge source is located closer to the neutral end
rather than the line end of the stator winding.
Figure 4 shows the measured surge plot for Motor 1
with the DIV curve super imposed. Surges above the
curve may give rise to PD whereas surges below the
curve will not give rise to PD.
As can be seen in [13], the DIV measured on Motor 1 is
low compared to the DIV levels measured on virgin
stators. This indicate that for aged windings, i.e.
windings subjected to real operating conditions, DIV is
more related to the surge magnitude rather than the rise-
DIV vs. Rise-Time For Motor 1
3.5
3
2.5
DIV [pu]
Via specialized instrumentation, SurgAlert , the
voltage surge environment applied to a stator winding
can be fully characterized. The instrumentation is
thoroughly described in [6,8] but will be briefly
discussed here.
By temporarily or permanently
attaching a low inductance resistive or capacitive
voltage divider to one of the motor terminals, the risetime and magnitude of each surge occurring within a
given time interval may be measured. Thus, the surge
environment experienced during normal on-line
conditions may be fully characterized.
2
1.5
1
0.5
For Motor 1, the average DIV was 2.38 per unit. A
curve of the DIV versus rise-time for Motor 1 is given
in Figure 5. The curve shows the DIV to decrease with
increasing rise-time. This is surprising as other curves
obtained on new stators prior to being put into service
0
0
100
200
300
400
500
600
Rise-Time [ns]
Figure 5: Discharge Inception Voltage vs. Rise-Time For
Motor 1.
1200
The Effect of
Humidity On DIV
DIV [V]
Rise-Time Humidity DIV Standard
[ns]
[%]
[V] Deviation
39
1012
61
50
969
24
967
60
957
8
86
929
32
1150
1100
1050
1000
950
900
514
39
60
75
89
1041
1003
996
989
97
15
25
16
850
800
750
967 ns Rise-Time
20
Table 2: Discharge Inception Voltage versus
Humidity Results
514 ns Rise-Time
Humidity [%]
700
40
60
80
100
Figure 6: The Effect of Humidity
time coupled with the probability for occurrence of a
free electron, which increases with increasing (slowing)
rise-time [13].
addition, the ambient temperature and humidity was
logged. Based on visual observations, the nature of a
PD pulse was described and visual observations
regarding the dynamic behavior PD were logged.
The Effect of Humidity
The results are given in Table 2 and represented
graphically in Figure 6. The results presented here are
for rise-times of approximately 960ns and 510 ns
respectively. The ambient temperature did not deviate
more than 0.3°C during the tests.
As mentioned earlier, the surge environment is
characterized during normal operating conditions, i.e. at
a given stator winding temperature and ambient
humidity. The output of the drive as well as the wave
impedance of the drive, the transmission cable and
stator winding defines the surge environment measured
at the motor terminals. Assuming the wave impedance
of these components does not change significantly with
temperature, the surge environment is not expected to
change significantly with operating temperature.
However, ambient humidity may have an impact on the
measured DIV and, thus, an impact on the interpretation
of the calibrated surge plot. To initially help assess the
impact of ambient humidity on DIV under surge
conditions, a simple experiment was performed. A
twisted pair sample was placed in a closed chamber
where the humidity could be controlled. Prior to
testing, the insulation specimen was cleaned thoroughly
with isopropyl alcohol and, then dried; the specimen
was introduced into the test setup with one strand
connected to the surge electrode and the other strand
connected to ground. Following this, surges were
applied to the specimen.
The surge magnitude was increased at approximately
200 volts per second from zero volts until a partial
discharge was observed. The surge magnitude was then
quickly decreased to zero volts. The procedure was
repeated seven times for each rise-time. Based on this,
the mean DIV was calculated for each rise-time. In
As can be seen, both rise-times show the DIV to
decrease with increasing humidity. Specifically, for a
967 ns rise-time surge, the DIV at 39% humidity was
1012 V whereas the DIV at 86% humidity was 929 V
corresponding to a decrease of 8%. For a 514ns risetime surge, the DIV at 39% humidity was 1041 V
whereas the DIV at 89% humidity was 989 V
corresponding to a decrease of 5%.
As the electrical field distribution between a strand of
two wires is defined by the spacial vector sum of the
Laplacian and Poissionian field, it may be argued that
the measured decrease in DIV with increasing humidity
indicates the influence of the Poissionian field over the
Laplacian field in the partial discharge process for this
type of insulation system. Specifically, the surface
conductivity of a strand will be the determination
parameter for the Poissionian field. As the humidity
increases and moisture is deposited on the magnet wire
surface, the conductivity is expected to increase. The
time constant for charge decay thus decreases and the
Poissionian field should decrease thus decreasing DIV
for this type of test.
From a practical point of view, these results show that
although a correlation between humidity and DIV
exists, the decrease in DIV due to increasing humidity
is minor. This would indicate that the DIV curves
obtained on stators in-situ for the purpose of assessing
the impact of the applied surge environment during
normal operating conditions do not change significantly
with changes in ambient humidity.
Conclusions
The DIV measurements performed on Motor 1 clearly
documented that the Discharge Inception Voltages,
under surge conditions, were lower than the max surge
magnitudes measured on-line during normal on-line
operations. This strongly indicates that the root cause
of the failures experienced were indeed the presence of
partial discharges.
The measurements investigating the effect of humidity
on DIV showed little decrease in DIV between 40%
humidity and approximately 90% humidity. This
indicates that the DIV curve, i.e. the relationship
between DIV and rise-time measured on a stator during
off-line conditions and super imposed on the surge plot,
does not change significant with ambient humidity.
At 100% relative humidity condensation will occur.
This may impact the DIV significantly.
It may thus be argued a DIV curve obtained at less than
90% relative humidity is a strong tool for interpretation
of the impact of the voltage surge environment with
respect to the presence of partial discharges during
normal on-line operations.
Future Work
A study investigating the influence of humidity and
temperature on DIV for various types of stranded
magnet wire under various conditions has been
initiated. The work will include studying the effects of
condensation on DIV.
References
[1] E.W. Boehne, Voltage Oscillations in Armature
Windings Under Lightning Impulses, Trans AIEE,
p1587 ff., 1930.
[2] M.T. Wright, S.J. Yang, and K. McCleay, General
Theory of Fast-Fronted Interturn Voltage
Distribution in Electrical Machine Windings, Proc.
IEE, Part B, p 245 ff., July 1983.
[3] A.L. Lynn, W.A. Gottung, D.R. Johnston, Corona
Resistant Turn Insulation in AC Rotating
Machines, Proc. IEEE Electrical
Conference, p 308 ff, Oct. 1985.
Insulation
[4] W. Yin, et al, Improved Magnet Wire for InverterFed Motors, Proc. IEEE Electrical Insulation
Conference, p. 379 ff., Sep. 1997.
[5] E. Persson, Transient Effects in Applications of
PWM Inverters to Induction Motors, IEEE Trans
IAS, p1095, Sep. 1992.
[6] G.C. Stone, S.R. Campbell, M.Susnik, New Tools
to Determine the Vulnerability of Stator Windings
to Voltage Surges form IFDs Proc. IEEE Electrical
Insulation Conference, p 149. Oct. 1999.
[10] C. Lanier, A Novel Technique For The
Determination of Relative Corona Activity Within
Inverter-Duty Motor Insulation Systems Using
Steep-Fronted Voltage Pulses, IEEE Conference
Record of The International Symposium on
Electrical Insulation, p 229 ff., June 1998.
[11] G.C. Stone, S.R. Campbell and S. Tetreault,
Inverter Fed Drives: Which Motor Stators are at
Risk?, IEEE Industrial Applications Magazine, p.
17 f.f. Sep. 2000.
[12] J. Pedersen and M. Fenger, Case Studies in
Measuring Surges From IFDs Applied To Random
Wound Motors, Conference Proceedings of the
2002 IEEE Cement Industry Conference, May
2002.
[13] M. Fenger, S. R. Campbell and G. Gao, The
Impact of Surge Voltage Rise-Time on PD
Inception Voltage in Random Wound Motors of
Different Design, 2001 Annual Report, Conference
Proceedings of the 2001 IEEE CEIDP.