Low Frequency Dielectric Response of Thermally Degraded Epoxy-Mica Insulation

B. Nageshwar Rao, A. Sudhindra and B. Ramachandra1

Central Power Research Institute, P.B.No. 8066,
Prof. Sir. C.V. Raman Road, Bangalore 560 080, India
1
P.E.S College of Engineering Mandya, India
Email: [email protected]
Fax: 91-080-23604435

Abstract: Dielectric spectroscopy measurement is as an additional tool for condition assessment of stator
essentially a dissipation factor measurement performed insulation a study was undertaken in the laboratory of
at multiple frequencies ranging from few mHz to kHz. CPRI. This paper presents and discusses the data on low
Low frequency dielectric spectroscopy measurement frequency dielectric response of thermally degraded
has been performed on number of sound and thermally epoxy mica composites and motor coils.
aged epoxy-mica representative laminates and on actual
motor coils. Dielectric loss angle, Power Factor, Principle of Dielectric Spectroscopy
capacitance was measured at frequencies ranging from 1
mHz to 1 kHz. The dielectric response exhibited by of During the process of thermal ageing of insulation or
thermally aged epoxy-mica insulation at lower due to ingress of moisture, polar species are generated
frequencies shows the predominant influence of in the insulation as by-products due to increased rates of
interfacial polarization on the total dielectric response of chemical reactions that occur at high temperatures. The
epoxy-mica layer. polar species, which vary in size and density with the
ageing, orient or move in different direction under the
Keywords: Epoxy-mica, dielectric spectroscopy, low influence of electric field and thus contribute to the
frequency dielectric response, interfacial polarization. losses in the insulation. Dissipation factor tests below 1
Hz are responsive to these polar species and this
INTRODUCTION response can be studied using the dielectric functions
[6]. The frequency dependence has been found to be
The stator winding of rotating machine is comprised of different in the different functions [7] and are based on
mainly mica with inorganic reinforcing, epoxy bonding basic equations
and impregnating materials. Mica minerals consisting of ε* = ε′ – j ε" ………..(1)
layers of silicates are separated by alternating layer of M* = M’ + j M" ………(2)
metal oxides and metal ions. Mica which has unique Z* = Z’ – j Z" …….. (3)
combination of excellent physical, thermal and Y* = Y′ + j Y" …… ..(4)
electrical properties find wide applications in high
voltage machines. Mica paper/synthetic resin tan δ = ε* /ε′ = M" /M′= Z′/Z" =Y’/Y" ……(5)
combinations (polyester, epoxy, silicone resins, etc.) are
employed almost exclusively in the form of tapes, ε* = 1/M* = 1 /jωCoZ* = jωCo Y* ……. .. (6)
which are wrapped around the conductor of stator
winding, impregnated and then cured. The stator where ε* is the complex permittivity, M* is the complex
winding during its normal operational service undergoes modulus, z* is the complex impedance, Y* is the
ageing and the insulation deteriorates due to various complex admittance and ω is the angular frequency.
operational stresses. The progressive deterioration of Insulation diagnostics is based on material
high voltage machine insulation is assessed through non characterization and therefore material models are often
destructive techniques like measurement of Insulation used. Impedance models are represented in polar and
Resistance, Polarization Index, Dissipation factor, rectangular co-ordinates. More often a complex
capacitance, Partial Discharge (PD) activity, etc. [1,2,3]. capacitance model describes insulation impedance as a
Methods based on Low frequency dielectric complex capacitance, where the imaginary part of the
spectroscopy are also reported for stator insulation capacitance represents the losses. The complex
condition assessment [4,5]. Central Power Research capacitance model is defined as follows:
Institute, Bangalore (CPRI), India has been carrying out Z= 1/jωC where C = C′ – jωC" ………. (7)
extensive diagnostic testing work to various utilities,
Industries in condition assessment of high voltage Complex C:
machines in service. For better understanding of the C′ = Re{1/jωZ} ……………. (8)
deterioration mechanisms due to thermal ageing and to ΔC′ = C′+k ……………...(9)
explore the possibility of using dielectric spectroscopy C" = -Im {1/jωZ} …………….(10)
Where ΔC′ is defined as the capacitance, C′, with an interfacial polarization on the total dielectric response of
arbitrary constant k. epoxy-mica layer.

Dielectric spectroscopy

Dielectric spectroscopy measurement is essentially a

dissipation factor measurement performed at multiple
frequencies ranging from few milli Hz to kHz.
Programma, IDA 200 make instrument measures
impedance at a specific frequency and amplitude.
Parameters such as capacitance, tan δ and cosφ (Power
Factor) are calculated from the measured impedance as
follows:

C′= Re {1/jωZ} …….. …...(11)

PF = cos φ = Re{Z}/ |Z| …….…(12)
Tan δ = -Re{Z}/Im{Z} ….……(13)
Fig. 1 Experimental setup
EXPERIMENTAL SET UP

The experimental investigations were carried out on Imp. ()

epoxy-mica based 6.6 kV motor coils and laminar 1
composites. The insulation investigated is a complex 5 4
dielectric consisting of three components, viz. Glass Sample 1
tape, mica paper and binding resin. Every component Sample 2
0.1 Sample 3
has its own volume fraction, permittivity and
Sample 4
conductivity. The coils/laminates were subjected to
Tan δ

Sample 5
accelerated thermal aging using thermal oven and its
3
response to low frequency (dielectric spectroscopy) was
0.01 2
studied on both unstressed and thermally aged samples. 1
Figure 1 shows the picture of the experimental setup
that was used in this investigation for measurement of
tan δ at power frequency and voltages upto 5 kV. Tan 0.001
delta and Capacitance measuring bridge, Doble make 0.001 0.01
0.1 1 10 100 1000
was used (Extreme right). Dielectric loss angle, and Frequency: (Hz)
capacitance was measured at frequencies ranging from 1 . Fig. 2a Tan δ vs freq. unstressed epoxy-mica
mHz to 1 kHz at 140 Vpeak using Insulation Dielectric
spectroscopy instrument (extreme left). Imp. ()
1
RESULTS AND DISCUSSION 1 Sample 1
5 Sample 2
4
Dielectric spectra of unaged specimens Sample 3
PF (cos φ)

Sample 4
0.1 Sample 5
In order to characterize the low frequency dielectric
response of thermosetting epoxy-mica insulation, the
dielectric losses were measured. Figs. 2a and 2b show 3 2
the variation of tan δ and power factor (cos φ) as a 0.01
function of frequency obtained on 1mm thick, 100mm x
100mm unstressed epoxy-mica laminates. It is seen
from these figures that the variation of tan δ and cos φ is
0.001
not significant above 1 Hz but there is wide dispersion
0.001 0.01 0.1 1 10 100 1000
in values below 1 Hz. A slight increase of the dielectric Frequency: (Hz)
loss was noticed below 1 HZ upto 0.001Hz signifying Fig. 2b PF (cos φ) vs frequency
the presence of polar species which vary in size and for unstressed epoxy-mica laminate
density contributing to the losses in the insulation.
The power factor is maximum at the lowest frequency Dielectric spectra of Thermally aged samples
of 0.001 Hz. The maxima in the tan δ curves and the
significant increase in capacitance at low frequencies The laminate samples were thermally stressed at 473 K
(figure not shown) confirm predominant influence of for 1000 hours. Fig. 3 shows a comparative plot of the
variation of loss factor as a function of frequency for
selective sound and thermally aged specimens. The Imp. ()
curves at the upper band show the variation for 1
thermally aged specimens while lower band of curves

PF (cos φ)
are for that of sound specimens. The degradation can
be distinctly seen with aged specimens showing higher
values (tan δ = 1 at 0.01 Hz) than the sound
specimen Sample 1
0.1
Sample 2
Imp. () Sample 3
Sample 4
1 Sample 5
Thermally aged at 473o K
for 1000 hrs
0.01
0.1 0.001 0.01 0.1 1 10 100 1000
Tan δ

unstressed Frequency: (Hz)

Fig. 4b. PF (cos φ) vs frequency for thermally aged .
. . samples (473 K, 2000 hours)
0.01

Results of 6.6 kV, Motor coils

0.001 Motor coils of 6.6 kV, were subjected to accelerated

0.01 0.1
1 10 100 1000 electrical and thermal stresses at 5 kV/mm and 453 K
Frequency: (Hz)
for a duration of 1000 hours. Dissipation factor, Power
Fig. 3. Tan δ vs frequency for sound and aged specimen factor and capacitance measurements were conducted in
the frequency range 1 mHz to 1 kHz. Figures 5a to c
show the variation of dissipation factor, power factor
(tan δ =0.12 at 0.01 Hz) and the trend similar to that of and capacitance as a function of frequency. The trend
sound specimen. The dissipation factor measured at 1 observed is similar to the trend observed in case of
kV, 50 Hz using Tan δ and Capacitance bridge laminates. One of the coil which failed during power
instrument was in agreement with the dielectric frequency tan δ measurement at 3 kV, has shown a
spectroscopy results. However the values measured at different trend when measured at very low voltage 140
power frequency were much higher at higher voltages Vrms and variable frequency as seen from figures 5a to c.
due to other effects like partial discharge activity etc. In case of a failed coil, the maxima in tan δ is observed
at 10 Hz and decreases below 10 Hz.
Figures 4a and 4b show plots of variation of tan δ and
power factors with frequency for samples thermally Imp. ()
stressed at 473 K for 2000 hours. From figure 4a it is
10
observed that there is a significant loss in dielectric (tan Failed coil
δ = 4 to 10 at 0.01 Hz) due to higher degree of
degradation of insulation compared to samples aged for
1000 hours (tan δ = 1 at 0.01 Hz). 1
Tan δ

Imp. ()
100 0.1
Sample 1 coil 1 coil 4
Sample 2 coil 2 coil 5
10 Sample 3 coil 3 coil 6
Sample 4 0.01
Sample 5 0.001 0.01 0.1 1 10 100 1000
Tan δ

1 Frequency: (Hz)
. Fig. 5a Tan δ vs frequency

0.1
The shape of frequency dependence of the loss factor
can be interpreted as follows. The dipole processes is
0.01 increasingly overlapped by the effect of conduction in
0.001 0.010.1 1 10 100 1000 the resin. At still lower frequencies the barrier effect of
Frequency: (Hz) the lower conductivity of the organic component
.Fig. 4a Tan δ vs frequency for thermally aged . . . manifests itself. Charge is accumulated at the interface
samples (473 K, 2000 hours) and this causes redistribution of the field strength in the
composite dielectric and decrease in the resulting
conducting current.
 Imp. () REFERENCES
1 Failed
coil [1]. Stark K.H., “ Assessment of the insulation
serviceability of Turbo generator stators and of HV
bushings”, Proceedings A, 1962 pp 71-88
PF (cos φ)

[2]. J.S.Simons, “Diagnostic testing of high voltage

machine insulation”, IEE Proceedings, Vol-127,
0.1
part B, No.3, 1980
coil 1 coil 4 [3]. Kelen A, “ Diagnostics of HV machine insulation
coil 2 coil 5 from inspection to continuous monitoring”,
coil 3 coil 6
CIGRE Symposium 05-97, Section 10-2, 1987
[4]. S. Cherukupalli et al, Application of some novel
0.01 non-destructive diagnostic tests for condition
0.001 0.01 0.1 1 10 100 1000 assessment of stator coils and bars following
Frequency: (Hz) voltage endurance tests”, Conf. Record of the 2002
.Figure 5b. PF (cos φ) vs frequency IEEE International Symposium on Electrical
Insulation, Boston, MA pp-565-69.
Imp. (F) [5]. H.G. Sedding “Application of low frequency
technique for stator insulation diagnostics” Iris
1E-7
Failed coil Rotating Machine Conference, June 2002, San
Antonio, TX
[6]. J. Ross McDonald, Impedance Spectroscopy
Capacitance

1E-8 Emphasizing Solid Materials and Systems, John

Wiley and Sons, New York, 1987
[7]. R. Gerhardt, “Dielectric and Impedance
spectroscopy J. Phys. Chem. Solids 55, 1994, pp-
1E-9
1491-1506
coil 1 coil 4
coil 2 coil 5
coil 3 coil 6
1E-10
0.001 0.01 0.1 1 10 100 1000
Frequency: (Hz)
. Figure 5c. Capacitance vs. frequency

CONCLUSIONS

The use of dielectric spectroscopic technique has

resulted in a better knowledge of the electrical
properties of thermally degraded epoxy–mica stator
winding insulation. The dielectric response exhibited by
of thermally aged epoxy-mica insulation at lower
frequencies shows the predominant influence of
interfacial polarization on the total dielectric response of
epoxy-mica layer. This method could be possibly used
to evaluate the condition of the bulk of the insulation
and its degree of aging.

ACKNOWLEDGEMENT

The authors thank the management of Central Power

Research Institute for giving permission to publish this
paper. Thanks are also due to Smt. Arunjothi and Miss
G. Latha for their assistance in carrying out the
experiments.