Technology for Detecting Wet Bars in Water-Cooled

Stator Windings of Turbine Generators

Y. Inoue, H. Hasegawa, S . Sekito, M. Sotodate, H. Shimada and T. Okamoto
Toshiba Corporation
Yokohama, Japan

-
Abstrad Stator bars of largecapacity direcl water+ooled
turbine generators may suffer from cooling water leak into the
The capacitance measurement technique, however, can not be
applied to bars which have in the coil end
g r o u d w f l hulation by galvanic corrosion in clip to pbm& insulation. For detecting wet bars Of we
brazed area. Once the cooline water leaks. the water accelerates developed a new... technique can be applied to such
insulation deterioration ofbars by bydrol&i which is cawed by designed bars I*'. The principle of the new technique is based
we1 and beat. Consequently early detection of the water leak is on the phenomenon that the potential of internal electrodes
necessary for safety operation and proper maintenance. increases because of the decrease of the impedance between
While common insulation of stator b a n are made Of only the conductor and the innermost internal electrode after
mica t a p lapped, tbere are bars with another design in which ocrmeat,on of water into the insulation, ne
some conducting layers (internal electrode layers) are inserted in
intermediate layer of ground wail insulation at coil end. The
kffectiveness of thenew method is demonstrated with ,+I
straightbars, a 11 scale bar and an machine,
internal electrode layers work fu a voltage divider 80 that the
elcrlric field at fbe surface of end portion of bars are sopprmsed.
While a capacitance measuremen1 metbod is commonly used for IL OF
detecting wet bars with the former design of bars, the method A typical cross section of a wet bar (cooling water
can not be applied to the latter design of bars. For detecting wet
bars of tbe latter design we have developed a new and sensitive
permeated bars) is shown in Fig. 1. Approximately a half
method. inner insulation layer is discolored where the insulation layer
After our ao many experiences of capacitance metbod is is remarkably degradedby wet heat deterioration.
simply described, the detall of newly developed metbod for the Dielectric properties of sound bars and wet bars in a
ban with internal eledrodes is described in tbis paper.
domestic generator operated for long-term were measured
I. INTRODUCTION with LCR meter. The properties of permittivity (E) and
dielectric loss (tans) as a function of tkquency are shown in
Stator bars of large-capacity direct water-cooled turbine Fig. 2 and 3.
generators may suffer from cooling water leak into the
ground wall insulation by galvanic corrosion in clip to strands In a sound bar (Bottom C), E and tan6 keep almost constant
brazed area Once the cooling water leaks, the water (E=-4.5 and tan&l-2%) from 20 Hz to IMHZ. On the
permeates into ground wall insulation of stator bars and then contrary, wet bars have strong frequency dependence in the
causes rapid deterioration of the insulation by wet-heat dielectric properties. The permittivity of wet bars (Top A and
deterioration that is due to chemical reaction of hydrolysis of C) increases as the frequency decreases, and it become larger
binder resin in the insulation "I. Therefore early detection of than 15 (more tban 3 times as normal). The tan8 has quite
the water leak is necessary for safety operation and proper large value and has the maximum at near 1kHz. The
maintenance. phenomenon mentioned above can be understood by using
equivalent circuit model of double layer dielectrics of normal
Typical method for detecting wet bars is capacitance and cooling water permeated layer which has the property
measurement between a conductor and the surface of a bar shown in Fig. 4. Fig.4 also shows the quite high frequency
end portion We have developed an automatic method as dependence, and both E and tan8 increase up to 60 and 500%
well as a manual method. In the latter method a capacitance
measuring electrode is applied by hand, on the other hand, in
the former method the capacitance is measured automaticall
with a specially designed robot without removing a rotor PI,.
We have abundant experiences on these techniques. In this
paper comparison of accuracy of both methods is described
at around 30Hz in spite of remaining at 20 and 100% at capacitance is often measured at near lkHz in case of
LOOOHz. This suggests the measurement at lower frequency Toshiba. The capacitance for a wet bar increases with
gives higher sensitivity statistical significance.
We have three methods for measuring the capacitance.
20 r One is a method in which the electrode is pressed on the
surface of a bar by hand, another is a method in which the

I
15 . ........... electrode is placed at the proper position and then pressed
pneumatically @y controlled air pressure) and the other is a
I 10

5
......... .............

1 method with a measuring robot in which the electrode is

pressed pneumatically.
An appearance of a test using a manual pneumatical
electrode is shown in Fig.6. And a test with the robot
0 specially designed for measuring capacitance of stator bars
which moves around a rotor end rin and measures
capacitance automatically is shown in Fig.7 I. 6

.
...
...
...
...
...
...
..

0
10 100 loa loo00 lm lMMo00
Frequency a13
Fig. 3 Frequency depcndmcc of -8 of sampled wet bars
andarwndbar

600

400 3
a
E

............ 200

0
10 100 1oM) low0 1 m lowoo
Frequncy (Hz)
Fig. 4 Fregucncy dependace of E and tansfa water
m d lays
Fig.6 Manual capacitance mapping tee

m.DETECTION METHOD BY CAPACITANCE

MEASUREMEST Comparison of accuracy among measuring methods

A method described in this section is capacitance Capacitances were measured by three methods (manual
mapping test m which a capacitance between an electrode test with electrode pressed by hand and by air pressure and
and a conductor is measured as shown in Fig. 5 and the data automatic inspection using the robot) on the same generator
for all bars are statistically processed. By the trade-off and occasion. The measured capacitance are arranged
between sensitivity and measurement accuracy, the statistically and transformed to deviation score (z (U) value).

1338
 IV NEWLY DEVELOPED DETECTION METHOD FOR
WET BARS WITH INTERNAL ELECTRODES
The meihod described in section 3 can be applied to bars
without internal electrodes but it can not be applied to bars
with internal electrodes in the insulation layer at the coil end
portion. As there is no method to clecmcally detect water
permeation into [he insulation so far, we have developed a
new methodl'l.
A. Reason why capacilance method Is dificult lo apply

An equivalent circuit of a bar end portion with three

internal electrodes is shown in Fig.9. The capacitance Ca has
a large scatter because of large scatter on insulation thickness
at this area. So the Ca is not suitable for inspection in many
cases. The capacitance between the conductor and the
electrode b is calculated as a series capacitance of Cb and C l .
Rg.7 Robotic capacitancemapping est
Cb (less than several tens pF) is much less tban Cl (more
than several thousands pF), so h e capacitance between the
The transformed data are plotted in Fig.8, where the data of
conductor and the electrode b is nearly equal to Cb and is
three methods arc good correlation Correlation coefficient scarcely innuenced by the change in CI. Electrode c is also
among these methods is shown in Table I which shows that
not suitable in the same reason as tbe elecwode b.
these methods have good enough reliability.
We have experience of inspection of 239 times by manual
method and 5 1 machines by the robot so far. These abundant Main wall insulation
experiences increase our reliabilityin detecting wet bars.

- lnvmal C l C h d e Kz,v2
Rg.9 An q u ~ l e n"lt
l of bar md pmon wth thrse bmal
el&

B. New detection methodfor wet bars with internal electrodes

(a) Principle of new method
Deviation score of capacitance (robot) Potential VI and V2 of internal electrodes are calculated
Fig.8 Comelatian basreen capacitance m-ed using the equivalent circuit in Fig.9 as follows:
mth robot and by hand
VI=Vappl*
zz+w
21+22+23
Table 1 Correlation coefficient
Method Robot Man-air press. Man-hand press.
Robot I
Man-air press. 0.87 I
Man-hand press. 0.82 0.78 1
where Z1,Zz and a are impedance (complex) of
parallel circuits of Ci and Ri (i=1.2,3)
Zi = Ril (l+jwCiRi )

1339
 The potential VI and V2 increase after water permeation In case of the setup (A) the narrower the air gap is, the
into the inner side of the insulation layer as shown in Fig.9 better the accuracy of measured potential is. The frequency
because the water permeation gives rise to large increase of between 20% and 60Hz is appropriate. In case of the setup
capacitance and dielectric loss between the conductor and the (B) the lower the measurement frequency is, the smaller the
internal electrode K1 and hence decrease of the impedance thickness dependence of measured potential is. Judging &om
ZI, which increases the potential VI and VZ by the experimental result, the potential of the internal electrode
under insulation layer can be measured with sufficient
(21+22+23) e/
(21+22+23) accuracy when appropriate .arrangement of the probe of
As explained above the water permeation can be detected surface potential meter and an appropriate measurement
by measuring the potentials VI and/or V2 of intemal fiequency is applied.
electrodes of bars.
to surface potential meter
(a) Principle of Non-contact SMjace Potential Measurement
Although the potentials VI and V2 of internal electrodes
can not be manxed directly, they can be measured with a (A) Insulation Layer (I")
non-contact surface potential meter over the surface of bars with various thickness
as shown in Fig.10. An induced voltage on a sensor electrode
of a probe from the measured surface is processed to measure
the potential in non-contact surface potential measurement
system. The induced voltage is usually modulated to ac
voltage by means of vibration of electrode or rotation of
grounded sector type eleckode between the m r electrode
and the m d slrrface.
h Fig. 10 where the bar has three internal electrodes. the
surface potential has three steps. Therefore, the position and probe of nufacc
the potential of each internal electrode can he obtained &om potential meter
the w e of the surface potential which is obtained by (B) Insulation Layerw)
measuring the potential along the su+e of the bar. with various thickness

meter

Chductor End portion of

$A
1C p d , , source stator bar

Fig.10 sfhematic diagram ofnew daccficu method

(c) Measurement occuracy of suface potential meter

FRP air total
The potential of an internal electrode under insulation +5.75 0.25 6.0
layer was measured with a surface potential meter by the +5.5 0.5 6.0
setup as Fig.11. In the setup (A) the distance between the +4.7 1.3 6.0
+3.0 3.0 6.0
probe and the electrode was kept constant (6mm) and the + 0 6.0 6.0
thickness of inserkd insulation plate (FRP:Fiher Reinforced
Plastic) was changed, while in the setup (B) the inserted FRP 0 20 40 Bo eo
plate was changed with the gap between the probe and the Frequency EIS
FRF' plate kept zero. Measured potential of the setup (A) is
Fig.12 Frqucncy dcpcndenm of measured potential
shown in Fig. 12 and that of the setup (J3) is shown in Fig. 13
of cxpaimcntal s h l p (A)
as a function of the fiequency of applied voltage.

1340
 100 I 1
100

3 80
v
-
8 60
.
20

0
20 30 40 50 60
Frequency (Hz)
Fig.13 Frequency depmdenec of measured potential of Fig.15 Measured potential of intonal electmds KI, K2 and
expaimmtal srmp I?) conductor (KO)

(e) Experimenion afull scale bar

(d)Experimentfor ‘the effectiveness veriication wiih siroighi
bar model A stator bar for water-cooled turbime generator with three
internal electrodes was manufactured. An ‘2“letter type
seaight bars with three were ( was on both upPR and si& surface
manufactured to demonstrate that the condition where a little
at bar end ) and a flat plate electrode( which was applied on
water permeates into the layer between a conductor and an
upper surface at bar end ) were applied on the surface of the
innermost intemal electrode could be detected. The outline
bar and the of the were measured by the
of the test is shown in Fig.14. A wet sample (left figure in
surfacepotential meter, The measured potentialis plotted in
Fig.14) in which an electrode with 105” long was applied
Fig.lS, The change of the electrodes is
on the wet bar ‘Top A” shown in Fig.2 and 3 and model
clearly appeared in the figure, which that the
straight bars with three internal electrodes that had leads for
can be measured on the surface of
connecting with outer circuit were prepared. AC voltage was
applied to conductors of bars from adjustable voltage and
frequency AC power supply unit and the surface potential fl Experiment on an actual machine
over the electrodes KO (conductor), KI and K2 was measured
with a non-contact surface potential meter before and after
In actual machines, measured potential may have larger
connecting a small wet bar. variation due to the variation of insulation thickness and
flatness of bars by vanish or suppOrting structure.
The measured potential is shown in Fig. 15 as a function of Surface potential of bars of an actual generator (2p
the frequency of applied voltage. The potential of the 740MVA) of which bars do not have internal electrodes were
conductor and the internal electrodes K1 and K2 is measured measured. The potential of bottom bars were measured
at 90 to 95 % of applied or calculated voltage. After through gaps between upper bars The measured potential is
connecting small wet sample the potential of KI and K2 shown in Fig.16 and Table 2. The potential of top bars bas
increased by 1.28 times which is large enough compared with smaller variation than that of bottom bars. Even in the
scatter of measured potential by different operators and bottom bars with larger variation, the coefficient of
different bars. When straight bar model is used, this variation is 3.61%. The value is evaluated for
detecting methcd has demonstrated to have high repeatability detection reliability in the next section.
and high accuracy.

(‘‘Top” A in Fig 2 and 3) model bar with intemal electrodes

Fig.14 E x p e r m ” w t h e mth mdcl rtrslght bar

1341
 ___ I I
Condition. So in this condition the amount of permeated
water is considered to be less than a half of the bar in Fig.1.
The calculated potential is plotted in Fig.17. The potential of
the internal elecmdes V1 and V2 increases by 1.6 times at
below the power frequency in the wet condition. From tbis
........ ...........
result, the surface potential along a bar can be calculated in
wet condition. The s d c e potential at 5OHz in dry condition
(same as Fig. 15) and wet condition is shown in Fig.18. The
potential of internal electrodes is considered to be measured
clearly.
Then, the amount of potential change as shown in Fig.17
and 18 by water permeation is compared with the variation of
-1W 5 0 0 50 100 150 200 measured value in Fig.16 and Table 2.
Distance 6vm tbe end of in"!elearode (mm)
The potential variation of bottom bars is the largest in
Fig.15 Mcanrnd surface potential on the stator bar with three Table 2. Even the value of three times of the standard
intnnalelatrodcs -
deviation of the potential 3*3.61%11%- is small enough
compared with the variation 60% a h water permeation as
100
shown in Fig. 17.
The above mentioned result suggests that the sufficient
-E- 90 ........... reliability is obtained to find bars in wet condition.

...........
=a
e .............

4
+Tap bars. uppr surface
-€-Top bars, side surface tl
6o
50'
I------- -&-Bottom bars, side surface
II
I
0 IO 20 30 40
No. of bars
10 100 1m 1Mm lwmo lMOw0
Fig.16 Meagurcd potential of &cWn of an actual generator Frequency (€Id
StatM

Fig.17 Estimated frequency dependence of i n d decuode

Table 2 Statistical paramctcrs of mcanrnd potential potential after watcr pmeation
I Measureduortion

Top bars, side surface I 85.5 I 1.69 I 1.98

Bottombars,sidesllrfacc I 86.7 I 3.13 I 3.61 ..............

Avg.: Average, S.Dev: Standard Deviation .............

CV: Coefficientof Variation (= S.Dev/Avg.)

V. DISCUSSION
The potential of internal electrodes are calculated using -100 -50 0 50 100 150 200
the equivalent circuit in Fig.9 in case that water permeated Distance h u the end of inmost el&cde (mm)
into ground wall insulation and the dielectric property
Fig.18 Measured surface potential on the stator bar with
between the innermost internal electrode K1 and the
h htcmal electrodes m dry condition (me&)
conductor has the same dielectric property as Fig. 2 and 3 and wet condition (estimated)
with the outer insulation layer over K1 maintained in dry

1342
 VI. CONCLUSION accidents and aids in planning optimum maintenance
programs.
We have developed capacitance mapping techniques for
detecting wet stator bars without intemal electrode in the REFERENCES
ground wall insulation. We can provide the techniques using
capacitance measuring robot without removing rotor io the [l] H. Be, S.Seldto, T. Tamhi, H. Shrmada and M Tart 'Wet-heat
Daaioration and Residual Life Esfimatian Tahnology on Watn Cooled
same acuracy as manual detection with rotor removed. We Statot Coils of M i n e G m m W , EPRI, CERE and IEEE Joint
have experiences of inspection 239 times by manual and with Collquium "Rotating Elcchic Machinery Colloquium", Scptmba 8-
the robot so far. These abundant experiences increase the 10,1999.
reliability of ow technique for detectingwet bars.
We have recently developed a new method for detecting
wet bars designed with internal electrodes in the ground wall 131 M. 'Tari,S. N&o, H. Hasegawa, K Tashko, S. M"i and I€
Shimada: 'Vevelopment of a capacitance messuring robot for 61s
insulation at the end portion. The efficiency and accuracy of watn-cwled stabx -ding of turbine generatas", IEEE OD
the method has been verified using model straight bars, a full EaergyConvusian,Vol.14,Na.4,pp.1366-1371,1999.
scale bar and an actual machine. We have named this test as [41 Y. Inwe, H. b e g a w S. Sekito, H. Ogawd and H Shim&?
PMT (Potential Mapping Test). Advanced Methods for D e t d u g Wet Bars in Wata- Cmled Stator
Windings of Turbine Genmm'*,EPRI I n t a n a t i d Conference on
We have convinced that we can detect water leak and Elsfrie Generatot Redictive Maintenance and ReWisb"t, EPRI ,
insulation deterioration on any designed stator windings Orlando,Ian., 2003.
using these techniques. and it is effective in preventing

1343