Testing High-Voltage: Electrical Insulation Rotating Machinery Direct

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

cause maximum available fault currents in on economy of line protection against light- However, one discusser indicates that

cates that the


many systems have risen well above the ning as well as effectiveness at all voltage paper presents a more pessimistic picture on
maximum tube current ratings now avail- levels. effectiveness of protection than warranted
able. In addition, high-speed carrier-cur- by well-designed tube installations. UIn-
rent relaying has been adopted on most fortunately this higher degree of protection
power company systems at transmission H. N. Ekvall: Tie discussions by F E. could not be detected from the replies to the
levels of 100 kv and above. A time of 2 Andrews, 0. Ackermann, E. Beck, and R. M. simplified questionnaire.
cycles must be allowed for an expulsion tube Butler supplement the paper in a conistruc- The estimates on tube life are very in-
to completely clear a fault before any line tive manner by presenting additional in- teresting and give additional information
relays are allowed to operate This would formation on the effectiveness of expulsioa which was not covered by the paper. An
offset the advantage of fast operating time lightning arresters and by giving an estimate important factor affecting tube life has been
of 1 to 11/2 cycles possible with carrier- of tube life. the use of a weather-resisting phenolic paper
currenit relaying. There appears to be no disagreement with covering over the fiber tube. Field ex-
In conclusion, this paper rightly implies the general conclusion in the paper that perience has shown this to be a definite pro-
that considerable emphasis should be placed tubes offer protection in all voltages classes. tection against weathering.

1949. It was reported that a decision


Testing Electrical Insulation ol Rotating had been made to start systematic rou-
tine maintenance testing of the insula-
Machinery with High-Voltage Direct tion of large rotating machines giving
particular attention to a method of
determining the condition of the insula-
Current tion by using high-voltage direct current.
The importance of establishing and
maintaining a record of the condition
G. LESLIE HILL of the insulation of large rotating ma-
MEMBER AIEE chines becomes apparent when the kilo-
watt and kilovolt-ampere capacity of the
Pacific Gas and Electric Company system
Synopsis: Field test data are presented 1. Testing with alternating current: high- is totalled as of January 1, 1952:
and a simplified high-voltage d-c method for potential acceptance, proof, and mainte-
determining the condition of the insulation nance testing.23,4' 7,9 58 hydro plants, combined capacity of:
in large electric apparatus is described. A 1,326,600 kw
2. Testing with alternating current: meas- 17 steam-electric plants, combined capacity
graphical presentation is shown whereby urement of dielectric loss and power factor.5
the characteristic shape of the curve of the of: 1,722,800 kw
d-c microamperes versus d-c kilovolts is 3. Testing with direct current: measure- 46 synchronous condensers, combined ca-
utilized as the indicator of the condition. ment of insulation resistance by Ohm's pacity of: 1,045,900 kva
This method has been under experimental law.3'4.5'6'"1 Total capacity: 4,095,300
observation for several years and the tech-
nique now has reached a stage where it has The testing of cable insulation and the Some years ago the Department of
been adopted as the primary means for measurement of cable insulation resist- Electrical Operation of this company
testing and determining the condition of the ance with direct current has been stud- noted that a high percentage of insula-
electrical insulation of large rotating ma- ied.
chinery. All indications point out that tors, which had been tested with low-
predictions as to the life expectancy of The need for a nondestructive test on voltage direct-current and showed a
electrical insulation, in large rotating ma- the insulation of large rotating machines high value of insulation resistance of the
chinery and other apparatus, may be possi- to determine its condition was pointed order of 500 megohms and above, would
ble after years of testing and recording the out in 1949 in an unpublished AIEE paper
trend of the characteristic curves. This fail after being put back in service or
should aid very materially in setting up a
by E. A. Crellin who raised the question would fail on an a-c high-potential test.
program of planned maintenance and help of how to determine the condition of the An investigation made soon disclosed that
to answer the question: when to rewind, insulation and when to rewind a big gen- the 500 to 2,500 volts d-c obtained from
repair, or replace old electric apparatus. erator, thus starting the work which re- the small hand-driven generators were
sulted in the methods outlined here. Mr. not high enough to reliably pick out
THE purpose of this paper is to give Crellin brought out the fact that the faulty insulators or the faulty individual
a summary of representative d-c problem becomes one of large monetary shells of a multipart insulator. A d-c
field test data which have been obtained magnitude under the following represen- voltage of 40,000 was found to be the
on stator insulation and other apparatus tative conditions:
Paper 53-3, recommended by the AIEE Rotating
to date and to present the method that 1. When the cost of a rewind job is con- Machinery Committee and approved by the AIEE
is being utilized to analyze these data in sidered. Committee on Technical Operations for presenta-
tion at the AIEE Winter General Meeting, New
order to determine the condition of the 2. In taking into account the losses in- York, N. Y., January 19-23, 1953. Manuscript
electrical insulation. submitted March 31, 1952; made available for
volved if a failure occurs at a time when a printing October 7. 1952.
The theory of dielectrics and the test- generator is badly needed or in the case G. LESLI.E HILL iS Supervisor of Research and Spe-
ing of insulation of large rotating ap- where water may have to be by-passed or cial Engineering Tests, Department of Engineering,
synchronous kilovolt-ampere capacity may Pacific Gas and Electric Company, Emeryville,
paratus have been given a great amount Calif.
be needed for voltage control.
of attention recently, with many papers The assistance of W. E. Sprague, E. W. Brunsman,
and C. M. Buck, Test Engineers of the Pacific Gas
published. The subject papers may be A discussion of this subject covered and Electric Company, and David Hendrickson
grouped in three general classes or a the testing of insulation on the Pacific of the East Bay Municipal Utility District for ob-
taining valuable test data is gratefully acknowl-
combination of the classes as follows: Gas and Electric Company system up to edged.

APRIL 1953 Hill-Testing Electrtcal Insulation 159~


fear of overvoltage field and acceptance IC I
tests on large rotating electric machinery.
These a-c tests required a truckload of
equipment which had to be transported
to and handled at the station as the
kilovolt-amperes required to charge the
capacitance and supply the losses may
reach values'0 of 50 to 300 kva or higher.
High-voltage power-factor measure- I
ments seldom were made because of the I =o 2ltE
instrumentation problem and the re- I
quirement of current and potential trans- I
I
w S tI
formers with very accurate ratio and
A phase-angle corrections. The same prob-
lem in transporting equipment is en-
countered as with the a-c high-potential
tests. Figure 2. Vector diagram of a-c tests
Figure 1. Hydraulic analogy of insulation
resistance D-C INSULATION TESTS
By the use of a hydraulic analogy and The molecular friction in the metal dia-
minimum that would give satisfactory reference to a vector diagram of a-c tests phragm represents the dielectric loss in
results. These results were given in an endeavor will be made to show the the insulation.
1947,1 and it is interesting to note that the basic analysis and reasoning for adopting In the a-c tests the charging current,
d-c current-voltage curves obtained at the d-c method of testing insulation. shown as I=27rfCE in Figure 2, is large
that time have similar characteristics to Figure 1 shows a closed hydraulic sys- when compared with the current I,,,
those obtained on the insulation of rotat- tem were the fluid in cylinder A is dis- in phase with E. The current I. in
ing machines. placed by the piston, in a reciprocating turn may be considered as the sum of:
motion. This displacement creates a Id, dielectric loss
Method of Testing pressure E1 and E2 which moves dia- h, conduction or leakage current passing
phragm D in cylinder B to positions D, through the insulation
In the electrical utility industry at and D2. The movement of diaphragm Is, ionization loss current
large, the various methods of determin- D is the hydraulic analogy of the elec- I., current over surfaces and stray loss
current
ing the condition of insulation have been trical stress impressed on the insulation
a subject for discussion at any meeting by the alernating current voltage in a The conduction or leakage current I,
where insulation has been under considera- closed electric system. passing through the insulation, in phase
tion. The d-c equivalent for an a-c In case diaphragm D is overstressed with E may be of the order of 1/103
specified test voltage has been a long- by pressure E1 or E2 it will rupture. This to 1/105 of the total alternating cur-
standing problem.'3.4,9"12 action represents applying an overpoten- rent I. This is the current we are
tial and puncturing the insulation. The primarily interested in and it is repre-
A-C INSULATION TESTS size of the pump required to move the sented by the fluid leaking through
When testing new equipment with diaphragm back and forth is analogous diaphragm D of Figure 1. It is verv
alternating current it was generally the to the kilovolt-amperes applied to charge difficult to separate the conduction cur-
practice to follow the AIEE Test Code the electric circuit and is large compared rent 1, from the total current I; conse-
for Synchronous Machines7 and the to the small pump that would be needed quently the test engineer, when testing
American Standards Association (ASA) to supply a small leakage through dia- with alternating current on nondestruc-
Standard C 50-1943.8 The voltage applied phragm D if deflected in one direction. tive testing, has no warning when near
on acceptance tests has been that re-
quired by paragraphs 3-200 and 3-2008
of the ASA Standard. Proof tests with (D) CURENT LIMING RESISTOR
alternating current, after maintenance (C) V.T. PRCTIFIE - i
repair work, have been performed with
available apparatus such as a voltage
build-up from another generator of the
same voltage, or an assembly consisting
of a step-up transformer and regulator.
When a step-up transformer was used
the applied alternating potential was
generally the rated rms volts of the ap-
paratus plus some overvoltage. The
overvoltage to apply, if any, always has
been a subject of discussion and gener- 3> (B) H.V. TRANS1OM
ally it has been considered a risk to go
too high with a-c test voltages. Recent (A) VARIALE TRANSFOR
papersS"4 covering research on insulation Figure 3. Simplified diagram of the high-voltage insulation resistance testing device used by
for large machinery have verified this the Pacific Gas and Electric Company

160 Hill-Testing Electrical Insulation APRIL 1 953


Figure 4 (left). One
of the original insu-
i lation resistance test-
ers about 20,000
volts d-c, 10 milli-
*, amperes

Figure 5 (right). In-


sulation tester 40,-
000 volts d-c, 10
milliamperes
the danger point and may damage the type and thickness of insulation and in
insulation or cause a puncture to occur, the same condition. The larger machine
whereas with direct current he has a will have one-fourth the megohms resist-
close and accurate control of the current ance of the smaller but its insulation is
Ir passing through the insulation. just as good. Conversely, if the larger tion and discharge curves also are plotted
The test engineer plots the d-c micro- machine had the same megohms as the from the d-c readings which generally
amperes versus the d-c kilovolts while smaller its insulation would not be as are taken at 5, 10, 15, 20, and 25 kv and
the test is in progress. The maximum good. some at 30 kv. These voltages are
direct voltage to apply is decided before Since starting the routine nondestruc- higher than the AIEE recommended
the tests starts but if a sudden increase of tive d-c testing of insulation the AIEE direct voltage of 500 and 5,000. It
current or a "knee" in the curve is ob- Standard Number 436 was followed as a was decided that the d-c test voltage
served the test is discontinued and steps guide. Also power-factor measurements should be a value at least equal to or
taken to determine the cause. The basic with a low-voltage bridge are made at above the peak of the rms rating of the
requirement of this method is to await the same time and under the same condi- apparatus before Standard Number 43
a stabilized reading of micro-amperes tions as the d-c tests. Dielectric absorp- was issued.
at each increment of applied d-c kilovolts
This, in effect, is the final current reading 4000-
10000 VW
in a series of absorption tests, each addi-
tional higher level of dielectric absorption
current being obtained by adding defi- 3000
nite voltage steps on some lower value. 15000 1DC
When the test engineer is desirous of
obtaining the d-c puncture value without
tl200U- -tt 20000 V1
puncturing the insulation, he plots the
d-c micro-amperes versus d-c kilovolts
up to the bend in the curve and then extra-
R IA,
/// ow~~ 25000 VDC

polates the curve. He then draws a


tangent to this curve and the point where [aoc #W ALL 0 TO FROM.
VWDO. TEMP.s 19 DM!. 0.
JMB. HMUDmIr 46
the tangent intercepts the X ordinate, CAPACITAGZ. 0.4n4 NGD.
POWER FACTOR: 2.5%
at 90 degrees, will be the predicted
d-c puncture value of the insulation.
This becomes a very valuable tool to
obtain the predicted a-c dielectric strength
of the insulation by applying the d-c/a-c
dielectric strength ratio. If the alternat-
ing voltage is increased to the puncture
value this method then is used to obtain
the d-c/a-c dielectric strength ratio of
the insulation under study.
The insulation resistance in megohms
also is plotted versus d-c kilovolts, but
experience has shown that a better method
of diagnosis of the condition of the insula- Figure 6. Typical

.1
DISCHAME TDE - HINUTES
set of insulation I.
tion can be obtained from the d-c micro- resistance curves
20
amperes versus d-c kilovolt curve. As an
example, let us assume two generators,
made on a new |10_
13.8-kv 115,-
one of which has four times the area of 000-kva turbo- I
In
I0 15 20
I
the other and each wound with the same generator D-. VOLIME - KY

APRIL 1 953 Hill-Testing Electrical Insulation 16)1


40
,6
~200
.x
I
'.4

I
O
1100
200

0 4 8 12 0 10 2 30 C0 10 2
D-C VOLTAGE - [V
D-C VMLTA - IV DC VOLTAGE - IV
Figure 7. Capacitance of winding section Figure 8. Capacitance of winding section Figure 9. Capacitance of winding section
under test: 0.399 microfarad; winding tem- under test: 0.017 microfarad; winding tem- under test: 0.018 microfarad; winding tem-
perature 22 degrees centigrade, ambient hu- perature 23 degrees centigrade; ambient hu- perature 23 degrees centigrade; ambient hu-
midity 59 per cent; power factor 11.9 per midity 71 per cent; power factor 12 per cent. midity 71 per cent; power factor 11.2 per
cent. Wilson synchronous condenser 11 ,000 Wilson synchronous condenser 11,000 volts, cent. Wilson synchronous condenser 11,000
volts, 25,000 kva 25,000 kva volts, 25,000 kva

Records are made of the following at and watts (power factor) rectifier, utilizing the capacitance of the
the time of the stator and rotor tests 11. Rotor insulation resistance apparatus under test to smooth the wave
which generally are performed during the out when dielectric absorption is reached.
All the test data, and curves plotted from A full-wave rectifier circuit would have
annual overhaul.
the test data, are kept on file so that any
1. Station and division change in the condition of the insulation required two filament transformers and
a step-up a-c transformer of larger size.
2. Apparatus under test, serial number, mav be detected. The final design only weighs about 85
kilovolt-amperes, volts, amperes, and so
forth. to 90 pounds and is contained in one box
Test Apparatus Used by Pacific which may be transported in any auto-
3. Class of insulation Gas and Electric Company
4. Age and history of the insulation
mobile.
5. Year apparatus was installed When it was established definitelv that Figure 3 shows the simplified wiring
the maximum direct voltage (2,500) ob- diagram where the main component
6. Winding temperature tainable from the small hand-driven parts are (A) variable alternating voltage
7. Humidity at time of test generator type of insulation testers was regulator, (B) step-up a-c transformer,
8. Surface and physical condition of stator not high enough to detect faulty insula- (C) rectifier tube, (D) current limiting
winding tion, which could be revealed only by a resistor, (E) high-voltage voltmeter,
9. Surface and physical condition of rotor much higher voltage, equipment was as- and (F) the microammeter used to meas-
winding sembled to obtain a higher direct voltage. ure the current through the insulation
10. Individual field coil amperes, volts, The circuit selected was a half-wave under test.

Table 1. Representative Values of Power Factor, Polarization Index, and Insulation Resistance
Age of Power Factor, Polarization Insulation Direct
Apparatus Insulation Per Cent Index* ResistanceVoltage
Generator, Station P, 115,000 kva, 13.8 kv
. . ..........160t3..................
.. 5.1
.............. 1 600 ,to 3,750 megs ....... 10,000 to 25,00
Syn. condenser, Wilson Substation, 25,000 kva ............ 25 years
11 kv #1 phase....... 7.5 ............. 3.0 825 to 1,000 .......... a,000 to 20,000
#2 phase....... 8.0 .............8. 2.3t ......... 665 to 714 ............. ,000 to 9,000
#3 phase......... ... .......... 3.6 560 to 660 .............10,000 to 20, 000
Syn. condensers, Herndon Substation, 25,000 kva,..1. -6 1 years 1.8 ....... .......... 2, 100 to 3,500 ........... 10, 000 to 25, 000
12 kv #2-6 years ............. 2... ........... 7.7 2, 200 to 4, 000 ........... 10, 000 to 25, 000
#3-1 years ............. 2... ........... 6.2 2,600 to 3,800 ........... 10, 000 to 25, 000
Syn. condensers, Salinas Substation, 10,000 kva ..... #....1-26 years ......... 8..3.1
5 .... .... 1 ..4 0. 87 to 125 . . 10,000 to 23,000
11 kv #2-23 years ... 9.5 . . 1.10 ....... 160 to 500 ............. 5, 000 to 21, 000
Generator, Kilarc Power House, 1,500 kva, 2.2 kv . ......... 47 years ......... 1 .85 .. . .................... 3, 300 to 8, 000 ........... 2, 000 to 5, 000
27 megs. 5,500
Generator, West Point Power House, 16,000 kva ............. new ......8............... ......... 2.01 ......... 8 to0 300 . 10 000 to 1, 000
11.5 kv
Syn. condenser, Calif. Ave. Substation, 20,000 20 years ..... 3.5 ............. 2.6 ......... 84 to 623 .,000 to 20,000
kva, 12.5 kv
Generators, Bucks Crk. Power House, 23,000 kva .........f #-24 years ........ Phase 1-4 .0 ........2 1,500 ..... 15,000
11 kv. #2-3.2 ...... 3.3 ......... 3,900 .... 15,000
#3-3. 0... No data§ . no data .1, 500 max.
#2-24 years... Phase #1-2.8. No data§ ....... 1,930§§. 2,900 max.§§
#2-3. 0 . 5 .8. 5,000. 15,000
#3-3.0 . 1.811tt. 1,200. 15,000
Syn. condenser, Station A Stockton, 10,000 kva ............. 28 years
12 kv
......... 9.0 ............. 2. 181¶ ........ 312 to 212 megs ......... 1O,000 to 23,000
Syn. condensers, Station G, El Crrito, 15,000 kva ......... #1-29 years ...... 4...4.0 .............2 181 ........ 450 to 280 megs ......... 5, 000 to 25, 000
11 kv #2-29 years . 3.8 ............. 2.32** ....... 380 to 105 megs ......... 10,000 to 19,000
* Polarization Index at 20,000 volts d-c except
§ notenotugh readings and ** at 15,000 volts d-c
t at 9,000 volts at13,000 volts
¶ at 20,000 volts
tt at 15,000 volts; no good at 20,000 volts
$ at 15,000 volts a-c §§ failure indication at 3,000 volts d-c

162 Hill-Testing Electrical Insulation APRIL 1 953


Table 11. Test Values for D-C/A-C Dielectric Strength Ratios

Ratio
Maximum Predicted A-C Puncture
Test Applied D-C D-C Kv D-C/A-C D-C/A-C
Station Number Test Kv Puncture RMS Peak RMS* Peak

Bucks Creek generator number 2; age of insulation 24 years; ... 1. 35,000 .... 38,000 ....... 17,400 .... 24,500 .... 2.18 .. 1.55
Class B 2. 34,000 ....... 36,000 . 15,300 . 21,600. 2.35. 1.66
3. 28,000 ....... 29,000 . 15,300 . 21,600 . .89 . 1.34
4. 34,000 ....... 36,000 . 22,200 . 31,400. 1.62 . 1.15
a.
6.
35,000 ....... 40,500 . 22,000 .
60,000 ....... 67,000 . 40,000 .
31,000.
56,500.
1.84.
1.69 .
1.30
1.20
7. 35,000 ....... 37,500 . 28,300 . 38,500. 1.33 . 0.98
Electra Power House generator number 1; age of insulation 3 1/2. 1. 35,000 . 37,000 . 29,800 . 42,000. 1.17. 0.83
years; class B 2. 36,000 . 37,000 . 26,300 . 37,000 . 1.41 ...... 1.00
3t .. 27,000 . 28,500t . 19,000 . 28,000. 1.50. 1.02
Santa Maria syn. condenser; age of insulation 20 years; class B . ........7......... 43,000 . 49,000 . 33,300 . 47,000 . 1.47 . 1.02
Wilson syn. condenser; age of insulation 24 years; class B . ............. 3......... 31,000 . 32,500 . 21,750 . 30,650. 1.50. 1.06
4. 20,000 . 20,500 . 14,000 . 19,700. 1.47. 1.04
8. 25,000 . 29,000. 25,800 . 36,400. 1.12 . 0.80
9. 15,000 . 17,000 . 9,78 . 13,800. 1.73. 1.23
Pardee generator number 2; age of insulation 10 years; class B .... 1......... 24,000 . 28,000 . 13,500 . 18,900 . 2.15 . 1.53
(composite) 1(a) .. .16,500 . 23,100
2. 24,000. 25,000 . 14,000 . 19,600 . 1.78. 1.27
3. 25,000 . 27,500 . 14,500 . 20,300. 1.90. 1.35
4t. 25,000 ..... 25,OOO .t .10,000 . 14,100
5 $ 22,000 ..... 22,000 . 5,500 .
. 7,700
* Average 1.72, minimum 1.12, maximum 2.35.
t Insulation on coil physically damaged but not punctured.
$ Discharge or puncture at this point-insulation may have been damaged.

The first device built used a mercury as well as the insulation of rotating approximately 75 per cent of the total
vapor rectifier tube to obtain 10,000 machines. The output is limited to kilovolt-amperes of rotating machines on
volts d-c. It was soon determined that 10 microamperes at 40,000 volts; with the system. During this testing 14
10,000 volts d-c was not high enough, current ranges of 0 to 100, 0 to 1,000, machines have been found with insulation
so one was built for 20,000 volts. Tests and 0 to 10,000 microamperes. considered to be bad and on the border
showed a higher d-c test voltage would be A d-c rectifier set also was assemnbled line of possible failure at an early date.
desirable, so one was built to obtain 40,- for experimental testing up to 150 kv d-c. Some representative test data which
000 volts d-c. The voltage and current in micro- have been obtained to date are given
Figure 4 shows one of the original in- amperes are obtained during the test and now.
sulation testers assembled from miscel- the resistance is calculated in megohms
laneous equipment to obtain approxi- or the readings are plotted for graphical STATION P STEAM PLANT NEW
mately 20,000 volts d-c. Figure 5 shows analysis, as covered in a preceding section 115,000-KVA 13.8-KV GENERATOR
one of the later devices built as a complete entitled "Methods of Testing." Figure 6 shows the curves obtained
compact portable unit for 40,000 volts when this new generator was tested.
d-c. Three of these sets have been built Results of Tests The curve of micro-amperes versus d-c
to date and two more are being built. kilovolts should be noted especially.
They are used for testing insulators, The stator insulation testing to date The low current passing through the
cable bushings, and other apparatus has covered about 3,000,000 kva which is insulation resulted in the high insulation

18I
(1) A-C I:AK BREAKDOWN -|
1z-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(2) A-C PEAK BRFA3tDOWN _l~ ... sn

80 20
:a i
l M:
620

ZL~ ~ ~ ~ ~ IS
PC
I
t
40 I
I x
0

Ir I
0 - 0 5 16 2/+
D-C VOLTAGE - KV D.C VOLTAGE - KX D-C VOLTAGE - KV
Figure 10. Capacitance of winding section Figure 11. Pardee generator number 2, East
Bay Municipal Utility District, 6.9 kv, 7,500 Figure 12. Pardee generator number 2, East
under test: 0.01 5 microfarad; winding tem- Bay Municipal Utility District, 6.9 kv, 7,500
perature 23 degrees centigrade; ambient hu- kva
kva
midity 71 per cent; power factor 10.5 per Test made on C phase only. After the d-c test
cent. Wilson synchronous condenser 11,000 C phase was divided into two sections and an Test made on A phase only. A-c and d-c test
volts, 25,000 kva a-c breakdown made on each section made on half section of A phase winding

APRIL 195D3 Hill Testing Electrical Insulation 163


Table Ill. D-C Predicted and Actual Puncture Values
Max. Applied Predicted Actual
Insulation Test Direct Voltage Puncture Puncture
Under Test INumber Before Prediction Voltage Voltage

Santa Maria............................ ..1 ...5 000 59,500 .... 57,000


2
50, 000 .... 54,000 ....3.. 1,500
I 3. 5,000 ..
55.... 3.. 9,000 58,000
4. 49, 52,000 50,500
: 31,000 35,000 34,000
6....-45,000 48,000 46,000

Wilson Substation syn. condenser.......... l,000 .... 17,200 .... 16,000


6. 1.... 2,000 13,000 14,000
7 12,000 13,200 13,000

Bucks Creek Power House genera ..... ........38 56,000... . 59,000... . 38,000
D-C VOLTAGE - KV
tor number 2 9 .3. ,00 6.1. 61, 000.... 58,000
Figure 13. Kilarc Power House generator
number 2, 2.2 kv, 1,500 kva
Test on total windings to frdme. Cdpdcitdnce: resistance of 1,600 megohms at 25,000 This test was not considered a satisfactory
0.059 microFarad; winding temperature 25 volts d-c. After 30 minutes discharge, comparison for d-c/a-c ratio so the wind-
degrees centigrdde1 ambient humidity 56 per when the starting point was 25 kv d-c, ing was cut up into smaller sections.
cent; power fdctor 1.85 per cent the winding still maintained a voltage Test 2(b), Figutre 7: Direct current
of 11,000. was applied, in increments of 500 volts,
It should be noted that these curves
120 (500 Micro.Amps.) at are very unifonn up to 20 kv d-c. Test
up to 6,000 volts; when the current
18 kv D-C irtez) u e had reached 4,000 microamperes it was
0 experience and study of these curves discontinued; however, by extrapola-
.l indicate that this insulation is in very
A 8C~ ~~Cuv
good condition. They are similar to
tion of the curve the direct current
would have reached 6,000 microam-
I
the curves obtained on all large genera- pere.s at 7 kv d-c. This section of
v
4°~~~Cuv15 CZ2 tors and synchronous or static condensers the winding punctured at 6,750 alternat-
when the insulation is in good condition. ing current rms or 9,500 peak alternating
0I current.
WILSON SUBSTATION
2 SCVOLT^GE - veV

Tests for Figure 7, 8, 9, and 10 were


The characteristic curve of this test
is shown as Figure 7. These values were
made on a 25,000-kva 11,000-volt syn- not used to obtain a d-c/a-c ratio as
0D LT 2 30 chronous condenser, class-B insulation, the high direct current current was
age 24 -ears. undoubtedly due to some of the lower
Figure 14. Station A generator number 5, A d-c insulation test revealed two phases coils having been partially immersed in
18,700 kva, 11,000 volts good up to 20,000 volts and one phase water remaining in the slots after the
good up to 10,000 volts d-c. The machine fire was extinguished. It was evident
Curves 1, 2, 3: phdses A, B, and C with was put back on the line and continued
neutral current transformer and 100 feet of that the d-c puncture point had not been
in sernice for about 9 months when it reached no was it possible to predict
cable. Curve 4: phase A with neutral cur-
rent transformer, no cable. Winding tempera- failed. The failure became a phase-to- the puncture point from the curve.
ture 24.5 degrees centigrade; ambient hu- phase fault, doing considerable damage,
so it became economically advisable to
Test 3, Figure 10; Test 4, Figure 8;
midity 55 per cent Test 9, Figure 9; Test 8 (no graph):
rewind the machine. It was decided to
make d-c/a-c comparison tests on the The characteristic curves of the first
600 three of these in Figures 8, 9, and 10 and
old winding before removal, by going up
hase No. 1
as high as the curve would indicate with
the predicted d-c puncture and the a-c
direct current and then remove the direct puncture values of the four tests are
current and puncture the winding with given in Table II. The d-c/a-c ratios
2400 also are given in Table II for these tests.
alternating current. A total of ten
tests was made as follows: Tests 5, 6, 7; The sections of the wind-
Test XVunmber 1: All the observed bad ings, in these tests, were punctured with
coils were cut out. At 4,600 volts the direct current. These values and the
b~~~QPase No. 2

current was 4,000 micro-amperes. The predicted value of d-c puncture, before
d-c test was discontinued and alternating d-c puncture occurred, are given in Table
ase No. 3 current applied but the kilovolt-ampere III. Tables I, II, and III give the test
capacity was not sufficient so it was values on this synchronous generator.
0 10 20 30
D.C VOLTAGE - KV decided to cut out some of the winding. PARDEE DAM
Figure 15. El Cerrito Substation, synchronous Test A-nuber 2(a): Direct current was The East Bay Municipal Utility Dis-
condenser number 2, 1 5,000 kva, 11,000 volts applied up to 4,600 volts when again it trict has two 7,500-kva 6.9-kv generators
Winding temperature 28.7 degrees centigrade; was discontinued due to excessive direct at Pardee Dam Figures 11 and 12, which
ambient humidity 35 per cent; power factor: current current of 2,800 micro-amperes. supply power to the system. They were
phase 1 3.1 per cent, phase 2 3.2 per cent, A-c failure was indicated by high kilvolt- flooded and, after dry out, number 1
phase 3 3.1 per cent amperes and by smoke at 6,625 volts. generator tested satisfactory and was

164 Hill-Testing Electrical Insulation APRIL 1 953


placed on the line, but number 2 genera- has allowed oil to flow down and over the 900
tor indicated poor insulation at 2,500 stator winding on several occasions. Pbase No. 2
volts d-c. Additional drying did not Curves 1, 2, and 3 show the test results
raise this value and two faulty coils on phases A, B, and C including the neu- | Phase No. 1
were replaced. The tests indicated poor tral current transformners and about 100 ,3600
insulation at 10,000 volts d-c on one phase feet of cable. Phase A, curve 1, has a
but the generator was placed on the line rapid rise of current, starting at 17 kv
and after several months failed. Coils (40 microamperes). At 18 kv the cur-
were repaired but after each repair job rent was 500 microamperes. Curve 4 300
additional coils failed on test until it was gives the current obtained without the
c? ,ae No. 3
decided to rewind the machine. cable for the A phase. The rapid rise
Several tests were made on the winding started again at 17 kv (10 microamperes)
that remained to get infornation simi- and at 18 kv it was again about 500 0 10 20 30
lar to that obtained on the 25,000-kva microamperes which shows definitely D.C VOLTAGE _ KV
condenser at Wilson Substation. All that the weak spot is in the A phase. The
the curves had the same characteristic cable then was eliminated from the test
Figure 16. El Cerrito Substation, synchronous
condenser number 3, 30,000 kva, 11,000 volts
shape, that is, at a given direct and phase A still showed a bad insula-
voltage the d-c micro-amperes would tion condition. The steam end of this Winding temperature 25.5 degrees centigrade;
start to rise very rapidly. The peak a-c machine requires considerable mainte- ambient humidity 65 per cent; power factor:
breakdown value was consistently lower nance repairs and the condition of the phase 1 10.0 per cent, phase 2 10.0 per cent,
than the d-c value, indicating insulation electrical insulation in the generator phase 3 10.0 per cent
failure, on this machine. Figures 11 will have considerable weight in the
and 12 show two of the typical test management's decision as to what ex- 300
curves obtained on the winding of this penditures to make. Curve No. 2
generator. Table II gives all the d-c
applied voltages and the a-c puncture EL CERRITO SUBSTATION
values on this generator. 8 J~~~~~1,2,&32
Three synchronous condensers: 200
KILARC HYDRO PLANT Number 1 15,000 kva, 11,000 volts, class-B Curve No, 3 g
02
Figure 13 shows the direct current insulation, 29 years old o~~~~o
curve obtained on a 1,500-kva 2.2- Number 2-15,000 kva, 11,000 volts, class-B
volt generator installed in a small hydro (composite insulation), 29 years
plant in 1903, the Insulation thus being Number 3-30,000 kva, 11,000 volts, class-B
47 years old. This machine has the (composite insulation), 27 years
original armature winding which is of a Number 1 Condenser: The characteris-
modified Class-B insulation. The in- tic shape of the curve of the stator wind-
sulation was dry and brittle at the time ing was found to be satisfactory up to
of test but the direct current curve has and includin, 95,000 volts d-c. The
the same characteristic shape of the other direct current was only 90 micro-amperes erator number 1, 1,875 kva, 2,300 volts
machines. The curve shows a uniform for the full stator winding, at 25.7 degrees Winding temperature 44 to 54 degrees centi-
characteristic up to 2,500 volts d-c with centigrade, and considering the age and Cuv No.
grade, ambient temperature 25 degrees centi-
only 0.2.5 micro-amperes at this value. size of the machine this was considered grade;
0 ambient humidity 36.4 per cent
c 306
At 5,000 volts d-c the current was onlv satisfactorv.
1.5 micro-amperes but at 5,500 volts
d-c the current rose to 200 micro-amperes Number 2 Conideniser: Figure 13 gives
for an increase of only 500 volts. Ob- the results obtained when the separate Winding tem Puncture44t54dgescni
viously the recommendation was to phases of the stator of this machine were grde
| __
abinthuidty31.4V pe1cn
.~~~~()-55 r3
order a new winding for this machine, tested. Phase 1 had an abrupt change of
slope starting at 16,000 volts d-c, and at 45 M-35(X)53KV
the decision being based on the test data
obtained and the age of the present wind- 20,000 volts d-c the current had risen D-CVOLTAGE - D-C
V
ing. The machine has been rewound and to 500 microamperes. This was the
is now on the line. maximum applied to this phase. The
30
curves of phases 2 and 3 are shown which
STATION A STEAM PLANT GENERATOR have satisfactory characteristics up to
NUMBER 5 and including 23,000 volts d-c. The
low-voltage power factor of this machine 0 5 30 4
Figure 14 shows an 18,700-kva 11,000-
volt generator, with the insulation 40 was: phase 1, 3.1 per cent; phase 2,
years old. This is a vertical steam- 3.2 per cent; and phase 3, 3.1 per cent.
D-C VOLTAGE - KV
turbine generator installed in 1912 and Nulmber 3 Condenser: The test on the
the last cleaning date was in 1948. full stator wvinding, Figure 16, showed Figure 18. Bucks Creek Powerhouse, gen-
No stator -winding failures have occurred this winding to have a rapid rise in cur- erator number 2, 25,000 kva, 11,000 volts
on this machine but at one time it operated rent starting at 13,000 volts. When the
at a high temperature, up to a smoking
Winding temperature 15.5 degrees centigrade;
phases were separated phase 1 had an ambient temperature 15.5 degrees centigrade;
condition, for a short time due to auxi- abrupt change of slope starting at 14,000 ambient humidity 45 per cent; power factor
liary apparatus failure. Bearing trouble volts d-c, and at 15,000 volts it had risen 2.6 per cent

APRrLl95'3 Hill-Testing Electrical Insulation 165


Table IV. Predicted A-C and Actual A-C Puncture Values, Generator Number 5, Big Bend 600
Power House
10,000 Kva, 11,500 Volts, Class-B Insulation. Age 28 Years-About 25 Per Cent of Coils I
Had Been Replaced
400 All__ _ _ _ __ _

Predicted Max. Applied A-C Puncture Ratio Phases


Max. Applied Puncture, Kv A-C Volts RMS, Volts RMS D-CI to
Test D-C Volts, First Second A-C
Number Kv D-C A-C Build-Up Build-Up RMS
_
1 . 32 .......... 33 ......... 18.8 punctured . ........... 1.70
A 200 _

2 . 30 .......... 31 ......... 18.2 .....20.8 punctured . ........... 1.49


3 . 38 .......... 41 ......... 24.1 ...... 23-kv. max........ 20.8 . ......... 1.97
373 .....3......... 22.9 ........ 19 .3 punctured.
37. .. ..... 2.02

I
4. Phase 1
5. 31 .2................ .. 18.8 ........ 17.1 punctured ................... 1.87
6. 37 ... 37.5 ......... 22.0 .. 21.3 punctured .. 1.76
......
0 Phase3
7. 32.8 .. 33.5 ......... 19.7 . 17.8 punctured .. 1.88 0 10 20 30
8. 37 .. 38 ......... 22.3 . 23-kv max. 27.4 . 1 l37
23-kv max. 20.0.......... 1.75 D-C VOLTAGE -KV
9 . 34 35 .. 20.6 ........ ..
0
Average 1. 75
60 __ _

All
to 700 microamperes, when the test Generator Number 2: After dryout a Thases
was stopped on this phase. The curve
of phase 2 had an abrupt change of slope
weak spot was found in this generator
at 7,000 volts d-c. This spot was located
I
AJ
starting at 13,000 volts d-c, and at 16,000 and repaired and, after additional dry- B 40

volts d-c the current was 900 micro- out, the generator was tested to 10,000
amperes. volts d-c and placed on the line. It is li
4-

It was recommended that one set of still running satisfactorily. It is interest-


winding coils be obtained for condensers ing to note that a dielectric absorption
number 2 and 3 as these machines have test at 5,000 volts d-c did not disclose
the same dimensions and ratings; then faulty insulation. No graphs are shown
locate the weak spots by high-voltage on this generator.
direct current and replace coils. If 20
Generator Nutmber 3: The d-c tests 0 10 20 30
this does not bring the insulation to a on the stator insulation disclosed result- D-C VOLTAGE -KV
satisfactory condition it will be neces- ant curves so similar to Figure 17 of
sary to rewind. generator number I that they could be
100 A
o~~~~~~~~~
Phases
AMERICAN RIVER POWER HOUSE mistaken as duplicates. One phase was
Tests for Figure 17 were made on three satisfactory and two phases were bad a75
2,300-volt generators, number 1, 1,875kva; starting at about 6,000 volts d-c. The >
windings were not punctured and it was $2
number 2, 2,115 kva; and number 3,
1,875 kva; class-A insulation. These decided to try the machine on the line. l
The generator stator winding failed while 50
generators were installed about 1919.
The stators of number 1 and number 2 being synchronized at the time of closing
were rewound in 1929. The stator of the circuit breaker. The winding was
number 3 was rewound in January 1952. separated into phases and, when tested, 9 25
This hydro plant was flooded up to the so many bad spots were located that it
base of the generators and after the water was necessary to rewind. The generator 0-~~~~~~~$
receded all generator stators were tested, was rewound in January 1952, tested, and
placed in service.
0 1 20 30
at the time other maintenance work was C D-C VOLTAGE -KV
performed, and the following results
obtained. BUCKS CREEK POWER HOUSE 100- -
Line
Generator Number 1: After dryout the Tests were perfonned on two 25,000- Haf
whole stator was tested with the results kva generators, 11,000 volts, with class- vt
shown as curve 1, Figure 17. The rapid B insulation 24 years old, as shown in I75
rise in current started about 5 kv and Figure 18.
punctured at 7,500 volts d-c. The phases
were separated and curve 2 was obtained 50
on phase 1, curve 3 on phase 2, and curve
Figure 19. Coalinga Substation, synchronous 5~ Neutral
condenser number 2, 9,000 kva, 12,000 volts
4 on phase 3. The difference in the 25~~~~~~~a
curves of phases 1 and 2 compared with A-First Test: windings dirty; winding tem- o
the curve of phase 3 should be noted. perature 14.2 degrees centigrade; ambient ci
A total of seven bad coils in phases humidity 76 per cent
1 and 2 were found and replaced or re- B-Second Test: after winding cleaned; 'II>/ 7i
paired. The machine was put on addi- winding temperature 16 degrees centigrade;
ambient humidity 58 per cent D D-C VOLTAGE - KV
tional dryout and then tested again and C Third Test: after varnishing winding;
the results plotted as curve 5. The winding temperature 18 degrees centigrade; ambi ent humidity 55 per cent
machine then was put on the line and is D Fourth Test: phase 2 separated in two parts ;; winding temperature 21 degrees centi-
still running. grade; ambient humidity 44 per cent
166 Hill Testing Electrical Insulation APRIL 1 953,
and each tested; phase 2 showed a simi-
lar characteristic curve as the curve of
IZ all phases, starting with 18 microamperes
Curve3 z
at 8 kv and rising to 500 microamperes
14~~~~~~~~~~~~~~~~~~~~~-
at 8.5 kv. Figure 19(A) shows these
Curve 2
results and the curves of phases 1 and 3
Curve 6 which were satisfactorv.
IX L The machine then was cleaned and
Figure 19(B) shows the low conduction
currents obtained and similar characteris-
'CCurve 1
tic curves for all phases. The analysis
D-C VOLTAGE - KV D-C VOLTAGE - KV was that there were cracks or voids or
dirty surface paths which had been elimi-
Figure 20. Typical curves obtained when Figure 21. Predicting life expectancy of insu- nated. This was verified when another
testing insulation lation test was made after varnishing and while
the varnish was fresh. Figure 19(C)
Generator Number 1: When the stator tests were made on the old winding. The shows phase number 2 again with a bad
of this machine was first tested it was results are tabulated in Table II when insulation condition starting at 18 kv.
found impossible to build the voltage punctured with alternating current and This phase then was separated, Figure
above 1,300. The winding was separated Table III when punctured with direct 19(D), and the bad half which was on
and the results on the individual phases current Figure 18 shows the curve of the line side then was placed on the
were: Phase number 1 was good up to direct current obtained up to the point neutral end of the winding. The machine
and including 20,000 volts d-c. Phase X (34 kv d-c). The direct current was was put on the line then. A test will
number 2 was good up to and including removed but the curve was extrapolated be made soon to check this machine
20,000 volts d-c. Phase number 3 was to obtain the predicted d-c puncture again to determine if the varnish has
only good to 1,500 volts d-c. The point Z (35.5 kv d-c) by projecting the eliminated the bad spot.
power factors were 4 per cent, 3.2 per tangent Y to the X ordinate. Alter-
cent, and 3.0 per cent respectively. nating current then was applied and the TYPICAL CURVES
peak or crest puncture of 31.4 kv was The test data obtained to date seem
Generator Number 2: The first test dis- obtained. Generator number 2 was re-
closed that 4,000 volts d-c was the limit to follow the pattern of the curves shown
wound on March 15, 1952. Tables I, in Figure 20. Curve 1 is the characteris-
that could be applied to this generator. II, and III give all the test values on
The stator winding was separated and tic shape obtained when testing new and
these generators. also old insulation when in good homo-
the results on the individual phases were:
Phase number 1-the current started to COALINGA SUBSTATION geneous physical condition with a mois-
rise very rapidly at 3,000 volts d-c, but ture content which would be classed as
Tests on a synchronous condenser of dry. To obtain the d-c dielectric
at 2,900 volts the current was steady at 9,000 kva, 12,000 volts, insulation 23
1.3 micro-amperes (1,900 megohms). strength the test values are plotted as the
years old were performed. This machine test progresses. The applied direct cur-
Phase number 2 tested good up to and was first tested with all phases tied to-
including 20,000 volts d-c. Phase num- rent is removed at point X and the curve
ber 3 tested good up to 15,000 volts d-c
gether at the neutral and a very rapid extrapolated. The tangent or asymptote
rise in current was noted to start after Y, to the curve 1, projected on the X or-
but the current started to rise abruptly the reading of 23 microamperes at 6 kv.
at a slight increase above this value. dinate will give the predicted d-c punc-
At 7 kv the current rose to 400 micro- ture value.
The power factors were 2.8 per cent, amperes. The phases were separated
3.0 per cent, and 3.0 per cent. Curve 2 is the type obtained when test-
Coils were ordered to rewind both
generators and it was decided to make rz
guinea pigs out of the old windings in
order to obtain data on the relationship 11 tf/I~~~~~
between d-c test voltage and a-c dielec-
I'11 kI
tric strength.
Both generators continued in service
until the following winter (1951-52) when
a mountain avalanche partially flooded
54
c
0
I
$4
0
I ii~~~~~~C #
the basement with debris and water. C?
After the cleanup the generators were put Curve 7
on dryout and returned to service. Gen-
erator number 1 was on about 6 weeks
when it failed in service, developing an
internal phase-to-phase fault with con-
siderable iron damage. The rewinding
job was started immediately and there
was no time for testing. er Service - Years Service -
When number 1 was returned to serv- Figure 23. Additional output gained by
ice number 2 was taken out and d-c Figure 22. Predicted maximum service planned maintenance

APRIL 1 953 Hill-Testing Electrical Insulation 167


ing faulty and badly deteriorated insula- aintenance test voltage would be limited AJEE alternating potential tests after
tion due to aging, mechanical damage, some value such as 25 or 30 kv d-c. the d-c test.
tape separation, and conducting creep- When a curve like number 6 was ob- Curve A, Figure 24, gives the results
age paths caused by dirt and oil contam- ined the test voltage would be limited obtained on one of the six good bushings
ination. The current will increase the point X and the curve extrapolated after it had absorbed moisture. The
more or less uniformly with the increase ) obtain the predicted limit at point Z. following statistics are the readings:
in applied voltage but at some value it Figure 22 shows a hypothetical curve
will rise very rapidly and a knee in the where the current through the insula- Kv, D-C Microamperes
curve is observed. The test is stopped at ,on, for any given voltage, is given for
this point, unless it is desired to puncture ears of service. The predicted maxi- 30.......................... ...... I
50 ....
the winding, and steps taken to deter- lum service would be obtained by ex-
-

64...................... ...... 10
13.5
-

mine the cause. Only two punctures, -apolating the curve from the point X 70......................
76 ............... 20
without warning, have occurred to date the point Z where the tangent Y is
-

82......................... .. ..70
when running this type of test, and the ien projected on the X axis to obtain 83 ......................... 80
100
88........................
action, in these cases, was similar to the ie predicted maximum service. 90.......................... .105
240
94..........................
discharge of a condenser. D-c test discontinued.
Curve 3 is the type generally obtained UTPUT GAINED BY PLANNED
when the insulation is wet. The cur- AJAINTENANCE Alternating potential was applied and
rent rises more rapidly than it should If a machine could be operated con- the bushing failed at 68 kv rms (96-
under normal conditions and may reach a inuously without any outages a straight kv peak).
value of several milli-amperes without ne output curve would be obtained for Figure 25 gives the results of direct
puncturing the winding. The d-c curve he years of service up to replacement current applied to three 4,250-kv 72,000/
of Figure 7 is a good example of a rela- ime. This is shown on Figure 23 and 6,600-volt transformers after a bushing
tively high current passing through a he difference between the planned main- failure in transformer A at Drum Power
winding without puncturing. enance curve and the unplanned main- House. The stator of one generator also
enance curve gives the net gain to be failed during this trouble and had to be
PREDICTING INSULATION LIFE EXPECTANCY
xpected if the insulation is tested and rewound. Curves Al, B, and C give the
It would be economically desirable to 7eak spots are repaired instead of waiting comparative currents obtained on the
predict the life of a machine or insula- Dr service failures to occur. three transfonners up to 40 kv d-c.
tion to the last day of operation, before As an example the synchronous con- Curve A2 gives the results on transfomer
it required replacement, but it does not .enser stator winding at Wilson Sub- A after cleaning and reinsulating and
seem that this goal will ever be reached. tation failed four times in service before drvout.
However, it is hoped to be able to come he fifth and final one when it had to be
very close to that prediction in the future ewound. Discussion of Results
by the method proposed here or some im-
provement of this method. ESTS ON BUSHINGS AND TRANSFORMERS
. Table I gives representative v-alues of
The curves in Figure 2 are hypothetical The curves shown on Figures 24 and 23 power factor, polarization index, insula-
but have the characteristic shape of zok very similar to others shown but tion resistance, age of insulation, and the
those obtained from test data. Curve 1 hey are on bushings and transformers direct voltages applied.
represents the results we would expect to nd not on the insulation of rotating Table II gives the maximum d-c test
obtain on new insulation and, if an ac- aachines. Figure 24 gives the tests on voltage applied, the predicted d-c punc-
ceptance test on a new 12- or 13.8-kv ix condenser-type bushings rated 69 ture value, and the a-c puncture values
machine, we will assume that it would be iv by the manufacturer. Curve B obtained during tests on the machines
carried up to 50 or 60 kv d-c. After rives the maximum current and curve C at Bucks Creek, Wilson Substation,
the machine had been operated we would ,iv-es the minimum currents obtained. Pardee Dam, and on coils from Electra
expect to obtain curves like numbers tt 100 kv d-c the maximum was 9 micro- Power House and Santa Maria Substation.
2 through 5 inclusive when tested over Lmperes and the minimum was 3 micro- The ratios of d-c predicted puncture
a period of years. The maximum applied tmperes. These bushing withstood the to a-c puncture values are given.
The d-c test values are given in Table
III. The direct voltage applied and the
100 1 Figure 24 (left). Curve A is a test on
240 Micro-Amps, at 94 Ky-DC one 69-kv condenser-type bushing
Predic. D-C failue - KY
100 with faulty insulation. Curves B and 4,00
50 C represent maximum and minimum
t75 results on a group of six 69-kv con-
denser-type bushings with good insu- $ 300
I
lation
l 50 CT-e A- uz

E-4 1 200
A-C Peak Breakdown /
97KV
Figure 25 (right). Tests on three
c 25 4,250-kva 72,000/6,600-volt trans- w 100
formers. Curves Al, B, and C of
Cue
Curv max. transformers A, B, and C after bushing
a1
_t C urve C_ -pMin. failure in A. Curve A2 after cleaning,
0 reinsulating, and dryout of transformer 40
0 .-

0 25 50 75 100 0 10 20 30
D-C VOLTAGE - KY A D-C VOLTAGE - KV

168 Hill-Testing Electrical Insulation APRIL 1953


predicted d-c puncture compared with is not detected because of the current volts (Figure 19) would not have shown
the actual d-c puncture are given. passing through this spot being a small a fault in the insulation. If the voltage
An interesting game was played in percentage of the total conduction cur- selected to apply was 10,000 volts it
trying to predict the d-c puncture values. rent. would have an index of 1, and the insula-
The values were obtained quite accu- This d-c method of testing gives a far tion would have been damaged or possibly
rately on insulation that looked fairly more useful tool to work with than a-c punctured.
good but on insulation that was old, testing. It gives a measure of the degree Another example is the polarization
brittle, damaged or eaten away by corona of insulation deterioration whereas with index of 2.3 of number 2 phase of the
the d-c puncture value was hard to pre- alternating current it is a "go" or "no Wilson Substation condenser compared
dict and several samples punctured with- go" gauge. As an example, if it had been with 3.0 and 3.6 of the other two phases.
out warning. Only two cases of puncture decided to test both El Cerrito Substa- This definitely did not disclose a faulty
without warning while testing large tion condensers (Figures 15 and 16) insulation whereas the d-c amperes/volts
machines have occurred. In both of with alternating current and 1.5 times curve did show the fault.
these cases the puncture point was located 11,000 a-c volts was applied the winding The test data show that the dielectric
about 112 inches out of the slot and rela- undoubtedly would have punctured with- absorption curves will not pick out a
tively easy to repair without removing out warning. With direct current it faulty winding if the voltage applied is a
the coil. has been determined that the insulation value below where the fault would be
Table IV gives the predicted a-c and is bad, but still operable, and steps will detected. As an example, a good die-
the actual a-c puncture values obtained be taken to repair or rewind. lectric absorption curve could be obtained
during a d-c/a-c test on generator num- Other examples could be cited and it on the station-A generator number 5
ber 5 at the Big Bend Power House. The can be stated that none of the machines (Figure 14) at 15 kv d-c but no curve
predicted d-c puncture values were ob- where the insulation has been tested and could have been obtained at 20 kv d-c.
tained by the method previously de- found to be good have failed in service,
scribed and the a-c puncture values pre- whereas 4 out of 14 machines have failed Conclusions
dicted by using the d-c/a-c ratio of 1.7 in service that had indications of bad
from Table II. or faulty insulation. 1. Experience and test results obtained to
It was intended to build up to the a-c Practically all a-c high-potential tests, date lead to the conclusion that the best
and most reliable indicator for determining
predicted value, trv to avoid a-c puncture, on large rotating machines, have been the condition of insulation, during non-
then back down to zero a-c volts, and discontinued except to obtain information destructive and maintenance tests on large
build up again. This is a difficult test on the comparison of d-c/a-c dielectric rotating machines, is by observing the char-
to perform but it will be noted that in strength ratio. acteristics of the direct current/direct volt-
tests 3 and 9 the winding punctured age curves obtained on the insulation of the
The condition of insulation with any machine under test.
below the first build-up voltage. The degree of certainty from power factor
average d-c/a-c dielectric strength ratio 2. No definite formula has been derived
values has not been determined but a nor has any rule been written as to the
for this generator was 1.75. study of Table I and other test data shows direct test voltage to apply on tests, but the
The method, shown in Figure 18, and a general trend of high power factor as- basic requirement is that the direct voltage
curve 1, Figure 20, for obtaining a com- sociated with low insulation resistance. should be as high or somewhat higher than
parison of d-c/a-c dielectric strength the alternating rms peak rated voltage of
However, the test data show that power- the apparatus, or the a-c peak of the speci-
ratios is considered a better method factor values do not pick out a faulty fied a-c fest voltage.
than testing a group of coils with direct winding. As an example, the power 3. The maximum direct voltage to apply
current and an adjacent group with factors of the six phases on two generators to the apparatus under test should be de-
alternating current. The method shown at Bucks Creek Power House only varied cided before the test starts but the test
on Figure 18 was used on the Wilson from 2.6 to a maximum of 4.0 per cent. engineer should determine how high a volt-
Substation, Bucks Creek, Pardee, and Another example is the El Cerrito age to impress on the insulation from the
characteristics of the current/voltage curve
Big Bend machines. Substation condensers where the power and stop before a dangerous value is reached,
It was interesting to note the similarity factors of all phases on machine number unless he wants to puncture the insulation.
of the direct current/voltage curves to 2 were approximately 3.1 per cent and the 4. When testing large rotating machines
one side of an equilateral hyperbola power factors on the number 3 machine the direct voltages generally are limited to
where the asymptotes become the X- Y were 10 per cent on all phases. the following values:
ordinate in some tests and a similarity The power factors on the bushings
to one-quarter of and ellipse in others. having their d-c amperes/d-c volts curve A-C Apparatus Rating
Old
Insulation
New
Insulation
The direct current/direct voltage curves shown on Figure 24 were all practically
have been found to be relatively easy to of the order of 2.6 to 3.2 per cent whether 11, 12, 13.8 kv ...... 20 to 25 kv .30 kv
analyze if the readings are taken in small of faulty insulation or good insulation. 6.9, 7.2 kv ...... 15 to 20 kv .25 kv
2.3, 2.5 kv ...... 6 to 7.5 kv .. 10 kv
increments and a rapid rise or an abrupt It was estimated (from curve A, Figure
change in the curve is noted while the 24) that the a-c test voltage would have It is generally considered that these test
test is in progress. These curves will to be above 35 kv a-c rms to detect the voltages will pick out faulty insulation but
definitely search out the faulty insulation faulty insulation by power-factor meas- if desired, and conditions warrant it, higher
in apparatus when all component parts urement. values may be used.
are separated. The polarization index does not give 5. It has been concluded that any a-c
It is generally advisable to separate any degree of success in the analysis of overpotential test such as twice the rated
out long cable runs when making the voltage plus 1,000 or a lesser amount on
the condition of insulation in large maintenance tests may damage the insula-
first test as company experience is that machines. As an example, the polariza- tion without the test engineer's knowledge.
if too much apparatus is included in the tion index taken on the insulation of the 6. Power-factor values, obtained at a rela-
over-all test there may be a bad spot that Coalinga Substation condenser at 6,000 tively lower voltage than the rating of the

APRIL 19.53 Hill -Testing Electrical Insulation 169


apparatus, are only an aid, and not a posi- sidered that any universally accepted Ill.), November 1947.
tive indicator, in determining the condition d-c/a-c ratio of dielectric strength will be 2. ROUTINE INSULATION TESTING OF SYNCHRO-
of the insulation. If made at the full rated attained for all types of insulation in vari- NOUS MACHINES, L. F. Hunt, J. H. Vivian. AIEE
voltage or at an overvoltage they may be ous types of apparatus but ratios will have Transactions, volume 70, part I, 1951, pages 756-62.
more helpful but the large amount of appa- to be determined for each insulation and for 3. ALTERNATING AND DIRECT VOLTAGE ENDUR-
ratus required for charging the full kilovolt- the different physical conditions, age, and ANCE STUDIES ON MICA INSULATION FOR ELECTRIC
MACHINERY, Graham Lee Moses. AIEE Tranis-
ampere capacity of rotating machines and application of the insulation. actions, volume 71, part I, 1951, pages 763-69.
the very accurate instrumentation required 11. The predicted a-c dielectric strength 4. A MAINTENANCE INSPECTION PROGRAM FOR
make the method impractical to use. of a given insulation may be determined to LARGE ROTATING MACHINES, John S. Johnson.
7. The Polarization Index is only a minor a high degree of reliability by first obtaining AIEE Tra7nsactions, volume 70, part I, 1951, pages
749-55.
aid in determining the relative condition the predicted d-c dielectric strength, as
of the insulation when the insulation is outlined in the methods of testing, and then 5. A PORTABLE INSTRUMENT FOR MEASURING
INSULATION RESISTANCE AT HIGH VOLTAGE, F. W.
in good, or fairly good condition, and is of applying the d-c/a-c dielectric strength Atkinson, R. B. Taylor. AIEE Transactionts (Elec-
no especial value or aid when the d-c test ratio for the insulation under consideration. trical Engineering), volume 64, April 1945, pages
voltage reaches a limiting value when test- 164-66.
ing faulty insulation. 12. The success of the d-c method, as out- 6. RECOMMENDED PRACTICE FOR TESTING INSU-
lined here on the Pacific Gas and Electric LATION RESISTANCE OF ROTATING MACHINERY-
8. The condition of the insulation of insu- Company system, warrants the continua- AIEE Standard Nuinber 43, April 1950.
lators, transformers, and bushings may be tion of the high-voltage d-c maintenance 7. TEST CODE FOR SYNCHRONOUS MACHINES.
determined by the characteristics of the testing of the insulation in large rotating AIEE Standard N,nunber 503, June 1945.
direct current/direct voltage curve in the machinery and the extension to the main- 8. ROTATING ELECTRICAL MACHINERY. A SA
same manner as the method used on large tenance testing of other high-voltage appa- Standard Numnber C50-1943, American Standards
rotating machines. ratus such as transformers, bushings, and Association (New York, N. Y.) March 29, 1943.
9. The d-c/a-c dielectric strength ratio of cable. 9. MAINTENANCE OVERPOTENTIAL TESTS FOR
ARMATURE WINDnNGS IN SERVICE, R. W. Wiese-
a certain insulation may be determined by 13. The field test data, and methods out- man. Genteral Eleci ic Review (Schenectady, N. Y.),
obtaining the direct current/direct voltage lined, and the conclusions are presented August 1940, pages 24-28.
curve, and by extrapolation (or by calcula- with the hope that they will aid the manu- 10. CAPACITANCE OF SYNCHRONOUS-MACHINE ARM-
tion) the d-c puncture value may be pre- facturer in his research studies of insulation ATURE WINDINGS DETERMINED FOR HIGH POTEN-
dicted and the same sample then tested with and that they will encourage other utilities TIAL TESTS, R. W. Wieseman. General Electric
Review (Schenectady, N. Y.), July 1947.
alternating current to puncture under the to investigate the use of high-voltage direct
same conditions. current to test the insulation of large elec- 11. DIAGNOSIS OF A-C GENERATOR INSULATION
CONDITION BY NONDESTRUCTIVE TESTS, A. W. W.
10. The average d-c/a-c rms dielectric tric apparatus. Cameron. AIEE Transactions, volume 71, part
strength ratio, determined by the method III, 1952, pages 263-74.
outlined in item 9, was found to be 1.72 on References 12. RESULTS OF A QUESTIONNAIRE COVERING
CURRENT PRACTICES IN ELECTRICAL TESTS ON
the tests performed up to the writing of this DIELECTRICS IN THE FIELD, AIEE Committee
report. The minimum ratio was 1.12 and 1. INSULATOR TESTING WITH HIGH-VOLTAGE D-C, Report. AIEE Tr anisacdionzs, volume 70, part I,
the maximum wi-as 2.35. It is not con- G. Leslie Hill. Electric Light anzd Power (Chicago, 1951, pages 986-89.

- Ar

Discussion whole is not mentioned, although other in-


vestigators have given evidence to show that
lowest possible voltage.
Furthermore, we may ask ourselves if we
for equal searching effect the d-c test is less are dealing with a phenomenon in which the
John L. Fuller (Reliance Electric and damaging than the a-c one. Engineers current avalanche can be predicted by the
Engineering Company, Cleveland, Ohio): responsible for the operation of important mere extrapolation of the curve, unless the
The author gives an excellent discussion of generators which have seen service may be current has reached a magnitude that
newly discovered phenomena in connection reluctant to accept d-c test voltages much borders on avalanche. Mr. Hill in Table
with possible nondestructive testing of high- above the a-c peak phase-to-neutral operat- III shows that he used voltages on an aver-
voltage equipment using the application of ing voltage until more conclusive evidence age of 95 per cent of the actual puncture
d-c overpotential. Does the author know of its harmlessness is presented. voltage to predict breakdown. The element
of any similar work which may have been According to the author of this paper, the of time of direct voltage application natu-
done on insulation systems for equipment direct voltage level to which the test should rally enters into the problem also, as we all
rated 600 volts and under? The author be carried is either some predetermined know from our experiences with testing
discusses a powerful method of insulation maximum or that which gives a sharp bend cable by high direct voltages. Where the
analysis, and it would be very helpful to in the curve. The slope of the curve, or its current increase is extreme at very small
know whether the same type of approach slope as compared to that otherwise known increments of voltage, there is no doubt of
is applicable in lower voltage equipment. as satisfactory, becomes a measure of reli- the efficacy of the method, provided
ability. He further mentions a method the test can be stopped before avalanche
by which the curve may be extrapolated to occurs. This is shown by several of the
E. B. Curdts (James G. Biddle Company, predict the strength of the insulation. The cases described in the paper.
Philadelphia, Pa.): This paper can be con- slope of the curve at some unpredictable In analyzing data from field tests made by
sidered an interesting review of cases where voltage level therefore becomes an extremely us in several instances, we have replotted
the presence of localized incipient faults in important factor in the interpretation of the the current-voltage values on log-log co-
rotating machinery insulation has been data, and to be of practical use this slope ordinates after first subtracting out the
detected by direct voltage tests which reveal must be determined at a harmless voltage linear part of the current. The non-linear
sharply rising currents as the voltages are level. It is obvious then that for the parts of the total current were found to be
increased in steps. The evidence presented method to be practical in the hands of the constant power functions of the ionizing
seems conclusive and indicates that d-c average test engineer a simple and well- voltages up to the maximum voltages used.
test voltages, when raised to certain levels, defined procedure for quickly determining The total current then became
can oftentimes reveal weaknesses that the critical slope of the curve must be
cannot otherwise be found except by actual devised and standardized. i=iC+in
puncture. Two separate curves of the same test = Vo/R+k( Jo- VI)m
The voltage levels used in the cases re- data can look very different as to slope due
viewed in this paper exceed in most in- to the choice of co-ordinates. It would where
stances the values which have been thought therefore seem necessary to think in terms of ic = linear conduction current, micro-
of in the past as nondestructive. Whether the true mathematical slope of the nonlinear amperes
or not the voltages used were considered in part of the curve so that the insulation in = nonlinear ionization current, micro-
any way damaging to the insulation as a conditions can be judged effectively at the amperes

170 10Hill-Testing Electrical Insulation APRIL 1 953


Figure 1. Actual test data I.nl
with extrapolations at m = 2.26 200
and m' = 3.00 (k = 0.01 5)

in
180

160

140

120

Q)
CA.
zlg
IR 100
qz.
Qt
.j

80

60

iC
40

20

0
70
KILOVOL TS

Vo = applied voltage, kv bordering the avalanche zone current approaches nearer to avalanche
V1 = ionization starting voltage (break from will decrease the predicted breakdown
linear current), kv din/d V= mk( Vo-V)tm ' = tan a voltage.
Vo-V1 Data from one of our tests have been
V=
Tan a expresses the rate of change of the extrapolated according to the foregoing
m =slope of the nonlinear part of the nonlinear part of the current with voltage,
current before run-away or avalanche approach to the problem, and are shown in
in this case microamperes per kilovolt. Figure 1 of the discussion. Here the d-c
R = low-voltage insulation resistance or If we choose a ratio of 10 microamperes per
specimen, megohms per 1,000 test was stopped at 29 kv with m 2.26 and
=

kilovolt, which is equivalent to a being k=0.015. Extrapolating at this slope, the


k = a constant establishing the magnitude approximately 84 degrees, then V, the pre-
of the nonlinear current for any predicted breakdown voltage would be
dicted breakdown voltage above VI, is V+ V1* 91+6-97 kv. This value is un-
particular specimen, in this case in
terms of microamperes with the m m
l/mk=V1O/mk
I
reasonably high, and indicates the possi-
V= V bility that the slope of the curve would have
voltage expressed in kv
This is, of course, only a fair approxima- increased had the voltage been carried
Assuming i, to be a current uniformly dis- tion at best since it is difficult to determine higher. Actually, the winding broke down
tributed throughout the specimen and not V1, k, or m accurately. It is also to be at 40-kv peak a-c at a rate of rise of about
contributing to breakdown, it may be emphasized that V + V1 is the maximum 1 kv per second.
ignored in this consideration. Then, fur- voltage predicted from the slope used. If we use this 40 kv to determine tan 6,
ther, taking some tangent of the curve as Any further increase in this slope as the we find a to be about 74 degrees or a rate of

APRIL 195-3 Hill-Testing Electrical Insulation 171


Figure 2. Ioniza-
tion voltage test
results

(j)
0.

KW.

lz~
K-

0~

0
K-
Nt
0

V-( Vo -6) KV and PEAK VALUES FOR A-C TEST

rise of 3.5 microamperes per kilovolt. Since increased for some voltage above 29 kv. where
the d-c breakdown voltage should be higher Such a further change in slope would
than the equivalent a-c, we have reason to introduce a third term in the current V2= the voltage at which the slope changes
believe that 3.5 microamperes per kilovolt is rn'=the new slope
equation so that
too low to be considered as bordering on Again assuming that 84 degrees borders
avalanche. This again indicates the likli- iict+i+ini
= the avalanche zone, the predicted break-
hood that the slope of the curve would have in, = k'( Vo -V2)m' down voltage above V2 becomes

172 Hill Testing Electrical Insulation APRIL 1 953


rnl'- 1/ resistance measurement. It would be ex- Canada): I am particularly grateful for
V= V 10/rl'k' pected that watt loss values would be very Mr. Hill's paper as it increases confidence in,
and the total predicted breakdown voltage is erratic and, therefore, useless in cases where and background experience on, the similar
the insulation has already been punctured test methods we are now using in our com-
VO= V V in service or by previous tests. pany. I would like to join Mr. Hill in
Throughout this paper Mr. Hill has made stressing that the success of such tests de-
To show how the third (second nonlinear) numerous statements deprecating the use pends on allowing adequate time for true
current term may affect the predicted of the low-voltage dielectric absorption conduction-current readings.
breakdown voltage value, we may assume test in determining the condition of rotating As a contribution to the understanding of
that, in the case discussed, the slope changed machinery insulation. He is, of course, the peculiar shapes of some of the current/
to m'=3.00 at 29 kv and k' -=0.015 as before right in maintaining that the low-voltage voltage curves, such as in Figures 7, 8, 10,
dielectric absorption test cannot reveal and 19(B) of the paper, I might mention
`o- 13+29944 kv conditions that only the application of high that we have found that generator insula-
voltage can bring to light. All that has ever tions, as well as many other composite di-
The extrapolation for this change of slope been claimed for such tests is that they electrics, have a characteristic marked
is also shown in Figure 1 of the discussion. offer a simple and reliable means of deter- hump in the resistance/voltage curve at the
The foregoing is presented merely as mining the extent of excessive losses lower voltages. In the case of asphaltic
a possible theoretical analysis of the problem throughout the insulation as usually caused bonded mica for 13.8 kv, this is on the de-
with no intention necessarily implied that it by dirt and moisture. No one to my knowl- scent at 25 or 30 kv d-c, and the resulting
can be put to practical use during an actual edge has ever expected such tests to dis- increase on current in this region can cause
test. However, since the true slope or rate tinguish between localized and generally needless alarm. The shellac-mica hump is
of rise of current seems important, it may be distributed losses. Figure 7 may be referred lower and rarely obtrudes itself in practical
necessary to resort to an analysis of this to again as an excellent example of how the tests. In this connection, it would be
kind during the actual test. This may low-voltage d-c test (particularly the 2- valuable to know the types of insulation
require the use of very small voltage incre- voltage dielectric absorption test) would represented by Mr. Hill's results, instead of
ments, plotting the data on log-log paper as have revealed excessive losses. just the classes.
the test progresses, and carefully watching In closing this discussion, I would like It may be wise not to look for too much
for slopes that experience may indicate as to ask Mr. Hill to describe the guard system accuracy in forecasts of remaining econom-
dangerous. Whatever the future holds for which was undoubtedly a part of the ical life. There are great differences in the
this test method will depend on our degree measurement circuit used in these tests. voltage and current surges to which ma-
of understanding of the phenomenon and This is an important phase of high-voltage chines in different locations are subjected,
our ability to make measurements that can testing technique and has a direct bearing on and the incidence and severity of these can
be rationalized into data that present a clear the accuracy of measurement. be predicted only ve:ry roughly.
picture of what is going on during the test. In our experience, old brittle stator in-
The high-voltage d-c tests we have re- sulation, which is still intact but ripe for
cently made were in some cases supple- E. H. Povey (Doble Engineering Company, sudden cracking and failure, has often very
mented by ionization voltage measurements Belmont, Mass.): Mr. Hill has presented a high resistance at high voltages. In com-
of the amplified high-frequency output of a rather comprehensive report on his method parison, normal healthy insulation has
capacitance-resistance divider in the test of testing stator insulation; there are, how- moderate resistance, and new insulation has
circuit. Ionization voltage measurements ever, some points on which additional in- low resistance due to remanent moisture.
were also made during comparable alter- formation would be helpful in judging the This is illustrated by Table I of the paper;
nating voltage applications. Similarities practicability of the method. the Kilarc machine and the two Bucks
in the characteristics of the ionization volt- The method is described as high-potential Creek machines, all on the verge of failure,
age and the nonlinear part of the direct and nondestructive. These seemingly con- had high insulation resistances on the phases
voltage-current curves give evidence of tradictory terms lead one to ask Mr. Hill not at the moment affected. Therefore,
their common cause. Results of the ioniza- if he has ever experienced unexpected or I wonder how reliable would be the proposed
tion voltage tests are shown in Figure 2 of unpredicted breakdown of the insulation life prediction by extrapolation of a curve
the discussion. during such high-potential testing. To like in Figure 22 of the paper.
A question arises as to whether or not it avoid errors of extrapolation in extending We are gaining some confidence in the
would be desirable to analyze the data in the current-voltage curve, it is necessary specific absorption criterion of expenditure
terms of the product of the nonlinear part to carry test voltage to the verge of break- of life. I agree with Mr. Hill as to the un-
of the current and voltage, so that the down, as Table III of the paper indicates. reliability of polarization index for this
condition of the insulation can be spoken Another point affecting practicability is purpose: it is a mixture of the two super-
of in terms of watts, or watt-seconds if the time element involved. It would be imposed but independent phenomena of
time of voltage application is considered. interesting to know the time required to absorption and leakage.
The practicability of this will assume that produce a normal set of curves, for example, I believe that extensive regular testing of
the nonlinear part of the currenlt is the the curves of Figure l9(B) of the paper. the type evolved by Mr. Hill, and experience
result of a localized condition which is We agree with Mr. Hill that power-factor of the trends in dielectric characteristics so
vulnerable to heat, and with future experi- tests made at low test voltages have a revealed, related to service histories, will
ence may provide a permissible upper limit limited value. We do not agree, however, teach us enough in years to come to make us
in terms of watts as well as voltage. that power-factor tests at full rated voltage reasonably reliable prophets of the lives of
It is interesting to note in studying the are impractical. The measurement prob- our machines. I trust that many will
various curves presented in Mr. Hill's lem is no greater than for the low-voltage follow Mr. Hill's pioneering work and so
paper that what may be considered as in- tests which Mr. Hill has made. Potential hasten that essential experience.
nocuous nonlinear currents usually result in and current transformers are not necessary
losses of about 1 watt or less at the maxi- for this type of measturement and in general
mum direct voltages used, whereas non- are not used. Measuring equipment to- G. Leslie Hill: Mr. Curdts questions
linear currents falling in the run-away gether with auxiliaries for energizing the whether or not the direct voltages used were
category range from 1 to 5 watts in general stator of a large generator from a 30- considered in any way damaging. The
and to 10 watts in a few cases. Since nearly ampere lighting circuit may be transported results of this particular phase of our testing
all of the machines are in the 15-kv class with in a station wagon. Power-factor tests were inadvertently omitted from the paper.
comparable ground wall insulation thick- made at voltages up to rated voltage have This was one of the first questions we
ness, these watt loss figures may be signifi- been found useful in following the general settled in our own minds at the beginning
cant. A curve such as that shown in condition of stator insulation from year to of the d-c test program. We repeated d-c
Figure 7 of the paper, taken from a machine year without subjecting the insulation to tests as many as five and six times on various
which was wet, and having a nonlinear loss excessive voltages. machines and individual coils and obtained
of about 16 watts, would not ordinarily be curves similar to curve 1, Figure 20 that
tested by the high-voltage d-c method since would repeat without any indication of
its condition of wetness would obviously be A. W. W. Cameron (Hydro-Electric Power damage to the insulation. When we tried
revealed by the usual low-voltage insulation Commission of Ontario, Toronto, Ontario, to repeat a-c build-up voltages to a point

APRIL l1953 A 3Hill Testing Electrical Insulation 173


near the puncture value, we found that statements in the paper deprecate the use of runs, and transformers are all tested on the
puncture would occur in some cases at a low-voltage dielectric resistance testing. same shutdown, and will take about 2 hours.
lower value than the previous build-up What we have tried to bring out is that Mr. Povey questions the reference in the
value. This was especially the case when until our high-voltage d-c testing method paper in regard to power-factor measure-
testing old insulation. was inaugerated we were almost completely ments. We recognize his organization has
An over-all answer to paragraphs 3, 4, in the dark as to a method that would enable some ingenious apparatus and instrumenta-
and 5 of Mr. Curdts' discussion will be us to diagnose the condition of the insulation tion requiring step-up transformers, special
attempted. First, the test engineer should in a large machine and tell us if we could put current transformers, special inductors, and
be well qualified and experienced; then he it on the line. Obviously the high-voltage so forth, to measure dielectric loss and power
has to use his own judgement as to the a-c tests were either "go" or "no go." factor of insulation in electric apparatus.
maximum d-c test voltage to apply. This Mr. Curdts should note that Figure 7 of the What he has not told us is how to diagnose
voltage is based on an interpretation of test paper was obtained after the machine had the power-factor values obtained and
results. The test data, as a general rule, failed and that part of the winding was wet determine the dielectric strength of the
are plotted to the same co-ordinate scale. because of water used in extinguishing the insulation under observation. Our experi-
We consider that Mr. Curdts is unduly con- fire. If he will refer to the text he will find ence leads us to the conclusion that some
cerned about avalanche currents. We that it states "a d-c insulation test revealed technique in a-c testing will have to be
have had only two punctures to date without two phases good up to 20,000 and one phase developed to separate the conduction cur-
a warning on the plotted curve. These good up to 10,000 volts d-c." This means rents, through the insulation, from dielectric
occurred at about 22 and 23 kv on 12-kv the knee of the curves started on two phases loss currents before a-c testing will give us
a-c synchronous condensers. Upon ex- at 20,000 and on one phase at 10,000 volts the information that d-c testing does.
amination, the puncture paths were found to d-c. No low-voltage d-c absorption test Mr. Povey will recognize the instrumenta-
be through voids or tape separation which would have shown this condition. Ob- tion problem when the small amount of
resulted in air path breakdowns. viously a d-c test that indicated 10,000 watts, from conduction currents, involved
We have performed sufficient tests so volts d-c as the limit gave a warning signal are considered in comparison to the volt-
we do not question the fact that we can to repair or rewind. The machine failed amperes required to charge a large machine.
build up to 95 per cent of the d-c breakdown, after 9 more months of additional operation This was brought out by Mr. Curdts in his
draw the curve, and predict the puncture resulting in considerable damage. It is discussion.
value. A great deal of patience must be thought that Mr. Curdts will recognize that In answer to Mr. Fuller's question in
exercised in obtaining this type of test the high-voltage d-c testing technique, as regard to similar work on 600-volt appa-
data and it is absolutely necessary to await a described in this paper, has opened up a ratus: 2,300-volt motors and generators are
stabilized current at each increment of new horizon and given us a new tool to the lowest rated machines we have tested.
voltage increase. work with. The guard circuit, as used, is We do not know of any work carried out at
The mathematical analysis given is similar to the typical circuits in low-voltage this low voltage.
mainly of academic interest. We have insulation testers whereby the surface Mr. Cameron questions the type of in-
derived an equation for calculating the leakage current is by-passed around the sulation in the machines from which were
breakdown (which differs considerably from microammeter so that it can not cause an obtained the curves shown in Figures 7, 8,
Mr. Curdts' equation) and have obtained error. 10 and 19(B). Both of these machines had
very good checks with actual values. Mr. Povey's first question is partly a composite insulation consisting of several
However, in the field the test engineer is answered in the preceding remarks but I layers of impregnated mica tape with outer
generally pressed for time so we rely mostly might again reiterate that an experienced layers of varnished cambric. The text
on the diagnosis of the curves plotted from test engineer may carry the d-c test voltage omitted the class in one case and stated
test data. The method used by Mr. to the verge of breakdown, without punc- class B in the other, which was an error.
Curdts in obtaining the curve "actual test ture, on homogeneous insulation. We agree with Mr. Cameron that a high
data" in Figure 1 of the discussion is not Mr. Povey is interested in the time to accuracy cannot be expected in forecasts
given so it would be difficult for us to perform the high-voltage tests on a certain of the remaining economical life of the in-
calculate the puncture value. Assuming machine such as a sy,nchronous condenser of sulation in old machines. His work has
he used the method as recommended in the 9,000 kva. The total time will vary with paralleled ours to a very great extent and
text of the paper, the puncture value we conditions, but a careful d-c test to obtain I am quite sure he will agree that the present
obtain is approximately 50-52.5 kv d-c. values for a complete curve on a machine, methods of prediction that both are using
We most likely would have carried this test large or small, will take about 30 to 45 will give us a better chance of prediction than
to at least 40 or 45 kv d-c before stopping, minutes. Obtaining the clearances and we have had in the past. Our first predic-
that is, assuming we wanted to puncture switching will generally take longer than the tion work was in the early part of 1950
with alternating current. test. As a general rule, a large generator and his first prediction work was in the
Mr. Curdts considers that numerous including the house generator, cable and bus early part of 1951.

174 Hill Testing Electrical Insulation APRIL 195-3

You might also like