2017 Electrical Insulation Conference (EIC), Baltimore, MD, USA, 11 - 14 June 2017

Asphalt Stator Winding Insulation

Clyde V Maughan
Maughan Generator Consultants
Schenectady, NY
cmaughan@nycap.rr.com

Abstract – Asphalt stator bar insulation was first used in electrical duty improvement. The asphalt insulation systems
about 1915 and quickly became universal in the industry for provided excellent performance. It was rugged, forgiving,
large generators until about 1950. Numerous tape migration not prone to mechanical vibration. Some early winding
(“girth crack”) problems occurred beginning in about 1950 on installed in hydro generators are still in service after 100
large units, >40 MW. The problems were so great that the
years of service.
industry generally abandoned asphalt in favor of thermoset
resins, e.g., Thermalastic, Micapal, Micadur. But asphalt was
not entirely abandoned; its use was continued on smaller coil But asphalt had recognized duty limitations and thus
windings until the early 1990s by at least one major OEM, designs were conservative: low electrical stress (45 volts/mil
General Electric. vs. 65-90 vpm on modern systems), low operating
temperature (while commonly rated Class 130, reliable
Most of the large, high speed generator with asphalt operation required designing at below Class 105
windings have either been retired or rewound with thermoset temperature), and low magnetic vibrational forces (in the
insulation, although there are perhaps 100 with their original range of 3 or 4 pounds/inch in the slot vs. up to 110 #/in. in
winding still in reliable service. There are many low-speed,
modern windings).
hydro generators in service with asphalt windings. There are
also 1000s of the smaller high-speed generators still in service,
and because they were designed to considerably higher duties, Still as MW rating evolved upward problems began to
they have also experience the migration problem. occur, e.g., puffing Photo 1

This paper will briefly discuss stator winding design

evolution but will focus on the exceptionally complex and
generally poorly understood design/manufacturing parameters
that result in insulation migration. The paper will also provide
recommended operation and maintenance guidelines to
maximize on-going winding reliability and to minimize
unnecessary maintenance. Photo 1

Keywords – stator windings, insulation systems, deterioration, asphalt bleeding, Photo 2,

failure, high potential test.

I. INTRODUCTION

Before the turn into the 20th century, insulating materials

were natural products: shellac, cotton, paper. The
rudimentary designs were at low voltage and low Photo2
temperatures, and apparently functioned fairly well as long
as duties were kept sufficiently low. With inevitable trends wide-spread partial discharge indications, Photos 3 and 4,
toward higher voltages and higher thermal and mechanical
duties, much better materials were required. Late in the 19th
century, mica flake was discovered to have remarkable
electrical and thermal properties. But still with shellac,
cotton and other relatively primitive materials incorporated
in the systems, troubles continued.
Photos 3 and 4

By the mid-teen years, 1915, it was discovered that by

using a vacuum-pressure cycle, mica/cotton tapes could be
impregnated with a hot asphalt compound to obtain a major

978-1-5090-3967-8/17/$31.00 ©2017 IEEE 135

Authorized licensed use limited to: Universidad del Valle. Downloaded on August 10,2023 at 15:24:32 UTC from IEEE Xplore. Restrictions apply.
 vulnerable to CO2 (“dry ice”) cleaning damage, Photo 5, temperatures the asphalt was near a liquid and gave no
mechanical stability to the groundwall insulation. Load
cycling would cycle the length of the bar, typically 1/4 to
3/8”, depending on core length. For reasons not well
understood, due to some of the 8 factors listed above, e. g.,
4, 5, 7, and 8, there was a slight tendency for the groundwall
insulation to follow copper length more closely during load
Photo 5. Bare copper exposed reduction than load increase. In worst measured case the net
migration of the top layer of tape was 11 millimeters/cycle,
and most damaging – tape migration, Photo 6, thus it is apparent that tape migration could be an aggressive
deterioration mechanism.

But migration only occurred in the slot portion of the bar,

thus causing the tape separation to occur at the core end
region, and giving the appearance of a crack, thus the mis-
Photo 6. interpretation, Girth Crack.

The latter, tape migration, was very troublesome and was But in the slot portion, the layer of tape directly against
extremely difficult to understand. It was also difficult the copper would be expected to remain un-migrated, and if
(almost impossible) to accurately reproduce under the spirally apart at the core ends did not occur at the same
laboratory conditions. As a result tape migration was not axial location of the bar on each layer of tape, then even
well understood in the industry. As will be discussed though the top-layer migration was great, >1”, the electrical
below, GE termed the phenomenon a mis-leading “girth integrity of the groundwall may be only slightly impacted.
cracks”; Westinghouse engineers understood the
phenomenon much better and termed the phenomenon more That this was true was proven on “badly deteriorated”
accurately, “tape separation”. [1] windings that when hipot tested at new-unit hipot value of
2E + 1 did not fail. [2] [3]
Description of Phenomenon

Probably the reason the tape migration phenomenon was II. BRIEF HISTORICAL SUMMARY
never well understood is because it was subtly dependent on
many variables (more or less in descending order of Experience on Large Generators
importance):
1. Load cycling (required in order to have differential Asphalt/mica systems served the industry well for 35
expansion between copper and core). years, although by the late 1920s the tape migration
2. Core length (thus the amount of differential expansion). phenomenon had surfaced. At General Electric, this
3. Cooling gas temperature (directly affects asphalt occurred on 2 very long, very large 4-pole generators. The
hardness). engineers mis-diagnosed the root cause of the condition as
4. Bar copper temperature (directly affects differential cracking of a “brittle” stator groundwall insulation, and thus
expansion and resulted in greater separations on top bars applied the term, “girth crack”, still used 90 years later.
which operated hotter than bottom bars). They believed the problem was due to too hard a system and
5. Selected asphalt softening point (directly affects asphalt lowered the softening point of the asphalt used in the
viscosity). groundwall. This was exactly the wrong change, since the
6. Load power factor (higher power factor gives hot iron “cracks” were actually resulting from the groundwall
at core ends, causing tape separations to concentrate at ends migrating in the slot toward the axial centerline of the core.
of core).
7. Tightness of the bar in the slot (tighter fit for unknown In retrospect the better solution would have been a higher
reasons correlated with increased tendency to migrate). temperature softening point. But somewhere around 1935
8. Direction of tape application. (On GE bars, tape for reasons not recorded, the GE engineers diluted the
separations were greater on top bars collector end, and asphalt with about 1/3rd linseed oil. Linseed oil has the well-
bottom bars turbine end). Taping direction was presumed to known property of becoming more viscous if held at
be the explanation. elevated temperature, thus if a GE generator was not load
cycled for 3 or 4 years after placing in service, the
The asphalts used typically had a flow (liquid) asphalt/linseed oil blend became more viscous
temperature around 100C. The copper temperature at load (mechanically stable), and insulation migration tendencies
was designed to be around 90C. Thus at load operating were much reduced.

136
Authorized licensed use limited to: Universidad del Valle. Downloaded on August 10,2023 at 15:24:32 UTC from IEEE Xplore. Restrictions apply.
 The net result of these 2 changes (lowering of asphalt Measurements of the groundwall on the bar that had failed
flow point and adding linseed oil) was that these GE site Doble test revealed necking of insulation from
generators over about 40 MW had serious tape migration insulation migration exceeding 40%, Photo 8.
problems if load was cycled immediately on being placed in
service, Photo 7. Otherwise, migration was minor or not at
all. Of the ~400 large asphalt-insulated generators built by
GE, about 1/3rd were rewound; the remainders are still in
service without migration or have been retired.

Examples of tape separations are seen in Photo 6 above

and 7 and 8 below.
Photo 8. Necking from tape migration.

While definitive information on this fleet of high-speed

asphalt windings and on low-speed asphalt windings is
limited, the available design and performance information
suggests reasons for concern.
Photo 7. Huge 2-5/8” separation after 19 months
operation, 11 mils/load cycle. Separation is inside the core
due to high power factor load. III. RECOMMENDATIONS

Westinghouse engineers better understood the tape Because of the intrinsic nature of asphalt windings to the
migration phenomenon, but weren’t sufficiently lucky to specific deterioration mechanisms discussed at the
have added linseed oil. By the mid-1940s, most of their beginning of this paper, certain fundamental operating and
stator windings on their larger generators were experiencing maintenance recommendations seem appropriate, and are
fatal migration. On a crash basis they developed an entirely summarized below.
new groundwall insulation system – Thermalastic. This
system used a polyester resin binder to replace the asphalt Operation and Maintenance
and was a remarkable accomplishment, particularly in view
of the relatively primitive resins available at that time. Since higher temperatures adversely affect insulation
stability, avoid as far as practical operating at overload
During the ensuing years, all OEMs developed resin- conditions or elevated cold gas temperatures.
based groundwall systems: Micapal (a polyester-like
epoxy), Micapal II (a true epoxy), Thermalastic-Epoxy (a Do not rewedge the winding as slot tightness correlates
true epoxy), Micadur (a true epoxy) . . . with tendency for the insulation to migrate. Bar vibration
cannot occur for two reasons: a) the vibrational driving
Experience on Smaller-size GE Generators forces are very low, and b) because the groundwall is soft,
bouncing from the downward mechanical driving forces
While on large generators use of asphalt windings was cannot occur.
discontinued by 1950 by Westinghouse and 1960 by GE,
asphalt was used on 2-turn coil windings until 1990 by GE. Inspection
These generators generally were rated less than 25 MW.
The asphalt used was the blend of asphalt and linseed oil In general, do not be overly concerned if some bar puffing
used by GE from the mid-1930s. To address the migration and some asphalt flow throughout the winding is observed.
issue, the designers added a 3” wide strip of glass clothe at These windings tend to be old and some puffing is
each core end to give resistance to migration of the tape. inevitable due to the nature of the groundwall. Unless the
But the designers also went to somewhat higher copper asphalt flow is extensive, the groundwall integrity should
operating temperatures. The windings apparently have given not be significantly impacted.
good operating performance.
However, if individual bars stand out as highly puffed and
A known exception. The writer was involved 10 years ago high asphalt flow, shorted strands may be suspected and
in a law suit where a 30-year-old asphalt coil winding had winding quality may be suspect.
failed during Doble test, line-to-neutral voltage. Laboratory
evaluation of several half-coils from this winding was Vertical shaft hydro units have the special considerations
conducted. Failures occurred at low hipot test values. The relating to the possibilities of the winding moving down
groundwall insulation was badly deteriorated, Photo 4. through the core.

137
Authorized licensed use limited to: Universidad del Valle. Downloaded on August 10,2023 at 15:24:32 UTC from IEEE Xplore. Restrictions apply.
 Test IV. CONCLUSIONS

Groundwall and copper resistance values should be There are many 1000s of asphalt stator windings still in
measured at every convenient opportunity. The groundwall service. A few would be the large GE 2-pole units. The
resistance will be primarily affected by moisture and/or behavior of these units is fairly well understood, and if still
contamination. If groundwall resistance is low, drying and operating after 65 years of reliable serve, those windings are
cleaning may correct the condition. But special care in unlikely ever to need replacement.
cleaning should be taken. Solvents can attack the asphalt,
and CO2 cleaning can easily damage the groundwall. Photo Of the 1000s of “newer” GE coil windings in service,
5. little is recorded as to their operating, maintenance history.
But the Recommendations above should be applicable to
High copper resistance may suggest failing connections. maximizing their ongoing reliability.
Low resistance is unlikely since the main cause would tend
to be shorted turns, and a coil with shorted turn should have And there are many hydro generators with asphalt
already catastrophically failed in service. winding still in service. So far as the author is aware, there
is no available recorded history of the performance of these
None of the common low voltage tests, e.g. powerfactor, units, but again, the Recommendations above should be
tipup, are likely to give useful maintenance information. applicable to maximizing their ongoing reliability.

By far the most definitive test for groundwall condition

would be hipot, even at a test value as low as 1.1E. But of
course there is the possibility of failure and this is likely to ACKNOWLEDGMENT
force major winding repair. Individual bars can normally be The author wishes to acknowledge with thanks several
replaced on 4-pole and hydro generators as end-arms are individuals who contributed information included in this
short. paper: Robert Hamilton, Consultant; Jim Lau, Siemens
Energy; Bert Milano, Consultant; James R Michalec, JR
On 2-pole units the long arm and compliant insulation Michalec Consulting; James Timperley, Doble Engineering.
allowed wide deviation of the bar from design location. But this is not to imply that all contributors would agree
Also there is the intrinsic removal/installation interference with all that is written above on this complex topic.
issue. GE had a system, termed “Conforming Micapal”, for
individual bar replacement, though it is doubtful this REFERENCES
complicated system is still available.
[1] J S Johnson and J C Botts, “Physical Effects of Thermal Cycling on
Stator Coil Insulation of Turbine Generators”, AIEE Paper 56-5.
On a hydro machine with a failed coil, corresponding
coils in each phase belt can be isolated and service [2] Maughan, C. V. and A. F. Bristor, ”High-Potential Testing of Large
continued. Turbine-Generator Stator Insulation” IEEE Conference Paper 70 CP196-
PWR, Available at http://www.generatortechnicalforum.org
A paper suggesting a method of ac over-potential testing [3] J. E. Timperley and J. R. Michalec, “Estimating the Remaining
to assure continued reliable service life was published in Service Life of Asphalt-Mica Stator Insulation”, IEEE Transactions on
1970 [2]. The basic premise is that high voltage insulation Energy Conversion, December 1994, Vol. 9, No. 4, pp. 686-694.
systems deteriorate more or less linearly with time. This
idea was successfully applied to allow postponing or
eliminating stator rewind on many generators. [3]

138
Authorized licensed use limited to: Universidad del Valle. Downloaded on August 10,2023 at 15:24:32 UTC from IEEE Xplore. Restrictions apply.