You are on page 1of 12

HYDROGENERATOR CORE CONSIDERATIONS AND CORE LOOP TEST

Bert Milano, April 4, 2008



Many small- to medium-size hydro generators (less than 100 MW) built before 1940 or so and many larger
machines built after 1940 or so are nearing the end of stator core life. One hundred years for smaller units and
60 years for larger units seem to be magic numbers people talk about, but no one wants to be quoted, as there are
many variables involved. In any event, some history and concerns in dealing with existing and new cores will be
discussed as more and more both large and small- to medium-sized cores may soon be approaching the end of life,
or at the minimum, approaching a time for serious assessment, thus there is a growing need for information in this
arena.


SOME GENERATOR DESIGN BACKGROUND
Given the hydro generator rating, mechanical and hydraulic engineers set the turbine diameter and generator speed.
The generator inside core diameter and number of poles is then essentially fixed, and it follows that the generator
core length is then predetermined. Something similar to this process sets the core length for steam units.

Designers usually keep the generator core tooth flux density below 1.5 Wb/m
2
(1.5 Tesla), and the back iron depth is
then selected to keep back iron flux below 1.15 Wb/m
2
(1.15 Tesla). Higher flux densities put excessive demands
on the rotor design (winding, ampere-turns, exciter, etc.). Further, with these limits the higher-grade magnetic steels
can be used to meet a 60 C temperature rise. Core steel is selected on loss and cost considerations, as a machine is
priced on a dollar-per-pound basis. Total core losses start to increase quickly above 1 Wb/m
2
(1 Tesla). Losses can
increase by 200 percent to 300 percent at 1.5 Wb/m
2
(1.5 Tesla). In addition, as the flux density starts to increase
above 1.25 Wb/m
2
(1.25 Tesla) the core steel begins to saturate (starts to look like wood!).



VERY BASIC CORE CONSTRUCTION
The core consists of a stack of electrically insulated punchings of magnetic steel. Thus, the core is basically a
laminated structure. The individually insulated laminations prevent large eddy currents from circulating in the core,
and thereby limit undesirable I
2
R losses and the associated core heating.

CORE QUALITY ASSURANCE
To ensure maximum service life and efficiency, machine cores are tested when new prior to installing the winding,
and also prior to installing a replacement winding. A core (or ring) test is performed to obtain assurance that the
core interlaminar insulation is serviceable; that is, to ensure there are no hot spots or general deterioration or
deficiency of the core interlaminar insulation.

Core tests will result in one of the following:

- The core is serviceable as is
- The core requires minor repairs
- The core must be restacked (major repair)
- The core must be reinsulated (quick fix in place of a time-consuming core replacement)
- The core must be replaced.

There are opportunities in replacing a core in that the designer can start with a clean sheet and adjust core
dimensions and winding layout to take advantage of modern materials and design advantages to optimize machine
capacity and/or efficiency. This is of particular advantage regarding older hydrogenerator core replacements.

WHY CORES DEVELOP HOT SPOTS/DEFICIENCIES
2
- Air gap rubbing due to objects/rotor
- Winding-related faults
- Burs and/or improper lamination stacking
- Thermal/mechanical wear, chevrons
- Moisture (old core plate)
- Winding tear-out (literally!) related damage
- Wedging/rewedging damage
- Thermal deterioration of laminar insulation
- High end-winding leakage flux producing damaging interlaminar insulation voltages in the core ends
- Vibrating stator teeth (non-uniform lamination thickness especially on older laminations)
- Core split issues
- Core warpage
- Loose core-related vibration.

CORE STEEL
Core steel lamination insulation is referred to as core plate. Up to the mid to late 1940s, hot-pack rolled steel was
used in generator cores. The steel was rolled out in packs or stacks. The oxidation fromprocessing causes a scale to
form which keeps the sheets from sticking to each other. Thickness varied by up to 25 percent, and this steel had
convex surfaces. Despite various punching and stacking solutions, and to ensure a tight core, very high clamping
forces were needed and a very strong clamping structure (frame) was required. This resulted in very rigid cores with
little vibration and/or movement-related problems. Prior to the 1940s the machines were not that large, thus there
were very few core stability issues (buckling, chevroning, waves). The design resulted in a constrained core stack.

The next development was directional (anisotropic) cold-rolled steel. This steel has a lower loss in the direction of
operating flux, and thus is ideal for use in transformers. It is not of much use in generators due to the flux path in
the teeth and back iron being at 90. In high-speed machines it is at an even greater disadvantage due to flux
orientation at the edges of the pole pitch. It is also a more costly option.

Next up was nondirectional (isotropic) cold-rolled strip steel. The electrical properties, grade for grade, were similar
to hot-pack steel. Thickness deviation was reduced from 10 percent to 25 percent to about 1 percent to 3 percent.
Surface peak-to-peak variations were reduced from 100 microns (hot pack) to 5 microns (cold rolled). Cold rolled
thus has a mirror-like surface, while hot pack has a dull matte surface.

About this time, generators started to grow in size and rating; cold-rolled cores became standard and core clamping
pressures were reduced. It was no longer possible to restrain a core, and thus cores and frames were designed to
allow for radial expansion and contraction per thermal cycling-related forces. If not designed properly, cores would
warp, become wavy, or develop chevrons at the core splits. Waves in the core usually occurred at the bottom of the
core where the weight of the core above and the frame construction prevented the slip/slide from occurring. Today,
on some very large machines the bottom packet or two are glued together to address this problem.

Modern, non-directional electrical steel has between 0.5 percent and 3.5 percent silicon and up to 0.5 percent
aluminum. Silicon decreases hysteresis loss, increases permeability, eliminates aging, and greatly increases
resistivity. Aluminum further increases resistivity and thus reduces eddy current losses. Shearing strains do
adversely impact the magnetic properties, but this is not a major concern involving large stampings. Silicon steel
use in the States started around 1903 (patent date). Prior to this the core steel did age and experienced a significant
loss in magnetic properties.

Lamination thicknesses (in inches) commonly used are 0.012, 0.014, 0.0185, and 0.025. Lamination thickness
tolerance from specified thickness is about 0.0015 inch. On a given lamination, the variance is much smaller.

GRADING OF ELECTRICAL STEEL LAMINATIONS
There are two systems in the United States (six or more world wide). The American Iron and Steel Institute (AISI)
created a system in 1947. Each grade of steel was assigned a number based on maximum loss. The system consists
of a letter M (for magnetic steel) and the following numbers which represent 10 times the maximum allowable core
loss in watts per pound for 29-gauge steel (0.014-inch thick), measured at 15 kgauss (1.5 Tesla) and 60 Hz.

3
The other system, American Society for Testing and Materials (ASTM), consists of the first two digits representing
the thickness of the material in millimeters x 100. A letter represents the class of material (F being fully processed)
followed by three digits being the maximum core loss times 100 at 15 kgauss (1.5 Tesla) and 60 Hz.

ASTM 36 F158, similar to M19 29-gauge, is 0.36-mm (0.014-inch) thick and has a maximum core loss of 1.58 watts
per pound at 1.5 T and 60 Hz.

CORE PLATE
Laminations are coated with insulation, referred to as core plate, which is designed to limit/reduce the flux-induced
currents in a core. These currents form a loop in the axial and radial direction creating heat losses commonly
referred to as eddy current losses. The ASTM Standard A-976 provides classifications and descriptions of various
insulating coatings. Classifications range from C-0 to C-6.

Many old cores have class B (130 C) stator winding and core plate type 3 (C-3 class interlaminar insulation)
insulation systems. C-3 is an organic varnish/enamel coating. Upgrading such a machine to class F (155 C)
winding insulation will not allow one to operate the machine at core temperatures above 90 C, as the typical C-3
begins to deteriorate when cores operate above 90 C for long periods of time. Maximum burnout temperature (for
global VPI windings) allowed for C-3 core plate is 360 C.

Today, most large machine laminations are coated with C-5 class insulation. C-5 is an inorganic coating. C-5
results in lower electrical losses and good high-temperature stability. Losses in cores have gone from 4 to 5 watts
per pound to less than 2 watts per pound (at 1.5 T) due to higher-quality steels, thinner laminations, and better core
plate. C-5 also can withstand burnout temperatures of about 400 C and stress relief annealing up to 840 C. Non-
oriented electrical steel is stress relief heated somewhere between 700 C and 790 C. Stress relief is performed to
reduce stresses due to shearing and punching operations and to increase magnetic properties. The stress zone
consists of the bur and a small distortion zone directly behind the bur.

Some core plates facilitate punching and also will withstand annealing.

The very old C-2 core plate used in very old generators consists of magnesium silicate and is referred to as mill
glass, which is formed by the solidification of a melt. The coating can be abrasive. It has a milky-white look and a
pearly luster. Magnesium silicate also is known as talcum powder. Some of the older coatings have been referred to
as water glass which is sodium silicate, so water glass may or may not be the correct term. The silicate coatings
can be very abrasive once they break down.
*


Laminations for very small motors and even large rotors usually are coated with type C-0 core plate. It is a surface
oxide-type film resulting from normal steel processing.

Core plates usually range in thickness from 0.5 to 1.5 microns. Older core plate usually was roller coated. The
roller marks can be easily seen on the laminations. Today, more modern application systems are probably in use.

Stator laminations usually are punched and thus have burs. The laminations are deburred, but substantial burs
remain, thus the need to stack the core with the burs all facing in the same direction. Otherwise, bur-to-bur contact
will short the lamination insulation, creating hot spots in the core. Reasonable bur length for electrical steel
laminations usually is limited to no more than 10 percent to 15 percent of sheet thickness. The core plate must be
capable of adhering to the punched edges, thus another reason to debur is to ensure proper adhesion of the core plate
at the lamination edges.

CORE TEST PURPOSE
The core flux test is performed to validate/check that a new core is correctly designed and assembled, that there is
adequate interlaminar insulation, that the burs have been properly dressed and aligned, and in general that there are
no damaged areas that will generate hot spots.


*
Grand Coulee
4
On older cores the test is performed to ensure the core is still serviceable, and if not, to locate areas that must be
repaired prior to installing a new winding. After repairs are made, another core test is performed to ensure the hot
spots were eliminated.

The core test also is important in that it may detect insufficiently tight core splits. This can be determined from the
existence of excessive noise. Loose joints can result in noise in excess of 85 dB to 120 dB at a distance of 6 feet.
However, loud core tests also could be due to excessive lamination vibration and/or core-to-dovetail bar vibration.
A generator may have loose splits that will go undetected during operation due to the rotor magnetic force pulling
the core radially inwards and/or due to thermal expansion-related forces at running temperature.

CORE TEST INTRODUCTION
A generator core usually is tested prior to installation of the winding. The core is energized by wrapping the core
with an exciting winding and energizing the winding with a voltage source. In normal operation, at rated terminal
voltage, the flux in the generator core induces a given voltage in each turn of the stator winding. This is the rated
volts per turn of the generator, after making minor correction for winding factors. If we energize the exciting
winding at one-half this rated volts per turn, we will obtain rated flux in the back iron of the core. Then, based on
the voltage source available, the number of turns in the excitation winding is calculated to match the voltage of the
source. Of most importance in performing a core test is to obtain the correct flux density in the back of the core.

The NI product (ampere turns) of the exciting winding is the magnetomotive force (akin to voltage). This force is
required to counter balance the magnetic potential drop (akin to an IR drop) around the magnetic circuit. That is, the
total magnetic potential drop around any path must equal the source magnetomotive force applied to that path. Said
another way, the exciting coil current is determined by the required field intensity (H) necessary to support the
required flux per the magnetic core steel B-H curve. The need to use the steel B-H curve instead of a value of
reluctance (akin to resistance) is a result and complication of the nonlinear characteristics of electrical steel.

SEARCH COIL
A single turn of #12 to #18 gauge insulated wire (the search coil) is wrapped around the core being tested. The
voltage induced in this coil is due to and proportional to the flux in the back iron of the core. A true reading RMS
voltmeter must be used. The search coil should encircle only the core. The coil should be placed as close to the
core as possible, and should lie in the bottom of a coil slot. This is to prevent induced voltages due to stray leakage
flux. Do not include the frame of the core as this can result in errors of 1 percent to 8 percent. There are almost
always places where the search coil wire can be snaked through to avoid including the frame. If this is not the case,
small access holes can be drilled through frame members or core clamping structures. Another alternative is to use
core air vents; however, the search coil reading would have to be corrected accordingly. The coil leads must be
twisted together between the core and test meter, again to avoid stray flux influences.

FLUX TEST LEVEL
The exciting winding is installed with the proper turns and excited with a given voltage to obtain a predetermined
back iron flux level. As the current and voltage involved can be substantial, the exciting winding is held away from
the core and frame via insulating spacers (usually 2x4 or 2x6 framing). If the voltage source is not adjustable, one
or two extra turns are wrapped around the core. These turns can be removed, which will increase the flux density
and thereby allow one to zero in on the correct flux level. Removing exciting coil turns also will increase exciting
coil current. These extra turns are necessary to ensure the core is not overfluxed on initial startup. Additional turns
beyond what originally was calculated usually are removed to obtain the required flux. This is because stray leakage
flux and the flux induced in the core frame cannot be accounted for in the calculations. If the voltage source is
adjustable, it can be raised or lowered to obtain the proper flux density in the core back iron without altering the
exciting winding.

Depending on the reference cited, the test flux level can be anywhere from 50 percent of rated flux all the way up to
that which produces rated terminal voltage on open circuit. Several references specify testing at 110 percent of rated
terminal voltage to cover operation under normal overexcitation levels.

ACTUAL MACHINE OPERATING FLUX LEVELS
It should be noted that 85 K-lines per square inch (1.3 T) appears to be the back iron operating flux density often
quoted in the literature. However, in many machines the actual rated flux is somewhere around 75 K-lines per
square inch (about 1.15 T).
5

The actual machine used as an example later on in this section operates at 74.4 klines per square inch. It is a 30-
MVA, 6.9-kV, 94.7-rpm generator that first went into service in 1942.

Possibilities for lower flux levels:

- Mechanical considerations involving the back iron depth
- Limits on tooth flux levels could result in a lower back iron flux density
- Iron losses are lower at lower flux densities
- May have been necessary in older designs due to the quality of the steel at hand
- Specified machine reactances also impact flux density design levels. These reactances are not that critical
these days, as fast-responding high-ceiling voltage regulators will force a desired response regardless of
machine reactances and time constants.
- But most probable: The output coefficient of a machine used to maximize rated kVA and minimize
machine volume ($ saved) peaks at an air gap density of about 0.5 Wb per square meter (0.5 Tesla).

THE TEST
Once the agreed upon flux density is achieved, a 1-hour-long heat run test usually is performed. These days infrared
video equipment is used to look for unusual heating and hot spots.

Typically a 5 C to 15 C hot spot rise specification is agreed upon. At this time, detailed standards regarding flux
and temperature rise limitations do not exist, and thus these criteria must be specified and/or agreed to in advance
between customer and supplier. Infrared (IR) scans should be performed by a licensed IR technician, and calibration
and scan details worked out ahead of time. The best equipment and personnel can have video temperature data that
are a couple degrees off from a thermocouple attached to the surface of the core. How hot spots over ambient or
background are determined also must be worked out in detail and ahead of time.

RETORQUING THE CORE AFTER A CORE TEST
After testing, the core must be retorqued, as the test causes shrinkage per seating-in of the laminations and core
plate. This is especially true of new cores, and some manufacturers require a much longer seating-in period than
afforded by the core test. Typical shrinkage for a new core is about 0.1 percent of stack height, but most of the
shrinkage occurs in the core plate and is about 4 percent of total core plate but could go higher. Thicker coatings
can result in up to a 0.3 percent core shrinkage, which can reduce clamping pressure substantially. When
Reclamation obtains a new core, we require new clamping nuts and washers, and all bolt threads cleaned and chased
with chaser dies. A commonly accepted practice of checking to see if a core is tight is to attempt to force a penknife
blade between laminations on the face of several teeth.

CORE TESTING SAFETY ASPECTS
The high voltage and current involved in the exciting circuit requires attention to associated hazardous energy issues
(high-voltage barriers, proper relaying, sizing of conductors, conductor insulation, job hazard analysis, etc.). The
interior of a core might be considered a confined space. Energization of a large core will result in very high noise
levels, especially if the core has splits. A recent test on one of our 174-MVA hydrogenerators having core splits
resulted in noise levels in excess of 120 dB. This is a much higher level than during normal operation due to the fact
that there is no mechanical damping (winding), no torque loading, nor magnetic pull-in associated with the field.
Whereas a core test on a 2.4-MVA station service unit (no splits) was relatively quiet at about 70 dB.

Another safety aspect associated with core testing has to do with induced voltage across the core stack. This is
essentially the search coil voltage. On long hydrogenerator cores the voltage can be in the range of 90 to 120 volts,
and even higher in longer high-speed steam-driven generators. One could be subjected to this potential should one
span a large portion of the core.

An extensive list of core testing safety considerations can be found in IEEE Standard 56, Guide for Insulation
Maintenance of Large Alternating Current Rotating Machinery.

CORE TEST CALCULATIONS
The first step is to calculate the actual operating volts per turn (V
PT
) of the exciting winding which is of the stator
winding volts per turn:
6


PT
V to equal is
d
K
p
NTK
Lg
E 2
Where
E
Lg
is rated line-to-ground voltage
N is the number of turns per circuit
T is the number of turns per coil
K
p
is the coil pitch factor
K
d
is the winding distribution factor

The factor of 2 accounts for the fact that the flux per pole splits left and right as it enters the back iron area.

The volts per turn is the single-turn search coil voltage that will be monitored to ensure rated core flux is obtained
via the exciting winding. When the calculated volts per turn is induced in the search coil, we will know that the core
is operating at rated flux in the back iron. It is this rated flux that induces the voltage in the search coil.

In our example, the generator is a 6.9-kV, 30-MVA, 94.7-rpm generator. Rated line-to-ground voltage is 3983.7
Vac. There are 40 coils per circuit and 2 turns per coil. K
p
is 0.997 and K
d
is 0.965 (see following sections for
calculation details). Thus volts per turn is 25.89 volts.

Next, the number of turns for the exciting winding must be determined based on the exciting voltage supply
available as follows:

For a 480-volt supply, the number of turns required is 480/25.89 or 18.54 turns.
However, if the number of turns is too many, a higher voltage source should be considered.

The calculated number of turns needed is approximate due to the fact that most exciting windings also include the
core frame, which will carry some of the flux. A rule of thumb is to increase the supply voltage by 10 percent or
decrease the number of turns by 10 percent. Thus, the required turns would be closer to 16 (non integer turns are not
possible), and to be safe, one could start with 17 or 18 turns and then adjust the voltage or remove turns to obtain the
desired search coil voltage.

Next, one must determine exciting coil current requirements to properly size the voltage supply KVA rating and the
exciting winding wire. There are a couple ways of doing this. A discussion by L. M. Domeratzky of General
Electric in the 1952 IEEE paper by H. R. Tomlinson suggests that due to flux in the frame and other hardware (being
as great as 8 percent), the current requirements would be in the order of 400 to 500 ampere-turns per foot of mean
back iron core diameter.

Therefore, in our example where we have 29.6 feet mean core diameter and 15 turns (actually used), the required
current is:

500 x 29.6/15 or 987 amperes

Actual current measured during a recent core test was 925 amperes.

A more accurate method would be to first calculate the flux per pole which is

2.25x10
7
(E
Lg
) (#of phases) (#of circuits) / (freq) (#of coil turns) (#of slots)K
p
K
d


And for our example =1.941 x 10
7
lines

From this the back iron flux density is

Flux per pole / (2 x core area)

For our example =74,438 lines per square inch

7
Note that the core area must be reduced by vent duct area and by 0.95 or so for lamination stacking and core plate
considerations. This factor typically ranges from 0.94 to 0.99.

From the B-H curve the ampere turns per inch can be obtained; however, this does not account for frame and
leakage flux. Thus we correct the flux by 10 percent (worse case based on experience), thereby resulting in a flux
density of 82 klines/square inch. From the B-H curve the AT/inch is closer to 13 than the 8 that would have been
used. Therefore,

Current = turns of umber diameter/n back mean / 13 ) ( - -t in AT

And in our example for 15 turns and a diameter of 356 inches, required current is 969 amperes, where the actual
current measured was 925 amperes.

Actual accuracy will depend on the quality of the B-H curve data available for the core.

There are even more detailed calculations to take into effect watts loss, but on old cores, obtaining the stator core B-
H curve and watts loss per pound information is next to impossible and may not be desirable or necessary as watts
loss in this example was 76 kW compared to the measured KVA of 400.

Next, the excitation winding cable must be sized. The rule of thumb:

Cable size =1000 circular mils per ampere (or one could use the National Electrical Code tables).

Therefore, 1000 amps requires 1 million circular mils. That is, a 1000-mcm cable is needed.

One last note: This core has a stack of about 1250 laminations. For a normal operating volts per turn of 25.9 volts,
that works out to 20 millivolts across each lamination at rated flux.

WINDING COIL PITCH K
p

The distance between the two legs of a coil is the pitch. When this distance is exactly the distance between two
poles, the winding is a full-pitch winding and the voltages induced in both legs of the coil are exactly in phase. If
the distance is less than the distance between two poles, then the machine has a fractional-pitch winding and the
voltage induced in the coil is less than for a full-pitch winding.

To calculate the true induced voltage, one must correct the subject equation by the pitch factor, K
p
, where

( ) pole per slots span coil 90 sin - =
p
K

Example: For our machine:

We have a 480-slot, 3-phase core with 76 poles. The coil pitch is 1 to 7. Therefore, the span is 7 minus 1, or 6, and
the slots per pole is 6.3158. Therefore,

( ) 997 . 0 or 6.3158 / 6 90 sin - =
p
K

WINDING DISTRIBUTION FACTOR K
d
When several coils in a group are connected in series, the actual group voltage is less than the voltage induced in
each coil times the number of coils in the group (due to relative coil displacements). Another correction factor thus
is needed and is the distribution factor, K
d
.

K
d
is equal to sin ( ) ( ) 2 / sin 2 / - d n nd

Where n is slots per pole divided by the number of phases
d is the number of electrical degrees between adjacent slots or 180/slots per pole

8
Thus per the previous example

n is 6.3158/3 or 2.1053
d is 180/6.3158 or 28.5
K
d
=sin ( ) ( ) 2 / 5 . 28 sin 1053 . 2 2 / 5 . 28 1053 . 2 - - or 0.9648

TEMPERATURE RISE CONSIDERATIONS
- Various references for what is considered a hot spot rise varies between 5 C and 15 C.
- The temperature reference for the hot spot rise is a key point/issue and should be agreed upon prior to
testing.
- The core test results in heat being produced in the back iron due to laminar, interlaminar, hysteresis,
anomalous, and magnostrictive type losses.
- During a core test the teeth carry very little magnetizing flux and therefore do not generate much heat.
However, the teeth do heat up due to conduction from the back iron. Due to the conduction transport lag
and the heat transfer capability of the teeth to the surrounding air, we see that during a test the teeth can be
typically 3 C to 5 C cooler than the iron in the bottom of the slot.
- In an infrared image the air duct face temperature will be seen to lie somewhere between the teeth face
temperature and the bottom-of-the-slot temperature.
- The bottom and top packets of the stack, having more air surface, also may sit at a 3 C to 5 C lower
temperature (in some instances).
- If the top and bottom core packets are at a higher temperature the core plate may have been damaged due to
end winding leakage flux inducing large interlaminar voltages, resulting in higher eddy currents (mostly a
problem on large steam units in the far under-excited region of operation). One should look for discolored
varnish around the clamping fingers.
- The heat from the test setup conductors may affect the core temperatures in the area of these conductors.
- There may be hotter or cooler spots around the core due to frame and/or air cooler- related circumstances.
On a new core or on one with little to no hot spots these small temperature differentials may be a minor
concern. Not so for a questionable core that may have some hot spots or a large area of overheating.
- Most hot spots on the tooth face are due to loose vibrating laminations or due to damage occurring during
winding replacement and/or winding removal.
- Many teeth may have hot spots on the tooth corners near the wedge groves due to installation and/or
removal of wedges and/or removal of the winding.
- Tooth face and tooth corner hot spots are fairly easy to repair. Not so for bottom-of-slot hot spots, but
those are not so common.
- Hot spots can and do show up at core splits (chevrons at the splits due to thermal expansion issues). These
usually can be repaired.
- Large hot spots in the back of the iron of significant size and of a significantly high temperature are a major
issue regarding the decision to return to service, restack, reinsulate, or replace the core.
- The average core ambient temperature is used by some from which the hot spot is obtained. Others use the
average tooth face temperature as a basis for tooth hot spots, and core bottom-of- the-slot temperatures for
the basis for back iron hot spots. As a result, maximum allowable hot spot rise should be adjusted
accordingly.
- To detect back-of-iron hot spots, some deenergize the core at the end of the hour-long test then monitor the
core for higher temperatures that would surface over time, thereby indicating back-of-core hot spot
problems. This is understandable in the days prior to infrared cameras (one could not easily measure
bottom-of-slot temperatures). However, with infrared cameras, the bottom-of-slot temperatures now can be
monitored easily. Therefore, the author finds this after-test temperature monitoring to be of little added
value. Back-of-iron depth is approximately equal to tooth depth, which tracks back-of- iron temperature
with only a 3 C to 5 C lag, thus one can argue the IR scan of the bottom of the slots is sufficient.

RING TEST LOSS RELATIONSHIPS (or where does the heat come from)
- The inter-laminar loss (I
2
R loss due to circulating eddy current through adjacent laminations) is
proportional to the square of the back iron depth and also to the square of the flux density and the square of
the frequency. However, the interlaminar loss typically is 5 percent of the normal eddy losses, so except
for hot spots we can ignore interlaminar heating effects.
9
- The laminar loss (I
2
R loss due to circulating eddy current within a lamination) is proportional to the square
of the thickness of the lamination and also proportional to the square of the flux density and the square of
the frequency.
- The hysteresis loss is proportional to frequency and approximately proportional to the square of the flux
density.
- In summary, for a given frequency the core loss primarily is due to hysteresis and laminar loss, both of
which are about proportional to the square of the flux density.
- If we replace an old core having standard 0.025-inch-thick laminations with 0.02 inch-thick laminations
(and having the same magnetic properties), the new core laminar losses will be reduced to 64 percent of the
old core.
- If we test this new core at 93 percent of rated flux, the new core laminar losses would be reduced further to
55 percent, and hysteresis losses would be reduced to 86 percent. Why 93 percent? If one uses a fixed
voltage source and a small number of exciting coil turns are used, then it is not always possible to obtain
the desired back iron flux density.

CRYSTAL POWERPLANT CORE REPLACEMENT
- Crystal Powerplant: The old core had 0.025-inch-thick (24-gauge) fully processed M19 laminations. At
1 Tesla the loss would be 0.91 W/lb and 1.8 W/lb (conservative number) at 1.5 Tesla.
- The core was replaced with new lower-loss 0.0196-thick laminations having a loss of 0.45 W/lb at 1 Tesla
and a loss of 1.09 W/lb at 1.5 Tesla.
- The old core at a rated flux of 1.3 Tesla (85 klines/square inch) would have a watts loss of about 1.45 W/lb.
- The new core at rated flux of 1.3 Tesla would have a watts loss of about 0.77 W/lb (see E-105 B-H curve
data sheet).
- The new core actually was tested at 93 percent of rated flux (1.2 Tesla) at which point it would have a watts
loss of about 0.64 W/lb.
- The watts loss of the new core tested at 93 percent of rated flux would be about 0.64/1.45 or 45 percent of
the old core tested at rated flux.
- For a normal 1-hour core test starting at an ambient air temperature of 20 C, one would expect to see a
15 C to 25 C temperature rise.
- The new core should see only 45 percent of this temperature rise or about 7 C to 11 C temperature rise.
The actual back iron temperature rise was about 11 C. The average core face rise (including bottom of slot
and tooth face) was about 7 C.
- Many machines are tested at 85,000 lines/square inch. This may be because it is about 110 percent of
typical back iron flux or it could be because a machine may have been designed to operate at a higher flux
density. The lower-loss modern-day steels may be one factor contributing to this. Over the years losses
per pound have dropped considerably with material processing advances.
- Many data sheets provide lamination losses only at 1 T and at 1.5 T. If a quality watts loss curve for a
given steel is not available, then a linear interpolation between the 1 Tesla and 1.5 Tesla data points will
provide reasonable results to obtain the required watts loss.

CORE REPAIRS
Discovering at the last minute (during a scheduled generator rewind) that a core needs to be replaced can put a
severe burden on a utility due to the need to greatly extend the outage period. To reduce the outage time there are
two options. The first option is to reinsulate the existing core laminations. However, this also is somewhat time
consuming. Another option is to restack the existing core. This is an ideal solution when only a small section of the
core has deteriorated or when a winding failure has caused extensive damage that is confined to only one or two
slots. In this instance, it is not unusual to find a burned out section of the core about the size of an apple.
Fortunately, this kind of damage usually occurs only in multi-turn windings. Single-turn (bar) windings rarely cause
core damage of any significance.

In cores having splits, mechanical forces can cause chevrons and wear patterns that could result in hot spots at the
splits. Restacking the overheated areas and installing an insulating shim is a typical solution.

For smaller areas needing repair, one usually is confronted with a tooth packet or a cluster of packets (as many as six
to eight) that are overheating. Sometimes the damage is limited to the tooth packet corners damaged from wedging,
winding removal, or bore cleanout.
10

REPAIR OPTIONS
- Sometimes a rubber mallet blow to the face of the tooth packet will break up the circulating current paths.
Then the packets are stabilized by some means to prevent the overheating from recurring.
- If the faces of the laminations are smeared or damaged, or the corners are distorted, grinding and etching
are oftentimes used to separate the lamination edges.
- In some instances the surface shorts may extend deeper into the tooth packet. In this case the core
clamping bolts are loosened and the packet laminations are separated by driving thin pieces of mica sheet
into the laminations. Two to five mica sheets spaced evenly into the packet typically are required.
- In some instances after unclamping the core a mica-epoxy paste or a weeping epoxy can be forced in to
reinsulate lamination edges. However, burs and/or smearing problems must first be addressed.
- In almost all cases a weeping epoxy applied after other repairs will provide additional insulation and also
mechanically stabilize the packet.

Exact procedures for repairing hot spots should be discussed with the contractor as each firm has fine-tuned
their own repair procedures.

CORE LOSSES
Core losses can be broken down as follows:

Hysteresis losses
Eddy current losses within individual laminations
Eddy current losses between laminations (interlaminar)
Magneto-striction losses (mechanical, vibration, hum)
Anomalous losses (whats left over).

Core plate usually is designed to limit interlaminar eddy current losses to 5 percent or less of eddy current
lamination losses.

Loss Formulas

P
(eddy laminar)
= ( )
3
6
16
10
2 2 2 2
cm w
B f t

t




Or

P
eddy laminar
2 2
B t

( )
2
density flux
2
) (thickness

Similarly

P
inter-laminar
2
B
2
d

( ) ( )
2
density flux
2
iron back of length radial

and from the Steinmetz equation,

P
hysteresis
( )
6 . 1
B f

However, it has been found that the Steinmetz exponent depending on the material varies between 1.5 and 2.5, and 2
appears in most references; therefore,
11

P
hysteresis
( )( )
2
density flux frequency

RECOMMENDED SOURCES
- ASTM A840/A840M-00 has some information on standard specifications for fully processed magnetic
lamination steel. (Probably not needed by utility engineers.)
- ASTM Ready Reference on Electrical and Magnetic Properties of Metals lists many electrical steel and
parameter values, but also is probably not of much use to utility engineers.
- The August 1952 IEEE paper by H. R. Tomlinson is a useful source.
- The ASTM Standard Classification of Insulating Coatings for Electrical Steels by Composition, Relative
Insulating Ability and Application (A976-03) is a good reference for core plate descriptions
- The American Iron and Steel Institute Steel Products Manual, J anuary 1983, is a valuable document for
reference.
- The Text Book Electrical Machines by C.S. Siskind is a good reference for detailed information on
corrected voltage of an alternator and for understanding and calculating K
d
and K
p
.

REFERENCES
1. Stone, Greg C., E. A. Boulter, Ian Culbert, and Hussein Dhirani, Electrical Insulation for Rotating Machines,
IEEE Press, 2004.

2. Siskind, Charles S., Electrical Machines, McGraw-Hill, 1959.

3. Beckley, Philip, Electrical Steels for Rotating Machines, IEE Power & Energy Series 37, 2002.

4. Walker, J . H., Large Synchronous Machines, Clarendon Press, Oxford, 1981.

5. American Society of Mechanical Engineers, Electrical and Magnetic Properties of Metals, 2000.

6. Amazeen, Paul G., Topics in Engineering Electromagnetics, Worcester Polytechnic Institute, September 1966.

7. American Society of Testing Materials, A840/A, Standard Specification for Fully Processed Magnetic
Lamination Steel

8. American Society of Testing Materials, A976-03, Standard Classification of Insulating Coatings for Electrical
Steels

9. Institute of Electrical and Electronics Engineers, Standard 56-1977, IEEE Guide for Insulation Maintenance of
Large Alternating-Current Rotating Machinery (10,000 kVA and larger)

10. American Iron and Steel Institute Electrical Steels Steel Product Manual, J anuary 1983.

11. Tomlinson, H. R., Interlaminar Insulation Test for Synchronous Machine Stators, AIEE Transactions,
August 1952.

12. POSCO Electrical Steel Catalogue, 2001.

13. Villa Industries Silicon Steels and Their Applications.

14. Anderson, Anthony, Generator Core Failures.

15. EASA Technical Note 17 Stator Core Testing.

16. National Material Company Product Catalog.

17. EASA Guidelines for Maintaining Motor Efficiency During Rebuilding.

12
18. Klempner, G., and I. Kerszenbaum, Operation and Maintenance of Large Turbo Generators, IEEE Press, Wiley-
Interscience, 2004.

You might also like