See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3353863

Core faults in large generators

Article  in  IEE Proceedings - Electric Power Applications · December 2005

DOI: 10.1049/ip-epa:20050102 · Source: IEEE Xplore

CITATIONS READS

32 3,461

2 authors, including:

P.J. Tavner
Durham University
167 PUBLICATIONS   12,532 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Laurence Scott & Electromotors Ltd View project

SRC large generator core project View project

All content following this page was uploaded by P.J. Tavner on 01 January 2018.

The user has requested enhancement of the downloaded file.

Core faults in large generators
P.J. Tavner and A.F. Anderson

Abstract: A core fault is a failure in the laminated core of a large electrical machine. Such faults are
relatively rare but when they occur, the value of the machine and its importance mean that the
investigation of the failure assumes a high priority. On many occasions the details of such failures
assume major commercial significance, therefore failure investigations have, of necessity, to be
handled in a confidential manner, touching as they do on the design, manufacture, operation and
insurance of large electrical plant. There has therefore been no published literature on core faults.
However, the scientific principles of the mechanisms at work have been studied in considerable
detail and papers published on those principles in the international literature. This paper brings
together that literature and those scientific principles, giving details of the underlying design,
constructional and operational factors which affect these faults. New evidence is presented of the
underlying factors which allow core faults to initiate and grow. The paper will allow engineers to
analyse such faults, draw rational conclusions on the causes for each occasion and devise suitable
repair/rebuild strategies applicable to that situation.

List of symbols r radial axis, in plane of coreplates, m

y circumferential axis, in plane of coreplates, radians
B flux density, Tesla z vertical axis, perpendicular to the plane of
H field strength, ampere-turns/m coreplates, m
J current density, amperes/m2
‘ length of the core fault cavity
L‘ inductance of length ‘ of the core and frame circuit
to Yoke Flux Linking 1 Introduction
Ri inner radius of the stator core
Ro outer radius of the stator core 1.1 What is a core fault?
Rc radius of core fault cavity in stator core Core faults can occur anywhere in the laminated steel core
Rb stator core radius at a background temperature of large rotating electrical machines but are more common
of Tb in the stator core. They can also occur in transformers but
R‘ resistance of length ‘ of the core and frame circuit are more unusual, probably because the transformer has
to the Yoke Flux Linking lower leakage fields and a more stable electrical, thermal
Z‘ impedance of length of the core and frame circuit and mechanical operating regime.
to the Yoke Flux Linking A core fault, Fig. 1a, initiates when coreplates are
r resistivity of the coreplate material, S electrically connected together, either because of an
s density of coreplate material, g/mm3 insulation failure between them, or as a result of physically
Ti melting point of core steel, 1C imposed short circuits, Fig. 1b, due to various causes
Tb mean temperature of the core, 1C including foreign bodies. These connections circulate
k radial heat transfer coefficient of laminated core additional currents in the core, coupling the main flux,
steel, W/m1C leading to additional losses and heating, further weakening
mo magnetic constant coreplate insulation and expanding the fault. The fault is
mr relative permeability of a permeable material then a zone in the core which is overheated, buckled and
d skin depth in permeable conducting material, mm electrically interconnected. This can expand, melting
do skin depth in non-permeable conducting material, material, leading to the catastrophic runaway of the fault
mm and creating a cavity in the core continuing until the main
g stacking factor of a laminated core conductor insulation is damaged and the machine is
x horizontal axis, in the plane of coreplates, m disconnected for an earth fault. Core faults generally occur
y horizontal axis, in the plane of coreplates, m some distance from the main conductors and final
disconnection frequently occurs after irreparable damage
r IEE, 2005 has been done to the core and conductors. Core faults still
IEE Proceedings online no. 20050102 remain an important issue, the authors have been involved
doi:10.1049/ip-epa:20050102 in investigating such faults on modern machines during the
Paper first received 26th March and in final revised form 23rd June 2005 past four years.
P. Tavner is with the School of Engineering, Durham University, South Road,
Durham, DH1 3LE, UK 1.2 Core construction
A.F. Anderson is an Independent Electrical Engineering Consultant Figure 2 is a photograph of a large stator core of a two-pole
E-mail: [email protected] turbogenerator under construction, showing some of the

IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005 1427
 stator core stator core
back building bar

hydrogen stator core

cooling ducts segment

stator core
frame member

stator core
frame member
Fig. 2 Laminated stator core of a 500 MW, 2 pole hydrogen-
cooled turbogenerator under construction
Clearly showing the segmental core construction, axial cooling ducts in
the coreplate, keybars and stator frame

lower the risk of interlaminar conduction, sometimes they

are reinsulated after deburring. For smaller (o8 MW)
stator cores the coreplates can be complete circles or annuli
but for larger machines, as in Fig. 2, they are always
stamped in segments and the core is built up in brick wall
a
fashion, with butt joints between each segment. Coreplates
are also stamped with holes and slits to create axial cooling
core fault ducts, also visible in Fig. 2.
The core cooling medium may be air for smaller machines
(o120 MW) or hydrogen particularly for larger two- or
fault current
R
four-pole machines (4150 MW). Stator cores can be cooled
by the axial flow of coolant gases but also by radial flow, in
Ri
which case the core is divided into packets separated by the
radial ducts for ventilation. These ducts are supported by
radial spacer beams. Many axially-cooled large cores also
have additional radial ducts near the core ends to assist the
cooling of that high loss part of the core. Coreplates near
the core end on large machines may receive multiple
b
reinsulation coats after deburring.
Coreplates are keyed to frame bars or welded to frame
Fig. 1 Laminated core with core fault sticks, providing support for the huge weight of the core
a Example of damage to the end of a laminated core after a core fault, and the necessary mechanism to resist the torque reaction
showing molten coreplate material running out over the stator core from the air-gap magnetic field. Welded construction is used
frame
for smaller machines, with complete circular coreplates, but
b Laminated stator core showing origin and spread of a core fault
for segmental cores the coreplates always interlock with
keybars attached to the inner bore of the stator frame. It is
constructional features described below. Cores are con- then the mechanical contact between the coreplate and these
structed from sheets or coreplates of electrical steel, alloyed so-called keybars that provides the torque reaction. There
with silicon to lower losses and insulated to prevent are various technologies associated with keybars:
conduction between them. A variety of proprietary insula-
tions have been produced for coreplates including:
They can be bolted or welded to the frame, as shown in
Fig. 2.

Oxide layers on the steel, produced during the rolling and
They can be manufactured of steel or a lower resistance
annealing process employed only on small machines, material such as bronze.
outside the scope of this paper.
The keybars can be insulated from the coreplates or the

Layers of paper insulation, an old technique no longer in frame to prevent the flow of current between core and
current use. frame but this insulation can be weakened by the torque

Coatings of inorganic material rendered insulating by reaction forces.
baking in the production process.
In the case above one keybar can be provided with a

Coatings of organic resin varnish cured during the conformable copper insert between the coreplate and the
production process. key, connected to earth. This is done to provide an earth
connection for the core. In some cases these conformable
The last method is almost exclusively used on large modern copper straps are fitted to all keybars.
electrical machines. Coreplates are usually stamped from
Low resistance keybars of bronze or conformable copper
insulated steel strip and may be deburred after stamping, to are sometimes connected together at each core end,
1428 IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005
 forming a conducting cage surrounding the core, with the on machines in California experiencing elevated stator
intention of reducing the leakage flux which can reach the core end region temperatures because of underexcited
core frame. operation. Winchester [7] in 1955 presented an analysis of
the end region field of a turbogenerator using the
electrolytic tank. This was the forerunner of many papers
The stator frame formed by these sticks or keybars is either using numerical methods to compute the field. Mason [8]
the main frame of the machine or, for larger machines, an followed in 1959 with operational experience on UK
inner frame housed in a larger outer casing, which in the turbogenerators. The first paper to describe explicitly the
case of hydrogen-cooled machines forms the coolant damage that can lead to a core fault was Oberretl [9] in
pressure vessel. For large two-pole machines the inner 1963, although his work was about a short circuit generator
frame will be isolated from the outer by tuned springs, to used for switchgear testing.
reduce external noise due to the twice frequency excitation By the early 1960s the construction of larger rating
of the core. machines required more attention to design and manufac-
The core is built up in an external frame, or for smaller ture. Walker, Rogers and Jackson [10] in 1964 studied
machines in an internal mandrel. Mechanical pressure is the pressure in the laminated core. By 1970 when large
applied to core packets successively by the use of axial 500–700 MVA machines were becoming common, Howe
hydraulic rams. This process continues until the full core [11] searched for design methods to manage the end region
length is reached, when the core is clamped between leakage demonstrating the factors determining leakage flux
clamping plates, which finally are firmly attached to the on the surface and within the stator core. Stoll [12] also
keybars or sticks. A ring flux test can then be applied to extended the work on butt joints. Vickers [13] in 1974
confirm that the core can withstand the full magnetic flux produced the most thorough review available of current and
without excessive overheating. proposed turbogenerator construction methods.
These processes are carried out in accordance with Jacobs [14] in 1978 published the first 3D analysis of the end
defined manufacturing procedures, which provide objective region field in a laminated stator core, although a number
evidence of quality control, usually in the form of records of manufacturers had demonstrated operational programs
of: without publication. Subsequent work used newly available

Electrical steel origin, thickness and grade; computer facilities to analyse stray fields in machines more
carefully, predicting the unusual effects observed operation-

Applied pressure; ally in earlier work.

Core tightness; Renew [15] measured the influence of transient machine

Personnel involved; disturbances on end region field, because it had become

Finally core ring flux withstand when the core is clear that pole slipping could be a core fault initiator.
complete. Tavner [16] in 1978 continued trying to define design office
rules to manage end region fields while Jackson [17]
analysed the effect of leakage field and eddy currents on
1.3 Published literature the voltages between coreplates and their possible contribu-
Problems of overheating due to stray load losses were first tion to the damage reported by Oberretl. Again this was
recorded in the stator cores of rotating electrical machines thought to be a particularly important contributor to core
at the end of the 19th century, when the technology of using fault initiation during transient operations. The discussion
insulated steel coreplates in the core of AC machines was at a seminal meeting at the IEE in 1980 [18] summed up the
becoming established. These problems generally arose concerns in UK at that time for external leakage field
because of inappropriate steel thickness or incorrect alloying calculations under high temperature and transient fault
content or poor insulation. The practice of using silicon conditions.
steel alloys in coreplates of 0.5–1 mm thickness, insulated Work then moved into a more practical and applied phase
with interleaved sheets of paper was generally being applied as experience of a number of core faults on very large
after 1903, when the first patent for silicon alloying of turbogenerators increased knowledge. Anderson [19] in
electrical steel was taken out. It is not clear when and where 1980 reported a breakdown phenomenon, thought to arise
core faults first occurred but it was certain that their cause in transient conditions and to be the initiator of a core fault.
was associated with stray losses, particularly towards the Anderson [20] measured the external leakage flux of a small
ends of the stator core. machine and Platt [21] measured internal leakage flux and
The first analysis of stray losses in machines appeared in interlaminar voltages on a full size machine under similar
Germany in 1924 from Rudenberg [1]. Rockwood [2] in conditions.
1928 devised more methods for calculating stray loss and The emphasis now shifted to the coupling of the external
Laffoon [3] followed in 1929 as the USA wrestled with leakage flux with the frame surrounding the core, because
the introduction of larger AC generating units, feeding evidence had suggested that these components contributed
transmission lines of great length, where the effect of to core faults. Tavner [22] offered a qualitative explanation
capacitive loading and underexcitation of the machine of the phenomena and Hilliar [23] provided detailed
was beginning to be realised. Phear [4] in 1936 initiated a measurements from a 500 MW turbogenerator showing
process of analysing the true MMF drop in the butt that currents as high as 12 kA circulated in the frame at
joints of stator cores, which were now so large they had to rated stator operating current but that these currents could
be constructed using coreplate segments. This started a be higher at lower rotor excitation. Bratoljic [24] in 1986
thread that will run right through the study of core loss and then used sophisticated 3D magnetic field analysis on a
failure. 360 MVA turbogenerator to predict the fields and currents
Post-war, the hydrogen-cooling pioneered in the USA in the surrounding a stator core, reaching values in excess of
1930 s started a rush to use the higher ratings, 4100 MVA 16 kA, while Mecrow [25] in 1989 reported the application
in other countries. In UK Richardson [5] demonstrated of similar 3D analysis to the field around and inside the
better methods of calculating stray loss on these larger stator core end region showing how the various design
machines. Estcourt et al. [6] wrote a seminal work in 1953 features of the machine affect the end region losses. This

IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005 1429
completed the work of [14] but applied it for design there is now adequate knowledge available in the public
purposes to practical turbogenerators. domain.
By 1983 Sutton [26] had developed a low-voltage alternative This paper attempts to rectify this omission by drawing
to the core ring flux test for detecting incipient core faults in the information together, presenting the results of new
large machine cores. This paper reported a device developed research and setting out the main causes and effects of core
by the CEGB and known as El CID, the Electromagnetic faults.
Core Imperfection Detector. The technique has become
widely used by manufacturers and operators as a quality 2 Design principles
control check on laminated stator cores on completion of
manufacture and in service. The following quantities are important to the design of an
Beckley in 2002 has traced in [27] the development of electrical machine and there is a fundamental relationship
modern electrical steels, explaining the benefits of various between them:
insulation coatings and thicknesses on the performance of a

Electrical loading, axial linear current density at the air
stacked core and describes the action of a core fault. Finally
in 2004 Mogi [28] has demonstrated a direct experimental gap, in A/mm;
method for measuring the rotational loss in a core with butt
Magnetic loading, air gap flux density, in Tesla;
joints.
Mechanical loading, rotor peripheral velocity, in m/s;

Thermal loading, temperature rise, in K.
1.4 This paper
The published work has traced many, but not all the Electrical design is a compromise between the electrical
factors that can initiate a fault, but there have been only and magnetic loading, allowing the maximum transfer of
two papers, Oberretl [9] and Anderson [19], describing power through electrical and mechanical components. The
mechanisms occurring in real core faults. This is in part resultant design incurs losses in machine components
because of the commercial confidentiality surrounding impacting on the thermal loading, relieved by cooling in
failures and the sensitivity of the information involved but the radial or axial ducts at the thermally stressed areas. The

Table 1: Comparison of design parameters for large machines

Label 1 2 3 4 5 6 7 8 9

Type of Air cooled Air cooled Air cooled H2-cooled Air cooled H2O and Air cooled Air cooled H2O and
Machine Induction Induction Turbogen- Turbogen- Hydroge- H2-cooled Hydroge- Hydroge- H2-cooled
Motor Motor erator erator nerator Turbogen- nerator nerator Turbogen-
erator erator
Stator winding Air Indirect Air Indirect Air Indirect H2 Indirect Air Indirect H2O Direct Air Indirect H2O Direct H2O Direct
Cooling
Stator core Air Direct Air Direct Air Direct H2 Direct Air Direct H2 Direct Air Direct Air Direct H2 Direct
Cooling
Rotor Cooling Air Direct Air Direct Air Direct H2 Direct Air Direct H2 Direct Air Direct Air Direct H2 Direct
S [MVA] 7.11 12.56 125 108.6 100 294.1 100 355.5 1111
P [MW] 6.14 10.75 100 92.3 91 250 90 320 1000
Poles 4 4 2 2 32 2 88 14 2
Rev/min 1790 1770 3000 3000 187.5 3600 68.2 428.6 3000
Core Length, 628 874 4450 2800 2500 3700 1280 2960 7200
mm
Rotor Diameter, 742 988 890 924 6482 1034 12 968 4934 1200
mm
Mechanical 69.5 91.6 139.8 145.1 63.6 194.9 46.3 110.7 188.5
Loading, Rotor
Peripheral
Velocity [m/s]
U [kV] 6.8 13.8 11 13.8 15 15.75 15.75 22 24
Magnetic 0.9 0.63 0.94 0.83 0.99 0.98 1.05 1.12 1
Loading, Air
Gap Flux
Density [Trms]
I [A] 604 538 6560 4543 3850 10 782 3666 9329 26 729
Power per unit 9.8 12.3 22.5 33.0 36.4 67.6 70.3 108.1 138.9
Core Length
[kW/mm]
Electric 93.0 97.0 116.3 146.8 58.6 155.7 62.5 96.2 255.3
Loading,
A [A/mm]
Current Density, 3.9 3.1 2.8 3.8 3.2 5.8 3.7 5.3 11.1
h [A/mm2]

1430 IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005
compromise achieved defines the rating of the machine in 1.75
MVA. recommended V/Hz
Table 1 is a comparison of the loadings of nine large IEC 34.4 test
rotating electrical machines, with ratings from 7 to
IEEE 115 test
1100 MVA, a range of over three decades. Figure 3 shows
a graphical comparison of these machines in order of 1.50
ascending power per unit core length kW/mm.

PU flux
1000

1.25

100
loading variable

10 1.00
0.01 0.1 1 10 100
minutes
a
1.0
4.5

4.0
0.1
induction motor
air cooled
induction motor
air cooled
turbogenerator
air cooled
turbogenerator
H2-cooled
hydrogenerator
air cooled
turbogenerator
cooled
H2O & H2-
hydrogenerator
air cooled
hydrogenerator
air cooled
turbogenerator
cooled
H2O & H2- 3.5

3.0

loss, W/kg
2.5

type of machine 2.0

S, MVA electric loading, A A /mm 1.5

magnetic loading, air power per unit pore 1.0

gap flux density, Trms length, kW/mm
0.5
mechanical loading,
rotor peripheral 0
velocity, m/s 0 0.5 1.0 1.5 2.0
flux density, T rms
Fig. 3 Variation of loading variables for different types of
machine, see Table 1 Stoll, calculated Eddy loss, half- Nakahara, unsegmented
lap 0.75 mm gap, peak in coreplate properties measured on
coreplate Epstein square
Mogi, calculated total loss in a
Core faults usually, but not always, occur in the stator and core, half-lap 0.8 mm gap, normal yoke flux density
average in coreplate
one can be forgiven for assuming that the only design factor
Mogi, measured total loss in a
of importance is the magnetic loading. But the saturation of core, half-lap 0.8 mm gap, guaranteed loss for this Grade of
coreplate at 1.5 T rms
electrical steels limits the variation of magnetic loading average in coreplate

between machines of different size and design. The magnetic b

loadings in Table 1 extend from 0.81 to 1.12 Tesla a range
of 139%. However, to study the effect of magnetic loading Fig. 4 Stator core: Graph of overflux capability for a typical
on a core fault consider Fig. 4a, which shows an overflux turbogenerator stator core
a 1 pu flux represents a normal magnetic loading. The points show test
capability for a typical turbogenerator stator, where 1 pu
values of overflux prescribed by IEC and IEEE
corresponds to the rated air gap flux density. Typically a b Half-lap core losses for 350.50 coreplate showing from Mogi’s
core can stand an overflux without core damage provided measurements and predictions how losses increase sharply as flux
the application period is short. The curve shows that this density in the yoke increases
manufacturer recommends that a 1.2 pu overflux can be
sustained almost indefinitely. The curve also shows overflux
test values specified in two international standards and it
should be noted that the manufacturer’s recommendation The mechanical loading in Table 1 extends from 46.3 to
lies between those two standards, the IEC test offering the 194.9 m/s, a range of 420%. Leakage flux and consequent
most arduous condition. Essentially any operation of a loss has an important influence on core loading. This is
large machine that takes the magnetic loading above the shown by [14, 25] to be related to the stator current and
recommended line in Fig. 6a could initiate a core fault. The therefore the electric loading. Table 1 and Fig. 3 show a
magnetic loading, or air gap flux density, is defined by two wide variation of electric loading from 58.6 to 255.3 A/mm,
dimensional magnetic analysis at the centre of the laminated a range of 436%. It is interesting to note from Fig. 3 that
core. But additional losses arise due to the segmented nature the electrical and mechanical loadings of machines over the
of the core and towards the core ends, where axial leakage whole range are closely in step with one another. Electrical
flux drives up eddy currents, adding to the thermal loading loading is affected by cooling, which allows conductors to
[9, 11, 17]. Core faults do not usually originate in the centre carry higher currents, the most striking example being
of the machine core. Figure 4b, using results from [12, 28], the largest machine, rated at 1000 MW, with an electric
shows how rapidly losses in a segmented core rise as loading of 255.3 A/mm, achieved by directly water-cooled
magnetic loading increases. stator windings. The current density in the winding, at

IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005 1431
11.1 A/mm2, is over 50% higher than any other machine in
Operational factors
Table 1. The measurements of [23] showed that the leakage  Overfluxing incident.
flux at the core end of a practical machine increased in
 Pole slipping or other severe transient operation of the
direct proportion to the stator current, thereby increasing
the flow of stator frame currents. generator.
Readers can conclude from Table 1, Fig. 3 and the above  Foreign bodies shorting coreplates, arising from within
that: the machine due to wear and tear, introduced during
operation.

Loadings are not necessarily proportional to rating.
Machines of low rating can have high electrical, 3.3 Shape
mechanical or magnetic loadings and vice versa. The initiation zone of the core fault is a circular defect

Magnetic loading ranges are limited by saturation and between coreplates, in the r–y plane. The defect may be
probably do not indicate the propensity for a core fault embedded in the core or at the core surface, in which case it
to grow but operation of the machine above the curve in will be part-circular. Figure 5a shows typical damage in a
Fig. 4a is very likely to initiate a core fault and Fig. 4b core fault at the heart of a core where:
shows how that could happen close to butt joints in a
large core. Limiting the magnetic loading to 1 T will limit
the risk of core fault but this limit may have to be
exceeded in larger machines.

Mechanical loading ranges are large and generally in step
with electric loadings and do not affect core fault
initiation or growth.

Electric loading ranges are large, depending on cooling,
and machines with higher electric loadings will have a
greater propensity for core faults to grow, once initiated,
and need to be designed to reduce this risk. Limiting the
electric loading to 100 A/mm will limit the risk of core
fault growth but again this limit has to be exceeded for
larger machines. In such cases it is necessary pay
attention to coreplate insulation quality and cooling
intensity to reduce core fault risk.

A critical factor in the propensity of a core to suffer from
core faults could be power per unit Core Length, in
kWmm. This will be discussed further.

a
3 Core fault theory and results

3.1 Fault process

There are two regimes for the core fault:

Initiation.

Growth.
These two processes do not necessarily follow one another
immediately. In fact the authors have inspected stators
where a core fault had initiated but remained dormant
without expanding into an irreparable condition. Such a
fault could remain dormant for days, weeks or even years,
until conditions for growth are favourable.

3.2 Initiation
A core fault initiates when coreplates become electrically
connected together by mechanical or electrical action caused
by one or a combination of the following:

Constructional defects
 Burrs on the coreplate, [27].
 Inadequate insulation of the coreplate.
b
 Incorrect construction of the core.
Fig. 5 Core fault in turbogenerator
 Introduction of foreign bodies, particularly metallic, a Photograph of a coreplate at the centre of a core fault in a 500 MW
into the core during construction, shorting coreplates. hydrogencooled turbogenerator
Note the 4 cm diameter cavity where melting has removed material,

Thermal factors and the radial nature of the thermal damage to the coreplate insulation
b Flux plot generated using the MEGA FE package showing the field
 Local hot spot in the core. pattern surrounding a core fault developed by Mechanism I, Yoke
 Restricted cooling in one place in the core. Flux Diversion

1432 IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005

Each coreplate was in intimate electrical contact with its
Mechanism I or Yoke Flux Diversion at the site of the
neighbour, fault. Circulating current in the core fault itself opposes

Current could flow axially between coreplates, the main core flux, diverting it around the fault. The
return current circuit is within the ellipsoid-shaped cavity

Current flow caused coreplates to overheat,
itself. In this case the circulating current only links the

Coreplate insulation had suffered thermal damage in a flux crossing the core fault. A flux plot of a circular cavity
circular zone, with Mechanism I at work is shown in Fig. 5b.

Coreplates had buckled due to local thermal expansion,
Mechanism II or Yoke Flux Linking between the core

Coreplates had locally melted at the centre of the fault fault and the core frame. Current flows axially in the fault
then radially from the fault through coreplates to the
This damage actually centred on a circular axial cooling frame, flowing axially in the frame itself, see Fig. 6. In this
duct and was believed to be initiated by debris in that case the circulating current links the main flux in the
duct. However, the heating and subsequent melting has machine yoke.
enlarged the cavity exhibiting the circular form character-
istic of most core faults. At the two axial tips of the The latter mechanism is the most powerful because it has
fault cavity, see Fig. 6, the circulating current is concen- the potential to develop a higher driving emf. The balance
trated. The fault expands axially until it meets an between the two mechanisms will be discussed in the
impenetrable barrier. The circular shape is dictated by Appendix.
effective cooling of the laminated core, which limits radial A core fault grows in an axial direction, driven by the need
growth of the fault. There will be a high temperature at the for the fault to increase coupling with the main core yoke
centre of a fault but there is a substantial variation in flux. Radial growth is limited by the effective cooling of the
temperature radially away from it. This is exemplified by the core. Axial growth is not limited by radial ventilation ducts,
colour changes observed in the coreplate insulation around since conducting steel duct spacers bridge these ducts,
the fault shown in Fig. 5a. The high temperatures in a core allowing the fault current to pass across. There is evidence
fault can cause local melting of the coreplate, in which case on some faults that the duct spacers focus the fault current,
the cylindrical defect becomes a cavity as in Fig. 5a. The forcing it axially through the spacer at particularly
initiation zone thus expands, radially and axially, to form a favourable radial positions near the tooth root on coreplate
narrow ellipsoid cylinder-shaped defect, see Fig. 6. Core butt-joints. Interestingly the use of spacers with insulating
faults are rarely regularly shaped, however, they can be properties would limit the growth of core faults in machines
approximated to this ellipsoid model, allowing some with ducted cores. The authors have no knowledge of this
analysis. approach in large machines. The spacer property of concern
to the designer is mechanical strength and this would be
expensive to achieve with insulating materials.
stator endwinding If the core is in contact with the core back, a tell-tale
indicator of damage within the core will be arcing and
rotor stator
2
3 sparking activity on the core back, most noticeably at the
core-fault
1 ends of core fault where the return fault current has bridged
the contact between core and frame. At this point the
contact may be good but will vary with torque reaction and
load, opening and closing with each half cycle of the mains
frequency, resulting in the arcing activity observed. The
building bar fault circuit is almost entirely in structural or electrical steel,
the resistivity of which is high (30–55 mS).
In Mechanism I, in the local region round the surface of the
ellipsoid cavity the skin effect could be fully developed and
LI
2Rc
the currents inductance-limited as they circulate. The skin
Ro-Ri depth in the core fault will alter when melting occurs and
core fault
the steel rises above the Curie Temperature, in which case
yoke core fault skin depth increases, fault resistance falls and losses fall. In
I
RI axial Mechanism II the skin effect is not necessarily fully-
2Rc
ventilation developed, therefore the fault current is generally resistance-
ducts limited.
fault current mechanism II fault current mechanism I
yoke flux linkage yoke flux diversion In either case, Mechanisms I or II, the fault current has
tooth relatively little effect on the driving flux, like the resistance-
limited circuit described by Stoll [31], and this is an
Fig. 6 Section through a laminated core important reason for the destructive power of a core fault,
Showing extended core fault and Mechanisms I and II for the
since, once initiated and growing, an increasing fault current
extension of the fault
does not significantly reduce the driving flux provided by
the main excitation on the rotor of the machine. The fault
current is in effect a single shorted turn, the main excitation
remaining unaffected until the fault links a major propor-
3.4 Growth mechanisms tion of the main flux.
Once a core fault is initiated irreparable damage does not The most serious damage a core could sustain would be if
occur until the fault grows. Growth is a complex interplay coreplates were shorted at the core bore, at the stator tooth
of electromagnetic, thermal, material and constructional tip or slot bottom for example, and the coreplates were in
factors. There are two principal growth mechanisms for a contact at the core back. This would permit the fault
core fault, as shown in Fig. 6, defined by the circulating current to link the complete flux threading the core yoke on
current in the fault: each cycle. Such surface defects are fairly common but in

IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005 1433
the authors’ experience rarely seem to develop into full core 4.2 Results
faults. Damaging core faults generally occur some way into The results show that in the core of a 500 MW
the core, radially from the slot bottom. turbogenerator carrying a flux of density 1.5 Trms, a fault
of the size shown in Fig. 5a, 4 cm diameter, with each
3.5 Predicted results coreplate in intimate contact with its neighbour, would
The interplay of electromagnetic, thermal and construc- carry the following currents and losses:
tional factors in a core fault cannot be fully described in
mathematical terms, however, a limited analysis has been
Mechanism I, Yoke Flux Diversion, could circulate
presented in the Appendix, to give an indication of the 1004 Arms in the fault. This dissipates 33 kW/mm or
major factors at work. It should be noted that the power 17 kW/kg in the core, less than a quarter of the power per
depends upon the length of the fault, as follows: unit Core Length for a machine of this size, see Table 1.
Eð‘Þ ¼ IZð‘Þ
Mechanism II, Yoke Flux Linking, could circulate
11541 Arms in the fault. This dissipates an enormous
where Zð‘Þ ¼ Rð‘Þ þ jX ð‘Þ
Eð‘Þ ð1Þ
I¼
Zð‘Þ 100 00 000
2
P ¼ I Rð‘Þ 10 00 000

Therefore it is satisfactory to consider only a 1 mm length of

1 00 000
the core fault. In reality there will be effects at the ends of
the core fault cavity where current is concentrated and

power, w/kg
10 000
specific resistances will rise but this is neglected in these
calculations. The results from (7), (10) and (11), for 1 mm 1000
axial length of fault in a core with the details of Table 2,
have been plotted out in Fig. 7 for core faults of varying 100
radius, Rc, and the losses normalised to W/kg by assuming
that the loss is distributed uniformly over the coreplate 10
material inside Rc. This allows the loss density to be
compared to the background core loss density predicted for 1
the coreplate. 0 10 20 30 40 50
radius of fault, Rc, mm

4 Discussion thermal power needed to electrical power in the core

melt steel in fault cavity back circuit & fault cavity, yoke
4.1 General flux linking mechanism
The theory and results show the limits of the effects at work electrical power in the fault background core loss
in a core fault but in real conditions these are modified by cavity, yoke flux diversion
mechanism
practical core conditions. The following discussion sets out
these results with the other factors which can affect a fault, Fig. 7 Variation of power in 1 mm of a core fault due to different
addressing detailed factors, which influence the initiation or driving mechanisms
growth of a core fault, not covered by the introduction and Calculated for a large turbogenerator core with details in Table 2. The
Appendices. dotted lines show changed power input when steel melts

Table 2: Values of parameters used in appendix

Parameter Symbol Value

Rating and Type of Machine S 588 MVA, H2O and H2-cooled Turbogenerator
Core outer radius Ro 139 cm
Slot bottom radius Ri 82 cm
Frequency f 50 Hz
Angular frequency o 314 rad/s
Flux density in the core back yoke, see Table 1 for the magnetic Brms 1.5 T rms
loading, which is the flux density in the air gap
Radius of the core fault Rc Initiating at 1 mm and growing to 20 cm
Resistivity of the coreplate r 51 mS
Permeability of the coreplate mr 4300 as given by the B–H curve until the temperature
exceeds the Curie point of 4001C, at that point it will fall to 1
Radial heat transfer coefficient of coreplate k 60 W/m1C
Density of the coreplate material s 7.86 g/mm3
Skin depth in steel of permeability4300 d 0.94 mm
Skin depth in conducting material of permeability 1 d0 42.71 mm
Temperature of the core background Tb 1001C
Temperature of coreplate at melting Ti 11001C

1434 IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005
 281 kW/mm or 148 kW/kg in the core, more than twice
the power per unit Core Length for this size of machine.
This amounts to more than 28 MW in a 10 cm long fault.

If these current values are compared with those measured or

predicted in [23, 24] it can be concluded that a fault could
circulate, within the core, currents of the same order as
those circulated in the frame by the leakage field, outside the
core, in normal operation.
Three important facts emerge from Fig. 7:

The thermal model shows the power needed to melt
coreplate material, when uniformly distributed over the
ellipsoidal fault and this falls dramatically as the fault
radius increases.

The power input from the Mechanism I increases with
fault radius but is relatively low at fault diameter below
1 cm. The power is insufficient to melt the core until the
fault diameter is more than 1 cm. The power input goes
up significantly when the fault melts.

The power input from the Mechanism II is very large at
small diameters and, although the power reduces as the
fault diameter increases, it is more than sufficient to melt
even the smallest fault.

In any core fault each of these three factors will play a Fig. 8 Plot of eddy currents in the plane of lamination segments at
greater or lesser part. However, it is very unusual for cores the quadrature axis
to fail in the catastrophic way described by the last point, Taken from Mecrow et al. The common site for core faults, at the
junction between segments, near the slot bottom, is highlighted
suggesting that the Mechanism II is not normally effective.
The authors have seen only one incident, exemplified by
Fig. 1, where the catastrophic damage observed and the
rapidity with which the faults grew, suggested that Yoke
Flux Linking was the primary mechanism. A reason for close to the terminals of the affected machine. This lead
Mechanism II being ineffective is likely to be the presence of to considerable investigation of the possibility that
high resistance contacts at the core back, dramatically transient flux conditions could initiate core faults and [15]
reducing the currents flowing in that fault circuit. Indeed showed that pole slipping on a machine does give rise to
any core constructed with keybars insulated from the substantial internal axial flux between coreplates, causing
coreplates or the frame cannot develop Mechanism II higher interlaminar voltages [17], which could lead to
unless that insulation is bridged. However, when the coreplate breakdown. The following is drawn from that
Mechanism II is operative the resultant core fault damage experience.
is large, fast and catastrophic.
4.4.2 Interlaminar voltages and break-
4.3 Thermal factors down: Coreplate insulation breakdown and spot welds
Since the 1920s engineers have known that the end region of between coreplates are observable after most core faults.
the stator core has higher losses and consequentially a They occur between coreplates, which prior to breakdown
higher operating temperature varying with load power were seemingly well-insulated and it is striking that:
factor [3]. The authors have seen a number of core faults

The worst spot welds occur between coreplates with
initiated in the core end region, close to butt joints near to
core slot bottoms where flux densities are highest. This nominally the best insulation.
suggests that thermal factors are an important initiating
Spot welds seem capable of penetrating a number of
mechanism. Losses have been shown to rise close to a butt coreplates at one time.
joint [4, 12] and later work on electrical factors [17] has
It has been shown experimentally on a sample ring core
shown that planar eddy currents increase close to a butt in keybars from a machine, that if current is passed
joint at the slot bottom near a core end, exemplified by the between adjacent keybars, agitation of the keybars causes
calculated laminar eddy current patterns shown in Fig. 8 interlaminar breakdowns even on the opposite side of the
taken from [25]. This is emphasised by the results in Fig. 4b, core.
taken from [12, 28], especially extrapolating the measured
results of Mogi. Reference [29] has suggested that such
elevated losses near butt joints could be due to local The core is behaving like a large parallel-plate capacitor,
buckling in the core and high friction at the site due to intermittently shorted by the keybars. At the site of the
vibration, although the authors have seen no examples of breakdown the insulation between coreplates is punctured
this in their experience on turbogenerators. and local melting takes place. Langmuir reported some of
the features of arcs in hydrogen including the enhanced heat
4.4 Electrical factors transfer produced by atomic disassociation [30]. Unusual
interlaminar breakdowns have occurred in hydrogen,
4.4.1 General: In the UK a number of core faults on described in [19], where the breakdown arc has migrated
large turbogenerators were associated with transient faults some distance over the coreplate surface, driven by the

IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005 1435
 4.5 Material factors
4.5.1 Cooling gasses: Large machines use air or
hydrogen as a cooling gas, which may have a profound
effect upon the growth of a core fault.
When carrying out postfault investigations it is noticeable
that air-cooled machines show a much higher degree
of carbonisation and fire damage than hydrogen-cooled
machines, because of the presence of oxygen. Oxygen
in cooling air oxidises any intermittent contacts made
between coreplates or at the interface between the core
and the keybars, reducing fault currents below predicted
values.
In hydrogen-cooled machines monatomic hydrogen
may have a role in the arcing processes of a core fault
because when monatomic hydrogen reaches a coreplate
surface recombination takes place dumping of energy into
the coreplate. The effect on small interlaminar breakdowns
is that dissociation takes energy out of the arc dissipating
it into neighbouring surfaces without causing further
damage [30]. It is likely that with energetic arcing in
the confined space of an axial vent duct, monatomic
hydrogen could be a vehicle for core fault growth.
Progressive breakdown of interlaminar insulation may not
be necessary to expand a core fault in a hydrogen-cooled
machine, plasma containing monatomic hydrogen in a
vent duct may do the job equally well. In addition a
hydrogen-cooled machine does not oxidise coreplate and
core back contact points so fault currents can reach
predicted values.

4.5.2 Coreplate insulation: The quality of insula-

Fig. 9 Photograph of a coreplate showing an unusual interlaminar tion has been considered important for coreplate integrity
breakdown in a large hydrogen-cooled turbogenerator subjected to a since the 1900s, indeed insulation tests are prescribed in the
core fault, as described in [19] material specifications. However, Beckley has shown [27]
that coreplate spacing is probably more important than
insulation for minimising interlaminar loss. High quality
coreplate insulation can concentrate the action of inter-
laminar voltages at the site of any insulation defect. The
interlaminate fluxes in the machine, a photograph is shown resultant powerful interlaminar arc does more damage than
in Fig. 9. would be possible between coreplates of poorly insulated
Industrial experience when surge testing transformers and material, where interlaminar currents are diffuse and weak.
electric motors has shown that the application of surge However, in the latter case the background losses due to
voltages to windings, to test interturn insulation, induces interlaminar currents are undoubtedly higher than would be
substantial surge voltages between coreplates. These the case with high quality insulation.
surge voltages can be large enough to breakdown
insulation, creating an electrical bond between coreplates. 4.6 Further work
Both transformer and motor manufacturers have reported The discussion has shown that some features of core faults
a rise in core losses after surge tests and this must be the can be predicted. An important objective for turbogenerator
result of these coreplate bonds. One motor manufacturer operators on completion of plant outages is to confirm that
reported that the core losses reduced to the pre-surge the core is sound using Core Ring Flux Tests in the High
test values, once the machine had been operating for some Voltage or Low Voltage (El CiD) form.
time, presumably as a result of thermal and vibration A valuable piece of work, based on this paper, would be to
factors breaking the coreplate bonds formed during the use electromagnetic and thermal finite element techniques to
surge test. predict the magnetic field and temperature at the bore of the
turbogenerator core during the High Voltage and Low
Voltage Tests, with circular defects of various sizes at
4.4.3 Frame currents: Considerable discussion positions as shown in Fig. 8.
has centred on the cause and route of core frame
currents, as it was realised how large these can be on 5 Conclusions
machines with a high electric loading [22–24]. A relationship
A core fault has the approximate shape of a cylindrical
was sought between the core back electrical activity
ellipsoid with its major axis perpendicular to the
observed after a core fault and the huge currents that
coreplate. Under extreme conditions melting of coreplate
are known to circulate in the frame during normal
material occurs within the ellipsoid and a cavity is
operation. To date no such relation has been made except
formed.
that these currents do cause frame heating and may
contribute to lowering the resistance between the coreback
A core fault may be initiated by:
and frame.  Constructional defects in the core
1436 IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005
  Thermal factors axial growth of the fault. This approach is as yet
 Electrical factors unknown in the industry and would involve expensive
materials.
 Operational factors, such as overfluxing, pole slipping
or foreign bodies introduced during operation.  Ensuring that the machine is not operated at an
excessive field or armature current and consequently

A core fault may not grow immediately it is initiated and its core flux density and that it is not subjected to pole
growth is influenced by a complex interplay of electro- slipping.
magnetic, thermal, material and constructional factors,
Early warning of core faults can be given by the use of
which may not be operative at the time of initiation. core monitors, see [33].

A core fault grows axially along its major axis. The radius
of the cavity increases with length but this is not the main 6 Acknowledgments
axis of growth because it is limited by core cooling.

A core fault grows axially until the heating at the The authors are grateful to R. Martin at Durham
circumference of the cavity damages the main winding University for predicting the magnetic field around a core
insulation, causing an earth fault of the machine and fault using the finite element package MEGA and to
tripping, or the core fault extends to the core end. colleagues at Brush Turbine Generator Division in
Loughborough, Ridderkirk and Plze$n for large machine

Three mechanisms have been identified which control design information.
fault growth: Permission to publish Figs. 1a, 5a, and 9 has been given by
Siemens Power Generation (UK), who emphasise that these
 Thermal cooling of the core limiting radial growth. photographs show core damage some 30 years old. Major
 Mechanism I, Yoke Flux Diversion, which circulates faults on large machines are rare nowadays because of the
current in the fault. improvements in design and constructional methods,
 Mechanism II, Yoke Flux Linking, which circulates described in the conclusions, and the widespread application
current between the fault and the core frame. of core monitors. The authors are grateful to Siemens for
their permission to publish these photos.

Core faults tend not to grow unless the initiating defect is
41 cm in diameter.
7 References

Yoke Flux Linking Mechanism II is the most powerful
for growing a core fault but appears to be active in 1 Rudenberg, R.: ‘Zus.etzliche Verluste In Synchronaschinen Uhd Ihre
relatively few faults. It can be limited or eliminated by Messun’, Elektrotech. Zeitschrift., 1924, 59, p. 37
2 Rockwood, G.H.: ‘Calculation of stray losses’, AIEE Trans, 1928, 47,
introducing a high resistance between the stator core and pp. 582–585
its frame. 3 Laffoon, C.M., and Calvert, J.F.: ‘Iron losses in turbine generators’,
AIEE Trans., 1929, 48, pp. 370–374
4 Phear, H.W., and Mallock, R.R.M.: ‘The effect of the joints in the

The currents flowing in a core fault can be as large as the plates of a laminated iron core on the DC ampere-turns required for
current circulated in the core frame by the core leakage its magnetization’, J. IEE., 1936, 79, p. 560
5 Richardson, P.: ‘Stray losses in synchronous electrical machinery’, J.
flux. IEE, 1945, 92, (Pt. 2), pp. 291–304

The tendency of a laminated core to be subject to core 6 Estcourt, V.F., Holley, C.H., Johnson, W.R., and Light, P.H.:
‘Underexcited operation of large turbine generators on the Pacific
faults can be reduced by: Gas and Electric Co.’s system’, AIEE Trans., 1953, 72, (Pt. iii),
pp. 16–22
 Maintaining the highest quality in the construction of 7 Winchester., R.L.: ‘Stray losses in the armature end iron of large
the core by: turbine generators’, AIEE Trans., 1955, 74, (Pt. iii), pp. 381–397
8 Mason, T.H., Aylett, P.D., and Birch, F.H.: ‘Turbogenerator
’ Ensuring correct coreplate insulation quality and performance under exceptional operating conditions’, Proc. IEE,
thickness. 1959, pp. 357–373
9 Oberretl, K.: ‘Streufelder, Wirbelstromverluste, Erwarmungen, Krafte
’ Deburring individual coreplates. und Eisenbrand im Stirnraum von Turbogeneratoren’, Elektrotech.
Maschin., 1963, 80, pp. 539–550
’ Avoiding the introduction of foreign bodies, 10 Walker, J.H., Rogers, G.J., and Jackson, R.L.: ‘Pressing and clamping
particularly metallic, into the core construction. laminated cores’, Proc. IEE, March 1964, 111, (3), pp. 565–577
11 Howe, D., and Hammond, P.: ‘Examination of the axial flux in stator
’ Applying uniform and adequate clamping pressure. cores with particular reference to turbogenerators’, Proc. IEE, 1976,
’ Carrying out recorded visual and electromagnetic/
123, (6), pp. 554–559
12 Stoll, R.L.: ‘The effect of butt joints in large AC machine cores’.
thermal inspection on completion of core build. ICEM, Vienna, Sept. 1976
13 Vickers, V.J.: ‘Recent trends in turbogenerators’, Proc. IEE, 1974, 121,
 Insulating the core frame from the core can eliminate (11R), pp. 1273–1306
the effect of the Yoke Flux Linking Mechanism once a 14 Jacobs, D.A.H., Minors, R.H., Myerscough, C.J., Rollason, M.L.J.,
fault is initiated. and Steel, J.G.: ‘Calculation of losses in the end region of
turbogenerators’, Proc. IEE, 1977, 124, (4), pp. 356–362
 Lowering the magnetic loading below 1 T to reduce the 15 Renew, D.C., and Tavner, P.J.: ‘Axial flux variation in a synchronous
effect of the Yoke Flux Diversion Mechanism, but this generator during pole slipping’, Proc. IEE, 1978, 125, (6), pp. 535–536
16 Tavner, P.J., Hammond, P., and Penman, J.: ‘Contribution to the
has to be consistent with other machine design study of leakage fields at the ends of rotating electrical machines’,
requirements. Proc. IEE, 1978, 125, pp. 339–349
17 Jackson, R.J.: ‘Interlamination voltages in large turbogenerators’,
 Lowering the electrical loading below 100 A/mm will Proc. IEE, 1978, 125, (11), pp. 1232–1238
reduce a risk of core faults, however, if this limit has to 18 Phemister, T.G., Smith, R., Hammond, P., Jackson, R.J., Tavner,
P.J., Penman, J., Stoll, R.L., Lorch, H.O., and Howe, D.:
be exceeded for larger machines, for design reasons, it ‘Contributions to discussion on interlamination voltages in large
will be necessary pay closer attention to coreplate turbine generators and influence of winding design on the axial flux in
insulation quality and cooling intensity to reduce core laminated stator cores and examination of flux distribution in
segmented stator cores’, Proc. IEE, 1980, 127, (Pt C), p. 118
fault risk. 19 Anderson, A.F., and Guile, A.E.: ‘An unusual type of interlaminar
breakdown found in an electrical machine’. 6th Int. Conf. Gas
 Providing radial cooling ducts in the core with Discharges and their Applications, Heriot-Watt University, 8–11
high-resistance/insulating duct spacers to resist the September 1980

IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005 1437
20 Anderson, A.F., Bedford, T., and Craddock, A.F.: ‘Transient leakage Ref. [32] shows that the current density, J, must be
flux in small universal motors’, Proc. IEE, 1981, 128, (Pt B) (5), p. 254 distributed sinusoidally around the cavity to preserve a
21 Platt, R., Kerr, L.C., and Anderson, A.F.: ‘Measuring flux and
interlaminar voltage in turbine generator end regions’. IEE EMDA uniform flux density, B0 oB, across the cavity. The
Conf., London , 13–15 July 1982, Conf. Pubn. 213, pp. 201–205 relationship between the sinusoidally distributed current
22 Tavner, P.J., and Penman, J.: ‘Currents flowing in the stator-core of sheet around the cavity and the circumferential flux density,
large electrical machines’, Proc. IEE, 1983, 130, (Pt C) (6), pp. 273–277
23 Hilliar, R.A., Jackson, R.J., Ward, D.A., and Bennett, R.B.: B, incident on the cavity is:
‘Measurements of the currents flowing in the stator frame of a 500
MW turbogenerator’, Proc. IEE, Jan. 1986, 133,(Pt C) (1), pp. 16–25 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
)ffi
24 Bratoljic, T.: ‘Currents induced in the inner stator frame of large u(
u
m Jrms d t d 4
turbogenerators by the end-zone field’, Trans. IEEE, Dec. 1986, EC-1,
(4), pp. 108–114
Brms ¼ 0 1 þ 02 ð3Þ
2 R2c d
25 Mecrow, B., Jack, A.G., and Cross, S.: ‘Electromagnetic design of
turbogenerator stator end regions’, Proc. IEE, Nov. 1989, 136, (Pt C)
(6), pp. 361–372
26 Sutton, J.: ‘El CID:An easier way to test stator cores’, Electr. Rev.,  
1980, 207, (1) 2 Brms 1
27 Beckley, P.: ‘Electrical steels for rotating machines’ (Peter Peregrinus, ‘ Jrms ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
v( )ffi ð4Þ
2002) d m0 u u 4
28 Mogi, H., Kaido., C., Minematsu, E., Hanzawa, K., Nakahara, A., t 1þ 0 d
Takahashi, K., Ide, K., Kaneda, J., Hattori, K., and Watanabe, T.:
‘Core losses in turbine generators: segment core evaluated by torque R2c d2
method’. Proc. 16th ICEM, Cracow, Poland, April 2004, pp. 13–14
29 Lehoczky, K.N.: ‘Relationship between stator core buckling, core
losses, vibration and local over-temperatures’, IEEE Trans. Eng.
Convers., March 1988, 3, (1), pp. 56–63  
pffiffiffi Brms 1
30 Langmuir, I.: ‘Flames of atomic hydrogen’, Ind. Eng. Chem., 1927, 19, v(ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(6, 667), pp. 141–158
and I1rms ¼ 2 2Rc
m0 u )ffi ð5Þ
u 4
31 Stoll, R.L.: ‘The Analysis of eddy currents’ (Clarendon Press, Oxford,
1974), p. 61 t 1 þ d0
32 Moullin, E.B.: ‘Principles of electromagnetism’ (Oxford University R2c d2
press, 1955), p. 91
33 Tavner, P.J., Ward, D.A., and Gaydon, B.: ‘Monitoring generators
and large motors’, Proc. IEE, 1986, 133, (Pt B), pp. 169–180 The power P1 in the current flow, per unit length of the
cavity can be calculated by integrating around the cavity
8 Appendix theory giving:
8.1 Parameters of core fault cavity 1
The core fault cavity to be investigated in this paper will be P1 ¼ rp2 Jrms
2
Rc d ð6Þ
in a large, 500 MW, two-pole, hydrogen-cooled, turbogen- 2
erator and the parameters used are given in Table 2, they
are identical to those used in [17] a machine between the   
rating of items 8 and 9 in Table 1. In the following sections Rc Brms 2 1
‘ P1 ¼ 2p r 2 ( ) W =m ð7Þ
the electromagnetic and thermal analytical factors are d m0 d4
considered. 1 þ 02
R2c d
8.2 Electromagnetic factors
8.2.1 Mechanism I, yoke flux diversion: Con-  
sider the fault shown in Fig. 6, the intention is to calculate P1 p2 r
‘ R1eff ¼ 2
¼ Ohms=m ð8Þ
the current which could flow axially between coreplates Irms 4d Rc
around that fault under various conditions of fault size and
skin depth of the currents concerned.
Currents will flow axially between connected coreplates at
the surface of the fault driven by the pulsating circumfer- 8.2.2 Mechanism II, yoke flux linking: The
ential component of the core back flux density, B, see fault current is not restricted to the cavity but also flows
Moullin [32]. Assume the fault is circular with a radius, Rc, radially to the core frame and links the main yoke flux of
and that the axial currents flow in the permeable core the machine, as shown in Fig. 6. In which case the emf
material. The degree of connection will depend upon the developed by the greater flux linked by the circuit will
state of the fault, initially it may not be an intimate circulate a much larger fault current than calculated in the
connection and the skin depth may not be fully developed. preceeding Section. This larger current will be superimposed
However, as the fault develops the coreplates will become on the current described above.
intimately connected and eddy currents, limited in a skin- The fault current flows in a circuit with an impedance, Z‘ ,
depth, d, will flow axially between them. As the fault consisting of an inductive reactance, L‘ , due to its self-
progresses the connection improves, the temperature rises, inductance with the main core, in series with a resistance,
local core melting occurs and a cavity forms. In this case the R‘ , for the fault cavity, see Fig. 6. For the purposes of this
coreplate temperature will rise above the Curie point and analysis the resistance of the coreplate radial paths, the
the skin depth, d, will increase to the value for non- contact resistances at the core back and the resistance of the
permeable material of the same conductivity, d0, and the frame members are neglected:
resistance will reduce. Where: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffiffiffiffiffiffi  2 
2r Z‘ ¼ R‘ þ o2 L2‘
d¼ ð1Þ
om0 mr
where the resistance R2 is the axial resistance if the cavity
sffiffiffiffiffiffiffiffiffi
2r r‘
d0 ¼ ð2Þ R‘ ¼
om0 2pRc d

1438 IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005
where the inductance L‘ , is the self inductance of the circuit oðRo  Ri ÞB2rms
shown in Fig. 3 P2 ¼ (   )
  4Rc Ro  Ri 2
Ro  Ri ‘mo mr rd 1 þ
L‘ ¼ d Ro þ Ri
Ro þ Ri p ! 
8pr Rc ðRo  Ri Þ2 Brms 2
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ‘P2 ¼ 2
u( 2     ) mr d3 m0
u r‘ Ro  Ri ‘mo mr 2
‘ Z‘ ¼ t þ o 1
2pRc d Ro þ Ri p (   ) W =m
4Rc ðRo  Ri Þ 2 ð10Þ
1þ
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d ð Ro þ Ri Þ
u(    )
r‘ u t 4Rc Ro  Ri 2
Z‘ ¼ 1þ
2pRc d d Ro þ Ri

Emf in the fault circuit: 8.3 Thermal factors

df d As described in the text, core faults are generally circular in
Erms ¼ N ; f ¼ N ð‘ðRo  Ri ÞBrms cos otÞ; N ¼ 1 the core body and extend axially in the form of an ellipsoid.
dt dt A simple thermal model of the fault can therefore assume
Erms ¼ ‘ðRo  Ri ÞoBrms sin ot circular symmetry. Heat transfer is radial, at the surface of
the fault, Rc, the coreplate steel temperature has reached
melting point temperature, Ti, and at some radial distance,
Erms ‘ðRo  Ri ÞoBrms Rb, from the fault the temperature is close to the mean
I2rms ¼ ¼ vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi temperature, Tb, of the core. Assuming that in the
Z2 u(   2 )
r‘ u t 1 þ 4Rc Ro  Ri cylindrical region of the core fault:
2pRc d d Ro þ Ri

The power input in the fault is much greater than the
2pRc odðRo  Ri ÞBrms background core loss,
I2rms ¼ vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u(   2 )
The power input is uniformly distributed over the fault
u 4R Ro  Ri
rt 1 þ
c region,
d Ro þ Ri
The cooling effect of the core ventilation in the fault area
  is negligible,
Rc ðRo  Ri Þ Brms
I2rms ¼ 4p
The radial heat transfer is that of the coreplate and is
mr d m0
assumed constant at all temperatures.
1
 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u(   2 ) Laplace’s equation in two dimensions for heat transfer
u 4R Ro  Ri
t 1þ c
shows that with a volumetric power input, P, the
d Ro þ Ri surrounding temperature is as follows:
The power P2 per unit length of the cavity will be: ðTi  Tb Þ4k
2 P¼ ð11Þ
P2 ¼ I2rms R2eff ðR2b  R2c Þ
r
where R2eff ¼ Ohm=m The distance, ðRb  Rc Þ, between the molten material, at
2pRc d the core fault radius, and the point of background
2 temperature in the core is of crucial importance to
note R2eff ¼ 3 R1eff ð9Þ
p understanding the growth of the fault.

IEE Proc.-Electr. Power Appl., Vol. 152, No. 6, November 2005 1439

View publication stats