2008 International Conference on Condition Monitoring and Diagnosis, Beijing, China, April 21-24, 2008

Intricate Aspects Of Turbogenerator Endwinding

Vibration Monitoring Based On Data Analysis
Ram Sewak,Rajesh Ranjan and A.K.L. Rao
Heavy Electrical Equipment Plant
Bharat Heavy Electrical Limited, Haridwar , India
Email : ramsewak@bhelhwr.co.in, rranjan@bhelhwr.co.in, , akluck47@yahoo.co.in

the machine at site with respect to endwinding vibration

Abstract--End winding vibrations of large turbogenerators are behaviour.
assuming a greater role in the reliability of the machines to avoid
forced outages. So, on line monitoring during test run as well II. BRIEF TECHNICAL PARAMETERS
during operations have been given utmost importance in recent
times.
Electromagnetic forces, at double the operational frequency a) Rating : 200 / 210 MW, 15.75kV, 2-pole, 50 Hz
are responsible for overhang vibrations besides other mechanical b) No. of Slot : 60
phenomena. These vibrations depend on factors like type of c) Cooling Medium : Hydrogen and water
insulation, slot fixing arrangement, overhang bracing and d) Insulation System : Resin Rich
support system, dynamics of core with overhang and natural
frequency of the endwinding basket. III. TEST PROCEDURE
This paper presents the comprehensive data generated over a A. Vibration Probes:
period of two decades on endwinding vibration on 200/210 MW 1. Piezo-electric Accelerometers:
turbogenerators and dwells on the analysis there on. The data
collected have been studied to examine the various intricate
Mostly, in BHEL piezoelectric accelerometers have been
aspects affecting these vibrations and the trends compared. used for end winding vibration monitoring as has been the
In order to study the requirement of a robust slot-wedge practice with other leading manufacturers [1]. But, lately
system to contain the overhang vibrations and validate the Fibre-Optic Accelerometers (FOA) were also used sparingly.
adequacy of the adopted slot fixing system, the paper presents the For our study, the stand-alone monitoring system was used.
experimentation done on a full-scale turbogenerator Slot Model. Generally, three neutral / near neutral points of generator
Special treatment has been made on the natural frequency test
on overhang system and its significant bearing on the endwinding
stator endwinding bar-joints are chosen on each side of
vibration touching upon some corrective measures. Turbine End and Exciter/Collector End for mounting six
piezo-electric accelerometers in radial and tangential /
Keyword-- endwinding vibration, natural frequency test, circumferential directions. The probe is fastened on to a
factory test-run patented epoxy insulated base-plate (Fig.-1) which in turn, is
glued to bar nose-joint with putty. The probe-cables are taken
I. INTRODUCTION out from inside of the generator casing to measuring
I t is a well established phenomenon that overhang basket of
a 2-pole turbogenerator is subjected to electromagnetic
pulsating forces with double the operational frequency
instrumentation set-up via specially designed Lead-in-Plates
(Fig.-2).
(100/120 Hz). These forces cause relative movement between
adjacent bars at blocking and bracing points. These relative
movements gradually abrade the insulation, leading to
"dusting", erosion of semi-conductive layer, increased partial
discharges, cracking of the insulation leading to a ground fault
and ultimately generator failure. Leading OEM suppliers
worldwide carry out stator endwinding vibration test at site or
during factory test run in the plant. This practice, in BHEL
was initiated first at sites way back in 1984. It was
periodically monitored to know the trend ofvibration levels of Fig.-1: Patented base-plate with probes Fig.-2: Lead-in-plate on casing
overhang basket and also to ascertain the effectiveness of
remedial measures taken on specific generators to contain the 2) Fibre Optic Accelerometers ( FOA) :
vibration levels.
This paper attempts to analyze the comprehensive data thus These probes (Fig.-3) are mounted either on neutral or/ and
generated both at site and plant and highlights the significance also on phase bar-joints (Fig.-4), since they can safely
of factory test-run data for in ascertaining the performance of withstand higher operational voltage upto 27kV. These probes
facilitated to monitor 18 points of bars on end winding (9
This work has been supported by Bharat Heavy Electricals points each on Exciter/ Collector End and Turbine End) during
Limited, India factory test-run of turbogenerators in the plant.
3) Vibration Test:

978-1-4244-1622-6/08/$25.00 ©2007 IEEE

 Traditionally, the endwinding vibration monitoring is done recorded along the length of the slot. It was of the range of 2-
on following regimes of factory test- run of the 5mm (pk-pk) at some points along the slot length. Similar
turbogenerators in the plant and at site. The data is recorded experience has also been reported by others [2][3].
for all the 12-points using portable vibration monitoring
system.
a) Mechanical Run
b) Short Circuit Characteristic (SCC) at rated current
c) Open Circuit Characteristic (OCC) at rated voltage
d) Frequency response test from 47.5 Hz to 51.5 Hz at
rated SCC and OCC conditions
e) Load condition at sites

Fig. 5: Slot model for the generator

Combination of side and top ripple springs, optimization of

wedge length and pressure in fixing the bars in the slot were
Fig.-3: Fibre-optic Accelerometer Fig.-4: FOA mounted on a bar-joint
found to have significant bearing in reducing the bar
vibration in the slots (from 2 -5mm in free condition to less
Of late, some Utilities/customers are demanding and than 10μm) as well as endwinding vibration levels on such
installing on-line endwinding vibration monitoring system in type of generators during operation at sites
the overall scheme of condition monitoring of the TGs.
Generally, these systems are equipped with several data V. OBSERVATIONS
logging , display and analysis features and alarm facilities.
Generally, 12-point based endwinding vibration monitoring
IV. SIMULATION STUDIES ON SLOT MODEL is carried out. The maximum displacement value (pk-pk Pm)
among 12-points, at double the operational frequency i.e.
Endwinding vibration levels, for a given design of a 100Hz was taken for analysis in each regime - rated SCC,
generator largely depend on bar fixing system in the slots, OCC & Loads. Owing to various constraints of
overhang support structure, natural frequency of overhang facilities/availability of machine, only limited vibration data
structure and above all, overall workmanship / human skill have been generated on load conditions.
during manufacturing.
In order to comprehend the bar bouncing forces in the slot VI. ANALYSIS
region and to study the slot fixing arrangement , a full length,
7-slot SLOT MODEL (Fig.-5) with three slot full winding Comprehensive data generated over 20 years have been
arrangement and built in monitoring features, for Type –B analysed under following regimes:
generator was designed , manufactured in BHEL and
following tests were conducted on this model: 1. Three-Phase Sustained Short Circuit Characteristic
1. Vibration monitoring in the slot portion along the (SCC) and Open Circuit Characteristic (OCC)
length under various short-circuit current regimes. 2. Load Condition
2. Generation of operation deflection shape (ODS) of
A. Three-Phase Sustained Short Circuit Characteristic
bars in the slot under short-circuit current regimes
(SCC) and Open Circuit Characteristic (OCC)
3. Monitoring of the associated parameters like electro-
a) Stator end windings are subjected to three different
magnetic flux, wedge pressure, stress/strain on
vibratory forces [4].
tension bolts, temperature of the windings
i) Pulsating forces during start-up, shutdown and
It is a well known fact that as bar bouncing forces in the
normal operation due to forced vibration -
slot region is proportional to current square, they pulsate
sinusoidally between zero and a maximum at twice the x of the entire end-winding system, induced by
bearing and stator frame vibrations (excitation
fundamental frequency. Its direction is always towards the
frequency depending on the rotor speeds)
bottom of slot when the current in bottom and top bars is in
phase [2][3]. x of the entire end-winding system induced by core
In order to gauge the effect of bar bouncing forces, winding vibrations (double frequency excitation)
bars, in one of the slots of the model were freely placed and x of each single bar, induced by electromagnetic
excited electrically under momentary short circuit condition, forces (double-frequency excitation)
using a high current transformer. The vibration levels were
 ii) Forces resulting from restrained thermal In the tangential direction, the unrestrained bar-head
expansion which depends upon generator load joint is most compliant in bending. Every bar in the
and coolant temperature. endwinding is acted upon by forces at double the
iii) Electromagnetic forces of double the system operational frequency which vary in magnitude, direction
frequency resulting from a high D.C. component and phase angle along with each bar arm and from bar to
arising from abnormal operating conditions bar. In order to counter these forces and to reduce the
especially during short circuit conditions and vibration levels, sufficient strength is imparted making a
during synchronization. ring of insulation putty at bar-head lug-joint. This has
effected on overall tangential vibrations of overhang as
Resulting endwinding vibration, largely, is a combination revealed by data obtained (Table-1).
of electromagnetic vibrations of windings (self excited
vibrations) and core induced magnetic vibrations which are c) In the recent past, endwinding vibration monitoring
independent to stator current. In most of the generators, core system based on fibre-optic accelerometer was introduced.
induced magnetic vibrations at rated voltage OCC regime are Its unique feature allows it to be installed safely on live
found to be dominant (Fig.-6) demonstrating a strong phase bar-joints for rated voltage upto 27 kV with one
minute proof voltage upto 56kV.
Since then, using this system, phase bar vibrations were
200 monitored in addition to neutral ones. The limited data
reveals comparative results of vibration levels on phase bar
DISP. (PK-PK MICON)

150 joints as well on neutral ones. This may be due to

100 comparative current levels in phase and neutral bar joints.

50
B. Load Condition:
0
i) Continuous efforts have been made to develop empirical
1 3 5 7 9 11 13 15 17 19 21 23 25
M ACHINE NO.
formulae to predict endwinding vibration levels at load
AT 1 0 0 % In S C C AT 1 0 0 % Vn OC C using factory test-run data [4]. The relationship was given
as,
Fig.-6: Vibration behaviour of generators during factory-run test
Gwmax  2Gwsc +G woc
coupling between core and endwinding. This was reported where,
by others as well [1][4]. Gwmax = Maximum endwinding vibration level at any load
Gwsc = Endwinding vibration level at rated current
The resin rich insulation system being thermosetting in G woc= Endwinding vibration level at rated voltage
nature offer very little or no damping once cured, to winding
system subjected to vibratory forces. Consequently, In another Reference [8], authors duo have proposed
relatively high endwinding vibration were encountered on multivariant calculations of vibration in an individual end
such generators. On the other hand, old bituminous arc considering, the overhang bar structure as a
thermoplastic insulation system, due to its inherent nature, homogeneous orthotropic curvilinear bar having elastically
offers greater damping ability and hence lower vibrations. damped attachment fittings. The vibrations are described
This was corroborated by the data gathered on one of the of by set of differential equations and shown that results
equivalent rating turbogenerators with bituminous insulation obtained are in “good agreement” with the actual data. It
where vibration levels recorded were nearly 1/3rd to that of was deduced that in “loaded mode the vibration
the generators with the thermosetting insulation system. displacements are roughly 1.5 times those in the three-
phase (sustained) short-circuit mode”.
b) The electromagnetic forces on overhang system are Notwithstanding above, in our experience, both these
reported to be higher in tangential / circumferential direction formulations mentioned above do not validate the practical
rather than radial direction [2][3] . However, number of data recorded at loads (Fig. -7). Indeed, the levels obtained,
generators have exhibited maximum endwinding vibrations either remained more or less similar to factory test-run
in radial direction only (Table-1) even though the levels in the levels or had gone down. The data were recorded at a
tangential direction are of comparable magnitude to those in particular time by a portable vibration monitoring system
radial direction. as well as by a ‘continuous on-line monitoring system’
available with the utility, and mostly, after 2 to 3 years of
Table-1: Direction-wise comparison of endwinding vibrations commissioning / modernization.
No. of TGs. 35
% of TGs recording maximum vibration level in In our experience, the above formulas may be exception
80 % rather than general rule when analysing the practical data.
Radial Direction in any regime
% of TGs recording maximum vibration level in The complex nature of both the forces and overhang
20 %
tangential direction in any regime
structure coupled with their boundary conditions, make the
 prediction of endwinding vibrations quite difficult. In our frequency of overhang system, degree of tightness, thermal
view, the vibration levels at full load remain more or less expansion of winding etc.
in the similar range as obtained during factory test-run
30
under normal operational conditions.

Disp.(pk-pk) micron
DISP.(PK-PK MICRON)

150 20

100
10

50
0
60 80 100
0
% Of Rated Cureent
1 2 3 4 5 6 7
M ACHINE NO.
Fig : 9 : Vibration trend during SCC regime
100%In SCC 100%In OCC FULL LOAD

Fig.-7 : Vibration behaviour of generators at load

Therefore, in view of aforesaid, factory test- run levels 150

Disp. (pk-pk) micron

can be considered as a reliable and “good indicator” for the
performance of machine at site with respect to endwinding
vibration. 100

150
DISP. (PK-PK MICRON)

100 50
50 100 150 200
50 Load in MW

0 Fig.- 10 : Normal Vibration trend during load

50 100 150 200
LOAD (MW)

Fig.-8 : Load vs vibration behaviour at certain times

VII. NATURAL FREQUENCY TEST (NFT)

ii) Sometimes, very peculiar endwinding vibration
behaviour had been encountered, where the levels were in As mentioned earlier, resonance phenomenon, locally or
reverse order globally, on overhang due to natural frequency matching with
with loads i.e. higher at lower loads and lower at higher double the operating frequency (100 Hz/120 Hz) of the
loads (Fig.-8). overhang basket, is generally recognised as to be wholly
This phenomenon can be related to thermal effect [6][11]. responsible for excessive endwinding vibrations and ultimate
As the load is increased, corresponding winding temperature failure of windings. Hence, natural frequency test (NFT) /
also increases. The increased temperature lowers the modulus bump
of elasticity of binder [10]. This in turn, lowers natural test/ impact test is normally carried out by most of the
frequency of endwinding and consequently vibration level leading power equipment manufacturers world wide.
goes down and gets stabilized at full load. Typical NFT / bump test on a generator overhang is shown
Also, tight windings are less affected by temperature in Fig.-11. Each bar joint is tested both in radial as well as
change since these are already well braced. On the other hand, tangential direction by stimulating the structure locally and
the comparatively loose windings due to winding expansion measure the corresponding response in that direction. This
reduces clearances between endwinding and bracing system process is called as local mode testing. Recently, global mode
thus lowering the vibration levels [6]. testing has also been initiated. In this test, no of response
iii) In practice, generally ‘square relationship’ between points are measured simultaneously around the periphery of
vibration amplitude and electrical parameters – current/voltage, overhang by stimulating at one location only. Normally, radial
has not been revealed by the data, both at factory test-runs direction response is sufficient [10].
(SCC/OCC regimes) and on loads at site as given in Fig.-9 and Natural frequency (NF) of overhang structure depends
10. largely on factors like stiffness/ rigidity of overhang support
It is quite evident then; the vibration levels are following system , temperature, no of bracing points from slot exit to bar
the electrical parameter nearly linearly and not squarely. This head-joint, clearances between bar seating and support etc.
has been reported by others as well [4]. and may vary from plant conditions to site conditions.
The deviation in theoretical relationship between
According to one study, natural frequency found to be
electromagnetic forces and corresponding response can be
shifted about 13 Hz downward from testing stage in the plant
attributed to factors like temperature effect on natural
to commissioning stage at site. There upon, at loads, it varied
cyclically in the range of around 15 Hz during operation and the performance of machine at full load under normal
stop. In view of this, it becomes then, necessary to keep off the operational conditions. The vibration levels at full
natural frequency beyond double the operational frequency. ii. load condition are likely to remain in the same range
Accordingly natural frequency domain from 94 Hz to 115 as at factory test-run.
Hz should be cleared off for a 50 Hz generator of specific iii. The factory test-run data on SCC/ OCC, however, do
rating [9]. not seem to be amenable to exact empirical
formulations and basing on these, exact or nearly
exact prediction of vibration behaviour on full load
may not be feasible as reported by others.
iv. The vibration levels in radial and tangential
directions are found to be of comparable amplitude
irrespective of type and rating of a generator. Also,
the vibration levels on phase bar-joints to that of
neutral ones are not at much variance.
v. Generally, in practice, vibration data not indicative of
a ‘square relationship’ between endwinding vibration
Fig.-11.- : Typical natural frequency test on overhang amplitude and electrical parameters.
vi. Inverse trend observed between load and vibration
However, the overhang structure being composite mechno-
amplitude, at certain times, may be due to thermal
electric in nature, adhering to natural frequency domain limit
effect on natural frequency, insulation and bracing
from double the operational frequency, may not be feasible, on
system of overhang structure.
certain occasions, due to various practical constraints. Rather,
vii. Natural frequency test / Bump test is necessary for
it should be determined through trials in operation for specific
ascertaining the dynamics of overhang basket. A
situation [10].
“sufficient gap” between the natural frequency of
overhang basket and double the operational
A. Problem of Dynamics of Overhang and Remedial Measures:
frequency is required in order to avoid resonance
phenomenon.
In order to shift the resonant natural frequency of a structure,
following widely known classical methods are employed:
IX.ACKNOWLEDGEMENT
x Change of mass
x Change of stiffness
Authors express their sincere thanks on record to the
x Change of damping
BHEL management for granting permission to publish this
paper and for all round support and assistance for the study.
First one requires design modification of overhang
X.REFERENCES
structure which may not be feasible on an operating machine.
Change of stiffness and to some extent damping, is generally [1]. S. Nagano, M.Tari, et al “ Early Detection Of Excessive Coil Vibration
resorted to as a remedial measures as shown in Fig.-12 , In High Voltage Rotating Electric Machines” IEEE, Vol. PAS-101, No.6
where gap between unsupported bus-bars have been tightly June, 1982
[2]. C. A. Bucci, A.D. Coggeshall, et al “ Control Of Electromagnetic Forces
bridged by appropriate insulation spacer and then securing On Stator Windings Of Large Turbine-generators.” 71-TP51-PWR
with neoprene sleeve. IEEE Winter Power Meeting, New York, Jan.31- Feb. 5 1971.
[3]. Charles H. Holley, David M. Willyoung “ Stator Winding Systems With
Reduced Vibratory Forces For Large Turbine – Generators” IEEE , Vol.
PAS –89, No.8 , Nov-Dec. 1969.
[4]. M.R. Patel & J.M. Butler "End-winding vibrations in large synchronous
generators." IEEE Vol PAS-102 No.5, 5,May 1983
[5]. D. Lambrecht, H. Berger “Integrated End-winding Ring Support For
Water-Cooled Stator Winding” IEEE Vol PAS-102 No.4, April, 1983
[6]. Jackson Lin “ Applying Stator End-winding Vibration Monitoring
Technology At J.H. Campbell Generation Plant” Iris Rotating Machine
Conference, Taxas , June, 2002
[7]. P. Richardson, R. Hawley , “ Generator Stator Vibrations” IEEE PES
Winter Meeting, New York Jan. 1973, page no.1263
[8]. V.A. Pikul’skii and V.A. Tsvetkov, “Analysis Of Vibration State In End
Sections Of Turbogenerator Stator Windings” Elecktrotekhnika, Vol. 47,
No.7, 1976
Fig.-12 : Remedial Measure between bus-bars [9]. Bai Yamin “Analysis Of Typical Fault On Stator Endwindings Of
Tirbogenerators And Available Counter Measures” Iris Rotation
VIII. CONCLUSIONS Machine Conference,2007 San Antonio , Tx

i. Endwinding vibration levels obtained during factory

test run serve as a reliable and “good indicator” for
[10]. Dr. Lutz Intichar “Eigenfrequenz- und Eigenformbestimmung an
Generator-Wickelköpfen”. Symposium Schwingungsdiagnose an
Kraftwerksturbosätzen, 2006
[11]. Dipl.-Ing. Detlev Frerichs, “Überwachung der
Wickelkopfschwingungen von Generatoren mit faseroptischen
Beschleunigungsaufnehmern Schwingungsdiagnose an
Kraftwerksturbosätzen, 2006
[12]. In-house reports generated time to time.