Professional Documents
Culture Documents
22
March/April 1999
Low, long-time
disturbance torques
Excitation mechanisms are:
• Negative-sequence current, caused by
unbalanced loading of the three current
phases in the power system, causing an al-
ternating air-gap torque to act on the gen-
erator with double the grid frequency. The
allowed negative-sequence current is lim-
ited by international standards (3). Tor-
sional vibration caused by the allowed
negative-sequence current must be low if
train designs are to be free of resonance.
• Sub-synchronous resonance, which can Figure 3. Torsional model of a low-pressure steam turbine rotor.
occur if the generator is connected to long
transmission lines (for example, 1000 km
long) and may excite the lower torsional bine shaft train. Special know-how is blade row is separately modelled by means
modes (4). needed, for example, to model the active of a beam model with about 10 elements.
The dynamic air-gap excitation torque acts part of the generator with its copper wind- The beam model is linked to the torsional
uniformly over the length of the active part ing and wedges. model of the shaft.
of the generator. The effective excitation The number of degrees of freedom of the
of a mode shape depends on the mode coupled model is increased by the addi-
shape in the region of the active part of
Calculating natural tional degrees of freedom of the blade rows
the generator (Figure 2). torsional frequencies and considered.
mode shapes
Calculating torsional The natural torsional frequencies and
Case in point
vibrations mode shapes are calculated using the tor- The coupled shaft-blade model of a steam
sional model (Figure 3) described earlier. turbine shaft line with four
The mass moment of inertia and torsional Damping is not included because of the low
stiffness of the turbine generator train are modelled last-stage blade rows shows two
damping of the train. The free-free vibra-
described by means of approximately 200 additional pairs of coupled first blade
tion system has a rigid body mode at 0 Hz.
one-dimensional finite elements (FE) with modes. For one pair of these first blade
For each mode shape of higher natural fre-
linear shape functions (Figure 3). These FE modes the blade motion is in phase (M1
quency, the number of vibration nodes is
elements have constant properties over and M2) with the shaft torsional vibration;
increased. Typical groups of mode shapes,
their length (5). for the other pair it is in anti-phase (M3
for example, for a 700-MW steam turbine
and M4).
The mass moment of inertia and torsional generator, are shown in Figure 4 and in
stiffness are described for each element Table 1. Figure 6 shows the calculated shaft mode
separately in terms of the outer and inner shapes in a comparison with the uncou-
diameters for mass and stiffness. Young’s pled shaft train model and the coupled
modulus is applied as a function of the ac- Vibration’s influence shaft-blade model with the last two blade
tual element temperature. The turbine ro- rows of the low-pressure turbines.
The torsional vibrations in the turbine gen-
tor blades are modelled as an additional The coupling of the vibrations in the long
erator train are coupled to the vibration
mass moment of inertia. turbine blades with the torsional vibration
occurring in the long turbine rotor blades.
The accuracy of the results depends largely of the shaft train can be neglected due to
To investigate this phenomena, a coupled
on the quality of the modelling of the tur- the LP turbines having stiff, drum-type ro-
shaft-blade model is used in which each
tors of welded ABB design. Separate vibra-
tion designs for the rotor trains and long
turbine blades provide the required results.
23
Machine, Plant & Systems Monitor
24
March/April 1999
on measured and calculated natural tor- stant in the upper speed range. This is due
sional frequencies to the ‘stiffening’ of the rotor windings and
ABB currently uses either sensitive four- wedges caused by the centrifugal force act-
arm strain gauges or accelerometers and ing on them as the speed increases. Agree-
telemetric signal transmission. Experience ment between the calculated natural tor-
with these methods has been very good. sional frequencies and the measured val-
ues is good, with differences of only about
3% being obtained.
Torsional measurement in In the following, a look is taken at meas-
urements carried out on turbine generator
the spin pit trains in thermal power plants under three
The torsional vibrations of a hydrogen- different kinds of excitation condition.
cooled 500 MVA turbo-generator in the
Figure 10. Calculated torsional mode
spin pit were measured using accelerometers
for the purpose of verifying the calculated
shapes (1-3) of a 500 MVA turbo-generator Torsional measurement
rotor (M = measurement plane).
natural torsional frequencies (Figures 8 & test with synchronisation
9). The shaft configuration in the spin pit
comprised an electrical drive, two-stage
ferent speeds and at short time intervals, failure
were transformed into amplitude spectra
gearbox, cardan shaft and the generator rotor.
for display in a Campbell diagram (Figure The unit was synchronised with the elec-
11). These spectra are collected as a func- trical grid with a small phase displacement
tion of the speed and are sorted in ascend- angle of about 5˚. This small electrical dis-
turbance caused the shaft train to be ex-
cited with a torsional impulse. The re-
sponding torsional vibrations fade away
after a fraction of a second (< 0.3 s). Only
the lower natural torsional frequencies re-
spond with detectable peaks in the ampli-
tude spectrum.
The machine tested using this procedure
was a 900-MW half-speed turbine genera-
tor train. The instrumentation was fixed
onto the shaft journal in the workshop.
Figure 8. A 500 MVA turo-generator Special attention was given to the fitting
equipped with two accelerometers at the Figure 11. Campbell diagram of measured of the rotating measuring devices, such as
hub of the axial fan. natural torsional frequencies for a 500 the strain gauges, wires and transmitters,
MVA turbo-generator rotor (a = vibration since they have to withstand high centrifu-
amplitudes; f = frequency; n = rotational gal forces (Figure 12).
speed). The measured amplitude spectra clearly
show the natural torsional frequencies up
ing order in the diagram. Campbell dia- to the single grid frequency. No significant
grams, in which the rotor speed is shown resonance peaks are visible in the upper fre-
along the abscissa and the frequency of quency range (Figure 13).
vibration as the ordinate, are commonly
used in analyses of blade dynamics. A small
vertical bar gives the amplitude of vibra- Ramp test
tion for a certain speed and vibration fre- The unit was disconnected from the elec-
quency. Only amplitudes of high magni- trical grid and a solid line-to-ground fault
tude are plotted, those of very low set up on the HV side of the generator (Fig-
Figure 9. A transmitter casing with magnitudes being neglected. ure 14). Afterwards, the generator rotor
integrated accelerometers. Three natural torsional frequencies of the was excited with a low excitation current,
generator rotor are found in the range up and the shaft train was run up slowly with
Figure 10 shows the calculated torsional to 250 Hz. The torsional frequencies in- steam. The dynamic torque produced in
mode shapes and the measurement plane. crease slightly with increasing rotor speed the generator air gap has a frequency equal
During the start-up with a low rotational in the lower speed range and remain con- to the rotating speed times the pole number
speed gradient, torsional vibrations were
excited by means of a mesh and pitch er-
ror in the gearbox.
Two accelerometers were mounted on the
fan hub at 0˚ and 180˚ of the circumfer-
ence. The measurement signals from these
accelerometers allowed compensation of
the lateral shaft vibration components. The
signals were transmitted by means of te-
lemetry to a data acquisition system, where
they were stored.
Figure 12. Four-arm strain gauge
The signals, which were measured at dif- bridge on a shaft journal.
25
Machine, Plant & Systems Monitor
Figure 13. FFT analysis of short-time steps. The test was conducted on a 900 MW half-
speed turbine generator train (a = vibration amplitudes; f = frequency; fG = grid frequency;
t = time; M1 & M2 = measurement planes; see Figure 1 for key to other notation).
of the machine. In this example, the speed is important to carefully monitor the tem-
of the shaft train (with a four-pole genera- perature of the generator windings during
tor) was increased and resonances of the this test.
natural torsional frequencies, correspond-
ing to four times the rotational speed, were
passed through. Torsional test with
The measured signals were transferred into random excitation from
the frequency domain by FFT and plotted
in the Campbell diagram (Figure 15). The the electrical grid
Figure 15. Campbell diagram for a 900
sharp resonance peaks at the natural tor- MW half-speed steam turbine generator (a
During normal operation, minor distur-
sional frequencies are clearly visible on the = vibration amplitudes; f = frequency; fR =
bances of the type commonly found in elec-
four times speed-line. The natural torsional natural frequency of train; fB = blade
trical grids produce continual, random
frequencies were measured at different frequencies; n = rotational speed; 1-4:
excitation of the natural torsional frequen-
speeds with different centrifugal force ef- multiples of speed).
cies of the turbine generator train. The
fects. Extrapolations were used to convert
measured signals are sampled over a pe-
the measured natural torsional frequencies
riod of 30 minutes or more. By averaging
to the rated speed.
some hundreds of FFTs, a final vibration Comparison of different
Curve veering is visible in the zoomed win- spectrum is obtained which contains all the test methods
dow in the Campbell diagram (Figure 15). natural torsional frequencies in the fre-
Natural blade frequencies, highly depend- quency range of interest. Table 2 compares different aspects of the
ent on the speed, cross a natural torsional various torsional measurement procedures
This test procedure was demonstrated by
frequency of the shaft. As has been ex- used in power plants. The method employ-
performing torsional measurements on a
plained, the first natural blade frequency ing random excitation from the electrical
600-MW half-speed turbine generator
is split into two pairs, which are clearly grid has been found to be highly efficient
train. The rotating part of the measuring
visible. and to yield good measurement results. For
equipment consists of strain gauges and te-
This test procedure allows excitation of the lemetry devices on one measurement plane. this reason, it is the method preferred by
specified natural torsional frequencies. It A location was selected on the shaft train ABB and the one currently used during the
design of turbine generator trains and for
experimental verifications.
Summary
By designing steam turbine trains with stiff,
ABB drum-type, welded LP rotors it is en-
sured that coupling of the blade vibrations
in large steam turbines to the torsional vi-
bration in the shaft train is negligible. Sepa-
rate vibration designs for the shaft trains
and long blades therefore provide adequate
Figure 14. Solid line-to-ground fault on the high voltage side of the generator (G = results. The natural torsional frequencies
generator stator, R,S & T = high voltage lines).
26
March/April 1999
27
Customer’s Purchase
Order Order Number
Form
PO Box 72, Chipping Norton, Oxford, OX7 6JU, UK
Tel: +44(0)1451 830261 Fax: +44(0)1451 870661
E-mail: mail@coxmoor.com Web: www.coxmoor.com
Title* Publication Price*
ISSN or ISBN Date Quantity
Position:
Address:
Total Payment:
Note: Price is inclusive of postage and handling for all
orders. Invoices for orders without prepayment will
additionally be charged for postage and handling costs.
Signature: