Dyrobes Rotordynamics Software

http://dyrobes.com

7th IFToMM-Conference on Rotor Dynamics, Vienna, Austria, 25-28 September 2006

Unbalance Response and Field Balancing of an 1150-MW

Turbine-Generator with Generator Bow

Edgar J. Gunter
RODYN Vibration Analysis, Inc.
1932 Arlington Boulevard, Suite 223
Charlottesville, VA 22903-1560
Fax: (434) 971-2695
DrGunter@aol.com

ABSTRACT

This paper presents some of the theoretical and experimental unbalance response characteristics of an 1150-MW
turbinegenerator. These large 11-bearing turbine-generators can operate through as many as 10 distinct critical speeds. The
balance condition of these units may change over time due to various influences such as turbine blade erosion, alignment and,
in particular, generator bowing. Over a period of time, the generator may developed a bowed condition due to the nature of
the electrical conductors. It is difficult to directly field balance a generator which is encased in a hydrogen environment.
Removal of the generator for high-speed balancing in a balance facility is exceptionally expensive. This study examines the
control of the turbine-generator response due to generator bow and turbine unbalance by various combinations of balance
weights placed along the turbines. Some experimental responses of the turbine-generator due to various trial weights along
the system are also presented. The 11-bearing turbine-generator, including foundation effects, was analytically modeled using
transfer matrix and finite element methods. It was determined that in order to correctly model a large turbinegenerator, the
foundation effects must be included. Foundation flexibility may cause as much as an 80-90% reduction in effective bearing
damping. Various unbalance response simulations were performed with this model with combinations of turbine unbalance
and generator bow. It is shown that system response due to turbine unbalance and generator bow at running speed may be
controlled by proper application of unbalance weights and couples placed along the couplings and the low pressure turbines.

KEY WORDS

Turbine-generator response, Multi-plane Balancing, Three-dimensional complex modes

1 INTRODUCTION

The 11-bearing 1150 turbine-generator systems represent some of the largest units in operation. One of the maintenance
problems of concern is that of balancing these units in the field. Over a period of time, the generator may develop a bowed
condition due to the nature of the electrical conductors. It is difficult to field balance a generator which is enclosed in a
hydrogen environment.
The site facility has two identical units. Vibration measurements were recorded at the bearings of both units. Trial
balance weights were placed along both rotors to obtain influence coefficients. Although the units are identical, the
experimentally obtained influence coefficients were not. The bearings have jacking oil for liftoff and shutdown. After full
speed is achieved, the jacking oil is turned off and the bearings become fully hydrodynamic. The bearing hydrodynamic
pressure may be measured. Examination of the operating pressures developed in each bearing showed that several of the
interior bearings could be underloaded due to pedestal sag. The vertical pedestal alignment thus strongly influences the
system dynamics and hence the balancing. The vertical alignment influences the bearing loading and hence the bearing
stiffness and damping coefficients. The bearing pedestal flexibility reduces the overall stiffness and may also cause a
dramatic reduction in effective damping by as much as 90%.

1 Paper-ID 331
 Figure 1a: LP Turbine in Crane Figure 1b: Closeup of HP Balance Turbine Ring

Figure 1 represents images of the LP turbine in the crane and a close up of the HP rotor. In Figure 1a, note the tiny figure
of a man in the lower left corner. T hese turbines represent some of the largest utility rotors in service. In Figure 1b, the
technicians are in the process of installing a balancing ring to the turbine assembly. Also, in this figure, note the coupling
at the left side of the rotor. W hen these couplings are bolted up, they form a rigid connection. Field balancing corrections
may also be applied at each coupling location.

2 TURBINE-GENERATOR M ODELING
2.1 Transfer M atrix M ethod
The first method of modeling the complete turbine-generator system was by means of the transfer matrix approach. A
modified transfer matrix method was employed with scaling of the matrices to improve computational accuracy. Figures
2a&b represent models of a HP turbine and generator using a transfer matrix program.

Figure 2b - Generator Showing Exciter Mode

Figure 2a - Model of HP Turbine Showing 1 st Mode

Figure 2a shows the HP turbine, which is the first turbine of the drive system. The HP turbine 1 st critical is relatively
unaffected by the coupling to the first LP turbine. The individual LP turbine critical speeds are meaningless as they are now
part of a system and cannot be computed as individual components.
Figure 2b represents the generator with the attached exciter. In general, the generator is the most flexible component of
the system and hence has the lowest critical speed. The attachment of the exciter and the LP turbine have little effect on the
generator first critical speed. The mode shown in Figure 2b is a local exciter mode and is unaffected by the LP turbines. The
attachment of the LP3 coupling has no influence on the local exciter mode.
The attachment of a very flexible exciter to a massive generator can lead to a very serious situation. It has been
demonstrated through numerical time transient analysis that a blade loss on LP3 can cause exciter damage leading to
catastrophic system failure. The presence of the very flexible exciter attached to the generator also leads to numerical
problems in convergence when computed by the transfer matrix method.
W hen the entire system of the four turbines and generator was assembled to form the complete model, there were
encountered a number of computational difficulties for the determination of all the system modes. The first problem is that
of missed or skipped modes. This is caused by the size of the search increment used in the analysis. The second difficulty
was that of convergence of the higher frequencies due to the extreme flexibility of the exciter.
The most serious difficulty with the transfer matrix program is the inability to properly handle foundation or pedestal
effects with included mass and stiffness. This is due to the generation of local or branched resonances which must be removed
from the search range. It was later determined that the pedestal effects are of major importance in the determination of the
effective stiffness and damping at a particular bearing location. The transfer matrix method also is not suited for time
transient blade loss analysis or the computation of damped eigenvalues including foundation effects. Hence, the transfer
matrix method is unsuited for the analysis of multi-bearing turbine-generator systems.

2 Paper-ID 331
2.2 Finite Element Analysis

A PC-based finite element program was utilized for the computations of the static deflections and undamped system
modes.

Figure 3: Finite Element Beam Model of 1150-MW Turbine-Generator

A finite element beam model of the turbine-generator system was developed using a standard structural FEM program. In
this model, several hundred beam elements were employed. The bearings are represented by simple springs and no disc
gyroscopic effects are included in the model. This model was useful in the determination of the static shaft displacements and
loads with various assumed alignments and also for the computation of the system modes. Modes are not skipped when
computed by a finite element method. However, all the commercial finite element structural codes are unsuited for the

Figure 4: Finite Element Model of 1150- MW Turbine-Generator

complete dynamical analysis of turbine-generators. These programs cannot incorporate generalized bearing stiffness and
damping coefficients, disk gyroscopics, or compute complex eigenvalues, synchronous unbalance response with shaft bow
or blade loss simulation.
The system was next modeled with a finite element rotor dynamics program that could also include the bearing, as well
as foundation effects. A model of the system is shown in Figure 4. Each bearing is represented by 8 bearing stiffness and
damping coefficients. These coefficients are determine from a fluid film finite element bearing code once the values of
bearing loading are estimated. Of greater difficulty is the determination of the pedestal properties of effective mass and
stiffness. Impact testing may be used to determine properties.
Figure 4 represents the turbine-generator model with 110 nodes. Since each node can have two displacements and two
rotations, the total model is represented by a system of 440 dof. Some rotor dynamic codes use a polynomial representation
for the system and attempt to extract the complex roots from the characteristic polynomial. This model far exceeds the
capabilities of such a technique. Transfer matrix methods also are not capable of analyzing such a system as shown in Figure
4. It should be important to note that the turbine-generator may be very accurately modeled with the finite element method.
In Figure 4, note that the bearings are connected to pedestal masses. The pedestals have a property of mass, stiffness, and
damping. These quantities of the foundation or pedestal effects are difficult to identify initially.
In addition to the pedestal characteristics, the vertical alignment of each pedestal determines the amount of loading on each
bearing. The bearing loading in turn, determines the bearing characteristics. Thus, identical bearings do not have identical
bearing coefficients. Also, identical units do not have identical vibration responses or influence coefficients. The experiential
data obtained from trial unbalance runs can be used to further refine the rotor dynamics model.

3 Paper-ID 331
3 BEARING LOADING AND FLUID FILM BEARING PROPERTIES

Figure 5 represents the pressure profile and bearing operating characteristics at 1,800 RPM . For the average bearing
loading as applied in Figure 5, the bearing is operating at a very high eccentricity ratio of 0.8 at running speed. Because of
these high operating eccentricity ratios, jacking oil is required on liftoff. At a suitable speed, the jacking oil is turned off.
At the oil feed location at the bottom of the bearing, pressure gauges monitor the oil pressure. At running speed, this oil film
pressure is generated by the hydrodynamic bearing action. It was observed that not all of the turbine bearings are generating
equal pressures. Table 1 shows the bearing pressures generated in the two units.

Figure 5: Turbine Bearing Characteristics, 1800 RPM Figure 6: Relationship Between Measured and
Maximum Oil Film Pressure

Figure 6 shows the relationship between bottom oil pressure, maximum pressure and bearing loading. This relationship
is relatively insensitive to assumed viscosity. By computing the load-pressure characteristics for each bearing, the actual
applied loading at each bearing may be estimated. Table 1 shows that the LP3 bearings of unit 2 are underloaded.
Table 1: Bearing Static Loading Versus Orifice Film Pressure M easured
N = 1,800 RPM , : = 10 centiPoise

Bearing Brg Unit 2 Cb Bearing Unit 3 Cb Brg

Locations No. kPa mm Load, N kPa mm Load, N
HP Outbrd 1 3,448 .48 .280E6 1,725 .48 .152E6

HP Inboard 2 2,760 .64 .367E6 2,070 .64 .286E6

LP1 Outbrd 3 4,830 .64 .777E6 650 .64 .735E6

LP1 Inboard 4 3,800 .64 .65E6 3,450 .56 .608E6

LP2 Outbrd 5 3,450 .64 .604E6 3,450 .56 .608E6

LP2 Inboard 6 3,450 .64 .604E6 3,100 .56 .551E6

LP3 Outbrd 7 1,380 .64 .26E6 1,030 .56 .213E6

LP3 Inboard 8 690 .51 .15E6 2,760 .64 .495E6

Gen Inboard 9 2,760 .64 .546E6 3,100 .51 .635E6

Gen Outbrd 10 3,100 .51 .632E6 1,380 .51 .307E6

Exciter 11 1,380 .165 .032E6 1,725 .165 .036E6

TOTAL 4.90E6 TOTAL 4.63E6

The LP3 bearing no 8 of unit 2 has a pressure of only 690 kPa as compared to the overloaded LP1 outboard bearing no
3 of the same unit of 4,830 kPa. Also note that the LP1 outboard bearing of unit 3 is underloaded with a pressure of only 650
kPa. These differences in measured hydrodynamic bearing pressures are caused by the differences in vertical alignment
between the two machines. Hence it may be expected that the experimental influence coefficients between the two units will
have difference values of amplitude and phase response to the trial weights.

4 Paper-ID 331
4 Undamped Natural Frequencies of Turbine-Generator System

The turbine-generator system may have as many as 10 distinct undamped modes in the operating speed range. However,
when fluid film bearings and foundation effects are considered, then up to 30 complex eigenvalues must be considered due
to bearing asymmetry and the presence of forward and backward modes.

Figure 7: 1st T-G Mode at 637 RPM - Generator Figure 8: 2 nd T-G M ode at 706 RPM

Figure 7 represents the first system mode. Since the generator is the longest unit of the system, the first critical speed is
the generator 1 st mode. This mode is relatively unaffected by the LP turbines. The mode is primary influenced by the generator
design and bearing coefficients. Figure 8 represents the 2 nd system mode at 706 RPM. In this mode, the mass centers of the
low pressure turbines LP1 and LP3 are out of phase to the second low pressure turbine, LP2. The generator motion in the
mode is very quiet.

Figure 9: 3 rd T-G Mode at 734 RPM Figure 10: 4 th Mode at 757 RPM

In Figure 9 for mode 3, the mass centers of LP1 and LP3 are moving out of phase and the second low pressure turbine
is quiet. As the speed is increased a small speed increment of 23 RPM, the mode changes in which the mass centers of the
three low pressure turbines are moving in phase. These three modes of the LP rotors take place in only a 50 RPM speed
range. When the LP turbines are well balanced before assembly to correct mass center unbalance, then these modes are not
usually observed during startup.
It is of considerable interest to note that these closely spaced critical speed mode shapes are not normally detected using
a transfer matrix method of calculation. A typical search speed increment of 50 RPM will cause these modes to be completely
missed. This situation of missing modes is not encountered when using a finite element. This is only one of many problems
encountered with using a transfer matrix method for computation of the dynamics of large turbine-generators.
In general, transfer matrix methods are unsuited for the computation of the dynamical characteristics of large turbine-
generators with pedestal effects. Problems with convergence are also encountered. The first four system modes indicate
the desirability of initially balancing the generator and turbine in low speed balancing equipment to correct initial mass center

Figure 11: 7 th T-G Mode at 1450 RPM Figure 12: 8 th T-G Mode at 1534 RPM

unbalance.
Figure 11 shows the 7th mode of the system at 1450 RPM. In this mode there is high exciter motion The generator has
a conical motion in which the ends are out of phase. As the higher system critical speeds are approached, the mass centers
of the turbines approach node points with little motion. It is of interest to note that the 7th mode shows 7 crossover points.
Figure 12 represents the system 8th mode at 1534 RPM. This mode is similar to the 7th mode shown in Figure 12 except
that there are now 8 crossover or node points along the turbine-generator centerline. The fact that the mass centers of the
generator and turbines all indicate node points at the rotor center of gravities has an important conclusion for high-speed field
balancing. It would be of little use to apply balance corrections to the rotor mid spans even if these locations could be reached.
Midspan corrections are of influence at the lower modes.

5 Paper-ID 331
 Figure 13: 9 th T-G Mode at 1662 RPM Figure 14: 10 th T-G Mode at 1849 RPM

Figure 13 represents the 9 th mode at 1662 RPM. Note the high exciter motion that is also seen in this mode. The number
of nodal points is 9 which is the order of the critical speed. As each higher critical speed is reached, the nodes or crossover
locations of zero amplitude also increase. Figure 14 represents the 10 th mode at 1849 RPM. This mode has 10 crossover
points. Again it is observed that the turbine mass centers become modal nodal points.
The next higher mode is above 2,400 RPM. This class of turbine -generator was originally designed for 50 Hz operation.
The required frequency of 60 Hz for operation in North America places the operating speed at 1,800 RPM. This introduces
two additional critical speeds of interest into the system. The observation of the high exciter motion represents a serious
situation. A failure of a last stage turbine blade on LP3 has been shown to cause an excitation of the 9 th mode leading to
exciter failure. This is conjectured to be the cause of failure of the Detroit System on Christmas day of over 10 years ago.
The system model here is of similar design to the Detroit unit.

5 Turbine-Generator Complex Influence Coefficients at 1,800 RPM

5.1 Experimental Influence Coefficients For Various Balance Planes

In order to perform field balancing, it is necessary to establish the system influence coefficients under applied unbalances
at various locations along the shaft. The measured responses along the shaft divided by the trial weight is referred to as the
influence coefficient response. It should be noted that for the case of large turbine-generators, the response is load dependant
as well as speed dependant.
The response of the system was recorded for trial weights placed along couplings 2, 3, and 4, and trial weights placed
at either end of the third low pressure turbine, LP3. Amplitude and phase measurements were taken at all 11 bearings. An
additional measurement between the 10 th and 11 th bearings on the exciter shaft would have been desirable due to the extreme
flexibility of the exciter shaft.

Figure 15: T-G Influence Coefficient Response at 1800 Figure 16: T-G Influence Coefficient Response at 1800
RPM W ith Trial W eight at 2 nd Coupling C2 RPM W ith Trial W eight at 3 nd Coupling C3

Figure 15 represents the T-G complex response with a trial weight placed at the second coupling. The complex response
was resolved into in-phase and out of phase components. A cubic spline curve fit was used to generate the shaft response.
Note the large response obtained at the third coupling C3. Fig 16 represents the complex influence response with a trial weight
placed at the third coupling, C3. Note the large responses of the system at couplings C2 and C3. Also note that there is very
little response of the mass centers of the turbine rotors.
Figure17 represents the influence coefficient response with the trial weight place at the number 4 coupling. This coupling
location connects the generator with the last turbine LP3. The trial weight placed at the generator coupling has a significant
influence on the generator response. There is a large out-of-phase 2 nd mode type of reaction on the generator. Generator
bowing in operation due to conductor shorting tends to create a bow resembling the first mode shape.

6 Paper-ID 331
 Figure 17: T-G Influence Coefficient Response at Figure 18: T-G Response W ith at 1800 RPM
1800 RPM W ith Trial W eight at 4 th Coupling C4 W ith Couple Unbalance on LP3

The influence coefficients were generated for individual trial weights placed on either ends of the LP3 rotor. By vector
resolution, the influence response was computed for a pure couple response on the LP3 rotor. As can be seen in Figure 18,
there is a substantial response on the generator.

5.2 Analytically Determined Influence Coefficients For Coupling C1 Balance Plane

Figure 19: Analytically Determined Influence Response at 1800 RPM For Balance W eight At 1 st Coupling - C1

In the balancing procedure, it was determined after many balancing calculations, that an additional balancing plane using
the first coupling location could be of benefit in balancing the turbine-generator at speed. The use of 6 or more balancing
locations seems logical since the system is theoretically operating through 10 critical speeds. It is normally preferred to have
at least the same number of balancing planes as critical speeds required to operate through.
Since no experimental data was available to generate the influence response of the rotor system for a weight placement at
the first coupling, the response was generated analytically using a finite element rotor dynamics program. However, before
this could be accomplished, numerous simulations of the known influence responses were computed to estimate the pedestal
effects. Although the LP turbine bearings are similar, there stiffness and damping coefficients vary due to the applied loads
as shown in Table 1.

6 M ulti-plane Field Balancing Calculations Based on Experimental and Analytical Influence Responses
M ulti-plane balancing corrections were computed assuming various balancing planes. All balancing calculations in this
study was performed using a modified least squared error method which could also incorporate weighting values in the data.
This provides some flexibility if one desires to emphasize the reduction of response at certain locations such as at or near the
generator. Table 2 represents the initial T-G bearing measurements at speed. It is important to note that all measurements and
influence coefficients responses must be based on the same generator power conditions. As generator output is varied, so is
the thermal bowing of the generator. Reliable data cannot be obtained if taken under various power conditions. After the
balancing corrections are computed, the balancing program predicts the new response after the application of the computed
weights.

7 Paper-ID 331
 6.1 Three-Plane Balancing W ith Couplings C2, C3, and C4
The first balancing prediction was based on the use of the influence responses at the second, third and fourth couplings.
The use of only these three couplings produces a reasonable system response. W ith this balancing combination, it is seen that
amplitude at the HP turbine bearing no 1 increases as well as bearing no 3 on LP2 rotor and bearing 7. The responses on the
LP3 rotor and generator are substantially reduced. It is noticed that the exciter amplitude has increased. Observation of the
exciter behavior indicates that it should be balanced last as a single plane. This situation then would represent a 4-plane
balancing solution.
Table 2 - Initial and Predicted Rotor Response
with Coupling Planes and LP3 Dynamic Couple Balancing
5 Planes - CPLS
3 Planes 2, 3, 4, & LP3
Probe Initial CPLS 2, 3, & 4 Dynamic 6 Planes
Location mm mm mm CPLS 1-4, & LP3

1 .055 .060 .043 .023

2 .045 .039 .030 .050

3 .124 .133 .055 .076

4 .138 .109 .090 .072

5 .107 .091 .098 .022
6 .090 .025 .065 .029

7 .058 .076 .034 .053

8 .108 .055 .083 .067

9 .088 .038 .065 .059

10 .056 .028 .055 .016

11 .075 .087 .070 .070

6.2 Five-Plane Balancing W ith Couplings C2, C3, C4 and Dynamic Couple on LP3 Turbine
The second balance computation involves the use of the three couplings and the two planes on the LP3 turbine. With this
balance calculation, the amplitudes through bearings 1 to 4 are reduced, however the amplitude at the generator coupling and
generator-exciter end bearing return to its original value of 0.055 mm. It should be noted that the balance values on the
generator may be greatly improved if the exciter motion is ignored in the least-squared-balance calculations. The exciter may
be single-plane trim balanced after reduction of generator and turbine vibrations.
In this balance calculation, the largest vibration is at bearing No. 5 with 0.098 mm. The desired vibration amplitude at
running speed is 0.076 mm ( 3 mils). This balance level is achieved at all locations except at bearings 4 and 5. As noted in
the previous balancing run with only the 3 coupling planes used, the required balance level may be obtained if the balance
calculations are based on ignoring the exciter. This is done by the assumption of a zero weighting factor for the exciter
readings. The exciter is then balanced separately by as a single plane. Because of the extreme flexibility of the exciter span,
exciter trim balancing has little influence on the response of the generator and turbines.
6.3 Six-Plane Balancing W ith Couplings C1, C2, C3, C4 and Dynamic Couple on LP3 Turbine
In the last balancing case, as shown in T able 2, the first coupling was added as an additional balance plane. Since no
experimental influence responses where measured, the required influence coefficients were numerically computed from the
analytic model. Figure 19 shows the theoretical unbalance response with a trial weight placed at the coupling. The first
coupling between the HP rotor and the first LP rotor was not originally considered as a possible balance plane to control
generator response. At the time, the operating engineers assumed that this location was too far removed from the generator
to have an influence on it.
W ith the addition of the first coupling as an added balance plane, the vibrations of the entire turbine-generator system may
be reduced to the desired level of 0.076 mm. W hen a least-squared-error method is employed in a multi-plane balancing
calculation, we often have the situation in which several locations are over corrected and insufficient reduction is achieved
at other locations. For example, bearings 5 & 6 are overbalanced with amplitudes of 0.022 and 0.029. The LP3 and inboard
generator bearing predicted response is 0.67 and 0.59 respectively. There are several ways to improve and enhance the
balancing at the generator and LP3 turbine. The first method is to ignore the influence of the exciter. The second method is
the use of weighting functions to place more emphasis on generator vibration reduction. This requires some trial runs but the
computations may be rapidly computed on a PC-based balancing program.

8 Paper-ID 331
 In the third method, the balancing is computed using linear programming theory. Further research is being conducted on this
technique as it does not require operator intervention with weighting factors applied.

6.4 Balancing Summary W ith Various Couplings and 2-Plane M oment Balance on LP3 Turbine
Figure 20 represents a plot of amplitude vs bearing number for the initial vibration and of the 3 cases shown with 3, 5, and
6 planes of balance corrections applied to the turbine-generator at running speed. Superimposed on the plot is a dotted red
line at 0.076 mm. This represents the desired level of vibration amplitude that it is preferred to operate at. The 6 planes of
balance correction clearly meet the desired response level using the 4 couplings and a moment correction on the LP3 turbine.
It is apparent from the balancing results that the unmodified least-squared-error balancing method overcorrects the balance
of the LP2 turbine. A finer level of balancing of the LP3 turbine and inboard generator bearing may be achieved by the
specification of weighting factors to reduce the influence of the amplitudes at the 5 th and 6 th bearings corresponding to the LP2

Figure 20 - Summary of Initial And Various Balancing Runs

Amp mm Vs Bearing Number
turbine and weighting factors to increase the influence of the No. 8 LP3 bearing and No. 10 , the inboard generator bearing.
In all cases, to adequately reduce the vibration levels of the exciter, a separate single-plane balancing should be performed on
the exciter. This is done after balancing of the main turbine-generator is accomplished. The balance along the entire turbine-
generator is improved when the exciter vibrations are initially ignored. This will result in a balance state in which the turbines
and generator will be well balanced but the exciter amplitude has increased. The exciter amplitude may then be reduced by
single-plane balancing.

7 DISCUSSION AND CONCLUSIONS

7.1 Discussion
It is highly desirable to develop field balancing procedures that can assist in balance corrections of large turbine -generators
on site. The application of arbitrary balance corrections can be a costly and time consuming process if conducted manually
without the aid of a computer. The balancing process is dependent upon the development of accurate response influence
coefficients. This requires that the data be taken at the same power levels for the initial and influence response runs. These
power levels need to be recorded. If the alignment of the intermediate bearings changes due to pedestal sag, then the
experimental influence coefficients will not be accurate.

7.2 Conclusions
1. To properly understand field balancing of large turbine-generators, it is desirable to develop analytic models of the complete
system including foundation effects.
2. The transfer matrix method is not suited for the analysis of larger turbine-generators. W ith the finite element method, it
is possible to compute the critical speeds, complex eigenvalues, unbalance response and time transient motion due to blade
loss.
3. A valid T-G model must include pedestal effects such as mass and stiffness. The flexibility of the pedestal supports may
result in over a 90% reduction of effective damping. Pedestal characteristics may be found by testing and by the simulation
of the experimental influence response motion with the analytic model.
4. Calculations indicate as many as 10 undamped modes exist in the operating speed range (and as many as 30 complex
eigenvalues, of which only a few are of interest). The undamped modes show that the lower modes involve motion of the
turbine and generator mass centers. Thus, low-speed balancing is useful in controlling these modes. For low-speed balancing
of the turbines in a balancing machine, the 2 planes of balance correction should be resolved into static and dynamic (couple
unbalance).

9 Paper-ID 331
Half of the computed static correction should be applied at the rotor center with the remainder placed at the end planes. Failure
to do this may result in a turbine that is not well balanced at speed.
5. The experimental unbalance response influence coefficients must be developed under repeated T-G load conditions. It
was found that for identical T-G units, the influence coefficients may be substantially different. The influence coefficients for
two identical units should not be mixed in the balancing computations.
6. The bearings of these large T-G sets are heavily loaded. Hence, jacking oil must be used to support the shafts under startup
and shutdown (loss of jacking oil during shutdown will lead to bearing failure). At speed, the jacking oil is turned off. The
hydrodynamic pressure generated at each bearing was recorded. This pressure measurement may be used to determine the
bearing loading. A settling of some of the interior bearing pedestals may cause these bearings to carry reduced loading.
This in turn will effect the influence coefficients. Two identical units may have different influence coefficients due to the
differences in vertical alignment and hence bearing loading. For the situation in which several interior bearings are lightly
loaded, the vertical alignment should be adjusted. From the measurement of the bottom bearing oil pressure at speed, it is
possible to determine the bearing loading and hence predict the bearing 8 stiffness and damping coefficients at speed for use
in dynamic analysis such as unbalance response and stability (damped eigenvalues).
7. At operating speeds, the mass centers of the turbines become nodal points. The motion of each turbine is a conical type
of motion. This conical behavior occurs after the fourth mode as shown in Figure 10. In Figures 11-14, which represent modes
7 to 10, it is seen that the turbines have conical motion. Therefore, it appears that the placement of static balance shots on the
turbine at either end in phase are not indicated. The turbines at speed better respond to dynamic balance couples in which two
weights are placed on a turbine at either end out of phase.
8. Satisfactory field balancing is difficult to achieve with fewer than four balance planes. A satisfactory balance condition
can be achieved with the four couplings and a two-plane moment correction to LP3 turbine. Using a least- squared-error
balancing procedure, it may be necessary to incorporate weighting values to interior measurements so as not to over correct
the interior turbines and not adequately reduce the generator vibrations.
9. Due to the extreme flexibility of the exciter, it should be balanced last as a single plane. The excessive flexibility of the
exciter not only leads to balancing difficulties, but also places the entire T-G system at risk. The loss of an LP3 turbine blade
will lead to exciter failure with catastrophic results.

8 REFERENCES
[1] Adams, M. L. (2001): Rotating Machinery Vibration: From Analysis To Troubleshooting, Marcel Dekker, Inc., New
York.
[2] Chen, W . J., and E. J. Gunter (2005): Introduction to Dynamics of Rotor-Bearing Systems, Trafford, Victoria, BC,
Canada.
[3] DyRoBeS - Dynamics of Rotor-Bearing Systems Users’Manual (2002): RODYN Vibration Analysis, Inc.,
Charlottesville, VA.
[4] Gunter, E. J., and C. Jackson (1992): “Balancing of Rigid and Flexible Rotors,” Handbook of Rotor Dynamics,
McGraw-Hill, New York.
[5] Gunter, E. J. and W . J. Chen, (2005): Dynamic Analysis of an 1150-MW Turbine-Generator, Proceedings ASME
Power, PW R2005-50142, Chicago, Illinois.
[6] Gunter, E. J., and C. G. Gaston (1987): CRITSPD-PC Undamped Critical Speed Users’ Manual, RODYN Vibration
Analysis, Inc., Charlottesville, VA.
[7] Gunter, E. J., and W . E. Gunter, (1986): Field Balancing 70 MW Gas Turbine-Generators, Proceedings, International
Conference on Rotor Dynamics, JSM E, IfToM M .
[8] Gunter, E. J., Z. Fang, and J. R. Henderson (1994): Static and Dynamic Analysis of a 1150-MW Turbine-Generator
System, Part I: Static Analysis, Proceedings, 18 th Annual Meeting, Vibration Institute, pp. 41-54.
[9] Krämer, E. (1993): Dynamics of Rotors and Foundations, Springer-Verlag, Berlin.
[10] MSC/PAL Users’ Manual Version 4, The MacNeal-Schwendler Corp., Los Angeles, CA.
[11] Nelson, H. D. (1980): A Finite Rotating Shaft Element Using Timoshenko Beam Theory, ASM E Journal of Mechanical
Design, 102, pp. 793-803.
[12] Nelson, H. D., and J. M . McVaugh (1976): The Dynamics of Rotor-Bearing Systems Using Finite Elements, ASME
Journal of Engineering for Industry, 98, pp. 593-600.

10 Paper-ID 331