SITE BALANCING OF A LARGE FLEXIBLE ROTOR

CONTAINING UNBALANCE ECCENTRICITY AND

PERMANENT RESIDUAL BOW
by
A. S. Maxwell
Maintenance Specialist, Balancing and Vibration

and
A. F. P. Sanderson
Maintenance Engineer, Mechanical System Maintenance Division
Ontario Hydro

Toronto, Ontario, Canada

apply and conventional balancing techniques are invalid. The

Stewart Maxwell was educated at the
Royal Technical College, Salford En­ theoretical explanation of the unbalance response of bowed
flexible rotors has recently been well documented. This paper
gland and is a registered professional en­
describes how a flexible rotor containing a significant bend was
gineer of the Province of Ontario. He
joined Ontario Hydro four years ago and optimally balanced over its complete speed range, by incor­
is presently responsible for providing ad­ porating bowed rotor response theory into the balance tech­
vice, guidance and assistance to person­ nique.
nel in fossil and nuclear stations and
other areas of Ontario Hydro operations, INTRODUCTION
on all aspects of maintaining plant and The primary objective of the authors in the execution of
equipment in satisfactory operating con­ site balancing work is an overall increase in reliability and
dition from the viewpoint of vibration and balancing. His re­ reduction in maintenance costs and outage times. The work
sponsibilities include development and implementation of that will be described is part of a more general program to
methods and techniques to ensure the corporation is kept at the improve accuracy and effectiveness of site balancing flexible
state of the art in these areas. steam turbine-generator rotors. It was felt, for example, that
After six years in the aerospace industry in the field of more use should be made of computers for such things as
stress and vibration analysis, he spent nine years with Wes­ optimizing weight corrections between several planes which
tinghouse Canada as a research and development engineer in might have conflicting individual balance requirements. Also,
experimental stress and analysis, noise and vibration and as a the advent of hand held computers or calculators should make
senior design engineer in the Turbine Engineering Department. hand graphical computations and constructions on polar graph
paper obsolete. Obviously, in such a system, the use of
Alan F. P. Sanderson was educated at stroboscopic lights would be inadequate for phase measure­
Newcastle-upon- Tyne Polytechnic, En­ ment. Although the limited number of measuring and balanc­
gland. He is a graduate member of the ing planes available would preclude true, in place, modal
Institution of Mec hanical Engineers. balancing, it was nevertheless felt that the full potential of site
Serving as a research engineer in the balancing was not being utilized.
Machine Dynamics laboratory of C. A.
Parsons and Company for five years, Mr. SITE BALANCING PHILOSOPHY
Sanderson was primarily interested in
If we accept the premise that it is theoretically not possi­
vibration problems associated with large
generator stators and flexible rotor ble to exactly site balance flexible rotors, then the practicing
engineer who is continually facing this dilemma must address
balancing. He was subsequently trans­
himself to the question: "What is the best that can be done
ferred to Canada to take responsibility for field balancing in
with the tools available?" He does not really care whether his
North America. Mr. Sanderson now works for Ontario Hydro,
solution is mathematically elegant or not, but whether it is
where he is involved in the maintenance aspect of vibration
possible for him to correct balance defects at site sufficiently
analysis on steam turbine generators and rotating equipment.
that the machine will be in a completely satisfactory condition
for unrestricted operation over its specified speed and load
range and duty cycle.
ABSTRACT
The procedure adopted by the System Maintenance Divi­
\Vhen turbine-generators operating at constant speed re­ sion of Ontario Hydro to achieve this consists of pertorming
quire site trim balancing, this is most commonly pertormed at optimum correction of unbalance by operating the machine
the synchronous speed using a conventional influence coeffi­ through its complete speed and load range and dealing with
cient method. In some cases, the vibration of flexible rotors each modal defect in turn.
during run-up through critical speeds is excessive and balance
can only be achieved by correcting in tum each model defect There are three essential ingregients in this prodedure:
throughout the speed range. Difficulties arise in the case of 1. A proven and rigorously followed technique with strict
rotors that contain a permanent residual bow. Classical adherence to sign convention and repeatable measur­
phase/speed relationships through a critical speed no longer ing conditions.

23
24 PROCEEDINGS OF THE FIFTH TURBOMACHINERY SYMPOSIUM

2. Precise and convenient measuring instruments and with the slow roll defect, and understanding of how optimally
transducers. to balance the rotor is of great benefit.
3. The ability to identify the model defects affecting each
rotor span and to make the appropriate corrections. THEORETICAL BACKGROUND
In most cases these procedures will produce acceptable It was as a direct result of the work contained in a course
results. However, difficulties still arise in the case of rotors that ( 1) given by Professor E. J. Gunter and his associates at the
have significant distortion of their elastic center, and it is with University of Virginia, that the response of bowed flexible
the resolution of this problem that this paper will deal. rotors was understood and described sufficiently clearly that
this effect was included in the balancing program with confi­
DISCUSSION OF THE PROBLEM dence.
In general, a bent rotor in a continuous train of solidly At this time a computer program was being written to
coupled rotors will result in at least one journal running eccen­ extend the conventional two-plane theory of Thearle, (2) to
trically during slow roll. This eccentricity may be small in four planes, and it was to this program that Dr. Gunter's work
comparison to the maximum eccentricity that would be ob­ was adapted.
served at the rotor's mid span if such a measurement were A brief summary of the theory is provided as follows:
possible, but in the practical case this is usually the only indica­
tion of the rotor slow roll shape available for direct comparison The general expression for rotor response due to unbal­
with the rotor shape during running. The relationship between ance and slow roll eccentricity is shown below where x iden­
the mid span eccentricity and the journal eccentricity will be tifies the bearing number, y the balance plane, and n, the run
complex depending on the concentricity of adjacent couplings number.
as well as relative rotor stiffnesses and weights and bearing
loadings. In order to completely balance the bent rotor it is not
L j [ ][ ] [ ]f]
Zx n = auxv �uv + arxv rx (I)
necessary to know what this relationship is. Theoretically, the
same balancing logic applies to the quite different cases of: X= 1, 4 y= 1, 4
n= 0, 4
1. A rotor that is significantly bent because of distortion
during service.
2. Straight rotors with significant coupling concentricity
[ J aux is a 4 x 4 matrix
errors, causing lateral offset of a rotor span without any
imposed bending. Zxn Complex shaft amplitude of bearing x during run n
3. Normally straight rotors with imposed bending �uy Complex unbalance eccentricity in Plane y
moments due to coupling face errors (coupling faces Complex influence coefficient relating unbalance in
not perpendicular to shaft axis). Plane y to the response at Bearing x
All three of the above cases result in constant bearing Complex slow roll eccentricity at bearing x
forces related to stiffness as well as dynamic forces caused by
unbalance. The forces due to any or all of the above effects arxy Complex influence coefficient relating slow roll ec­
cause a constant slow roll eccentricity and the contributing centricity at Plane y to the response at Bearing x
components of each need not be known. The balancing process Study of Dr. Gunter's work shows that in a rotor that has a
is aimed at removing only the dynamic forces. This can be done high shaft stiffness compared to its modal mass, the residual
quite successfully even with a combination of the above listed bow or slow roll eccentricity complex influence coefficient ap­
defects. proaches unity and becomes independent of speed. It has been
It is axiomatic among most balancing engineers that at­ our experience that this assumption is valid for large steam
tempts to balance a particular rotor with known misalignment turbine rotors with mid span bow eccentricities of up to .005
will result in either moving the high vibration to an adjacent inch and speeds of up to 3,600 RPM. It agrees with our basic
bearing or will achieve reduction of vibration at the chosen objective in balancing bowed rotors which is to balance only
balancing speed at the expense of higher vibration at other the elastic deflection to zero and not to impose any internal
speeds, especially at criticals. We believe that the above ob­ forces or moments on the rotor. Again, the aim is to achieve a
servations occur because part of the permanent slow roll eccen­ condition where the shape of the bowed rotor at all speeds is
tricity has been compensated for during the attempted balance the same as its shape during slow roll.
and that this is done unknowingly and without an understand­ Following the assumption that the residual bow coefficient
ing of the relationship between slow roll eccentricity and elas­ is equal to unity, then the residual bow may be subtracted from
tic unbalance. This is the case particularly when balancing is the rotor amplitude. Thus for the initial run at balancing speed,
done based on bearing housing vibration. The fact is that there from Equation ( 1).
is a limit to what can be acheived by balancing a rotor with a
permanent slow roll eccentricity. This limit is the removal of
dynamic forces only, leaving unaffected the eccentricity that
was measured during slow roll at each journal. The ideal result
where f } [ } [ } [��}
zxn = auxv �uv +

would then be vibration equal to the constant value of slow roll

eccentricity throughout the speed range of the machine.
It should always be remembered, however, that in spite of
then [ ] �J[ }
zxo grx = ux �uv (2)

the above assertions, the preferred solution to the problem of X= 1, 4 y= 1, 4

slow roll eccentricity is to rectify the fault directly at its source. "'

However, where the only recourse is to balance a machine 8rx is determined experimentally by running the machine
until the defect can be corrected at an overhaul or where it is at a very low speed where shaft elastic deflection may be con­
judged to be cost effective to operate the machine indefinitely sidered negligible.
 SITE BALANCING OF A LARGE FLEXIBLE ROTOR CONTAINING UNBALANCE ECCENTRICITY
AND PERMANENT RESIDUAL BOW 25

360
For the calibration runs:

L
Zxn - 6,x ] [ }[ Uuxy euy + �CAL J {3)
--

270
<I>
"'
LLJ
0:
"'
"'
a
BEFORE
n = 0, 4 1, 4 y = 1, 4
I
X= "' BALANCING "'
"" 180 ...J
"' "'
where �CALy = Complex Calibration Weight 0.
0
- _,,.-, __
z
""
Eccentricity in Balance Plane Y. .... "'
<I>

By subtraction: Zxn - Zxo Uu xy �CALy 90 �

"'
<l
"'
= 0.

0.

or Zxn - Zxo "'

(4) i oL-�--- ---�--� 0

'@'CALy
2000 3000 4000
I SHAFT SPEED- REV./MIN.
"'
a

The Z terms are measured or derived during the calibra­ :::>

....
--- AMPLITUDE

tion runs, 1l'cAL is known, therefore aux

::; ----- P HASE
can be 0.
y y ::E
<l
I 360
calculated and equations (2) solved. z I
0 I
;:: 6
In practice, since each term in equation (2) is complex, <l I <I>
"'

I 270 �
0:

they are separated out into real and imaginary parts in order to "'
> "'
obtain eight linear simultaneous equations for programming I "'
a
-' I '
AFTER BALANCING
convenience. ;'! 4 I "'
z
"' / 180 �
::E "'
In extending this theory to four planes and multi-span -- ,. �
;3 ,.
z "' "'
rotor systems it also has to be extended to accept the concept of :::>
... 2 .,...
--- - ---"" 90 it
<I>
<(
modal correction. This is done by maintaining a coherent sys­ /
tem ofZ terms and '8" terms. If balancing is concerned with
rx
/

correcting p articular modes that have been isolated from

2000 3000 4000
analysis of the vibration vectors from each rotor span, the SHAFT SPEED- REV./ MIN.
theory can still be applied if the modal components are used
consistently throughout the computation in the same way that Figure 2. Typical Response of Rotor Test Kit.
individual vectors are normally used. The modal components
of the terms used are substituted for the terms themselves, and No. 1 and 3 were similar. These tests were done primarily to
the solution obtained is a solution for the relevant mode only. check the computer program and to confirm that Dr. Gunter's
theory had been correctly transposed to our instrumentation
and computer system. The reason that the results achieved
DEVELOPME NT OF THE METHOD here are not exact is mainly that very small weights were in­
The method described above was first used on a rotor test volved and the required accuracy in obtaining the precise
kit shoV\.n in Figure 1 having a deliberately bent and unbalanced weights at the precise angular location was not pursued. Also,
rotor, three bearings and three balance planes. The results for there was some problem with resonances in the test kits' foun­
Bearing No. 2 are shown in Figure 2; the response at Bearings dation.

Figure 1. Test Rotor Containing Permanent Bow and Unbalance Eccentricity.

26 PROCEEDINGS OF THE FIFTH TURBOMACHINERY SYMPOSIUM

360
APPLICATION OF THE METHOD ---

TO A LARGE TURBINE ROTOR 12

The addition of the residual bow effect in the work de­ 270
10
scribed was not done solely out of a desire for academic
thoroughness. There was the suspicion that a permanently
180
""

bowed rotor existed in one of Ontario Hydro's large turbine­ <(

"'
0. •
generator units. Shortly after the test kit work was done it was
�
reported that a 500 MW turbine-generator was experiencing "" 4
<(
excessive vibration at the intermediate pressure rotor bearings "'
0. 2
during run up through its first critical speed and that steady �
state vibration levels at sychronous speed were at the allowable i o�l-oio�
o ------ --- ---- 2o- r oo--------------J� oo-o--���
limit. This machine consists of a rigid high pressure rotor, and
SHAFT SPEED REV.IMIN.
intermediate pressure rotor operating between its first and
-- AMPLITUDE
second critical speeds, two low pressure rotors in tandem, and
----PHASE
a generator rotor each operating above its second critical ------- ..... 360
speed. The intermediate pressure rotor that was suspect, con­ z
0
;::
tains a site accessible mid span balance plane as well as rotor <(
"'
12
body end planes and balance planes in the couplings. The rotor N2 4 BEARING <II
IXl
270�
> 10
is shown in Figure 3; its weight is 45,000 pounds; the distance
between journal centers is 200 inches, the journal diameters
-'
;:! 8
�I

are about 16 inches and its maximum diameter is 53 inches.

z
"' 180 �
�
0
6 �
z

FIRST SITE BALANCING RUNS

:::>
"- 4

The machine was allowed to cool out and cold datum re­
sponse curves of amplitude and phase were obtained during a
0 �-r----.---�--� o
slow, controlled run to speed. The amplitude and phase re­ 1000 2000 3000 3000
sponse curves for both journals are shown in Figure 4. All SHAFT SPEED· REV./MIN.

vibration levels quoted refer to shaft seismic or absolute val­

ues. The Nyquist plot, shown in Figure 5, is a polar plot of the Figure 4. Amplitude and Phase Response with Speed of the
vibration vectors at Bearings 3 and 4 over the speed range Bowed Rotor as Found.
1,200 RPM to 3,600 RPM. Only the quadrants 0°, 90°, 180°,
and 270° have been shown for clarity. The numbers at the
various points refer to the speed of the machine in hundreds of
RPM. Thus, point 12 on the full line graph represents a vibra­
90
tion of 2.0 mils at a phase angle of 180° at 1,200 RPM. Simi­
larly, point 22 shows a vibration vector of 12 mils at 280° and
2,200 RPM, and so on. All the information on amplitude,
"
speed and phase for the two bearings that is shown in the more 180 -----""?'-<';:.
' . --..,;."'\., ;?- '' -1---------f':C';n,,_..:.;;._o:;:::------ 0
•

conventional form in Figure 4, is contained in these Nyquist ,

I
I
plots. The critical speed can be seen as the maximum vibration '"
I
vector occurring at the point of maximum rate of change of
curvature and at about goo phase change. The relative mag­ I 1120

nitudes and phase at each bearing can be compared directly, I

,. I I
I I
\ I
\ I
\ I
\ I
' ...
/
' 22

--i4_--
•,,

24

22

BEARING N2 4
BEARING Ni 3
270 SCALEf'\.., f------1• 2 MILS PEAK
TO PEAK

Figure 3. Intermediate Pressure Rotor of a 500 MW Turbine­

Generator Containing Pennanent Bow and Unbalance Eccen­ Figure 5. Nyquist Plot of the Bowed Rotor Showing Severe First
tricity. Mode Defect.
 SITE BALANCING OF A LARGE FLEXIBLE ROTOR CONTAINING UNBALANCE ECCENTRICITY
AND PERMANENT RESIDUAL BOW 27

and it is clear that both ends of the rotor are sensibly in phase corrected, the evidence of this is the dip close to zero
and that these indicate a considerable first critical speed mode amplitude at 2,950 RPM, above the critical speed. Although
defect. this condition of an undercorrected bend was suspected at the
time of balancing, any further increase in mid span first mode
Prior to this date two balance corrections applied to the
correction weights produced intolerably high vibration at 3,600
mid span balance plane had been made by the turbine man­
RPM. Subsequent analysis of the effect of the coupling weights
ufacturer, but these had been aimed at reducing the response
that had been applied during previous balancing showed that
at synchronous speed of 3,600 RPM only. Some balance
these weights were the reason that a better compromise be­
weights had also been added to a coupling plane outboard of
tween first mode defect and operating speed balance defect
the main rotor span. This latter practice is an expedient that is
could not be achieved.
sometimes preferred because of ease of access, but we will
discover later how this approach is incompatible with the con­
cept of modal defect correction, and in fact can distort the rotor LOW SPEED BALANCING
response to such a degree that it can make modal separation DURING OVERHAUL
and optimum correction impossible until these weights are
removed. At this stage, the rotor was only suspected of having Mid span run out readings made while the rotor was in a
a bend and an attempt was made to balance what the response low speed balancing machine showed a total indicator reading
indicated to be a severe first mode defect, using the mid span (TIR) of .009 inch and, of course, the angular location of the
plane. high spot. At this time the rigid body modes were corrected in
the balancing machine with corrections distributed between
The result of this exercise is shown in Figure 6. A balanc­ the mid span and end planes according to a formula provided
ing speed of 1,700 RPM was chosen and mid span corrections by the manufacturer. It was not expected that this procedure
for the first mode defect were applied. Here, the very high would completely correct the rotor because its bowed shape
peak at the critical speeed has been eliminated but the vibra­ and response in its own bearings would be significantly differ­
tion at synchronous speed has increased. At this point the ent to that in the balancing machine, and because no elastic
behavior of the rotor was puzzling; logic insisted that the first first mode deflection could be achieved. Balancing was done at
mode correction should not have upset the balance at 3,600 200 RPM. The low speed balancing was also useful in allowing
RPM to the extent that it did. Couple corrections were applied balance weights to be installed in the factory planes, leaving
to reduce vibration to acceptable limits at 3,600 RPM and this the field access balance planes available.
level further reduced with load. The machine was temporarily
left in this condition since it was due for a major overhaul, at The response of the rotor after low speed balancing is
which time the extent of any bow would be measured accu­ shown in Figure 7. The dip in the curve at 3,000 RPM repre­
rately and remedial procedures would be applied at that time. sents a soaking period to allow an increase in rotor bore tem­
The results obtained were analyzed in the meantime. The perature to reduce stress levels. This was not necessary during
shape of the response curve is similar in form to a bowed rotor other runs. This response curve is included to show the
response curve of Professor Gunter's where the bow is under- amplitude/phase relationship at 1,950 RPM and because it is

360

12
II)

210 �
"'

10 10
13
0
' Nl 3 BEARING
... "' � 8
180 6 �
::'l
.. . � o•
...
1:?
"4 ---­ � 4
"'
.. ��
..
.,.. ..... -----­

�z .,.,.,.- ..... -�

io1--T=========::::::;:::::===-���r---
�
1 1000 2000 3000
----J3600o �o,l_�---
1000 2000 3000
-=====;�:':::::=::;::Jo
3600
SHAFT SPEED - REV. I MIN.
� SHAFT SPEED-REV./MIN.
:::>
... AMPLITUDE
::; -- AMPLITUDE
.. PHASE
::1 ----PHASE
..
360
�
II)
"'
� 12
0:
4 BEARING "' "'
Nf
270 ffi
>10
-�
­ 1!: .... .... N• 4 BEARING
..J �� "'
i!a i'!s :
z � z
"'
,.-- ..... '
... 180 � lE I
.. .. 6 "'
,.., ..... _,

I· 0 I
z I
:::> "'
"'4 ... .

oL--,----�==���---.-0
1000 2000 3000 1000 2000 3000 3600
SHAFT SPEED- REV./ MIN. SHAFT SPEED- REV./ MIN.

Figure 6. Response of Bowed Rotor After Attempted Conven­ Figure 7. Response of Bowed Rotor After Low Speed Balanc­
tional Site Balancing Procedure. ing During Overhaul.
28 PROCEEDINGS OF THE FIFTH TURBO MACHINERY SYMPOSIUM

the actual run up after low speed balancing. It shows this time, types of transducer. One advantage of this instrumentation
and overcorrected bend according to Gunter's work since the system may not be readily obvious but is of major importance.
response dips to zero below the critical speed. More impor­ There is no phase lag introduced into the measuring system by
tantly, it shows how incorrect it would be to balance only at the instrument. It gives true phase lag angle from a predeter­
speeds up to 1, 900 RPM because of the subsequent response at mined angular datum to the rotor physical high spot at all
higher speeds. The Nyquist plot in Figure 8 shows this dramat­ speeds. This means, for example, that slow roll bend eccen­
ically. Data from this run were not used for the balancing tricities which were measured with a dial gauge and referred to
computations because of the transient conditions. Repeatable the phase datum manually, can be directly compared to the
conditions at balancing speeds of 1,700 RPM and 3,600 RPM displacement and phase angle shown on the meter. They are
were established for balancing purposes. the same measurement and will read the same angle at
machine speeds that produce negligible elastic shaft deflection.
INSTRUMENTATION The instrument phase angle reading is an absolute value that
allows the physical position of the rotor high spot to be
In preparation for inclusion of the residual bow effect, visualized during the tests.
proximity probes were mounted at the rotor Bearings No. 3
and 4 (and also No. 5, the inboard low pressure turbine bearing
adjacent to Bearing No. 4). The decision and opportunity to
include the slow roll eccentricity effect was made only two days
before balancing was to commence; consequently the test fix­
ture design is not very sophisticated. Throughout the balance
runs, journal seismic response was measured by mounting a
velocity transducer on top of the supervisory transducer which
observed shaft motion via a spring loaded shaft rider. In the
case of the slow roll eccentricity measurement, a proximity
probe was mounted on the bearing cap to observe a target that
had the same motion as the shaft riding probes. This arrange­
ment is shown in Figure 9.
A commercial digital phase meter containing some minor
custom features was the primary read out device. This is shown
in Figure 10. This instrument employs two synchronous track­
ing filters and provides a digital readout of speed, amplitude
and phase of the once per turn vibration component. The ve­
locity signals were integrated to read out in mils peak to peak
and the phase convention and display was the same for both

90

90

Figure 9. Arrangement of Transducers for Slow Roll Eccentric­

(,;• ity and Dynamic Vibration Measurement.
'"x.,24
...
1
180-------h-::i-----0
18 .. 20

BEARING N• 3

270

270

SCALE '\.1 1------l • 2 MILS PEAK TO PEAK

Figure 8. Nyquist Plot of Rotor Response After Low Speed

Balancing Showing Divergence of Response Above 2,300 rpm. Figure 10. Read Out Instrumentation.
 SITE BALANCING OF A LARGE FLEXIBLE ROTOR CONTAINING UNBALANCE ECCENTRICITY
AND PER'vfANENT RESIDUAL BOW 29

A portable vibration meter was used to provide signal FINAL BALANCING PROCEDURE
conditioning and power Sllpplv to each proximity probe and to It was {(lund that only Bearing No. 4, had signiHcant slow
allow probe gap setting and gap/voltage calibration. This is
shown in Figure 11. In addition to the direct proximity probe
roll eccentricitv . 0.5 mils at 180°) and excess he vibration, so
the correcti011 to be applied was based primarily on this re­
inputs, the instrument allowed six hearings to he switch sponse. A mid span calibration weight was applied and its ef
selected and monitored during a nm to speed. Phase rekrence f(,ct at the chosen and repeatable balance speeds of 1,700 RP\1
was obtained hom the 111achine's own supcrvisorv phase pulse and :3,600 H.P\1 was mcasmed. It should be noted here that if
and this was the datum used during run out checks in the low these data are put into the computer program directly, along
speed balancing machine, during slow roll, and throughout the with the slow roll eccentricity vector oJ a cliflerent solution
tests. The oscilloscope was used to nwnitor continually the will be predicted l(Jr each speed. The reason for this is that the
1mflltered \·ihration wave shape. thcorv. is based on the response of a model having a single IU<iss
at the center of a flexible shaft. 'With this conflguration, only a
first mode can he signiflcantly excited and all vibration data
rd�:r to this n1ode: furthern1ort>, in Gunter's model the re­
sponse was llleasured a t mid span making the system
additional!\ insensitive to anv ' second mode dfC:ct. To make
use of the J;rograrn, therd(m', only the in phase components of
vibration response at each hearing should lw used. The in
phase \'ectors \\Pre isolated and these data were used with the
slow roll ecccntricitv · vPctor to obtain the single correction
weight predicted h)· the computer program. Figure 12 shows
that the single weight produces an efit'd that converges to a
single point at both speeds and that this point of convergr,nce is
close to the ideal value of' the slow roll eccentricity of L'5 mils
at .180" for Bearing "io. 4 and close to zero f(lr H<"aring 1\:o. .').
The llnal response of the rotor is showu in Figure 1.3.
Numlwr 4. Hearing is seen to he HTV . close to the ideal caw of
constant slow rol l "ccentricitv ' vect< >r level over its complct<'
speed rang<'. No evidence or the critical speed at 2 . .350 HPl\-1
can he s<'CII at Bearing No. ,J although this can still h<: t!etccted
in the much lower Bearing :3 response. Again the �yquist plot
Figure 1 1. Usc of Purtablc Vi[Jrrttion !\[cterj(JI' Calihration and in Figure 14 is tl>e most graphic. Here the first critical speed
l'owcr SuJ•ply. can .,till he dearly seen al Bearing "io . .'3 in spite of tile low
90

360

12 U)
w
270 LLI
a:
10 ' "'
/ w
/ 0
'
3600 REV /MIN
,."''/ 160 lj
"'
"'
" ;�.'"� N2 3 BEARING z
w
6
' /
"'
0 "J
w
U)
0. 90 �
,_
0-
"'
<t 2
w
180 -·----".c---- -----+"'<-- --------- 0 0.
"'o L_-,------------�==r===�=:==���-r--.C�-�;�
.../ 1000 2000 3000 360�
i SHAFT SPEED- REV. I MIN.

AMPLITUDE

PHASE
1700 REV I MIN
360

z
0
� 12 U)
w
"' 270 w
a:
S: I O "'
w
-- 0
.../ ,.,.- ---------- '
SCALE .J' f-------1 = I MIL PEAK TO PEAK I! a -----
z
w ---- !80 �
"'
4 BEARING z
:I 6 N2
"
0
z w
:0 "'
... 4 "
90
:t:
a.
270

BEARING N2 4
�3 s�
,L__ '----------- -----T_ ______________ _ r�----
BEARING N2 3 1oo o 2 ooo 3ooo
SHAFT SPEED- REV./ MIN.
Figure 12. Conrcrgcncc of' Vectors at 1, 700 rpm and 3,600 rpm
Resulting From the Predicted Single Mid- Span Correction Figure 13. Final Rotor Condition Showing Flat Response at All
Weight. Speeds.
30 PROCEEDINGS OF THE FIFTH TURBOMACHINERY SYMPOSIUM

90
CONCLUSIONS
A large flexible turbine rotor containing a significant per­
manent bend has been successfully balanced. Further conclu­
sions regarding the technique for achieving this are as follows:
l. The amplitude/phase/speed response of the rotor must be
recorded over its complete speed range.
2. The first mode response must be isolated and only this
effect included in the balance computations.
18

- , :"t
22
180 ----------+-,----- :---f ..,.,--
-- -- 3. A precise measuring system and a meticulously applied
26
method for obtaining reliable and repeatable datum conditions
and calibration runs must be used.
4. As Professor Gunter's work predicted, the classical 90°
phase change at resonance did not occur for some combinations
of unbalance and residual bow. This effect compounds the in­
adequacy of basing correction weight location on this assump­
tion alone.
SCALE"-' t-------1 • 2 MILS 5. Balancing corrections based on a single speed are consid­
PEAK TO PEAK
erably more inadequate in the case of a bowed rotor than they
-- BEARING N• 4 are in the case of a normal flexible rotor.
----- - BEARING N2 3

270
REFERENCES
Figure 14. Nyquist Plot of Final Response Showing Residual
Small Unbalance at Bearing No.3 and Dominance of Slow Roll l. Gunter E. J., "Dynamic Response and Balancing of Rotor­
Eccentricity at Bearing No. 4. Bearing Systems," University ofVirginia, Short Course. Au­
gust 1974.
level, but Bearing No. 4 is dominated by the constant slow roll 2. Thearle E. L., "Dynamic Balancing of Rotating Machinery in
eccentricity vector which is in phase with the peak level of No. the Field," Transactions ASME Volume 56, 1934, pp. 745,
3 Bearing. 753.