HIGH SPEED BALANCE PROCEDURE

by
Dr. Robert Ehrich
Manager, Development Turbomachinery
Borsig GmbH
Berlin, West Germany

Robert Ehrich received the degree

"Diplom-Mathematiker" from the Tech
nical University of Berlin, Gennany,
where he also received his Ph.D. in 1960.
The subject of his doctoral thesis was
"Temperature Distribution in Steam
Turbine Casings". He was with AEG/
Berlin in turbine engineering and devel
opment for 16 years. For the past 10
years Dr. Ehrich has been with Borsig/
Berlin and has specialized in vibration of
turbine blades, reciprocating compressor pulsations, and rotor
dynamics of turbomachinery.

ABSTRACT
The paper describes elastic balancing procedures and
experimental results for turbomachine rotors. The commonly
known criteria of static and dynamic balancing no longer apply Figure 1. Vacuum Bunker for Balancing and Overspeed
in this case. Testing.
The rotating elements are balanced at full speed in their
contract bearings. Dynamic bearing forces are measured with
respect to direction and magnitude. Balancing weights are
attached to the rotor in several planes and the thermally caused
deflections, nonsynchronous vibrations and natural frequen
cies of bearing supports have to be accounted for. The rotor
sensitivity can be determined by attachment of test unbal
ances. Dynamic bearing forces and shaft vibrations are
measured and the shock excitations and log decrement calcula
tions yield the stability limits.

INTRODUCTION
Balancing turbomachinery rotors at low speed is a com
monly used practice. It is, however, very often the case that
such balanced rotors do not have the required running
smoothness after startup. To achieve low vibration amplitudes
and dynamic forces during operation, it is better to balance the
rotors at full operating speed after overspeed testing (high
speed or elastic balancing).
A bunker with thick concrete walls, as illustrated in
Figure 1, is required. This bunker must be sealed off, air-tight Figure 2. Control Room.
and evacuated to a vacuum level of approximately 5 mbar. 1. Polar diagram for bearing forces.
There are two main reasons for this. One is to keep the energy 2. Absolute value for bearing forces.
requirement at a low level; the other is to avoid non-allowable
3. Oscillograph for shaft vibrations.
heating resulting from ventilation losses. It is necessary to run
the rotor in its contract bearings in order to warrant a similar 4. Shaft vibration indicator.
vibrational behaviour as occurs in the machine. Also, in order 5. Recorders for speed, shaft vibrations, bearing housing
to detect existing residual unbalances and required corrective vibrations.
balancing weights, the indication of dynamic bearing forces as 6. Speed control.
to amplitude and phase in the rotating system is required. 7. Control and monitoring board for oil and air.
Measurement of shaft vibration must also be possible.
8. Monitor with test object.
The operation, monitoring and measurement are effected
by means of a closed circuit television transmission to a 9. Control for stroboscopic lighting of bunker.
separate control room (Figure 2). The installation, as shown, is 10. Window to testbed.

25
26 PROCEEDINGS OF THE NINTH TURBOMACHINERY SYMPOSIUM

suitable for balancing rotors up to 1.80 m in diameter, 6 m in BALANCING LIMITS

length and 15000 kg in weight, with speeds up to 22000 rpm. The correct approach to the balancing procedure is to
clarify beforehand the criteria of assessment and to determine
WHY IS HIGH SPEED the tolerances which are to be applied. Limits for residual
BALANCING NECESSARY? unbalances, U, for low speed balancing are specified in
international standards [1, 2, 3]. The test object has mass m,
When balancing at low speed, the rotor behaves as a rigid
thus the eccentricity of the center of gravity (c.o.g.) is U/m.
body. By applying corrective unbalance weights, Uc, at two
The speed of the c.o.g. is limited to a value of G. For turbo
planes selected at random (I and II in Figure 3a), specified
rotors, G is between 1 and 2.5 mm/s.
balance qualities can be achieved. The axial location of the
If w is the angular velocity at full speed, the following limit
residual unbalance, U0, remains unknown and is not neut
applies for low speed balancing:
ralized at the point of occurrence. If the rotor later deforms
elastically at high speed (Figure 3b), then the corrective
u
weights could be useless. The attempt is made to eliminate this w-<G (1)
disadvantage by incremental balancing during the assembly of m
the rotor.
Based on model laws, it can be shown that the dynamic stresses
in the rotor and the bearings are the same for all machine sizes
if this limit is adhered to. Defining n as the operating speed in
I Un II rpm, m in kg and U in g-mm, the following is derived from
equation (1) (see Figure 4):

9550 G m
U < ---- (2)
n

8,0

6,0
U /m
I

Qmm
l<g t
4,0

3,0
'

2,0
""
Figure 3. Low Speed Balanced Rotor.
1,5
a: Rigid b: Elastica[
G2,5
Each part is shrunk separately onto the shaft. Preferably
after each shrinkage operation, but at least after each second 1,0
""
one, the shaft is balanced at low speed. Corrective weights may
only be applied to the last parts which are mounted on the 0,8

shaft. These then can only be located on that part of the shaft to
coincide with the residual unbalances. 0,6

This procedure provides a reasonable level of balance, but
not 100% assurance for smooth running of the completely G=1
assembled rotor at full speed. It cannot be ruled out that some I"
parts shrunk onto the shaft will be displaced as a result of the 3000 4000 6000 8000 10000 1/min 20000
centrifugal force. There is then a change in the balance
condition as opposed to the condition at the start. If shrinkage Figure 4. Allowable Unbalance for Low Speed Balancing as
is inadequate at certain points, some parts will loosen or per ISO.
internal friction will result. A loosening of parts could result in
unbalances varying with time and could thus lead to vibrations
of a higher amplitude. Internal friction can cause instabilities, If the rotor behaves elastically, individual unbalances
meaning nensynchronous vibrations which also have higher distributed along the axial length lead to different deformations
amplitudes. It is not possible to carry out incremental balanc which depend on the speed. It is no longer sufficient to limit
ing on rotors which are manufactured in one piece, and such the magnitude of the unbalance according to the equations (1)
rotors are quite numerous. or (2). Neither is it possible to compensate the effect of existing
It is advisable to check and correct the rotor under those natural unbalance in two levels selected at random.
conditions at which it is later to be operated. This is the full The following values can be used for measurement criteria
operational speed. and tolerances when balancing rotors at high speeds:
 HIGH SPEED BALANCE PROCEDURE 27

Equivalent Unbalance [4] The disadvantage of this procedure is that vibrations and
forces occurring during the later operation at the site are not
Residual unbalances are distributed at random over the
axial length of the rotor. Deflections result at the rotor bearings directly limited. We do not apply this procedure due to its
at a certain speed, n, and take the form of vibrations or forces. obvious disadvantages. We have gained sufficient experience
These are measured in a coordinate system rotating with speed with the method described in the following sections.
and represented by the vector A in amplitude and phase
Shaft Vibration Amplitudes
(Figure 5a). Then, a test unbalance, UT, is attached, and its
effect depends on the flexural form of the rotor. Thus, the axial The objective of the balancing procedure is to obtain low
location ofUT is to be adjusted to the speed in question. amplitudes of vibration over the entire length of the shaft.
The size of UT is selected as per equation (2) with, for Limits for vibration amplitudes at the bearings can be found in
example, G 1. With the same speed as in the case of the first
= the standards [5, 6] and are also shown in Figure 6. In addition
measurement, a deflection results which is r3>resented by the to the size of the unbalances, the vibrations are influenced by
vector B (Figure 5b). The difference B - A is the effect of the bearing type and by the bearing supports as well. We use
the unbalanceUT on the deflection (Figure 5c). Then, as will the contract bearings for high speed balancing so that this
be shown later, corrective unbalance weigpts, Uc, are to be influeuce is precluded.
placed in such a way that the deflection ICI remains smaller We discovered, nevertheless, that the magnitude of the
than IB- AI (Figure 5d). By applying this procedure, only pure vibration amplitudes measured in the vacuum bunker does not
unbalances are compared to each other; thus having the always coincide with the measurements recorded on the
advantage of being independent of the boundary conditions testbed or in the plant. We do not use these vibrations when
resulting from the bearings and the measuring device. assessing balance quality. They are, however, subjected to
regular checking in order to detect any nonsynchronous
vibrations.

A Amplitude of Bearing Housing Vibrations [4]

If one observes the pure vibration amplitudes of the
bearing housings, one discovers that these are considerably
influenced by the bearing support stiffness. In order to
a compensate for this we apply the measured support spring
deflections as follows.

Dynamic Bearing Forces

First the vibration amplitudes of the bearing pedestals are
measured. Multiplied with the coefficients of the test springs
the bearing housing is supported on, the dynamic bearing
forces F result. These are exerted from the rotor to the
bearings.

b
With B as the bearing width and D as the bearing
diameter, the dynamic bearing compression is P F/(B D).
=

The applicable limit for use is,

P,;;; 0. 3 MPa (3)

This value approximately corresponds to the dynamic force for

each bearing,
( A
n
F ,;;; 0. 1 x mg -- N (4)
6000

where m is the rotor mass in kg, n the maximum operating

speed in rpm and g =9. 81 m/s2, the gravitation constant.
The smaller limit, as per equations (3) and (4), is applied as
the final limit for the high speed balancing procedure. This is
the assessment of measurement variables at full operating
speed. One obtains lower dynamic values if the bearing forces,
that have been measured during the low speed balancing
procedure, are extrapolated to the operational speed. Looking
back on the last ten years of experience in this field, we
d
x discovered that the required vibration values as shown in
Figure 6 were always met during plant test runs and startups,

CJ:js-Al provided of course, that the bearing forces in the bunker were
limited as per equations (3) and (4).
One disadvantage of the selected limitation is the occur
rence of a natural frequency resulting from the measuring
Figure 5. Equivalent Unbalance. device.
28 PROCEEDINGS OF THE NINTH TURBOMACHINERY SYMPOSIUM

50
I\ I I

m \ - Mpa
1\

r
40
I\
2A
1\
I \

30
1\ 1\
\

-n
5000 10000 1/min 15000

Figure 7. Change of Unbalance During Overspeed Testing.

20
cases, modal balancing is carried out in order to meet the limits
I" .
for the dynamic bearing forces.

3000 4000 6000 8000 10000 .

1/min 20000

Figure 6. Allowable Shaft Vibrations Acceptable to API. General Remarks on Balancing Procedures [7]
Elimination of the effects of the unbalance in any of the
axial planes of the rotor is done as follows:
In Figure 8a the natural unbalance is represented by
the bearing force X.
BALANCING AND The bearing force Y represents the natural force plus a
OVERSPEED TESTING test unbalance UT. The resulting vector Y X is the
-

First of all, every rotor is balanced at low speed. If effect of the test unbalance.
possible, incremental balancing is carried out as stated earlier. In order to compensate for the residual unbal
This is done to simplify the elastic balancing procedure. After ance, a corrective unbalance De is to be attached.
low speed balancing, the completely assembled rotor is
subjected to examination.
.
Its size is UT y , its direction is to be turned by the

angle oc as opposed to that ofDT (see Figure 8b).

Overspeed Testing
The rotor is accelerated to overspeed level. This speed is Balancing in the First Mode
15% above the maximum operating speed. In this case, 1. 8 This balancing is to be applied for speeds where the
MPa is the limiting value of the dynamic bearing compression, natural form of the rotor is predominantly determined by the
and this is six times the value of the final balancing procedure. first natural frequency. The test and/or corrective unbalance
Figure 7 shows, as an example for a compressor rotor, the weights shall be applied near the shaft center. This correction
behavior of the dynamic bearing compression versus the speed does, however, change the balancing quality achieved during
before and after the overspeed testing. A worsening of the the low speed balancing procedure. If this must be avoided,
balanced condition can be clearly seen here. This behavior is three unbalance weights shall be applied (Figure 8c) instead of
not always the case. Generally, unbalance changes only slightly the unbalance only in the shaft center. In addition to U in the
or not at all during the overspeed testing procedure. shaft center, two unbalances are attached near the bearings for
lf, after overspeed testing, the normal dynamic bearing test as well as for corrective unbalances. Their sizes are U/2
forces are not exceeded, the rotor is delivered. In all other each and their phase angle is at 180 to U [8].
 HIGH SPEED BALANCE PROCEDURE 29

BEARING 1

b BEARING 2
lx I
Uc= UTIY-xl

u
c
.
.x 13000
U/2 U/2
Figure 8. Compensation of Unbalance Effects.

Balancing in Higher Modes

This balancing procedure makes sense for speeds where
the mode of the rotor is determined by higher natural Figure 9. Polar Diagram of Bearing Forces in the Bunker.
frequencies. In this case, test and corrective weights are to be
attached in at least two planes.
Two additional planes are necessary if unbalance at low
speed balancing is to be limited in addition to limiting bearing
forces at full speed. In this case, a number of test runs are
necessary. Calculation of corrective weights based on the
results of the runs can only be carried out by sophisticated
formulas. There are programs for pocket calculators for this.

ACCOMPANYING PHENOMENA
Up to now, the principle of balancing was discussed.
There are, however, other accompanying phenomena which
nmax
are discussed as follows: --n
-
0 ------+---4-- ----------L
5000 10000 1/min 15000
Natural Frequency of the Measuring Device ------- Bearing 1
The bearing housing is spring .supported in order to
measure the bearing forces. Natural frequencies occur here
400 0,1mg. odo -------+---+-

which are independent of those of the rotor. The level of N

natural frequency depends on the bearing test stands used as 300
I \
well as on the weight of the rotor. Figures 9 and 10 compare
measuring results for the same rotor, in the bunker and on the
200-------------------4--1--
1 -----'----
testbed. The first natural frequency of the rotor is 7500 rpm;
the second is above the operating speed of 16500 rpm. The
amplitude increase at 13000 rpm is attributable to the bearing
stands of the bunker. It does not exist if the rotor is running in
the compressor.
Generally, we try to balance in such a way that, despite 5000 10000 1/min 15000
the natural frequency mentioned above, the limits of the ------Bearing 2
bearing forces can be met. The bearing forces, when the
machine is in full operation, are therefore usually smaller than Figure 10. Bearing Forces in the Bunker and Shaft Vibrations
those recorded in the bunker. During the Test Run.
 PROCEEDINGS OF THE NINTH TURBOMACHINERY SYMPOSIUM

!
30

1,0
Instabilities
0,8
The vibrations of the bearing pedestals and the resulting
0, 6
dynamic bearing forces are measured in a coordinate system
rotating with speed. Nonsynchronous vibrations are not 0,4
recorded. These can be caused by the oil film in the radial
0,3
bearings or through internal friction resulting from inadequate
ly shrunk parts. For this reason, we monitor the shaft
0, 2
vibrations near the bearings and sometimes also at the shaft
center in addition to the dynamic bearing forces. Instabilities
are indicated by high amplitudes with nonsynchronous fre
0,1
quencies.
0,08 800mm
As an example, Figure 11 shows the vibration records of
an instability. Synchronous vibrations with a low amplitude 0,08

occur up to 10000 rpm; at 10400 rpm there was a considerable

increase in the amplitudes. The vibrational frequency was less 0,04
o,. 100 200 300 400 500 600 700
than half of the rotational frequency. Mter a reduction of speed 40 80 120 160min 200
to 9500 rpm the instability had disappeared. ---0-1

Thermal Deflections Figure 12. Vibration Decay of Thermally Caused Shaft Deflec
tion with Respect to Time for Various Diameters. h0 initial
If turbomachinery rotors are not turned during cool
=

value at point of commencement zero.

down, different temperatures can result across the shaft
diameters. This leads to deflections and thus, subsequently, to
incorrect balancing results. One degree Celsius temperature
EXPERIMENTAL TESTING
difference over the diameter on the entire axial length leads to
a deflection in the range of 0.1 mm. This is about 100 times In addition to tests directly contract connected, we also
more than the tolerance on balancing allows. carry out experimental testing in our vacuum bunker.
It is therefore necessary to he sure of the steady state
conditions of the test values for every test and check run.
Unbalance Vibrations
Otherwise, the rotor is to be turned until such time as steady
state conditions have been achieved. Figure 12 shows the These measurements serve the purpose of studying the
gradual reduction of shaft bending caused by thermal condi sensitivity of the rotors, particularly within the natural fre
tions, over time t, for various diameters (9]. quencies. Figure 13 shows shaft vibrations measured on a

8,um

WVV\ \J

}J)JVJWMJ)J)JAA 13,um

LJ----0, 1_s======:3>1
I--E-
10000 10400 1/min 9500
Figure 11. Vibration Recordings.
 HIGH SPEED BALANCE PROCEDURE 31

50
1\
;Urn A X= e -tlt COS WE t
40
i 111/T

30

20

10 T

1,0
MPa n 8500 1/min
1\
0,8 p

0,6 i
0,4

0,2

20 40 60 "1 60 1/8 100

5000 10000 1/min 15000
Figure 13. Vibrations Measured at the Compressor Shaft with -n ngr

an Unbalance of G = 7.5.
Figure 14. Test of System Damping.

compressor rotor and dynamic bearing compression versus the

speed. A test unbalance was attached to the rotor as per
equation (2) with G 7.5.
= REFERENCES
1. Balance Quality of Rotating Rigid Bodies, International
Decay Function Standard ISO 1940.
Above a certain speed, rotors in journal bearings can 2. Beurteilungsmaflstiibe fur den Auswuchtzustand rotieren
become unstable, meaning, nonsynchronous vibrations can der starrer Korper, VDI 2060.
occur. This speed limit depends on the damping f.L of the
system which is made up of the rotor and the bearings. f.L is 3. US Navy Specification MTL-STD-167.
determined from the decay rate of the vibrations after a shock 4. Balancing Criteria for Flexible Rotors, ISO-Entwurf
excitation of the rotor. The quotient f.LN is the log decrement 8 Isorrc 108 1 sc 1 1 we 2.
and V is the natural frequency. f.L or 8 0 characterizes the
5. American Petroleum Institute Standards 612 and 617.
=

speed above which there are instabilities. Figure 14 shows

relevant test results. 6. Shaft Vibrations of Turbosets for Industrial Services,
Measurements and Evaluation, VDI 2059.
SUMMARY 7. Moore, L. S. and Dodd, E. G., "Mechanical Balancing of
High speed balancing of turbomachinery rotors provides a Large Rotors," Parsons Journal, 12, No. 69, (Summer
better assurance for later operation than low speed balancing. 1976), pp. 49-59.
The dynamic bearing forces have proven themselves as mea 8. K. Federn, "Grundlagen einer systematischen Schwing
surement criteria of high speed balancing. Additional natural ungsentstorung wellenelastischer Rotoren," VDI-Bericht
frequencies of the measuring device can lead to incorrect Bd. 24, 1957.
results. These influence the balancing operation only as
regards higher level accuracy. Shaft vibrations at the bearings 9. R. Ehrich, "Abklingen thermischer Durchbiegungen bei
are to be additionally monitored in order to detect nonsyn Teildrehzahl von Dampfturbinen," BWK, 17, No. 9,
chronous vibrations. Modal balancing can also be necessary in (1965), pp. 427-432.
order to maintain specified dynamic bearing forces. There 10. J. Glienicke, "Schwingungs- und Stabilitiitsunter
must be no differences in temperature along the rotor before suchungen an gleitgelagerten Rotoren," Motortechnische
each test or check run. Zeitschrift, 33, No. 4, (1972), pp. 135-139.