Sei sulla pagina 1di 9

Proceedings of the ASME 2009 International Design Engineering Technical Conferences &

Proceedings of the ASME 2009 International


Design
EngineeringinTechnical
Conferences
&
Computers
and Information
Engineering
Conference
Computers and Information in Engineering
Conference
IDETC/CIE
2009
August 30 - September 2, 2009, San Diego,IDETC/CIE
California,2009
USA
August 30 - September 2, 2009, SanDiego, California, USA

DETC2009-87836
DETC2009-87836

ONE-SHOT BALANCING OF RIGID ROTORS: A CLOSED FORM SOLUTION

Enrique S. Gutirrez-Wing
Departamento de Ingeniera Mecnica
Centro Nacional de Investigacin y Desarrollo Tecnolgico
Cuernavaca, Morelos, C.P. 62490
Mxico
Email: esgw@cenidet.edu.mx

Jorge E. Aguirre-Romano
Gerencia de Turbomaquinaria
Instituto de Investigaciones Elctricas
Temixco, Morelos, C.P. 62490
Mxico
Email: jear@iie.org.mx

ABSTRACT

FP1, FP2

This article presents a closed-form solution to the computation of


unbalance correction masses for rigid rotors mounted on flexible
bearings. The solution requires measurements of the response of the
rotor-bearing system to the unbalance that is to be corrected, as well
as knowledge of its total mass or other stiffness or damping
parameters that are readily deducible from the mounting
configuration. No detailed mathematical models or information from
previous balancing runs are required.

G
gij

Frequency-domain forces at the balancing


planes
Auxiliary computation matrix
Elements of matrix G

Vectorised version of matrix G

i
I
ICG

The computation is based on the relationships that exist between


some spatial and modal parameters of rotor-bearing systems. Such
relationships are derived in this article, and their use in the
computation of balancing masses is presented and verified with
results from an experimental test rig.

K
l
lCG
l1
M
mT
m1, m2, m3
mP1, mP2

This work demonstrates the feasibility of balancing rigid rotors


without trial runs, detailed models or historical balancing records.

NOMENCLATURE
Symbol
A
C
d
D1, D2
e
f
f1, f2
F
{F}PLANES

Definition

r
rP1, rP2

Coordinate transformation matrix


Damping matrix
Distance between balancing planes
Matrices of eigenvector normalisation
coefficients
Natural logarithm base
Vector of time-domain forces
Time-domain unbalance forces
Vector of frequency-domain forces at arbitrary
points
Vector of frequency-domain forces at the
balancing planes

t
U, V, W
uij, vij, wij
x
X
x1, x2

x
x

1
Identity matrix
Longitudinal moment of inertia of the rotorbering system with respect to its centre of
gravity
Stiffness matrix
Effective rotor length, i.e. distance between the
points on which the response of the rotor is
measured
Location of the centre of gravity
Location of left hand bearing
Mass matrix
Total mass of rotor-bearing system
Elements of matrix M
Correction masses to attach at the balancing
planes
Mode count index
Radial distances from the rotor axis to the
correction masses on each of the balancing
planes
Time
Auxiliary computation matrices
Elements of matrices U, V and W
Time-domain displacement vector
Frequency-domain displacement vector
Displacements
Velocity vector
Acceleration vector
Matrix of eigenvalues

Copyright 2009 by ASME


1, 2

1 , 2

H
T

Although one-shot balancing is the ultimate aim of any


balancing process, a mathematical model of a rotor is seldom readily
available for balancing tasks, and developing one is a considerably
time-consuming process. Furthermore, calibration of the model with
test data is likely to be required. Some balancing methods based on
the use of finite element and simplified models have been reported
[13, 14].
On the other hand, historical balancing records are not always
available and their usability is restricted to cases in which the
balancing setup can be exactly reproduced to match the conditions in
which the records were taken. Thus, although one-shot balancing of a
rotor is desirable due to the potential savings in resources that are
attainable through the elimination of trial runs, it is only feasible
when a reliable model of the rotor is available or when the behaviour
the rotor at hand is already known from previous tests. One-shot
balancing of a rotor of unknown dynamic characteristics is
impossible using the existing balancing schemes.
Although the balancing of rigid rotors is considerably simpler
than that of flexible rotors, it is by far the most common balancing
scenario [15]. Many of the balancing techniques applicable to
flexible rotors are applicable to rigid rotors. However, some
difficulties arising in the use of rigid-rotor methods on flexible rotors
have been pointed out in reference [16].
This paper presents a one-shot modal balancing method for
rigid rotors, based on some simple measurements such as the mass of
the rotor-bearing system at hand. Other data that are required are
merely geometrical, such as the distance between bearings and the
distances between the balancing planes and the bearings, which are
readily obtainable in the field. With the method that is presented
here, one-shot balancing of a rigid rotor is possible without using an
elaborate mathematical model or historical balancing test data.

State-space matrix of eigenvalues


Eigenvalues
Matrix of non-normalised eigenvectors
Non-normalised eigenvectors
State-space eigenvector matrix
Matrix of normalised eigenvectors
Frequency of excitation / vibration
Complex conjugation
Hermitian operator
Transpose operator

INTRODUCTION
Rotor unbalance is one of the most common causes of vibration
in rotating machinery. It has its origin in the uneven distribution of
mass around the rotation axis of a rotor, and is corrected by the
process of balancing.
Balancing is carried out by adding/eliminating masses to/from
discrete locations on the rotor. The magnitudes of these correction
masses, as well as their angular and longitudinal locations, are chosen
so that their effect on the vibration of the rotor is opposite to that of
the original unbalance, and thus cancels it out.
To determine these masses and their locations, conventional
methods based on influence coefficients [1-3], modal analysis
schemes [4-8] and combinations of both [9,10] require unbalance
response data from an initial run, in addition to data from one or
more trial runs. In the initial run the response of the rotor to the
unbalance forces that are to be cancelled-out is measured at one or
more speeds of interest, and at one or more stations along the rotor.
The trial runs are carried out to determine the effects of known
masses on the vibration of the rotor, with the aim of determining an
arrangement of masses that will bring this vibration down to within
acceptable limits [11].
In each trial run a known mass, or arrangement thereof, is
attached to the rotor and its unbalance response is measured. Using
linear relationships between the known trial masses and the effects
that they produce on the response of the rotor, it is possible to
determine a combination of such masses that will cancel out the
effects of the original rotor unbalance.
Theoretically, the number of runs that are required to determine
the balancing masses depends on: i) the number of stations along the
rotor and the number of rotating speeds at which the vibration is to be
reduced, ii) the number of modes that contribute to the observed
response of the rotor within the speed interval of interest, and iii) the
number of balancing planes on the rotor.

RIGID ROTOR BALANCING


Although a rotor may behave as a flexible body when mounted
on its bearings, it is common practice to balance a rotor on what is
known as a soft rig, in which the rotor is supported on flexible
bearings so that it vibrates in one plane, unbent, in response to
unbalance forces. In this type of balancing the rotor practically
behaves as a rigid body. Its vibration is due to the deformation of the
rig bearings, which in turn is caused by the unbalance forces that are
transmitted to them through the body of the rotor.
The dynamic analysis of rigid rotors takes advantage of some
simplifications that are not applicable to flexible rotors: i) only two
modes contribute significantly to the response of the rotor, ii) the
response at any point along the rotor may be determined from
response measurements made at any two reference points, and iii) the
resonances of the rotor-bearing system occur at low speeds of
rotation, so the resonant response data that is needed for modal
balancing methods is relatively easy to obtain, compared to the data
from the usually high-speed resonances of flexible rotors.
Rigid rotor balancing is usually performed as a first stage in the
balancing of flexible rotors, with the aim of reducing unbalance
forces so that higher rotation speeds are attainable once the rotor is
mounted on its actual bearings. A fine tuning stage of balance is then
possible, which is based on information from the flexible, highfrequency, modes of the rotor.
Since the response of rigid rotors is dominated by the
contributions of only two modes of vibration, only two balancing
planes are required to balance them.

Trial runs and one-shot balancing


Trial runs are performed to determine the effects that masses
attached to the different balancing planes have on the vibration of a
rotor. These effects vary according to the plane to which the masses
are attached. A balancing plane is said to be characterized when the
effect that a unit mass attached to it has on the vibration of the rotor
is known. In traditional balancing procedures, such as those based on
influence coefficients, as many trial runs are needed as there are
balancing planes to be characterized.
One shot balancing techniques do not require trial runs. The
unbalance correction masses are determined from the vibration
measured in the initial run of the rotor. These techniques use either
mathematical models [12-14] of the rotor at hand or historical
balancing data in order to characterize the balancing planes and
determine the correction masses.

Copyright 2009 by ASME

l
lCG

x1(t) , f1(t)

x2(t) , f2(t)

mT , ICG
C.G.

Bearing 2 equivalent
SDOF system

Bearing 1 equivalent
SDOF system

Figure 1 SCHEMATIC OF A RIGID ROTOR - SOFT BEARING RIG

m m2

M = 1
m m ,

3
2

DYNAMICS OF A RIGID ROTOR


Figure 1 shows a diagram of the typical setup of a rigid rotor on a
soft balancing rig. The rotor is supported on two bearings, which only
allow it to vibrate in one plane. Each of these bearings behaves as a
single-degree-of-freedom (SDOF) system with its own mass, damping
and stiffness characteristics. The rotor-bearing system has two DOFs
and hence exhibits two modes of vibration during the balancing runs.
The vibration of the system is caused by the unbalance forces that
ensue when the rotor spins. In general, unbalance forces are distributed
throughout the rotor but may be represented through two resultant
forces that act, for example, on each of its balancing planes [11].
The equation of motion of the rotor-bearing system shown in
Figure 1 is:
M { x ( t ) } + C { x ( t ) } + K { x ( t ) } = { f ( t ) }


Damping and stiffness matrices


The damping and stiffness characteristics of a rigid rotor-flexible
bearing system are determined by the supports and the bearings. For the
case in which the rotor has only two bearings and, as is usually the case,
the vibration of the system is measured at these components, the
damping and stiffness characteristics are represented by diagonal
matrices. If the rotor is supported by more than two bearings, or if the
vibration is measured at points of the system other than the two bearings
then the matrices are symmetric, but not diagonal. If the resultants of the
unbalance forces are computed for a coordinate system other than that
considered for the vibration measurements, then the damping and
stiffness matrices are generally non-symmetric.
In the following analysis, the DOFs of the rotor-bearing system
will be taken as the left and right ends of the rotor, and the resultants of
the unbalance forces will be computed for these locations. Through
further refinements these forces will be computed for the balancing
planes.

(1)

Mass matrix
With respect to the displacement coordinate system formed by x1
and x2, and to the force coordinate system formed by f1 and f2 , the mass
matrix of the rotor-bearing system shown in Figure 1 is:

1 mT ( CG ) + ICG

M = 2
m
CG ) ICG
T CG (

mT

CG

mT

CG

CG

) ICG

+ ICG

Eigenvalues and eigenvectors of the rotor-bearing system


During a balancing run, the rotor cruises slowly through its
resonances, achieving a steady-state of vibration at each rotation speed.
The unbalance forces are then harmonic functions of time, as is the
vibration of the system, and the frequency of both is equal to that at
which the rotor spins. The force and response vectors are represented in
the frequency domain as:
(5)
{ f ( t )} = { F ( )}ei t

(2)

This matrix satisfies the Equation of motion that the rotor-bearing


system would have if the support stiffnesses and dampings were null,
i.e. if:
M { x ( t )} = { f ( t )}

(4)

It will be shown in a later Section that this equation provides a


critical scaling parameter for the computation of unbalance correction
masses for a rigid rotor.

where M, C and K are real, symmetric, matrices of dimensions 2x2 that


represent the inertial, viscous damping and stiffness characteristics of
the system, respectively, { x(t) } = { x1(t) x2(t) }T is the time-domain
unbalance response vector and { f (t) } = { f1(t) f2(t) }T is the vector
of unbalance force components in the plane of vibration of the rotor.

mT = m1 + 2m2 + m3

{ x ( t )} = { X ( )}ei t

(3)

Using Equation (3) it can be deduced that the sum of the entries of
the mass matrix is equal to the total mass of the rotor-bearing system,
mT, i.e. :

(6)

Substituting these expressions in Equation (1), we obtain the


equation of motion of the system in the frequency domain [17]:
2M + i C + K { X ( )} = { F ( ) }

(7)

Copyright 2009 by ASME

For conciseness, in the rest of the paper the vectors X() and F()
will be expressed respectively as X and F.
The state-space representation of Equation (4) is [18]:

0
i

M M

C 0

i X 0
=


X
F

D12T + *D22 * H = M 1
D12 T + *D22 H = 0
1

(8)

D12 1T + *D22 ( * )

The modal properties of the system are obtained from the solution
to the following eigenproblem:

r M

M M

C 0


r r

0

= ,

r
0

1r 4

M ** 0 M

C
* 0 * 0

(9)

= 1
0

2

U =

** T

C
*

= I 0

0 I

0
0 **

=
*
*
K

( 22 )

{(

( 23 )

{(

2T ) H ( * ) H
( 24 )

1
{( T T ) H *H }
T

1T ) H ( * ) H

( 25 )

( 26 )

and thus the following expression is derived:


m2 = (u11g12 + u12g22 ) = (u21g11 + u22g21 )

( 27 )

or, alternatively:

( 13 )

{u21
where g = { g11 g21 g12 g22 }

( 14 )

u22 u11 u12 } g = 0

( 28 )

is a vectorized representation of G.

Similarly, using Equations (20) and (21), the following equations


on the unknown entries of G may be established:

{v21
{w21

Using Equations (12)-(14) the following relationships that exist


between the modal and spatial property matrices may be established:
2

( 21 )

u11 u12 g11 g12 m1 m2

u
m m
u g
g
3
21 22 21 22 2

RELATIONSHIPS BETWEEN SPATIAL AND MODAL


PARAMETERS

D122 T + *D22 ( * ) H = M 1CM 1

WG = K

( 12 )

Using this definition, a set of normalized eigevectors can be


determined that satisfies the bilinear transformations that result from the
following matrix equations:
*

( 20 )

The matrices U, V and W can be determined from the modal


analysis of the unbalance response of the rotor-bearing system measured
during its initial run. In contrast, G cannot be determined from these
measurements since it is a function of the unknown elements of matrix
D1.
The matrices U ,V ,W ,G ,M ,C , and K in Equations (22)-(25) are
of dimensions 2x2. An analysis of each of these equations will now be
carried out with the aim of determining the elements, gij , of the matrix
G.
If the entries of each of the matrices in Equation (19) are written
explicitly, this equation may be expressed as:

( 11 )

It is possible to determine multipliers for each of the eigenvectors,


and to assemble these multipliers in diagonal matrices, D1 and D2, such
that a normalized version of the eigenvectors is obtained through [17]:

T
*

VG = C

{( T T ) H * H }

W =

Normalized eigenvectors

( 19 )

1
V = {( T T ) H *H }

where 1 , 2 , 1 and 2 are scalars.

* *
**

= D1 0

* 0 D2

( 18 )

UG = M

G = ( D12 T

This equation is only valid for real M, C and K matrices, as is the


case here. The fact that the eigenvalues and eigenvectors occur in
complex conjugates is evidenced by the structure of the eigenvalue and
eigenvector matrices.
For a 2DOF rotor-bearing system, the submatrix blocks and
may be defined as:
,

H = K 1

where:

0 0
0 **

( 10 )
=
*
0 0
K

1 1

=


2
1

( 17 )

where [0] represents a 2x2 matrix of zeros. These relationships can be


written in the following form:

The system possesses four sets of eigenvalues and eigenvectors,


which form two complex conjugate pairs due to the fact that the system
matrices are real. The modal parameters may be assembled into a matrix
so that the following equation can be established in order to relate the
spatial and modal parameters of the rotor-bearing system:
0

( 16 )

v22 v11 v12 } g = 0

( 29 )

w22 w11 w12 } g = 0

( 30 )

The last three equations form an underdetermined, homogeneous,


system in which the unknowns are the four entries of G. One further
equation is necessary to establish a system with a unique solution. This
equation comes from Equation (4), which expresses that the total mass

( 15 )

Copyright 2009 by ASME

of the rotor-bearing system is equal to the sum of the elements of its


mass matrix. Using this equation together with Equation (26) the
following expression may be derived:
m1 + 2m2 + m 3 = (u11g11 + u12g21 ) + 2(u11g12 + u12g22 ) +

(u21g12 + u22g22 ) = mT

coordinate system. If it is required to compute a set of equivalent forces


which acting on the balancing planes would produce the observed
response, then a transformation matrix, [A], must be used:

{ F }PLANES

( 31 )

u12

2u11 + u21 2u12 + u22 } g = mT

( 32 )

This equation, together with Equations (28)-(30), form the


following linear system, the solution of which yields the elements of G:
u21 u22
u11
u12 g11 0

v
v

v
v12 g21 0
21
22
11

w11
w12 g12 0
21 w22


u
u12 2u11 + u21 2u12 + u22 g22 mT
11

( 33 )

Equation (8) is the frequency-domain form of the equation of


motion of the rotor-bearing system. To aid in the following derivations,
it is repeated here without modification as Equation (34):
0

i X 0


X = F

* *

1 T
0
i I
F

0
i I * H F
*

1 * ( i I * )

*1 G 1 { F }

(34)

{ F }PLANES

1 * ( i I * )

1
*1 { X }

( 39 )

1
d

+d

+d

( 40 )

2m r
r
0
m

p1
= 2 p1 p1
= 2 p1
( 41 )
m
mp2rp2
0
r
p2
p2 UNBALANCE

UNBALANCE

( 35 )
The correction masses may then be computed as:
1
m p1

1 rp1 0

=
A G

2 0 rp 2
mp2
CORRECTION

1
1
( i I ) 1 * ( i I * ) *1 { X }

( 36 )

( 42 )

where a negative sign has been added so that the correction masses
oppose the effects of the original unbalance forces.

or, alternatively:

{ F } = G ( i I

F
= P 1
FP 2

Lastly, the computation of unbalance correction masses must take


into account the fact that the centrifugal forces are proportional to the
square of the speed of rotation, , and the radial distances, rp1 and rp2,
from the central axes of the balancing planes to the locations at which
balancing masses are to be attached. These parameters are related in the
following way:

Using the relationships between the normalized and nonnormalized versions of the eigenvectors given in Equation (12), and the
definition of the matrix G given in Equation (22), the relationship
between the unbalance forces and the response may be expressed as:

{ X } = ( i I

[A] =

Using the relationships between the modal and spatial parameters


of the system given in Equations (13) and (14), the following expression
that relates the response to the excitation forces, i.e. the unbalance
forces, is derived:

i X =

then [A] is derived by determining the matrix that relates the force
vectors ( F1, F2 ) and ( FP1, FP2) under the condition that both pairs of
forces yield the same resultant total force and moment applied to the
system. For example, in the case of the rotor-bearing system shown in
Figure 2, the matrix [A] is defined as:

COMPUTATION OF CORRECTION MASSES

M M

C 0

{ X } = X1 , { F } = F1 , { F }PLANES

As it will be shown in the next Section, the G matrix forms the


basis for the computation of the unbalance correction masses for the
rotor-bearing system.

0
i

( 38 )

In this equation, the entries of [A] express the relationship that


exists between the representations of the unbalance forces in two
coordinate systems: that whose base vectors are unit forces applied at
the points in which the vibration response is measured and that whose
base vectors are unit forces applied at the balancing planes. If the
vibration is measured at the balancing planes then the transformation
matrix is an identity matrix. If the following definitions of the force and
response vectors are used:

or, alternatively:

{u11

1
1
= A G ( i I ) 1 * ( i I * ) *1 { X }

Additional equations

( 37 )

There are cases in which Equation (33) becomes underdetermined


due to the singularity of the coefficients matrix. This is the case, for
example, of systems with proportional damping, for which Equations
(19)-(21) become linearly dependent because of the fact that the
damping matrix is a linear combination of the mass and stiffness
matrices, thus making Equation (20) linearly dependant of the other two.
For such systems, additional Equations are required in order to compute
the unbalance correction masses. These equations may be obtained
based on the construction of the rotor-bearing system at hand. Two cases
are noteworthy because they represent the common set-up for rigid-rotor
balancing:

In this equation, the matrices of eigenvalues and non-normalized


eigenvectors are determined from the initial unbalance response of the
rotor, whereas the matrix G is computed by means of the procedure
described in the previous Section. Therefore, if the vector of unbalance
responses, {X}, is known for each frequency, , the magnitude of the
unbalance forces, {F}, can be determined directly as a function of the
frequency of rotation.
One should bear in mind that the derivation of the previous
equation is based on the assumption that the displacements of the
system and the forces acting on it were defined with respect to the same

Copyright 2009 by ASME

FP1

FP2

X1, F1

X2, F2

rp1

rp2

l1

d
l

Figure 2 ALTERNATIVE FORCE COORDINATE SYSTEMS

shape vectors are real, and that the damping matrices of systems with
real mode shapes may be suitably approximated by a linear combination
of the mass and stiffness matrices.

i) Systems with a diagonal damping matrix. For these systems, Equation


(20) may be used to replace Equation (29) with:

{v21

v22

0 0} g = 0

( 43 )

EXPERIMENTAL VERIFICATION

and by adding this expression to Equation (33) the rank of the system is
increased by one, so a unique solution to it can be determined.

Description of the experimental rig

ii) Systems with a diagonal stiffness matrix. The fact that the offdiagonal elements of the stiffness matrix are null can be expressed
through the following relationship, taken from Equation (21):

{w21

w22

0 0} g = 0

For the experimental verification, the rig shown in Figure 3 was


used.
The rotor is supported on roller bearings, which in turn are
mounted on flexible supports. Each of the supports consists of a steel
beam that is fixed at one end and is supported by rubber mounts on the
opposite end. The bearings only allow the rotor to vibrate vertically.
The rotor is spun by a d.c. motor with an operating speed range of 0-900
r.p.m. The speed of the rotor is controlled through the voltage fed to the
d.c. motor.
The rotor consists of a 660 mm long, 12.5 mm diameter shaft,
onto which are attached two discs which act as balancing planes. The
rotor with bearings weights 4425 gr. Each of the disks has bored holes
drilled around a circumference close to its outer edge, as shown in
Figure 4. The radius of this circumference is 45 mm for both disks. The
rotor rotates counter-clockwise when viewed from the side of the
nearest disk in Figure 4, which is the left disk in Figure 3. The
tachometer is placed on the left side of the rotor, when viewed as in
Figure 4.
In Figure 3, the distance between the bearings is 310 mm, that
from the right bearing to the right disk it is 90 mm and that from the
right bearing to the left disk it is 260 mm. Due to the slow speeds
attainable with the d.c. motor, the only resonances which are crossed
during operation of the rig are the two rigid-body modes of the rotor on
the flexible bearings.
The response of the system is measured using two accelerometers,
respectively mounted on top of each of the bearing housings. Upward
accelerations are of positive sign. A tachometer is used to detect the
passing of a mark placed on the outer rim of the right disk, giving
pulses of 12V each time the marker passes in front of it. The passing of
this marker is used for the computation of the speed of the rotor and is
also used as a phase reference for the tracing of the response diagrams.

( 44 )

The importance of these two cases lies in the fact that in soft
balancing rigs the response of the rotor is usually measured at its
supports, and that these are the only elastic elements in contact with the
rotor. Therefore, the principal directions of stiffness coincide with the
directions of the displacements of the rotor at the supports and the
stiffness matrix becomes diagonal for this coordinate frame. The same is
true for the damping matrix: since damping of the rotor-bearing system
only comes from the supports because the rotor is rigid- the principal
directions of damping coincide with the displacements of the supports.
The damping matrix is then diagonal for a coordinate frame whose base
vectors match the displacement vectors of the supports.
There may be cases where the response is not directly measured at
the supports, and in these the stiffness matrix is not necessarily diagonal.
However, for a rigid rotor it is always possible to determine the
displacements of the supports when those at any two reference points are
known, using a simple coordinate transformation.
Since the response of the system is usually measured at the
supports, Equations (43) and (44) represent the norm rather than the
exception for soft-balancing rigs, and so it is possible to compute the
unbalance correction masses even when Equations (19)-(21) are linearly
dependent. Other information, such as knowledge of the supportstiffnesses or dampings and in general of any element of the mass,
damping and/or stiffness matrices may be used to set-up additional
equations for the computation.
The following experimental example illustrates the use of the
algorithm that was described in previous Sections to compute the
unbalance correction masses for a rigid rotor with, assumed, proportional
damping. The assumption is based on the fact that both of the mode

Copyright 2009 by ASME

Figure 3 Two-DOF BALANCING RIG


Figure 5 UNBALANCE RESPONSE AMPLITUDES (DISPLACEMENT).

Experimental procedure
The objective of the experiments was to evaluate the accuracy with
which the unbalance of the rotor could be determined using the algorithm
described in the previous Sections.
In order to achieve this, the rotor was firstly balanced using
traditional modal methods. Once the response was small enough to be
undistinguishable from instrumental noise, an unbalance mass of 4.1 g
was attached to the right disk at a location 90 degrees from the
tachometer mark in a direction opposite to that of rotor spin.
The response (displacement) is plotted in the Bode diagram
presented in Figure 5. Figure 6 presents the polar response plot for each
of the accelerometers. From Figure 6 it can be deduced that the
unbalance force passes through the upward direction at 0 degrees, at the
instant in which the tachometer mark is detected, since this is the
direction in which positive displacements are measured. This matches the
location of the known unbalance mass.
Modal analysis of the response using curve-fitting algorithms [17]
yields the values for the modal parameters shown in Table 1.
Since the mode shapes are real, it is assumed that the damping is
proportional. The coefficients matrix in Equation (33) is rank deficient.
Taking into account the fact that the response is measured in the
directions of principal stiffness, Equation (44) is used instead of Equation
(30) to compute the coefficients matrix. Equations (33) and (44) yield the
following system of equations in the elements of G, gij:
4.38 104 69.4 104 62.1 104
0

g11

2.80 104


4
4
4

361 10
16.4 104
404 10
0

g21

25.6 10

=
i

404 104 16.4 104


0
0
0

g12

62.0 104 2.80 104 129 104 75.0 104


4425

g22

Figure 6 POLAR RESPONSE PLOT: DISPLACEMENT [m],


o
RESPONSE LEAD [ ].

Table 1 MODAL PARAMETERS OF THE ROTOR-BEARING SYSTEM

MODE,
r

(45)

1
2

UNDAMPED
NATURAL
FREQUENCY,
r [Hz]

VISCOUS
DAMPING
RATIO,

11.23
13.12

0.0219
0.0291

MODE SHAPE,

{ }r

{1.0 2.0}
{1.0 -0.32 }

Solution of this Equation (45) yields the G matrix:


2.185 0.1545
5

G =
0.1545 0.1545 10 i

(46)

Using the distances between bearing and disks, the coordinate


transformation matrix, [A], is:
1.529 0.294

A =
0.529 1.294

(46)

Figure 4 SIDEVIEW OF SUPPORTS AND BALANCING DISKS.

Copyright 2009 by ASME

With these data, Equation (42) can be used to determine the


unbalance correction masses for the response data at each frequency of
rotation, . The result is the plot presented in Figure 7.

accuracy, without the need for balancing runs or detailed


geometrical/dynamical models.
It was demonstrated that the total mass of the rotor, as well as
knowledge of some elements of the damping and/or stiffness matrices
can be used to set-up a system of linear equations that leads to the
computation of the unbalance masses.
As in any balancing method, the more information that is gathered
about the rotor prior to the balancing task, the best chances one has of
minimizing the number of balancing runs. It has been shown here that it
is theoretically possible to balance a rigid rotor on soft supports if the
total mass of the rotor-bearing system is known. Exceptions are
proportionally-damped systems, but these may be dealt with by using
the algorithm described here, under the condition that the response
measurements be made, or be transformed to, a coordinate frame in
which the damping and/or the stiffness matrices are diagonal.

ACKNOWLEDGEMENTS
The authors acknowledge the General Direction of Higher
Technological Education (DGEST) of Mexico for supporting this work
through research projects UR.CEN2-07 and 534.07-P.

REFERENCES

FIGURE 7 COMPUTED BALANCING MASSES, CONTINUOUS LINE =


RIGHT DISK, DOTTED LINE = LEFT DISK

[1]

T.C. Rathbone, Turbine vibration and balancing, Transactions


of the American Society of Mechanical Engineers, Part I, no.
51, 1929, pp. 267284.

As it can be seen in Figure 7, the computed correction masses


change with frequency. There are several possible causes for this:

[2]

i) Errors in response measurements. Random and systematic errors due to


noise or other experimental factors have an impact in the observed
response, as seen in Figure 5. The erratic appearance of the response
curve is evidence that the response measurements present a significant
amount of measurement noise. These errors affect the computation of
correction masses with Equation (42).

E.L. Thearle, Dynamic balancing of rotating machinery in the


field, Transactions of the American Society of Mechanical
Engineers, Journal of Applied Mechanics, no. 56, 1934, pp.
745753.

[3]

T.P. Goodman, A least-squares method for computing balance


corrections, Transactions of the American Society of
Mechanical Engineers, Journal of Engineering for Industry,
vol. 86, no. 3, 1964, pp. 273279.

ii) Errors in the identified modal parameters. These have their origins in
measurement errors which result in less-than-perfect curve-fitting results.

[4]

R.E.D. Bishop, The vibration of rotating shafts, Journal of


Mechanical Engineering Science, vol. 1, no. 1, 1959, pp. 50
65.

[5]

R.E.D. Bishop, G.M.L. Gladwell, The vibration and


balancing of an unbalanced flexible rotor, Journal of
Mechanical Engineering Science, vol. 1, no. 1, 1959, pp. 66
77.

[6]

A.G. Parkinson, K.L. Jackson, R.E.D. Bishop, Some


experiments on the balancing of small flexible rotors: Part I
Theory, Journal of Mechanical Engineering Science, vol. 5,
no. 1, 1963, pp. 114128.

[7]

A.G. Parkinson, K.L. Jackson and R.E.D. Bishop, Some


experiments on the balancing of small flexible rotors: Part II
Experiments, Journal of Mechanical Engineering Science,
vol. 5, no. 2, 1963, pp. 133145.

[8]

A.G. Parkinson, An introduction to the vibration of rotating


flexible shafts, Bulletin of Mechanical Engineering Education,
no. 6 1967, pp. 4766.

[9]

Parkinson, A. G., Smalley, A. J., Badgley, R. H., Darlow, M.


S., An introduction to a unified approach to flexible rotor
balancing, American Society of Mechanical Engineers, Gas
Turbine Conference and Exhibit and Solar Energy

DISCUSSION

iii) Non-linearity. It is possible that the soft rubber supports installed


below the flexible beams be non-linear.
In spite of errors in the computation of the correction masses, the
obtained results are close to the expected values of 4.1gr@180o in the
right disk and 0 gr in the left disk. As in standard balancing procedures,
the angular location of the balancing masses is measured from the
vibration transducer, with the tachometer mark placed in front of its
detector. That means that in the experimental example discussed earlier
the computed balancing masses should be placed opposite to the known
unbalance mass on the right disk (at the bottom-side of the disk with the
tachometer mark lined-up with the detector) and on the top side of the
left disk with the same mark alignment.
This second predicted mass is erroneous, as it is known that the
unbalance was placed on the right disk only. However the predicted
correction masses are close to the correct values, thus demonstrating the
applicability of the algorithm to practical situations.

CONCLUSIONS
It has been shown that knowledge of the mass of a rigid rotor
allows one-shot balancing may be carried out with a good degree of

Copyright 2009 by ASME

Conference, Mar. 12-15, 1979, San Diego, CA


[10]

Tan, S.G. and Wang, X.X., A theoretical introduction to the


low speed balancing of flexible rotors: unification and
development of the modal balancing and influence coefficient
techniques, Journal of Sound and Vibration, vol. 168, no.3,
1993, pp.385-394.

[11]

Rao, J.S., Rotor Dynamics, Section 12.3 - Balancing of rigid


rotors, 3rd edition, New Age International, 1996, pp. 369-375

[12]

J.V. LeGrow, Multiplane balancing of flexible rotors A


method of calculating correction weights, Paper No. 71-Vibr52, Toronto, Canada, September 1971.

[13]

Preciado-Delgado, E., Bannister, R.H., Balancing of an


experimental rotor without trial runs, International Journal of
Rotating Machinery, vol. 8, 2002.

[14]

Aguirre-Romano, J.E., Cervantes-Contreras, J., GutierrezWing, E.S., Modal balancing of turbo-generator rotors using
simplified spatial models, Proceedings of the ISMA
International Conference on Vibration and Noise Engineering,
Leuven, Belgium, 18-20 September 2006.

[15]

Wowk, V., Machinery vibration: Balancing, Ch.2: History of


balancing, McGraw-Hill, Inc., 1995.

[16]

Darlow, M.S., Balancing of high-speed machinery: theory,


methods and experimental results, Mechanical Systems and
Signal Processing, vol. 1, no.1, 1987, pp. 105-134.

[17]

Ewins, D.J., Modal Testing: Theory, practice and application,


2nd edition, Ch. 2: Theoretical Basis, Research Studies Press
LTD, Hertfordshire, 2000.

[18]

He, J., Zhu, Z.F., Modal Analysis, Butterworth-Heinemann,


Ch. 6 - Modal Analysis of a damped MDoF System, Oxford,
2001, pp. 123-139.

Copyright 2009 by ASME

Potrebbero piacerti anche