Estimating Unbalance and Misalignment of A Exible Rotating Machine From A Single Run-Down

ARTICLE IN PRESS
JOURNAL OF
SOUND AND
VIBRATION
Journal of Sound and Vibration 272 (2004) 967989
www.elsevier.com/locate/jsvi
Estimating unbalance and misalignment of a exible rotating

machine from a single run-down
J.K. Sinhaa, A.W. Leesb, M.I. Friswellc,*
a
Vibration Laboratory, Reactor Engineering Division, BARC, Mumbai 400 085, India
b
School of Engineering, University of Wales Swansea, Swansea SA2 8PP, UK
c
Department of Aerospace Engineering, University of Bristol, Bristol BS8 1TR, UK
Received 6 December 2002; accepted 24 March 2003
Abstract
Earlier studies have suggested that the reliable estimation of the state of unbalance (both amplitude and
phase) at multiple planes of a exibly supported rotating machine from measured vibration data is possible
using a single machine run-down. This paper proposes a method that can reliably estimate both the rotor
unbalance and misalignment from a single machine run-down. This identication assumes that the source
of misalignment is at the couplings of the multi-rotor system, and that this will generate constant
synchronous forces and moments at the couplings depending upon the extent of the off-set between the two
rotors, irrespective of the machine rotating speed. A exible foundation model is also estimated. The
method is demonstrated using experimental data from a machine with two bearings and a exible coupling
to the motor. A sensitivity analysis has also been carried out for the proposed approach with perturbation
errors in the rotor and bearing models, to conrm the robustness of the method.
r 2003 Elsevier Ltd. All rights reserved.
1. Introduction
Many rotating machines, such as power station turbogenerators, may be considered as
consisting of three major parts; the rotor, the bearings (often uid bearings) and the foundations.
In many modern systems, the foundation structures are exible and have a substantial inuence
on the dynamic behaviour of the complete machine. These rotating machines have a high capital
cost and hence the development of condition monitoring techniques is very important. Vibrationbased identication of faults, such as rotor unbalance, rotor bends, cracks, rubs, misalignment,
*Corresponding author. Tel.: +44-117-928-8695; fax: +44-117-927-2771.
E-mail addresses: [email protected] (J.K. Sinha), [email protected] (A.W. Lees),
[email protected] (M.I. Friswell).
0022-460X/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jsv.2003.03.006
ARTICLE IN PRESS
968
J.K. Sinha et al. / Journal of Sound and Vibration 272 (2004) 967989
uid induced instability, based on the qualitative understanding of measured data, is welldeveloped [1] and widely used in practice. However the quantitative part, the estimation of the
extent of faults and their locations, has been an active area of research for many years. Over the
past 30 years, theoretical models have played an increasing role in the rapid resolution of
problems in rotating machinery. Doebling et al. [2] gave an extensive survey on the crack detection
methods. Parkinson [3] and Foiles et al. [4] gave comprehensive reviews of rotor balancing.
Muszynska [5] gave a thorough review of the analysis of rotorstator rub phenomena. Edwards
et al. [6] gave a brief review of the wider eld of fault diagnosis. The study of the rotor
misalignment has been limited to qualitative understanding of the phenomenon [713] but no-one
has suggested a method for the direct quantication of misalignment faults.
In spite of the success of model-based estimation of faults, the construction of a reliable
mathematical model of a machine on exible foundations is still not feasible. Often a good nite
element model of the rotor and an adequate model of the uid bearing [14] may be constructed.
Indeed several nite element based software packages are available for such modelling. However a
reliable nite element model for the foundation is difcult, if not impossible, to construct due to a
number of practical difculties [15]. Inclusion of the foundation model is very important since the
dynamics of the exible foundation also contributes signicantly to the dynamics of the complete
machine. Many research studies [1622] have been carried out to derive the foundation models
directly from the measured machine responses but more research is needed on their practical
application. Hence a complete mathematical model of a machine is still not available for
condition monitoring in many cases.
Given the above limitations, an alternative method has been suggested by Lees and Friswell [23]
for the reliable estimation of multi-plane rotor unbalance. Unbalance is one of the most
frequently encountered problems in rotating machines that needs to be corrected at regular
intervals with the minimum possible down time. The method of Lees and Friswell [23] has the
potential for fast and reliable unbalance estimation. The method uses the measured vibration data
during a single run-down (one transient operation only) of a machine, along with models of only
the rotor and bearings, to estimate the multi-plane unbalance in a single step. The foundation
parameters are estimated as a by-product, and hence the method also accounts for the dynamics
of the foundation. Lees and Friswell [23] demonstrated the method on a simple simulated
example. The method has been further validated on a small simple experimental rig [24]. In both
cases the number of modes of the system were less than the measured degrees of freedom in
the run-down frequency range. However in many systems, for example turbogenerator sets, the
number of modes excited may be more than the measured degrees of freedom and thus
the estimated unbalance may not account for all of the critical speeds. The identication method
has been modied further by splitting the whole frequency range into bands so that the banddependent foundation models account for all of the critical speeds. The advantages of the
suggested approach have been demonstrated on a complex simulated example and on an
experimental rig [2527].
However, in a multi-rotor system there is always the possibility of rotor misalignment and this
may inuence the machine response and hence the quality of the unbalance estimation. In this
paper, the method described above has been modied to estimate both the rotor unbalance and
misalignment. The misalignment is assumed to affect only the rotor, whereas for systems with
uid bearings and rigidly coupled shafts the misalignment will also affect the distribution of static
ARTICLE IN PRESS
969
loads in the uid bearings. The general perception and observation is that misalignment in multicoupled rotors generates a 2 (twice the rotating speed) component in the response of the
machine [8,9] and the effect on the 1 component is assumed to be small. However, Jordan [28]
conrms that the misalignment initially affects the 1 response resulting in an elliptical orbit, but
in the case of severe misalignment the orbit plot may look like a figure eight due to the appearance
of a 2 component in the response. These features are usually used for the detection of the
presence of rotor misalignment [28]. In practice both the 1 and the 2 response will be affected
by misalignment, and the physical source of these effects may be modelled as a rotor bend
and rotor asymmetry, respectively. In this paper the identication of rotor unbalance and
misalignment uses the 1 response at the bearing pedestals of the machine from a single rundown, even though the direct inuence on the 1 response due to misalignment may be small.
Here it is assumed that the source of the rotor misalignment is the couplings of the multi-rotor
system. Such a misalignment is assumed to generate constant synchronous forces and moments
at the couplings depending upon the extent of the off-set between the two rotors and irrespective
of the machine rotating speed [7]. The force vector in the equation of motion is assumed to consist
of both the unbalance forces and the constant forces and moments at the couplings, and these
parameters are estimated along with the foundation model. The method is demonstrated using
experimental data from a machine with two bearings and a exible coupling to the motor. In a
real machine, accurate models of the rotor and uid bearings may not be available. Thus a
sensitivity analysis for the proposed approach has been performed, with perturbation errors in the
rotor and uid bearing models, to conrm the robustness of the method.
2. Theory
Fig. 1 shows the abstract representation of a rotating machine, where a rotor is connected to a
exible foundation via exible bearings. The equations of motion of the system may be written
Fig. 1. The abstract representation of a rotating machine.
ARTICLE IN PRESS
970
[1,26] as
ZR;ii
6
4 ZR;bi
0
ZR;ib
ZR;bb ZB
ZB
9 8 9
38
r
>
>
R;i
< fu >
=
<
= >
7
ZB 5 rR;b 0 ;
>
> > >
% F : rF;b ; : 0 ;
ZB Z
0
where Z is the dynamic stiffness matrix, the subscripts i and b refer to internal and bearing
(connection) degrees of freedom, respectively, and the subscripts F ; R; and B refer to the foundation,
the rotor and the bearings. r are the responses and f u are the force vectors, which are assumed to be
applied only at the rotor internal degrees of freedom. The dynamic stiffness matrix of the foundation,
Z% F ; is dened only at the degrees of freedom connecting the bearings and the foundation. In practice
this will be a reduced order model, where the internal foundation degrees of freedom have been
eliminated [1]. The dynamic stiffness matrix of the bearings is given by ZB : It has been assumed that
the inertia effects within the bearings are negligible, although these could be included if required.
Solving Eq. (1) to eliminate the unknown response of the rotor gives
% F rF ;b ZB P1 ZB IrF ;b ZB P1 ZR;bi Z1
2
Z
R;ii f u ;
where P ZR;bb ZB ZR;bi Z1
R;ii ZR;ib : It is assumed that good models for the rotor and the
bearings, ZR and ZB ; are known a priori and rF ;b is the vector of measured responses during a
single run-down of a machine. Thus, the only unknown quantities in Eq. (2) are the foundation
% F ; and the force vector, f u : The force vector may be dened as
model, Z
f u f un f m ;
where f un is the vector of unbalance forces and f m is the vector of forces and moments at the
couplings.
2.1. Modelling unbalance forces
Although the unbalance will be distributed throughout the rotor, this is equivalent to a discrete
distribution of unbalance, provided there are as many balance planes as active modes. Suppose
the unbalance planes are located at nodes n1 ; n2 ; y; np ; where p is the number of planes. The
associated amplitude of unbalance (dened as the unbalance mass multiplied by distance between
the mass and geometric centres) and phase angles are un1 ; un2 ; y; unp T and yn1 ; yn2 ; y; ynp T ;
respectively. These amplitudes and phase angles can be expressed, for the ith balance plane, as the
complex quantity uni exp jyni er;ni jei;ni : Hence, the unbalance forces, f un ; in the horizontal
and vertical directions, can be written as [1,26]
f un o2 Te;
where e er;n1 er;n2 ?er;np ei;n1 ei;n1 ei;n2 ?ei;np T and T is a selection matrix indicating the location of
the balance planes.
2.2. Modelling misalignment forces and moments
It has been assumed that the rotor misalignment occurs at the couplings between the multirotors. The nature of the rotor misalignment could be parallel, angular or combined as shown in
ARTICLE IN PRESS
971
Fig. 2, but all of them would generate forces and/or moments. Suppose that there are c couplings
in the rotor located at nodes m1 ; m2 ; y; mc : The associated amplitude of the forces and moments
are em fz;m1 fy;m1 My;m1 Mz;m1 fz;m2 fy;m2 ? fz;mc fy;mc My;mc Mz;mc T ; where the
subscripts y and z represent the horizontal and vertical directions and f and M are forces and
moments, respectively. Hence, the misalignment force, f m ; can be written as [1]
f m Tm em ;
where Tm is the transformation matrix indicating the location of the couplings [1].
The proposed method will estimate the forces and moments due to the rotor misalignment.
However, it is often more intuitive to consider displacements and angles rather than forces and
moments, although this requires the stiffness matrix of the coupling. Suppose that the stiffness
of the ith coupling is Kc;i ; then the linear misalignment, Dyi and Dzi ; and the angular
misalignment, Dyy;i and Dyz;i ; at the ith coupling in the horizontal and the vertical directions may
be calculated as [1]
8
8
9
9
Dzi >
fz;i >
>
>
>
>
>
>
>
>
< Dy >
< f >
=
=
i
y;i
Kc;i 1
:
6
>
>
Dyy;i >
My;i >
>
>
>
>
>
>
>
>
:
:
;
;
Dyz;i
Mz;i
Fig. 2. Schematic of rotor with misalignment at a coupling.
ARTICLE IN PRESS
972
2.3. Parameter estimation

Substituting Eqs. (4) and (5) into Eq. (2) produces
( )

e
2
% F rF ;b ZB P1 ZR;bi Z1
Z
ZB P1 ZB IrF ;b :
R;ii o T Tm
em
To identify the foundation parameters and forces in a least-squares sense, the foundation
% F;
parameters are grouped into a vector v: Suppose that the foundation dynamic stiffness matrix, Z
is written in terms of mass, damping and stiffness matrices. If there are n measured degrees-offreedom at the foundation-bearing interface, then v will take the form
8
v k%F ;11 k%F ;12 ? k%F ;nn c%F ;11 c%F ;12 ? c%F;nn m
% F ;11 m
% F;12 ? m
% F;nn T ;
where the elements in v are individual elements of the structural matrices. With this denition of v;
there is a linear transformation such that
% F rF ;b Wv;
9
Z
where W contains the response terms at each measured frequency [1]. For the qth measured
frequency
Woq W0 oq W1 oq W2 oq ;
10
where, if all elements of the foundation mass, damping and stiffness matrices are identied,
3
2 T
rF ;b oq
0
?
0
7
6
7
6
0
rTF;b oq
0
k6
7;
11
Wk oq joq 6
7
^
^
&
^
5
4
0
0
? rTF;b oq
for k 0; 1; 2: Eq. (7) then becomes
8 9
>
< v >
=
Woq Roq Rm oq e
Qoq ;
>
: >
;
em
12
where the form of R; Rm and Q may be obtained by comparing Eqs. (7), (11) and (12), as
Roq o2q ZB oq P1 oq ZR;bi oq Z1
R;ii oq T;
13
Qoq ZB oq P1 oq ZB oq IrF;b oq ;
14
Rm oq ZB oq P1 oq ZR;bi oq Z1

R;ii oq Tm :
15
Clearly there is an equation of the form of Eq. (12) at every frequency. The equations generated
may be solved in a least-squares sense directly, although the solution via the singular value
decomposition (SVD) is more robust. Such an equation error approach does not optimize the
error in the response directly, and thus the accuracy of the predicted response is not assured. The
great advantage is that the equations are linear in the parameters. However a non-linear
optimization (output error) may be performed, starting with linear estimated parameters, if a
ARTICLE IN PRESS
973
more accurate prediction of the response is required [1,22]. In the present paper, only the equation
error approach has been considered in order to concentrate on the inuence of frequency range
subdivision. Furthermore, the unbalance and misalignment seems to be estimated robustly by the
equation error approach, even if the foundation is relatively inaccurate [24].
2.4. Splitting the frequency range
The number of degrees-of-freedom of the foundation model estimated by the above method
would be equal to the number of measurement locations during the machine run-down. Often,
especially for large machines, the number of critical speeds of the system in the run-down
frequency range, is more than the number of measurement locations. This requires splitting the
entire frequency range into smaller bands and the foundation models have to be estimated for
each frequency band [22,26]. The splitting of the frequency range may be performed by a visual
inspection of the measured responses, together with some experience and a trial and error
approach. However, the rst trial itself will give a good estimate if the number of frequency peaks
seen on the measured responses (both in the vertical and horizontal directions) are less than the
total number of degrees-of-freedom of the foundation model to be estimated in each frequency
band.
Suppose that the frequencies at which the response is measured are oq ; q 1; y; N: Assume
that the run-down frequency range is split into b frequency bands. The vectors of the foundation
parameters are identied in each frequency band, and are denoted v1 ; v2 ; y; vb : Hence combining
the frequency-band-dependent foundation models and the global unbalance and misalignment, in
a similar way to the unbalance estimation [26], gives, from Eq. (12)
8 9
>
> v1 >
>
>
9
>
> 8
2
3>
>
>
>
0
?
0
Rband 1 Rm;band 1 > v2 >
Wband 1
Qband 1 >
>
>
>
>
>
>
>
>
<Q
< >
=
= >
6
0
Wband 2 ?
0
Rband 2 Rm;band 2 7
band 2
6
7 ^
:
16
6
7
>
4
5>
vb >
^
^
&
^
^
^
^ >
>
>
>
>
>
>
>
>
:
> >
;
>
>
0
0
? Wband b Rband b Rm;band b >
Qband b
e >
>
>
>
>
>
:e >
;
m
2.5. Regularization
Eq. (16) is a least-squares problem, and its solution is likely to be ill conditioned [1,26]. The
following three types of regularization were used to solve the least-squares problem.
2.5.1. Regularization Method 1
In solving the least-squares problem, generally two types of scaling, namely row scaling and
column scaling, may be applied [29]. Column scaling is necessary because of the different
% F and K
% F matrices, and the scaling factors used here were 1,
% F; C
magnitudes of elements of the M
2
o
% is the mean value of the frequency range. The scaling of the
% and o
% respectively, where o
columns of R and Rm depend upon engineering judgement based on the unbalance and
misalignment magnitudes expected. Truncated SVD was used to solve the equations [30].
ARTICLE IN PRESS
974

Other physically based constraints may be applied to the foundation model to improve the
conditioning in addition to Regularization Method 1. For example, the mass, damping and
stiffness matrices of the foundation may be assumed to be symmetric, therefore reducing the
number of unknown foundation parameters. Other constraints could be introduced, such as a
diagonal mass or damping matrix, or block diagonal matrices if bearing pedestals do not interact
dynamically.
Depending upon the physical constraints applied to the foundation model, the vector vc of the
foundation parameters to be estimated in each frequency band may be dened as
vc Tc vc ;
17
where c 1; 2; y; b and Tc are the transformation matrices. Substituting Eq. (17) into Eq. (16)
produces an equivalent set of equations for the unknown parameters, with the important
difference that the number of parameters to be estimated is reduced. This will improve the
conditioning of the estimation problem at the expense of a larger residual in the tted equation.
The partitioning of the system dynamic matrices in Eq. (1) includes the sub-matrix ZR;ii ; which
is the dynamic matrix of the rotor where the degrees-of-freedom at the bearing locations have
been xed (that is the rotor is pinned at the bearings). The estimation equations require that this
sub-matrix be inverted and this will clearly lead to problems near the natural frequencies of the
rotor with pinned supports, where the matrix ZR;ii would be close to singular. This ill-conditioning
is particularly bad for lightly damped rotors, and the errors introduced result in meaningless
parameter estimates. The approach adopted in this paper is to remove the run-down data within a
small frequency band close to these problem frequencies.
3. Experimental example: the small rig

Fig. 3 shows a photograph of the experimental rig installed in the Dynamics Laboratory at the
University of Wales Swansea (UK). Each foundation of this rig consists of a horizontal beam
500 mm 25:5 mm 6:4 mm and a vertical beam 322 mm 25:5 mm 6:4 mm made of
steel. The horizontal beam is bolted to the base plate and the vertical beam to the bearing
assembly as seen in the photograph. A layer of acrylic foam 315 mm 19 mm 1 mm was
bonded to the vertical beam and thin layers of metal sheet 315 mm 19 mm 350 mm added to
increase the damping. Modal testing conrms that the damping of the foundation increased from
0.6% to 1.4% for the rst lateral mode, due to this damping layer. A 12 mm outside diameter (d)
steel shaft of 980 mm length is connected to these exible supports and is coupled to the motor
through a exible coupling. The Youngs modulus of elasticity and the density of shaft material
are 200 GN=m2 and 7800 kg=m3 ; respectively. The translational stiffness kt and rotational
stiffness ky of the exible coupling are estimated to be 27 kN=m and 25 Nm=rad; respectively [1].
One exible foundation is connected at 15 mm and another at 765 mm from the right end of the
shaft through self-lubricating ball bearings. The shaft also carries two identical balancing disks of
ARTICLE IN PRESS
975
Fig. 3. The small rig at the University of Wales Swansea (UK). (a) Photograph of the small rig (b) FE model of the
rotor (c) Modelling of the coupling stiffness in x-z plane.
75 mm outside diameter and 15 mm thickness and placed at 140 and 640 mm from the right end
of the shaft. Disk A is dened as the one nearest to the motor.
3.1. Run-down experiments
Different run-down experiments were performed with the rotor speed reducing from 2500
r.p.m. to 300 r.p.m. for different combinations of added mass to the balance disks A and B listed
in Table 1. Runs 1 and 4 were residual run-downs, i.e., without any added mass to the disks. Order
tracking was performed using the VSI Rotate software [31] such that each set of the run-down
data consisted of the 1 component of the displacement responses in the frequency range from
5.094 to 40:969 Hz in steps of 0:125 Hz:
3.2. Rotor unbalance and misalignment estimation
A nite element model was created for the rotor using two-noded EulerBernoulli beam
elements, each with two translational and two rotational degrees of freedom. The nite element
model of the rotor is shown in Fig. 3(b) and the modelling details of the exible coupling in xz
ARTICLE IN PRESS
976
Table 1
Estimation of both the rotor unbalance and misalignment from the experimental run-down data for the small rig
Run Disk Actual
unbalance
(g at deg.)
Estimated
unbalance
(g at deg.)
1.84
1.46
1.99
1.44
2.48
2.33
2.75
2.96
2.54
1.34
4
2
3
5
A
B
A
B
A
B
A
B
A
B
Residual
Residual
Residual
Residual
0.76 at 180
0.76 at 0
1.52 at 180
1.52 at 0
0.76 at 45
0.76 at 225
at
at
at
at
at
at
at
at
at
at
130
350
127
341
144
354
158
1
114
323
Estimated misalignment
Added unbalance (g at deg.)
Hori., Dy Vert., Dz Hori. angle Ver. angle With respect

(mm)
(mm)
Dyz (deg.) Dyy (deg.) to run 1
With respect
to run 4
0.10
0.21
0.40
0.69
0.09
0.18
0.32
0.20
0.14
0.20
0.75
0.45
0.14
0.18
0.42
0.16
0.08
0.21
0.24
0.17
0.82
0.88
1.41
1.55
0.95
0.66
at
at
at
at
at
at
181
4
191
11
79
236
0.82
0.98
1.46
1.67
0.76
0.45
at
at
at
at
at
at
190
14
202
18
76
228
plane in Fig. 3(c). Sinha [1] gave details of the dynamic characterization of the rig by modal tests
and nite element analysis.
Four critical speeds (two in the horizontal and one each in the vertical and axial directions) of
the machine were present in the run-down frequency range of 5.09440:969 Hz: The unbalance
and misalignment estimation was performed using the suggested method for individual runs
assuming misalignment forces and moments at the coupling of the rotor with the motor. All three
regularization methods were used during estimation process by assuming symmetric system
matrices for the foundation parameters and the removal of a small frequency band of 1:25 Hz on
either side of 38:37 Hz (the natural frequency of the rotor with pinned supports at the bearings)
from the machine run-down. The frequency range was split into three bands; 5.09412:094 Hz;
12.09427:469 Hz; and 27.46940:969 Hz based on the observation that the estimated responses
were a close t to the measured responses using these bands. The estimated results are listed in
Table 1 and Fig. 4 compares typical measured and estimated responses.
Fig. 4 shows that the t of the estimated response is a close t to the measured data. Table 1
also shows that the estimated unbalance is excellent and close to the actual values and the
estimated misalignment in the rotor at the coupling is quite consistent for each run. The order of
the estimated misalignment (approximately 0.1 and 0:2 mm in the horizontal and vertical
directions and their related angular misalignment is 0.4 and 0:2 ; respectively) is very small
and such a misalignment is quite possible during the rig assembly. The small deviation in
the estimation may be because of noise in the measurements, however the estimation seems to
be quite robust. Hence this experimental example conrms that an accurate estimation of
both the rotor unbalance and misalignment is possible using measured responses from a
single run-down of a machine. However, the sensitivity analysis with perturbations in the rotor
model was carried out on this experimental example to check the robustness of the proposed
method.
ARTICLE IN PRESS
977
_2
Horizontal Displ., m
10
_4
10
_6
10
_8
10
10
15
20
25
30
35
40
30
35
40
Frequency, Hz
_3
Vertical Displ., m
10
_4
10
_5
10
_6
10
_7
10
10
15
20
25
Frequency, Hz
Fig. 4. Measured and estimated response at bearing A, for run 2 for the small rig (dotmeasured, lineestimated).
4. Sensitivity analysis
As discussed earlier, it is often difcult to get accurate values for the parameters of the rotor
and uid bearing models for rotating machines such as turbogenerator sets. Hence the estimation
process was repeated with perturbations in the rotor and uid bearing models. Initially the
estimation process was repeated on the above experimental small rig example with rotor
modelling errors. The estimation and sensitivity analysis was then performed on another
experimental example with four uid bearings [1,26] to give a comparison between two different
types of rig.
4.1. Experimental Example 1: the small rig
Fig. 3 shows the rig, which is supported by ball bearings. In the frequency range considered the
bearings provide a rigid connection between the translational degrees of freedom of the rotor and
the foundation at the location of the bearings. Thus no error is expected in the bearing model.
Although the rotor may be modelled reasonably accurately, some small errors are expected. Here
errors will be assumed in the shaft diameter d and the material properties, namely the modulus
of elasticity E and the density (r). Furthermore, the mechanical coupling stiffnesses, the
translational stiffness (kt ) and rotational stiffness (ky ), are also the part of the rotor model and
hence are equally important. Errors in these parameters were introduced in turn before estimating
the rotor unbalance and misalignment. The errors introduced in the different parameters were
75% in shaft diameter d in steps of 0.5%, 710% in the modulus of elasticity E and density
(r), and 720% in the coupling stiffnesses, kt and ky ; in steps of 1%. Simultaneous perturbation in
ARTICLE IN PRESS
978
all these parameters was also considered, although such a combination may be highly unlikely in
practice.
The unbalance and misalignment estimates due to the perturbations in the rotor modelling
parameters, for the individual experimental runs, are shown in Table 2. Fig. 5 gives a typical
graphical representation of the variation in the rotor unbalance estimation from the subtracted
unbalance of runs 1 and 2 (with respect to the initial estimates given in Table 1), as a function of
the perturbation in individual rotor modelling parameters. For a normally distributed random
perturbation error in all of the modelling parameters simultaneously, with a standard deviation of
5%, Fig. 6 shows the variation in the unbalance estimate. Figs. 5 and 6 show that the phase
estimation is more robust than the unbalance amplitude. Usually the unbalance is estimated with
an error of less than 4050% with respect to the initial estimate in Table 1, except in a few cases
where it is about 7075%. Table 2 summarizes the maximum deviation in the estimated unbalance
amplitudes and phases. Fig. 7 shows a typical variation in the estimates of linear and angular
misalignment with the rotor modelling error for run 3. The maximum effect on the misalignment
50
% Error in Amplitude
80
40
0
40
80
5
25
0
25
50
75
100
10
250
250
200
200
150
100
50
0
50
10
10
0
8
% Error in Elasticity
40
50
40
20
0
20
40
8
20
0
20
40
20
10
15
10
% Error in Density
250
250
200
200
150
100
50
0
8
% Error in Density
10
15
20
10
15
20
% Error in Coupling Stiffnesses
Phase (degree)
Phase (degree)
100
50
10
60
50
10
150
% Error in Rotor Diameter
60
10
Phase (degree)
Phase (degree)
10
150
100
50
0
50
20
15
10
% Error in Coupling Stiffnesses
Fig. 5. Typical deviation in the rotor unbalance estimation due to errors in the rotor model of the small rig for
subtracted unbalance runs 1 and 2 (square for disk A, circle for disk B).
ARTICLE IN PRESS
75
100
50
0
_50
_100
1
11
13
15
17
19
50
25
0
_25
_50
_75
1
21
250
250
200
200
150
100
50
0
3
(a)
11
13
11
13
15
17
19
21
15
17
19
21
Sample Index
Phase (degree)
Phase (degree)
Sample Index
_50
1
979
15
17
19
Sample Index
150
100
50
0
_50
1
21
11
13
Sample Index
(b)
Fig. 6. Typical deviation in the rotor unbalance estimation due to 5% random errors in all rotor model parameters of
the small rig (square for disk A, circle for disk B). (a) Substracted unbalance for runs 1 & 2 (b) substracted unbalance
for runs 1 & 3.
Table 2
Unbalance estimation for the small rig, with perturbations to the rotor model
Rotor model
Balance disk Aunbalance
Balance disk Bunbalance
Perturbation
in parameters
Error (%)
Max. % error
in amplitude (%)
Max. phase
error (deg.)
Max. % error
in amplitude (%)
Max. phase
error (deg.)
Shaft diameter d
Elasticity E
Density r
Coupl. Stiff., kt and ky
5% Random error to all
75
710
710
720
70
55
60
40
75
45
35
30
20
20
60
38
30
25
70
30
25
25
25
20
estimation is seen when the rotor stiffness decreases by 610%, otherwise the variation in the
estimates is small.
The above sensitivity analysis has been performed using all of the regularization methods
discussed in Section 2. The effect on the estimated unbalance and misalignment of regularization
methods 1 and 2 is not usually dramatic. However, the use of regularization method 3 is essential,
particularly when the rotor model contains some error. Fig. 8 shows a typical example for the
subtracted estimated unbalance for runs 1 and 2, as a function of the error in the rotor model,
when regularization method 3 was not used. The deviation in the estimated unbalance is now very
high, and demonstrates that data at the identied problem frequencies must be removed.
4.2. Experimental Example 2: the Aston rig
Fig. 9 shows the test rig at Aston University, Birmingham, UK. The rig is much larger than the
small rig and it consists of a solidly coupled, two-shaft system mounted on four oil lubricated
journal bearings. The bearings sit on exible steel pedestals bolted onto a large lathe bed which
ARTICLE IN PRESS
980
Linear Misalignment, mm
0.4
0.3
0.2
0.1
_010
_8
_6
_4
_2
10
10
Angular Misalignment, deg.
% Error in Rotor Stiffness

1.5
0.5
_010
_8
_6
_4
_2
% Error in Rotor Stiffness
Fig. 7. Typical deviation in the rotor misalignment estimation due to errors in the rotor model of the small rig for run 3
(square for horizontal direction, circle for vertical direction).
500
400
300
200
100
0
_100
_10
_8
_6
_4
_2
10
10
Phase (degree)
400
300
200
100
0
_10
_8
_6
_4
_2
Fig. 8. Typical deviation in the rotor unbalance estimation due to errors in the rotor model of the small rig for
subtracted unbalance runs 1 and 2, without regularization method 3 (square for disk A, circle for disk B).
ARTICLE IN PRESS
981
rests on a concrete foundation. The rotor itself consists of two steel shafts 1.56 and 1:175 m long,
each with a nominal diameter of 38 mm and coupled through anges of 150 mm long and 100 mm
diameter at the connecting end of both shafts. The Youngs modulus of elasticity and density of
the rotor material are 200 GN=m2 and 7850 kg=m3 ; respectively. At either end of the shafts are
journals of diameter 100 mm; and in the centre of the shafts are machined sections which take
balancing disks, three for the long rotor and two for the short rotor. Each balancing disk is made
of steel and is 203:2 mm in diameter and 25:4 mm thick. The balance disks were placed at 457.2,
746.8, 1016, 2351 and 2630 mm from the left end of the shaft. The dimensions of the rotor at each
station are given in Table 3. The bearings are circular, have a length-to-diameter ratio of 0.3, a
radial clearance of 150 mm and contain oil of viscosity 0:0009 Ns=m2 : One bearing is located at
either end of the shaft and the remaining two bearings were located at 1372 and 2046 mm from
the left end. Smart [32] and Sinha et al. [26] gave a more detailed description of the rig.
Since the rig has uid bearings the sensitivity analysis was performed for the perturbation errors
in the rotor and bearing models in turn. Finally errors in both of the models simultaneously were
considered and compared with the experimental example of the small rig with ball bearings. A
nite element model (shown in Fig. 10) was created for the rotor with 51 two noded Timoshenko
beam elements, where each node had two translational and two rotational degrees-of-freedom
[1,26]. The experimental run-down responses were obtained between 55 and 5 Hz in 200 steps
[1,26]. The sensitivity analysis for the unbalance estimates was performed for the subtracted
experimental runs (2-1) and (3-1) listed in Table 4, which also gives the estimated unbalance with
and without the use of regularization method 3 [1]. Sinha [1] and Sinha et al. [26] gave the details
Fig. 9. Photograph of the Aston rig.
ARTICLE IN PRESS
982
Table 3
The physical dimensions of the rotor for the Aston rig
Long rotor properties
Short rotor properties
Station
Length (mm)
1
2
3
4
5
6
7
8
9
10
11
12
13
203.2
100
14
203.2
38.1
15
101.6
103
16
177.8
38.1
17
101.6
109
18
177.8
38.1
19
101.6
117
20
203.2
38.1
21
203.2
100
22
50.8
38.1
23
25.4
78.1
24
12.7
38.1
25
Coupling between the two rotors through coupling anges
(see Figs. 9 and 10) of 300 mm length and 100 mm diameter
Disc
Bearing
Diameter (mm)
1
1
Station
4
2 Coupling
Length (mm)
Diameter (mm)
6.35
25.4
50.8
203.2
177.8
50.8
76.2
76.2
76.2
50.8
177.8
203.2
38.1
77.57
38.1
100
38.1
116.8
38.1
109.7
38.1
102.9
38.1
100
5
4
y
Foundation
F1
F2
F3
F4
Fig. 10. Finite element model of the rotor of the Aston rig.
of the unbalance estimation for this example and these are not repeated here. Since the shaft was
aligned, only unbalance estimation was considered.
4.2.1. Error in the rotor model
Using a similar approach to the small rig example, the sensitivity analysis has been performed
using perturbations in the parameters of the rotor model. Fig. 11 shows a typical graphical
representation of the variation in the estimated rotor unbalance due to a perturbation in the
parameters of the rotor model for run (2-1). Table 5 summarizes the maximum deviation in the
estimated unbalance amplitudes and phases. Fig. 11 shows a smoother variation in the estimated
unbalance amplitudes and phases, compared with the variation found in the small rig example
with rigid bearings. In general the phase estimation is robust and usually invariant with the
modelling error, and in some cases the estimated phase at disk 2 for run (2-1) is found to be closer
to the actual phase. Although the Aston rig is much bigger than the small rig, the improved
robustness to rotor modelling errors is probably because of the higher damping in the uid
bearings.
ARTICLE IN PRESS
983
Table 4
Unbalance estimation from the experimental run-down data by subtracting runs for the Aston rig
Run
Disk
2-1
Actual unbalance
(Amplitude g m at phase (deg.))
2
5
1
5
3-1
1.7
1.7
1.7
1.7
at
at
at
at
Estimated unbalance
(Amplitude g m at phase (deg.))
105
180
225
315
20
0
_20
_40
_10
With regularization
method 3 [1]
1.529
1.759
1.943
1.849
1.70
1.68
1.77
1.65
at
at
at
at
15
17
19
21
15
17
19
21
at
at
at
at
51.42
176.10
293.02
51.40
_8
_6
_4
_2
30
20
10
0
_10
_20
_5
10
_4
_3
_2
% Error in Density
200
Phase (degree)
Phase (degree)
_1
200
150
100
50
_010
77
152
304
48
40
40
Without regularization
method 3 [26,27]
_8
_6
_4
_2
150
100
50
0_
5
10
_4
_3
_2
% Error in Density
_1
100
25
0
_25
_50
_10
_8
_6
_4
_2
50
0
_ 50
_100
1
10
13
200
Phase (degree)
Phase (degree)
11
Sample Index
200
150
100
50
_010
_8
_6
_4
_2
10
150
100
50
0
1
11
13
Sample Index
5% random errors in all rotor parameters

Fig. 11. Typical deviation in the rotor unbalance estimation due to errors in the rotor model of the Aston rig for
subtracted runs 1 and 2 (square for disk 2, circle for disk 5).
ARTICLE IN PRESS
984
Table 5
Unbalance estimation for the Aston rig, with perturbations to the rotor model
Balance disk 2unbalancea
Balance disk 5unbalance
Max. % error
in amplitude (%)
Max. phase
error (deg.)
Max. % error
in amplitude (%)
Max phase
error (deg.)
Run (2-1)
Shaft diameter d
75
Elasticity E
710
Density (r)
710
5% Random errors in all
15
40
38
60
15
50
25
45
37
35
25
80
10
20
20
20
Run (3-1)
Shaft diameter d
Elasticity E
Density (r)
75% Random errors in
15
22
45
45
7
7
15
35
9
12
12
20
5
5
7
8
Rotor model
Perturbation
in parameters
Error (%)
75
710
710
all
Disk 1 for run (3-1).
Table 6
Unbalance estimation for the Aston rig, with perturbations to the uid bearings model
Fluid bearing model
Perturbation
in parameters
Error (%)
Balance disk 2unbalancea
Balance disk 5unbalance
Max. % error
in amplitude (%)
Max. % error
in amplitude (%)
Max phase
error (deg.)
Max. phase
error (deg.)
Run (2-1)
Bearing forces
720
Oil viscosity (m)
720
Clearance (e)
720
6
0.8
12
12
5
2
7
3
27
1.5
40
12
20
2
20
20
Run (3-1)
Bearing forces
720
Oil viscosity (m)
720
Clearance (e)
720
12
1.8
26
8
7
10
25
20
17
1.8
22
13
5
5
10
20
Disk 1 for run (3-1).
4.2.2. Error in the fluid bearing model

Short bearing theory was used to model the uid bearings, and although this theory is
reasonably accurate [2527], the model depends on the bearing static load, bearing clearance and
oil viscosity [14]. The effect of errors in these parameters on the unbalance estimates was
examined, by introducing the errors one at a time. First, identical variations in the parameters
were assumed for all of the bearings in the machine. Second, random perturbations to all of the
bearing parameters were applied. Table 6 gives the summary of the results and Fig. 12 shows
typical graphical representations. The following observations may be made.
ARTICLE IN PRESS
1.5
30
20
10
0
_10
_20
_16
_12
_8
_4
12
16
20
1
0.5
0
_0.5
1
_1.5
_20
_16
_12
_8
% Error in Bearing Forces
Phase (degree)
Phase (degree)
100
50
_16
_12
_8
_4
12
16
12
16
20
12
16
20
15
17
19
21
15
17
19
21
50
_16
_12
_8
_4
% Error in Oil Viscosity
15
100
_020
20
50
40
30
20
10
0
_16
_12
_8
_4
12
16
10
5
0
_5
_10
_15
1
20
11
13
Sample Index
% Error in Bearing Clearance

200
Phase (degree)
200
Phase (degree)
150
% Error in Bearing Forces
150
100
50
_020
200
150
_10
_20
_4
% Error in Oil Viscosity
200
_020
985
_16
_12
_8
_4
% Error in Bearing Clearance
12
16
20
150
100
50
0
1
11
13
Sample Index
10% random errors in all parameters

Fig. 12. Typical deviation in the rotor unbalance estimation due to errors in the uid bearing model for subtracted runs
1 and 2 for the Aston rig (square for disk 2, circle for disk 5).
(1) The unbalance estimates are not sensitive to the bearing oil viscosity.
(2) Bearing static load and the bearing clearance have some effect on the unbalance amplitude
estimation; however, the phase estimation is once again quite robust and the error is always
less than 20 :
(3) In general, the errors in the unbalance estimates due to errors in the uid bearing model are
less than those due to errors in the rotor model.
Hence this study indicates that the estimation of rotor unbalance can accommodate more error in
the uid bearing models than in the rotor model.
4.2.3. Error in both the rotor and fluid bearing models
Perturbations in both the rotor and bearing models were introduced simultaneously. A
normally distributed random variation was used for all of the modelling parameters with a 10%
ARTICLE IN PRESS
986
50
100
50
0
50
100
11
13
15
17
19
25
0
25
50
1
21
Sample Index
11
13
15
17
19
21
15
17
19
21
Sample Index
350
Phase (degree)
200
Phase (degree)
150
100
50
300
250
200
150
100
50
0
1
11
13
15
17
19
21
0
1
11
13
Sample Index
Sample Index
Run (2-1)
Run (3-1)
Fig. 13. Deviation in the rotor unbalance estimation due to random errors in both the rotor and the uid bearing
models for the Aston rig (dot for disk 1, square for disk 2, circle for disk 5).
standard deviation for the uid bearing parameters and a 5% standard deviation for the rotor
model parameter. Fig. 13 shows the rotor unbalance estimated for a small number of samples.
The error in the estimated unbalance amplitudes is usually less than 45% except for some
combinations of parameters for run (2-1) when the error is up to 70%. However the phase
estimation is robust and the maximum phase error is found to be 40 : It should be emphasized
that the combinations of errors in the parameters are unlikely to occur in practice.
5. Conclusions
A method to estimate both the rotor unbalance (amplitude and phase) and the misalignment of
a rotorbearingfoundation system has been presented. The estimation uses a priori rotor and
bearing models along with measured vibration data at the bearing pedestals from a single rundown or run-up of the machine. The method also estimates the frequency-band-dependent
foundation parameters to account for the dynamics of the foundation. The suggested method has
been applied to a small experimental rig and the estimated results were excellent.
A sensitivity analysis for rotor unbalance and misalignment estimation was performed by
introducing errors in the parameters of the rotor and uid bearing models, to check the robustness
of the proposed method. It has been observed that the maximum error in estimated unbalance
amplitude is usually less than 45% for simultaneous random errors of 5% in the rotor model and
10% in the uid bearing models. However, the phase estimation is quite robust which is very
promising for practical rotor balancing. In practice, the authors consider that the errors in the
rotor model should be much less than 5% for many machines. The variation in the estimated
linear and angular misalignment due to errors in the rotor modelling was also found to be small
for the small rig. In general it has been observed that the estimation method is far less sensitive to
errors in the uid bearing modelling compared to the rotor modelling error, and rotor systems
ARTICLE IN PRESS
987
having uid bearings are also less sensitive to modelling errors, compared to systems having rigid
bearings. The importance of regularization method 3, which eliminates data near the natural
frequencies of the rotor on pinned supports, has also been demonstrated. Hence the proposed
method seems to be robust and gives reliable estimates of the rotor unbalance and misalignment.
Acknowledgements
The authors acknowledge the support of the EPSRC through grant GR/M52939. Jyoti K.
Sinha acknowledges Mr R. K. Sinha, Associate Director, RD & DG of his parent organization
B.A.R.C., India for his consistent support and encouragement. Prof. Friswell acknowledges the
support of the Royal Society as a Royal Society-Wolfson Research Merit Award holder. The
assistance of Prof. John Penny and Andrew Glendenning in obtaining the measured data for the
Aston rig is gratefully acknowledged.
Appendix A. Nomenclature
K; C; M
Z
e
em
f un
fm
j
r
rF ;b
T; T ; Tm
v; v
x; y; z
o
Dy; Dz
Dyy ; Dyz
Subscripts
R; B; F
stiffness, damping and mass matrices

dynamic stiffness matrix, o2 M joC K
vector of the rotor unbalance parameters
vector of the forces and moments at the couplings in multi-rotors
rotor unbalance force
rotor misalignment
force
p
unit imaginary, 1
responses of a machine
responses at the foundation connected to bearing degrees-of-freedom i.e., at the
bearing pedestals
transformation matrices for forcing degrees-of-freedom
vectors of the foundation parameters
rotor axes, horizontal and vertical directions
rotor running speed in rad/s
Horizontal and vertical linear misalignment at a coupling in multi-rotors (mm)
Angular misalignment at a coupling in multi-rotors in the horizontal and vertical
directions (degrees)
rotor, bearing, foundation
References
[1] J.K. Sinha, Health Monitoring Techniques for Rotating Machinery, Ph.D. Thesis, University of Wales Swansea,
Swansea, 2002.
ARTICLE IN PRESS
988
[2] S.W. Doebling, C.R. Farrar, M.B. Prime, A summary review of vibration-based damage identication methods,
Shock and Vibration Digest 30 (1998) 91105.
[3] A.G. Parkinson, Balancing of rotating machinery, IMechE Proceedings CJournal of Mechanical Engineering
Science 205 (1991) 5366.
[4] W.C. Foiles, P.E. Allaire, E.J. Gunter, Review: rotor balancing, Shock and Vibration 5 (1998) 325336.
[5] A. Muszynska, Rotor to stationary element rub-related vibration phenomena in rotating machineryliterature
survey, Shock and Vibration Digest 21 (1989) 311.
[6] S. Edwards, A.W. Lees, M.I. Friswell, Fault diagnosis of rotating machinery, Shock and Vibration Digest 30 (1998)
413.
[7] C.B. Gibbons, Coupling misalignment forces, in: Proceedings of the 15th Turbomachinery Symposium, Gas Turbine
Laboratory, College Station, Texas, 1976, pp. 111116.
[8] D.L. Dewell, L.D. Mitchell, Detection of a misaligned disk coupling using spectrum analysis, Transactions of the
American Society of Mechanical EngineersJournal of Vibration Acoustics Stress and Reliability in Design 106
(1984) 916.
[9] F.F. Ehrich (Ed.), Handbook of Rotordynamics. McGraw-Hill, New York, 1992.
[10] A.S. Sekhar, B.S. Prabhu, Effects of coupling misalignment on vibrations of rotating machinery, Journal of Sound
and Vibration 185 (1995) 655671.
[11] G. Simon, Prediction of vibration behaviour of large turbo-machinery on elastic foundations due to unbalance and
coupling misalignment, Proceedings of the Institution of Mechanical Engineers, Part CJournal of Mechanical
Engineering Science 206 (1992) 2939.
[12] M. Xu, R.D. Marangoni, Vibration analysis of a motor-exible coupling-rotor system subjected to
misalignment and unbalance, Part I: theoretical model and analysis, Journal of Sound and Vibration 176 (1994)
663679.
[13] M. Xu, R.D. Marangoni, Vibration analysis of a motor-exible coupling-rotor system subjected to misalignment
and unbalance, Part II: experimental validation, Journal of Sound and Vibration 176 (1994) 681691.
[14] B.J. Hamrock, Fundamentals of Fluid Film Lubrication, McGraw-Hill, New York, 1994.
[15] A.W. Lees, I.C. Simpson, The dynamics of turbo-alternator foundations, in: IMechE Conference on Vibrations in
Rotating Machinery, Paper C6/83, London, UK, 1983, pp. 3744.
[16] A.W. Lees, The least squares method applied to identied rotor/foundation parameters, in: IMechE Conference on
Vibrations in Rotating Machinery, Paper C306/88, Edinburgh, UK, 1988, pp. 209216.
[17] G.A. Zanetta, Identication methods in the dynamics of turbogenerator rotors, in: IMechE Conference on
Vibrations in Rotating Machinery, Bath, Paper C432/092, 1992, pp. 173181.
[18] N.S. Feng, E.J. Hahn, Including foundation effects on the vibration behaviour of rotating machinery, Mechanical
Systems and Signal Processing 9 (1995) 243256.
[19] A. Vania, Estimating turbo-generator foundation parameters. Politecnico di Milano, Dipartimento di Meccanica,
Internal Report 9-96, 1996.
[20] R. Provasi, G.A. Zanetta, A. Vania, The extended Kalman lter in the frequency domain for the identication of
mechanical structures excited by sinusoidal multiple inputs, Mechanical Systems and Signal Processing 14 (2000)
327341.
[21] M.G. Smart, M.I. Friswell, A.W. Lees, Estimating turbogenerator foundation parametersmodel selection and
regularisation, Proceedings of the Royal Society Series A 456 (2000) 15831607.
[22] J.K. Sinha, A.W. Lees, M.I. Friswell, R.K. Sinha, The estimation of foundation models of exible machines,
in: Proceedings of Third International ConferenceIdentification in Engineering Systems, Swansea, UK, 2002,
pp. 300309.
[23] A.W. Lees, M.I. Friswell, The evaluation of rotor unbalance in exibly mounted machines, Journal of Sound and
Vibration 208 (1997) 671683.
[24] S. Edwards, A.W. Lees, M.I. Friswell, Experimental identication of excitation and support parameters
of a exible rotorbearingfoundation system from a single run-down, Journal of Sound and Vibration 232 (2000)
963992.
[25] J.K. Sinha, A.W. Lees, M.I. Friswell, Estimating the unbalance of a rotating machine from a single run down,
in: Proceedings of 19th International Modal Analysis Conference, Kissimmee, Florida, 2001, pp. 109115.
ARTICLE IN PRESS
989
[26] J.K. Sinha, M.I. Friswell, A.W. Lees, The identication of the unbalance and the foundation model of a exible
rotating machine from a single run down, Mechanical Systems and Signal Processing 16 (2002) 255271.
[27] A.W. Lees, J.K. Sinha, M.I. Friswell, The identication of the unbalance of a exible rotating machine from a
single run-down, in: Proceedings of the American Society of Mechanical Engineers Turbo Expo Conference,
Amsterdam, Netherlands, 2002, ASME paper GT-2002-30420.
[28] M.A. Jordan, What are orbit plots, anyway? Orbit, December (1993) 815.
[29] G.H. Golub, C.F. Van Loan, Matrix Computations, 3rd Edition, The John Hopkins University Press, Baltimore,
MD, 1996.
[30] P.C. Hansen, Regularisation tools: a MATLAB package for analysis and solution of discrete ill-posed problems,
Numerical Algorithms 6 (1994) 135.
[31] VSI Rotate User Manual, 2000. http://www.vold.com.
[32] M.G. Smart, Identication of Flexible Turbogenerator Foundations, Ph.D. Thesis, University of Wales Swansea,
Swansea,1998.

Estimating Unbalance and Misalignment of A Exible Rotating Machine From A Single Run-Down

Uploaded by

Copyright:

Available Formats

Estimating Unbalance and Misalignment of A Exible Rotating Machine From A Single Run-Down

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Estimating Unbalance and Misalignment of A Exible Rotating Machine From A Single Run-Down

Uploaded by

Copyright:

Available Formats

ARTICLE IN PRESS

Estimating unbalance and misalignment of a exible rotating

Fig. 1. The abstract representation of a rotating machine.

Fig. 2. Schematic of rotor with misalignment at a coupling.

2.3. Parameter estimation

Qoq ZB oq P1 oq ZB oq IrF;b oq ;

Rm oq ZB oq P1 oq ZR;bi oq Z1

2.5.2. Regularization Method 2

3. Experimental example: the small rig

Added unbalance (g at deg.)

Hori., Dy Vert., Dz Hori. angle Ver. angle With respect

% Error in Coupling Stiffnesses

% Error in Rotor Diameter

% Error in Rotor Diameter

% Error in Coupling Stiffnesses

Balance disk Aunbalance

Balance disk Bunbalance

Angular Misalignment, deg.

% Error in Rotor Stiffness

% Error in Rotor Stiffness

Fig. 9. Photograph of the Aston rig.

Short rotor properties

% Error in Rotor Diameter

% Error in Rotor Diameter

5% random errors in all rotor parameters

Balance disk 5unbalance

Disk 1 for run (3-1).

Balance disk 2unbalancea

Balance disk 5unbalance

Disk 1 for run (3-1).

4.2.2. Error in the fluid bearing model

% Error in Bearing Forces

% Error in Oil Viscosity

% Error in Bearing Clearance

% Error in Bearing Forces

% Error in Oil Viscosity

% Error in Bearing Clearance

10% random errors in all parameters

stiffness, damping and mass matrices

rotor, bearing, foundation

You might also like

Qoq ZB oq P1 oq ZB oq IrF;b oq ;