You are on page 1of 18

Journal of Sound and Vibration 332 (2013) 6216–6233

Contents lists available at ScienceDirect

Journal of Sound and Vibration


journal homepage: www.elsevier.com/locate/jsvi

Modal balancing of flexible rotors with bow


and distributed unbalance
M.B. Deepthikumar b, A.S. Sekhar a,n, M.R. Srikanthan b
a
Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai 600036, India
b
Chemical Technology Group, BARC, Mumbai, India

a r t i c l e in f o abstract

Article history: Unbalance and bow are found to be one of the most common causes of synchronous
Received 15 August 2012 machinery vibrations in rotating systems. Concentrated lumped mass models are adopted
Received in revised form in most of the finite element approach for modeling unbalances and subsequent balancing
4 April 2013
in rotating systems. But this assumption may not be appropriate for long slender
Accepted 27 April 2013
rotors with unbalances distributed along the length of the rotor. A polynomial curve for
Handling Editor: L.G. Tham
Available online 1 August 2013 eccentricity distribution with finite element modeling is used to identify the distributed
unbalance. The unbalance eccentricity distributions are estimated using the measured
vibration responses at a speed below the balancing speed. Modal correction mass required
to balance a rotor at its first bending critical speed, having both distributed unbalance and
bow is computed knowing the amplification factor at critical speed. The rotor is balanced
at its first bending critical speed using modal balancing method in a single trial run and
using a single balancing plane. The method thus avoids multiple trial runs required for
modal balancing of flexible rotors. This method is verified on an experimental rotor having
both bow and unbalance. The concept of quantifying the distributed unbalance using
‘Norm’ of eccentricity polynomial function is also introduced for the first time.
& 2013 Published by Elsevier Ltd.

1. Introduction

Rotating machinery often encounters excessive vibration due to various excitation sources. Among others, the synchronous
excitation due to rotating unbalance and residual bow is considered to be one of the major sources of synchronous vibrations in
rotating systems. Conventionally, unbalances are modeled as lumped mass, which is adequate for thin disks, but not for long shafts,
where in the unbalances are continuously distributed along its length. Practically, a long slender rotor has a combination of
distributed unbalance and residual shaft bow during manufacturing and assembly process.
The balancing techniques use lumped mass models based on three approaches such as modal balancing method [1–3],
the influence coefficient method [4–6] and the unified approach [7,8]. Nicholoas et al. [9,10] made extensive investigation
on bowed rotor response and observed that the rotor bow can significantly affect the amplitude and phase angle response of
a rotor, than is found with ordinary mass unbalance. Important conclusions were drawn from these tests on the rotor
amplitude and phase relationship for various combination of unbalance and bow. The differences in whirl undergo by
a rotating shaft subject to bow and mass unbalances are described by Parkinson et al. [11].
The continuously distributed mass unbalance of shaft was proposed by Lee et al. [12] using transfer matrix method. Yang
and Lin [13] formulated a method for estimation of distributed unbalance of rotor using polynomial curves with finite
element method for eccentricity identification of long continuous rotors.

n
Corresponding author. Tel.: +91 44 2257 4709.
E-mail addresses: sekh_as@yahoo.com, as_sekhar@iitm.ac.in (A.S. Sekhar).

0022-460X/$ - see front matter & 2013 Published by Elsevier Ltd.


http://dx.doi.org/10.1016/j.jsv.2013.04.043
M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233 6217

Nomenclature Um vector of global eccentricity coefficients


v linear nodal displacement in y direction
a coefficients of the local eccentricity curve V Force at each node
in x–z plane
A coefficients of the global eccentricity curve in Greek letters
x–z plane
Acr amplification factor for a bowed rotor β direction of rotating whirl vector
b coefficients of the local eccentricity curve γ resultant single unbalance correction mass
in y–z plane location
B coefficients of the global eccentricity curve in θ angular nodal displacement
y–z plane λi eigenvector for ith eigenvalue
C overall damping matrix including μ mass per unit length
gyroscopic effect ξ s/l, 0 oξo1
e eccentricity of the shaft element Φ phase angle of the unbalance of shaft
f overall unbalance force vector ψ shape function matrix
K overall stiffness matrix ω angular speed of the rotor
l length of a shaft element
P unbalance correction mass Subscript and superscript
M overall mass matrix
Mx moment at each node about x axis
c, s cosine term, sine terms
My moment at each node about y axis
d total degree of freedom
q overall nodal displacement vector
e element
Qc single equivalent unbalance modal
i degree of polynomial
correction vector
l local eccentricity
s local coordinate in a shaft, 0≤s≤l
L global eccentricity
S transformation matrix
n total no. of elements
Ssym system matrix
r resultant of x and y
u linear nodal displacements in x direction
um vector of local eccentricity coefficients

In the present study, the method developed by [13] is used for modal balancing of a flexible rotor system with distributed
unbalance and bow. An experimental rotor system is used to validate the results. Also, the concept of quantifying the
distributed unbalance is introduced. In this study, the conclusion made by Parkinson et al. [11] is used for modal balancing of
rotors having both unbalance and bow. That is, in case of rotors having unbalance and bow, it is effective to balance through
several flexural critical speed based on the net whirl than the total whirl. Initially the modal correction mass required
to balance the rotor is estimated by considering only the disturbed unbalance and this correction mass is amplified depending
on the critical speed amplification factor. The interest here is primarily for balancing of flexible rotor systems in a single trial
run and using a single balancing plane. A good mathematical model of the rotor system is a prerequisite to this method.

2. Modeling of rotor bearing system

A typical rotor bearing system is composed of shaft, discrete disks and discrete bearings. In using finite element method,
the rotor shaft segments are modeled by Euler–Bernoulli beam element. It is assumed that for the shaft the shape of the
cross-section, dimensions and materials constants are uniform in each element.
Fig. 1 shows the coordinate system used in a shaft element. The fixed frame is denoted by XYZ, and the rotating frame is
denoted by xyz. The rotor rotates about the z-axis with the angular velocity ω. The displacements of the rotor in the X, Y
directions are denoted by u, v and the rotational angles about X and Y axes are denoted by θx, θy. Fig. 2 shows the mass center
Me, geometric center Ge, eccentricity e and phase angle of unbalance Φ.
The system governing equation for computation of unbalance response with n shaft elements is

Mq€ þ Cq_ þ Kq ¼ f (1)

where M is the overall mass matrix, K is the overall stiffness matrix including the support stiffness, overall matrix C includes
damping and gyroscopic effects, q is the overall nodal displacement vector and f is the overall unbalance force vector
as given by Nelson and McVaugh [14]. The order of overall mass matrix (or overall stiffness matrix or overall damping-
gyroscopic matrix) of the entire system is 4(n+1)  4(n+1) and that of overall nodal displacement vector (or overall
unbalance force vector) is 4(n+1)  1 with n denoting the total number of shaft elements.
6218 M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233

Fig. 1. Model of a shaft element.

Me

ωt
Ge

u X

Fig. 2. Transverse section of a shaft element.

The nodal displacements and unbalance forces are defined relative to x–z and y–z planes. Following the procedure adopted by
Yung and Tin [13], the steady-state responses of the lateral vibration of a shaft with elliptical orbit can be expressed as

u ¼ U c cos ωt þ U s sin ωt; v ¼ V c sin ωt þ V s sin ωt (2)

The nodal displacement of the rotor is given by

q ¼ qc cos ωt þ qs sin ωt; where qc ¼ fU c1; V c1; θxc1; θyc1 …U cðnþ1Þ; V cðnþ1Þ; θxcðnþ1Þ; θycðnþ1Þ gT

qs ¼ fU s1; V s1; θxs1; θys1 …U sðnþ1Þ; V sðnþ1Þ; θxsðnþ1Þ; θysðnþ1Þ gT (3)

The synchronous external unbalance force acting on the rotor is

f ¼ f c cos ωt þ f s sin ωt; where f c ¼ fV xc1 ; V yc1; M xc1; M yc1 …V xcðnþ1Þ ; V ycðnþ1Þ; M xcðnþ1Þ; M ycðnþ1Þ gT

f s ¼ fV xs1 ; V ys1; M xs1; M ys1 …V xsðnþ1Þ ; V ysðnþ1Þ; M xsðnþ1Þ; M ysðnþ1Þ gT (4)

Substituting Eqs. (3) and (4) into Eq. (1) we get

Ssym q ¼ f (5)
" # " # " #
K−ω M2
C qc fc
where Ssym ¼ ; q¼ ; f¼
−C K−ω2 M qs fs

The order of Ssym is 8(n+1)  8(n+1) and that of q and f is 8(n+1)  1.


M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233 6219

3. Balancing methodology

3.1. Estimation of distributed unbalance

The eccentricity of the rotor system is defined in local and global coordinate system. The local eccentricity curves are
defined for each of the element and a global eccentricity curve is common throughout the rotor.
Assuming that the eccentricity curves are finite, piecewise continuous, and of m-degree polynomial, the local eccentricity
curve for each shaft element in x–z and y–z planes are respectively expressed with variable ξ (Fig. 1) as
m m s
xðξÞ ¼ ∑ ai ξi ; yðξÞ ¼ ∑ bi ξi ; ξ¼ ; 0 ≤ξ ≤1 (6)
i¼0 i¼0 l

Thus the eccentricity distribution is


e ¼ xðξÞ þ jyðξÞ; (7)
pffiffiffiffiffiffiffi
where j ¼ −1, ai and bi are the coefficients of local eccentricity curve, s is the local coordinate in a shaft and l is the length
of the shaft element.
The global eccentricity curve is written as
m m
XðzÞ ¼ ∑ Ai zi ; YðzÞ ¼ ∑ Bi zi (8)
i¼0 i¼0

where X(z) is the projection of the global eccentricity curve on the x–z plane, and Y(z) the projection of the global
eccentricity curve on the y–z plane.
If the eccentricity curve is assumed to have m-degree polynomial in space and projected into x–z and y–z planes, there
will be 2(m+1) real unknown coefficients totally. 8(n+1) equations can be found by modeling the shaft with n elements
in Eq. (5). So, a necessary condition in modeling the shaft is 8ðn þ 1Þ≥2ðm þ 1Þ.

3.1.1. Relationship between local and global eccentricity coefficients


The shaft element having four degrees of freedom at each node has four shape functions to satisfy the displacement
fields. For a polynomial form for the shape functions we must have a cubic functions (m ¼3). Hence, the global eccentricity
curve is written as
XðzÞ ¼ A3 z3 þ A2 z2 þ A1 z þ Ao ; YðzÞ ¼ B3 z3 þ B2 z2 þ B1 z þ Bo (9)
and the local eccentricity for each element is given by
xðξÞ ¼ a3 ξ3 þ a2 ξ2 þ a1 ξ þ ao ; yðξÞ ¼ b3 ξ3 þ b2 ξ2 þ b1 ξ þ bo (10)
Since each of the four constants in the local eccentricity curve is to be expressed in terms of global coordinates,
we require four boundary conditions. The boundary conditions considered here are that the eccentricity values and its
derivatives of the local and global eccentricity curve at the nodes of the element are required to be equal [13].
The four boundary conditions for each of the element in x–z plane is
Xðl1 Þ ¼ xð0Þ; Xðl2 Þ ¼ xð1Þ
X′ðl1 Þ ¼ x′ð0Þ; X′ðl2 Þ ¼ x′ð1Þ
Similarly, the four boundary conditions for each of the element in y–z plane is
Yðl1 Þ ¼ yð0Þ; Yðl2 Þ ¼ yð1Þ
Y′ðl1 Þ ¼ y′ð0Þ; Y′ðl2 Þ ¼ y′ð1Þ
Applying the boundary conditions,
Xðl1 Þ ¼ xð0Þ ¼ a0
a1
X′ðl1 Þ ¼ x′ð0Þ ¼
l
a3 a2 a1
Xðl2 Þ ¼ xð1Þ ¼ 3 þ 2 þ þ a0
l l l
3a3 2a2 a1
X′ðl2 Þ ¼ x′ð1Þ ¼ 3 þ 2 þ
l l l
Yðl1 Þ ¼ yð0Þ ¼ b0
b1
Y′ðl1 Þ ¼ y′ð0Þ ¼
l
b3 b2 b1
Yðl2 Þ ¼ yð1Þ ¼ 3 þ 2 þ þ b0
l l l
3b3 2b2 b1
Y′ðl2 Þ ¼ y′ð1Þ ¼ 3 þ 2 þ (11)
l l l
6220 M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233

where l ¼ l1 −l2 ; l1 ; l2 are the z-coordinate of the nodes of the element, and ‘′’ denotes the derivative with respect to
z-coordinate.
Eq. (11) in matrix form is given by
8 9 2 0 0 0 1 0 0 0
3
0 8a 9
>
> Xðl1 Þ >> >
> 3>>
>
> >
> 6 0 07 > >
>
> X′ðl1 Þ >
> 6 0 1=l 0 0 0 0 7>>
>
>
>
>
> >
> 6 7>> 2>
a >
>
>
> >
> 6 1=l3 1=l2 1=l 1 0 0 0 07 >
> >
>
>
> Xðl2 Þ >
> 6 7>> >
a 1 >
>
> > 6 7> >
< X′ðl2 Þ >
= 6 3=l3 2=l2 1=l 0 0 0 0 7 >
0 7 a0 =
< >
6
¼6 7 (12)
>
> Yðl1 Þ >> 6 0 0 0 0 0 0 0 1 7> b3 >
>
> >
> 6 7>>
>
>
>
>
> Y′ðl1 Þ >
> 6 07 >b > >
>
> >
> 6 0 0 0 0 0 0 1=l 7>>
> 2>>
>
>
> > 6 7> >
> Yðl2 Þ >
> > 6 0
>
3 2
17
>
>
> b >
>
>
> > 4 0 0 0 1=l 1=l 1=l 5>> > 1 >
: Y′ðl Þ >; 3 2 :b > ;
2 0 0 0 0 3=l 2=l 1=l 0 0

Inverting the matrix in Eq. (12) results in


8 9 2 3 8 9
>
> a3 >
> 2 l −2 l 0 0 0 0 >
> Xðl1 Þ >>
>a >
> > 6 >
> >
>
>
> 2
>
>
> 6 −3 −2l 3 −l 0 0 0 07 7
>
>
> X′ðl1 Þ >
>
>
>
> >
> > 6 7 >
> >
>
> a1 > > Xðl2 Þ >
>
>
> > 6
> 0 l 0 0 0 0 0 07 >
> >
>
> >
< > 6 7 >
> >
>
a0 = 6 6 1 0 0 0 0 0 0 07 7
< X′ðl2 Þ =
¼6 7 (13)
> b3 >
> > 6 0 0 0 0 2 l −2 l 7 > Yðl1 Þ >
>
> >
> 6 7 >>
>
>
>
> b2 >
> > 6 −l 7 > Y′ðl Þ >>
>
> >
> 6 0 0 0 0 −3 −2l 3 7 >>
> 1 > >
>
>
> > 6 7 > >
> b1 >
> > 4 0
> 0 0 0 0 l 0 05 > >
> Yðl Þ >
>
>
>
> > > 2 >
:b > ; : Y′ðl Þ >
> ;
0 0 0 0 0 1 0 0 0 2

Eq. (13) can be written in short as


ue3 ¼ Sel αe (14)

Extending the procedure to polynomials of degree m and assembling the element equations give the equation for all the
local eccentricity coefficients.
um ¼ Sl α (15)
where um S1 and α are the assembly of element matrices ue3 Sel e
and α respectively. The order of matrices um and α is
2n(m+1)  1 and that of S1 2n(m+1)  2n(m+1).

um ¼ fueT eT eT eT
m1x um1y …umnx umny g
T

uemnx ¼ famn ; …; a1n; aon gT ; uemny ¼ fbmn ; …; b1n; bon gT

aij and bij are the local eccentricity coefficients of degree i in element j, i ¼0–m, j¼ 0–n.
( e )
αx ðlj Þ
e e T e
α ¼ fα ð0Þ…α ðln Þg ; α ðlj Þ ¼
αey ðlj Þ

αex ðlj Þ ¼ fXðlj Þ; X′ðlj Þ; …; X ðkÞ ðlj ÞgT


αey ðlj Þ ¼ fYðlj Þ; Y′ðlj Þ; …; Y ðkÞ ðlj ÞgT

k ¼(m−1)/2 and m is an odd number.


If m is an even number, the derivatives up to order of m/2 of the local and global eccentricity curves are matched at one
node of the element while only the derivatives up to order of m/2 are matched at the other node.
Similarly, the values and derivatives of the global eccentricity curves at the node of the shaft element are
3 2
Xðl1 Þ ¼ A3 l1 þ A2 l1 þ A1 l1 þ Ao
2
X′ðl1 Þ ¼ 3A3 l1 þ 2A2 l1 þ A1
3 2
Xðl2 Þ ¼ A3 l2 þ A2 l1 þ A1 l2 þ Ao
2
X′ðl2 Þ ¼ 3A3 l2 þ 2A2 l2 þ A1
3 2
Yðl1 Þ ¼ B3 l1 þ B2 l 1 þ B1 l 1 þ Bo
2
Y′ðl1 Þ ¼ 3B3 l1 þ 2B2 l1 þ B1
3 2
Yðl2 Þ ¼ B3 l2 þ B2 l2 þ B1 l2 þ Bo
2
Y′ðl2 Þ ¼ 3B3 l2 þ 2B2 l2 þ B1 (16)
M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233 6221

Eq. (16) in matrix form is


2 3
8 9 l
3
l1
2
l1 1 0 0 0 0 8 9
>
> Xðl1 Þ >> 6 12 7>> A3 >
>
> > 6 3l 07 > > >
>
>
> X′ðl1 Þ >
>
>
>
6 1
6 3
2l1 1 0 0 0 0 7>
7>
>
> A 2>
>
>
>
>
> >
> 6 l 2 7 >
> >
>
>
> Xðl Þ >
> 6 2 l2 l2 1 0 0 0 0 7>> A >
>
>
> 2 >
> 7>> 1 >
>
< > 6 3 > > >
X′ðl2 Þ = 6 3l
6 2 2l2 1 0 0 0 0 0 7 A0 =
7 <
¼6 7 (17)
>
> Yðl1 Þ >
> 6 0 0 0 0
3
l1
2
l1 l 17 > B3 >
> >
>
> >
> 6 7> >
>
> Y′ðl1 Þ >
> 6 7>>
> B2 >
>
>
>
> >
> 6 2
07 >
> >
>
>
> > 6 0 0 0 0 3l1 2l1 1 7> >
>
>
> Yðl2 Þ >> 6
>
> 3 2 7>>
> B >
>
>
> > 6
: Y′ðl Þ ; 4 0 0 0 0 l2 l2 l2 1 7> 1 >
5>:B > ;
2 3 0
0 0 0 0 3l2 2l 1 0

Eq. (17) can be written as

αe ¼ SeL Um (18)

where αe contains the values and derivatives of the eccentricity at the nodes, and Um is the coefficient vector for the global
eccentricity curve.
Extending the procedure to the polynomial of degree m and assembling from the element equations yield the equation
for the global eccentricity coefficients,
α ¼ SL Um (19)

The order of matrix SL is 2n(m+1)  2(m+1) and that of Um is 2(m+1)  1, where

α ¼ fαe ð0Þ…αe ðln ÞgT


( )
Umx
Um ¼ ;
Umy

Umx ¼ fAm ; Am−1 ; …; A2 ; A1 ; Ao gT


Umy ¼ fBm ; Bm−1 ; …; B2 ; B1 ; Bo gT

From Eqs. (15) and (19), a relation between the coefficients of local and global eccentricity curves can be obtained as
um ¼ Sl SL Um (20)

3.1.2. Unbalance force in terms of eccentricity coefficients


The forces due to mass unbalance are modeled with the beam shape functions by following the normal finite element
procedure. In the fixed frame the forces are
Z ! ! !
e 2 T
xðξÞ −yðξÞ
f ¼ μω ψ cos ωt þ sin ωt ds (21)
yðξÞ xðξÞ

μ is the mass per unit length, ω is the angular velocity, ψ is the shape function matrix and x(ξ), y(ξ) are the local eccentricity
functions,
e e e
f ¼ f c cos ωt þ f s sin ωt (22)

where
2 m 3 2 m 3
Z ∑ a ðξÞi Z − ∑ bi ðξÞi
l 6i¼0 i 7 l 6 7
e 6 7 e 6 i¼0 7
fc ¼ μω2 Ψ T 6 m 7ds; fs ¼ μω2 Ψ T 6 7ds
4 i5 4 m
5
0 ∑ bi ðξÞ 0 ∑ ai ðξÞi
i¼0 i¼0

" #
N1 0 0 N2 N3 0 0 N4
ΨT ¼
0 N1 −N2 0 0 N3 −N 4 0
2 3 2
N 1 ¼ 1−3s2 =l þ 2s3 =l ; N2 ¼ s−2s2 =l þ s3 =l
2 2 3
N 4 ¼ −s2 =l þ s3 =l ; N 3 ¼ 3s2 =l þ 2s3 =l
6222 M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233

The relationship between the coefficients of the eccentricity curves and the unbalance force for each element can
be obtained by Eq. (22) is given by [13]
2 3
N1 0
6 7
6 0 N1 7
6 7
6 0 −N 2 7
6 7" #
Z l 6 7
6 N2 0 7 a1 ξ3 þ a2 ξ2 þ a1 ξ þ a0
μω2 6 7
e
fc ¼ 6 N3 ds (23)
0 6 0 7 3 2
7 b3 ξ þ b2 ξ þ b1 ξ þ b0
6 7
6 0 N3 7
6 7
6 0 −N 4 7
4 5
N4 0

2 l l 3l l
3
28 15 20 2 0 0 0 0 8 9
6 7 > a3 >
6 0 0 0 0 l l 3l l 7 >>
>
>
>
6
6
28 15 20 2 7 >
2 7
>
> a2 >>
>
>
6 0 0 0 0 −l2 −l2 −l2 −l 7 >
> >
>
6 105 60 30 12 7 >
> a >
1 >
6 l2 7 >> >
>
6
2
l
2
l
2
l
0 0 0 0 7 < a0 >
> =
e 6 105 60 30 12 7
f c ¼ ω2 6 3l 7
6 14 4l 7l l
0 0 0 0 7 > > 3 >
b >
6 15 20 2 7 >> >
>
6 l 7 >
> b2 >>
6 0 0 0 0 3l 4l 7l
2 7 >
> >
>
6 14 15 20 7 >
> >
>
6 −l
2
−l
2
−l
2
−l
2 7
> 1 >
> b >
6 0 0 0 0 7 >
> >
4 2 42 30 20 12 5 :b > ;
l l2 l2 l2 01
42 30 20 12 0 0 0 0

e
f c ¼ ω2 Fec ue3 (24)
e
Similarly for fs ,
2 −l
3
0 0 0 0 28 − 15
l
− 20
3l
− 2l 8 9
6 l 7 > a3 >
6 28 l 3l l
0 0 0 0 7 > > > >
6 15 20 2 7 > >
> a2>
>
>
6 −l2 2 2 2 7 > > >
>
6 105 −l −l −l
0 0 0 0 7 > >
6 60 30 12 7 > >
> a1
>
>
>
6 7 > >
6 0 0 0 0 −l2 −l2 −l2 −l2 7 < a0 >
> =
e 26 105 60 30 12 7
fs ¼ ω 6 l −7 
6 0 0 0 0 3l 4l 7l 7 > >
> b 3 >
6 14 15 20 2 7 > >
6 3l 7 > >b > >
6 14 4l 7l l
0 0 0 0 7 > > 2>
> >
>
6 2 15 20 2 7 > > >
>
6 −l −l2 −l2 −l2 7 > > b1 >>
6 42 0 0 0 0 7 > > >
4 30 20 12 5 :b > ;
2 2 2 2 0
0 0 0 0 − 42
l
− 30
l
− 20
l
− 12
l

e
f s ¼ ω2 Fes ue3 (25)

Combining Eqs. (24) and (25),


( e
) " #
fc 2
Fec
e ¼ω ue3 (26)
fs Fes

Hence,
e
f ¼ ω2 Fe um (27)

In general for n element and m degree polynomial,

f ¼ ω2 F u m (28)

The order of matrix f is 8(n+1)  1 and that of F is 8(n+1)  2n(m+1)


Substituting Eqs. (28) and (20) into Eq. (5)

Ssym q ¼ ω2 F Sl SL Um (29)

Eq. (29) relates the coefficients of the global eccentricity curves to the vibration responses. Hence in terms of q,

q ¼ ω2 S−1
sym F Sl SL Um (30)

In Eq. (30), Ssym is obtained from FEM model. F, Sl and SL are obtained from Eqs. (28), (13) and (17). q is the measured
vibration response.
M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233 6223

Eq. (30) is simplified as

q ¼ ω2 S U m where S ¼ S−1
sym F Sl SL (31)

Eq. (31) is a linear algebraic equation which is to be solved for Um, by singular value decomposition method. Thus, the
coefficients of global eccentricity curve Um are computed and the eccentricity distribution functions X(z) and Y(z) are
calculated.
If the rotor system also contains discs having an unbalance, the eccentricity coefficients of discs need to be identified
along with the shaft coefficients as explained by Yang and Lin [13]. In the present study, it is assumed that unbalance
distribution is present only in the shaft and the discs are individually balanced. The discs properties are lumped into the
overall system matrix at the respective nodes.
In rotating machines, not all nodal locations along the length of the rotor can be measured. Hence, only the row
corresponding to the measured degree of freedom is retained in S. Thus the capacity to yield reliable result depends on
the condition number of reduced matrix S which in turn depends on the position of measurement. Also the number of
measurement locations should be ≥2(m+1), so as to completely define the m degree polynomial function in two planes.

3.2. Quantification of distributed unbalance

The distributed unbalance polynomial function expressed by Eq. (9) can be quantified by defining the ‘Norm’ of the
polynomial function in the following way:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z l
Norm of polynomial function in x−z plane jjXðzÞjj ¼ ½A3 z3 þ A2 z2 þ A1 z þ Ao 2 dx
0
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z l
Norm of polynomial function in y−z plane jjYðzÞjj ¼ ½B3 z3 þ B2 z2 þ B1 z þ Bo 2 dx
0
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Resultant Norm jjRðzÞjj ¼ jjXðzÞjj2 þ jjYðzÞjj2 (32)

The single numerical interpretation of unbalance distribution using ‘Norm’ helps in quantifying the distributed unbalance
in rotors. Thus, in case of rotors having distributed unbalance, this type of quantification is more realistic than the 2-plane
unbalance, normally in use.

3.3. Estimation of single equivalent unbalance correction

One of the main objectives of this study is to balance the first bending critical speed of the rotor using single balancing
plane. As the bearing stiffness is small, the rotor will have high vibration amplitudes at the bearing locations as shown
in Fig. 6(c). N plane modal balancing procedure is adapted, with emphasis to limit the vibration amplitude at the middle
plane at the expense of higher vibration amplitude at top plane. The rigid modes (cylindrical and conical mode) vibrations
are controlled by proper support damping.
The distributed unbalance is replaced by equivalent lumped masses along the length of the rotor. The equation to balance
ith mode using a single correction mass at location c is given by [15]
N N
Px mðcÞλi ðcÞ ¼ − ∑ fm2 ðzÞXðzÞλi ðzÞgp =Acr ; Py mðcÞλi ðcÞ ¼ − ∑ fm2 ðzÞYðzÞλi ðzÞgp =Acr (33)
p¼1 p¼1

where

Px and Py are the single equivalent unbalance correction mass required at location c in x–z and y–z planes respectively.
m(c) is the mass at location c;
λi(c) is the eigenvector for ith mode at p distinct locations;
m(z) is the mass for ‘p’ distinct locations along the rotor length z;
X(z)/Y(z) is the unbalance eccentricity from Eq. (9) for p distinct locations in x–z/y–z planes along the rotor length z;
λi(z) is the eigenvector for ith mode for p locations along the rotor length z;
Acr is the amplification factor.

maximum amplitude response at the natural frequency


Acr ¼ (34)
amplitude response at near zero frequencyðvery low speedÞ

From this definition, assuming the maximum allowable response at the first bending natural frequency, and measuring the
response at near zero speed, the amplification factor Acr is estimated for use in Eq. (33).
6224 M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233

Hence, the resultant single equivalent unbalance correction mass Pr at location r required to balance the rotor at its first
bending critical is given by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  
Py
Pr ¼ P2x þ P2y to be located at γ ¼ tan −1 (35)
Px

3.4. Estimation of unbalance response

As explained in Section 2, the steady-state response is split into sine and cosine components. The synchronous unbalance
force f, and the steady-state nodal displacement q given in the system equation is expressed as
" # " !#−1 " #
qc K−ω2 M C fc
¼ ⋅ ; q ¼ Ssym f (36)
qs −C K−ω2 M fs

From Eqs. (20) and (27),


f ¼ ω2 F Sl SL Um (37)
After the addition of the single equivalent correction mass at location or plane c, the vibration amplitude responses are
measured at measurement points (or nodes). These measurement readings are used to simulate the response after mass
addition and are explained below.
Substituting these vibration amplitudes in Eq. (31), the new the coefficient matrix Um of global eccentricity curve
is estimated. Substituting Um in Eq. (37) the matrix f is evaluated and on back-substitution in Eq. (31), the nodal
displacement vector q is obtained.
The elliptical steady-state orbit of a node is expressed as
u ¼ U c cos ωt þ U s sin ωt; v ¼ V c cos ωt þ V s sin ωt (38)
Expressing in complex form as follows:
rðtÞ ¼ u þ iv (39)
The complex exponential forms for the sine and cosine functions are as follows:
i 1 iωt
sin ωt ¼ − ðeiωt −e−iωt Þ; cos ωt ¼ ðe þ e−iωt Þ (40)
2 2
Substituting the components of Eq. (40) into Eq. (38) and then substituting the results into Eq. (39) we get,
rðtÞ ¼ 12 ½ðU s þ V c Þ þ ið−U c þ V s Þeiωt þ 12½ðU s −V c Þ þ iðU c þ V s Þe−iωt (41)
In terms of two rotating vectors, r(t) is expressed as follows:
rðtÞ ¼ Ff eiðωtþβ1 Þ þ Fb e−iðωt−β2 Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ff ¼ 12 ðU s þ V c Þ2 þ ð−U c þ V s Þ2 ; Fb ¼ 12 ðU s −V c Þ2 þ ðU c þ V s Þ2
   
ð−U c þ V s Þ ðU c þ V s Þ
β1 ¼ tan −1 ; β2 ¼ tan −1 (42)
ðU s þ V c Þ ðU s −V c Þ
The elliptical orbit r(t) in Eq. (42) is thus expressed in terms of two rotating vectors, one rotating in direction of rotation
with radius Ff called forward whirl, and another rotating opposite to the direction of rotation with radius Fb called backward

ω
Fb
r
v
Fr

Fig. 3. Elliptical orbit: the sum of two counter whirling circular orbits.
M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233 6225

whirl as shown in Fig. 3. These two vectors are positioned by their respective angles β1 and β2. The shape and direction of
the orbit depends on the magnitude of Ff and Fb. For example, if Fb is zero it is a circular whirl, or if jF f j 4 jF b j the direction of
orbit is forward, otherwise it is backward.
The balancing methodology is explained in the flow chart, Fig. 4.

Start

Setup the Finite Element Model


and define rotor specifications

Assume a global and local eccentricity


distribution polynomial function

Establish a relationship between global and


local eccentricity coefficients of the polynomial

Express unbalance forces in terms of eccentricity


global eccentricity coefficients

Measure the Response at any given speed, and


estimate the global eccentricity coefficients

The unbalance distribution in x-z and y-z planes are


found and is quantified by the ‘norm’ of polynomial
function

The single equivalent unbalances correction mass to


balance the rotor at its first bending critical speed is
estimated taking into effect the Amplification factor
at critical speed

The Correction Mass is added and the Rotor is


balanced at Critical Speed

End

Fig. 4. Flow chart of balancing methodology.


6226 M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233

BEARING-1
1

DISC

3
M
E
4 A
S
U
5 R
STEPPED E
SHAFT M
6 E
7 N
T

8 P
R
O
9 B
E
S
10

11

DISC

12

DRIVE KEY PHASOR


MOTOR
BEARING-2

Fig. 5. A schematic of the rotor bearing system.

Table 1
Properties of rotor system.

Element no. Element nodes Element length (m) Outer diameter (m) Inner diameter (m)

1 1, 2 0.018 0.062 0.0546


2 2, 3 0.147 0.057 0.0546
3 3, 4 0.147 0.057 0.0546
4 4, 5 0.147 0.057 0.0546
5 5, 6 0.147 0.057 0.0546
6 6, 7 0.01 0.06 0.0546
7 7, 8 0.141 0.057 0.0546
8 8, 9 0.141 0.057 0.0546
9 9, 10 0.141 0.057 0.0546
10 10, 11 0.141 0.057 0.0546
11 11,12 0.018 0.057 0.0546
Bearing-1 at Node 1 Bearing-2 at Node 12
k1 ¼ 1500 N/m; c1 ¼ 24 Ns/m k2 ¼ 8000 N/m; c2 ¼ 55 Ns/m
Disc properties at Node 1 Disc properties at Node 12
Mass¼ 0.350 kg Mass ¼0.540 kg
Outer diameter¼ 0.083 m Outer diameter¼ 0.079 m
M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233 6227

4. Simulation

A vertically mounted rotor system as shown in Fig. 5 is considered to illustrate the procedure of estimation of distributed
unbalance. The rotor consists of a stepped steel shaft with two end discs, mounted on fluid film bearings on flexible
supports. The shaft is discretized into eleven ‘two noded Euler–Bernoulii beam elements’ with discs at nodes 1 and 12. The

Table 2
Rotor natural frequencies.

Cylindrical mode Conical mode First bending Second bending

Natural frequency (Hz) 7.1 17.8 161.4 471.8

Fig. 6. Mode shapes at first four natural frequencies. (a) Cylindrical mode; (b) Conical mode; (c) First bending mode; (d) Second bending mode.
6228 M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233

shaft is 1270 mm long and weighing 3.16 kg. The measurement probes are located at node 1, 4, 7, 9 and 12. The reference key
phasor is located at bottom of the shaft. The balancing plane to balance the rotor at its first bending critical is located at
node 1. The properties of rotor system with equivalent bearing stiffness and damping parameters are shown in Table 1.
The data used for this simulation are from the actual experimental rig (see Section 5). Simulations have been carried out in
MATLAB (Ver. 9) for the rotor model.
For the system discretized into eleven two noded beam elements, the total degree of freedom of the system is 48.
Therefore the mass, stiffness and gyroscopic matrices are of order 48  48. Hence, the state-space matrix Ssym of order
96  96 is evaluated from Eq. (5) and eigenvalues and eigenvectors are computed. The eigenvalues are shown in Table 2,
while the mode shapes are given in Fig. 6.
In general, the bow of a rotor is taken as the synchronous vibration response along the rotor length, when the rotor
is rotating slowly. The responses measured at 20 Hz is taken as the measure of bow and the net response, i.e. the response at
a particular speed, say 100 Hz subtracted vectorally by the bow response is taken as the measure of unbalance. Response
at 20 Hz is chosen as the measure of bow, as this is the minimum speed at which the rotor steady-state response can be
measured after its passes the two rigid critical speeds. The rotor whirl responses at five sensor locations for residual bow
and unbalances are shown in Table 3.
A three degree polynomial function is assumed for local and global eccentricity curves. The transformation matrices F,
Sl and SL of order 96  88, 88  88 and 88  8 respectively are computed from Eqs. (28), (15) and (19). Then the matrix Ssys
of order 96  8 is estimated. As the vibration response corresponding to nodes 1, 4, 7, 9 and 12 are only measured, the rows

Table 3
The rotor whirl responses for bow and unbalance.

Vibration amplitude in micron(0-peak)/phase angle in degree

Node-1 Node-4 Node-7 Node-9 Node-12

Residual bow (20 Hz) 81/254 169/263 109/278 130/322 192/336


Unbalance Response 50/322 43/68 94/110 124/123 111/120

Fig. 7. Unbalance eccentricity distributions in x–z plane.

Fig. 8. Unbalance eccentricity distributions in y–z plane.


M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233 6229

Fig. 9. Simulated synchronous response before balancing.

Table 4
Rotor synchronous whirl responses after adding the balancing mass.

Speed (Hz) Vibration amplitude (0-peak) in micron/phase angle in degree

Node-1 Node-4 Node-7 Node-9 Node-12

20 245/305 206/307 147/310 153/341 159/22


100 274/290 182/299 123/314 121/13 116/82

Fig. 10. Simulated synchronous response after balancing.

Table 5
Comparison of ‘Norm’.

Norm in X–Z plane Norm in Y–Z plane Resultant Norm

Before balancing 0.00353 0.00817 0.0089


After balancing 0.00514 0.00456 0.00687

corresponding to these nodes in the matrix Ssys are only retained and from Eq. (31), Um, the global coefficient matrix of order
8  1 is evaluated.
The eccentricity curve (before balancing) is given by

XðzÞ ¼ 0:00568z3 −0:01076z2 þ 0:00516z−0:00042

YðzÞ ¼ 0:02085z3 −0:04028z2 þ 0:01909z−0:00136


6230 M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233

The plots of unbalance eccentricity curve (before balancing) along the length of the rotor in x–z and y–z planes are given
in Figs. 7 and 8 respectively.
Using the whirl responses at 20 Hz and 100 Hz recorded in the experimental rig (Table 2) and the rotor properties
(Table 1), the coefficient matrix Um of global eccentricity curve is estimated. Substituting Um in Eq. (37) the matrix f
is evaluated and on back-substitution in Eq. (31), the nodal displacement vector q is obtained as explained in Section 3.4.
Hence the unbalance whirl responses before balancing are simulated and are shown in Fig. 9. A dip in the vibration response
is observed due to the presence of bow, which in the present case counteracts with the mass unbalance. Very high vibration
amplitude of 588 μm at first bending critical speed corresponding to middle plane/sensor 7 is observed.
The rotor is required to be balanced at its first bending critical natural frequency of 161.4 Hz, using a single correction
mass at balancing plane located at node 1. The emphasis is to limit the vibration amplitude at node 7 (middle plane) which
corresponds to maximum vibration amplitude during first bending natural frequency.
The amplification factor is computed using Eq. (34). The amplitude response from Table 3 corresponding to node 7
at 20 Hz is 109 μm and the limiting allowable response corresponding to node 7 at first bending frequency is 200 μm. Thus,
Acr ¼1.834, from Eq. (34). N plane modal balancing method is used to balance the rotor at its first critical speed. As plane 1
is easy to approach for balancing mass addition, on a vertically mounted rotor (enclosed inside the casing, with opening only
at top), and due to manufacturing difficulties, it is chosen as the balancing plane instead of plane 7 at the center. The single
unbalance correction mass at balancing plane 1 (i.e. at node 1) as per Eq. (35), required to balance the rotor at its first
bending frequency is 260.52 g mm at 2561.
The synchronous whirl response at 20 Hz and 100 Hz after the addition of correction mass is recorded in the experi-
mental rig, and is shown in Table 4. Fig. 10 shows the simulated unbalance whirl response, after the addition of single
unbalance correction mass. The responses are simulated as explained in Section 3.4 and using the responses after adding the
balancing mass (Table 4). The low speed responses after the addition of correction mass, increases with respect to node/
sensor 1 and 7 and decrease with respect to node/sensor 12. The maximum amplitude corresponding to node 7 at first
bending critical speed (161.4 Hz) is 368 μm which is substantially lower than before balancing. It may be noted here that
only support damping is considered and rotor material damping is not considered.
The eccentricity curves after the addition of correction mass in balancing plane/node-1 is

XðzÞ ¼ −0:00849z3 þ 0:01572z2 −0:00697z þ 0:000426

YðzÞ ¼ 0:01145z3 −0:02225z2 þ 0:01072z−0:000814

To quantify the distributed unbalance, the ‘Norm’ of the eccentricity distribution polynomial curve is calculated from
Eq. (32) and values are compared before and after balancing as shown in Table 5. The ‘Resultant Norm’ after balancing is less
by 23 percent, compared to that before balancing. The ‘Norm’ value before balancing in Y–Z plane is higher, when compared
with X–Z plane. This indicates that the unbalance distribution is dominant in Y–Z plane. This aspect is substantiated by
comparing Figs. 7 and 8.

5. Experimental results

To demonstrate the accuracy of the simulation results, the presented procedure is applied to the experimental rotor
system which is having inherent mass unbalance and bow during its manufacturing process (Fig. 11). The configuration of
the rotor system is provided in Table 1, which is same as that used in simulation studies. Concentrated inertial properties are
included for discs at nodes 1 and 12. The discs are individually balanced and then assembled on to the stepped shaft. The
two bearings are located at nodes 1 and 12.
The rotor system is instrumented with five eddy current displacement sensors, the sensitivity of which is 8 mV/μm.
An additional sensor is used to generate the tracking signal to filter the signals from these five sensors with respect to the
synchronous tracking signal. The notch on the bottom disc is used as the trigger for the tracking filter. All the displacement
sensors including the tracking sensor are arranged in the same vertical plane and positions of displacement sensors are at
node 1, 4, 7, 9, and 12 (Fig. 5). The rotor is run at very low acceleration rates and steady-state signals are measured.

Fig. 11. Experimental rotor.


M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233 6231

Casing

Vibration
Sensor

Vibration
Analyzer

Fig. 12. Rotor system casing with vibration instrumentation.

Fig. 13. Experimental synchronous amplitude response before balancing.

Rotor is housed inside a casing as shown in Fig. 12, which is maintained at negative pressure with respect to atmosphere
to avoid drag on the system. These sensors are mounted on the casing and measure the absolute vibration of the rotor
in terms of zero to peak amplitude. The bearing assemblies are housed inside the casing. The casing is bolted to the massive
structural steel base. The rotational speed is controlled by a asynchronous non-contact electric motor controlled by
a variable frequency drive. The motor is bolted and keyed to the same base to which the bottom bearing is housed. The
objective of this experimental rig is to run the test rotor to the desired operating speed of 180 Hz. For this to achieve, the
rotor has to traverse two rigid mode frequencies and one bending mode frequency. The rigid critical speeds are traversed by
proper damping of the support system. To traverse the first bending frequency, the rotor has to be balanced using the single
balancing plane available at top disc.
6232 M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233

Fig. 14. Experimental synchronous amplitude response after balancing.

The bow in the rotor system is approximated by measuring the synchronous whirl response at very low speed (20 Hz).
This low speed response is subtracted from the measurement signal at 100 Hz to get the unbalance response.
The synchronous response before balancing is shown in Fig. 13. The responses are having a decreasing trend initially due
to the presence of bow which in the present case opposes the mass unbalance. The rotor is stopped at 115 Hz as the
response at node/sensor-7 starts increasing. The comparison of simulated synchronous response before balancing (Fig. 9)
and the experimental synchronous response before balancing (Fig. 13) follows similar trend especially for station-7.
The vibration amplitudes in simulated results are higher compared to the experimental results as material damping is not
considered in the simulation.
The single equivalent correction mass of 260.52 g mm at 2561, as estimated in Section 3.3 is added at node 1/balancing
plane and the rotor is run to the full operating speed of 180 Hz. The response plot is shown in Fig. 14. A dip is observed in the
response curve corresponding to sensors located at node 7 and 12 which is similar to that is seen in simulated response of
Fig. 10. The maximum amplitude of 185 μm (less than the allowable limit of 200 μm) with respect to node 7/middle plane
is observed at 159.2 Hz. This maximum amplitude is well below the simulated response value (368 μm), as rotor material
damping is not considered in the simulations.
Comparing Figs. 13 and 14, it is observed that, the amplitude of responses increased at low speeds after the addition
of correction mass at node/balancing plane-1. This increase is substantial in case of node-1, wherein the response at 20 Hz
increased from 81 μm to 245 μm after the addition of correction mass. However, the response of 245 μm is less than the
allowable limit of 400 μm at node-1. These vibration limits are arrived based on the experience of balancing similar rotor
system in in-situ conditions. In addition the main objective of the present study is to pass the rotor critical speed with
middle plane amplitude less than 200 μm. With vibration amplitude level of 185 μm after balancing, the rotor is able to pass
smoothly the first bending critical with acceptable amplitude levels.

6. Conclusions

The present study experimentally validates the procedure of distributed unbalance in the modal balancing to balance
flexible rotors having both unbalance and bow. With this method, the rotor is run to a speed, well below the bending critical
and vibrations responses are measured. The correction mass required is computed taking into consideration the rotor mass
unbalance and bow. With the addition of computed correction mass the rotor is run safely passing the critical speed to reach
its operating speed.
A factor (‘Norm’) to quantify distributed unbalance helps to substantiate the amount of reduction in the unbalance
eccentricity. In the present study, a reduction of 23 percent (in simulation) in the ‘Resultant Norm’ helps in passing the rotor
through the first bending critical. The factor to quantify the distributed unbalance is first of its kind and needs further study
for different types of rotor system.

References

[1] R.E.D. Bishop, G.M.L. Gladwell, The vibration and balancing of an unbalanced rotor, Journal of Mechanical Engineering Science 1 (1) (1959) 66–67.
[2] W. Kellenberger, Should a flexible rotor be balanced in n or (n+2) planes? Journal of Engineering for Industry ASME 94 (2) (1972) 548–560.
[3] S. Saito, T. Azuma, Balancing of flexible rotors by the complex modal method, Journal of Vibration, Acoustics Stress and Reliability in Design ASME 105
(1983) 94–100.
[4] T.P. Goodman, A least-square method for computing balance corrections, Journal of Engineering For Industry Series ASME 86 (3) (1964) 273–279.
[5] J.W. Lund, J. Tonnesen, Analysis and experiments on multi-plane balancing of flexible rotors, Journal of Engineering for Industry ASME (1972) 233–242.
M.B. Deepthikumar et al. / Journal of Sound and Vibration 332 (2013) 6216–6233 6233

[6] J.M. Tessarzik, R.H. Badgley, W.J. Anderson, Flexible rotor balancing by the exact point–speed influence coefficient method, Journal of Engineering for
Industry Series B ASME 94 (1) (1974) 148–158.
[7] A.G. Parkinson, M.S Darlow, A.J. Smalley, A theoretical introduction to the development of unified approach to flexible rotor balancing, Journal of Sound
and Vibration 68 (4) (1980) 489–506.
[8] S.G. Tan, X.X Wang, A theoretical introduction to low speed balancing of flexible rotors: unification and development of the modal balancing and
influence coefficients techniques, Journal of Sound and Vibration 168 (3) (1993) 385–394.
[9] J.C. Nicholas, E.J. Gunter, P.E. Allaire, Effect of residual shaft bow on unbalance response and balancing of single mass flexible rotor. Part I—unbalance
response, Journal of Engineering for Power ASME (1976) 171–181.
[10] J.C. Nicholas, E.J. Gunter, P.E. Allaire, Effect of residual shaft bow on unbalance response and balancing of single mass flexible rotor. Part II—balancing,
Journal of Engineering for Power ASME (1976) 182–189.
[11] A.G. Parkinson, M.S. Darlow, A.J. Smalley, Balancing of flexible rotating shafts with an initial bend, AIAA Journal 22 (5) (1984) 683–689.
[12] A.C Lee, Y.P. Shih, Y. Kang, The analysis of rotor linear rotor bearing system: a general transfer matrix method, Journal of Vibrations and Acoustics ASME
115 (1993) 490–497.
[13] T. Yang, C. Lin, Estimation of distributed unbalance of rotors, Journal of Engineering for Gas Turbines and Power ASME 124 (2002) 976–983.
[14] H.D. Nelson, J.M. McVaugh, The dynamics of rotor-bearing systems using finite element, Journal of Engineering for Industry ASME 98 (1976) 593–600.
[15] M.S. Hundal, R.J. Harker, Balancing of flexible rotors having arbitrary mass and stiffness distribution, Journal of Engineering for Industry ASME 88 (1966)
217–223.

You might also like