You are on page 1of 16

Journal of Sound and Vibration 362 (2016) 276–291

Contents lists available at ScienceDirect

Journal of Sound and Vibration


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

Instability of asymmetric shaft system


R. Srinath, Abhijit Sarkar, A.S. Sekhar n
Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai 600036, India

a r t i c l e in f o abstract

Article history: In the present work, parametric instability of asymmetric shaft mounted on bearings is
Received 10 August 2015 studied. Towards this end, four different models of increasing complexity are studied. The
Received in revised form equations corresponding to these models are formulated in the inertial reference frame.
5 October 2015
These equations involve a periodically varying coefficient. This is similar to classical
Accepted 11 October 2015
Mathieu equation but in a multi-degree of freedom context. As such, under suitable
Handling Editor: L.G. Tham
Available online 29 October 2015 parameter combination these systems result in growing oscillation amplitudes or
instability. For wider generalization, the equations and results are presented in a non-
dimensional form. The unstable parameter regimes are found using the Floquet theory
and perturbation methods. These results are also corroborated with existing results in the
literature. The nature of the stability boundary and its dependence on various system
parameters is discussed in elaborate detail. The stability boundary can be used to deter-
mine unstable operating speed ranges for different asymmetric shaft cross-sections.
Further, material, geometry and bearing selection guidelines for ensuring stable opera-
tions can be inferred from these results.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Parametric excitation in rotor dynamic systems frequently arises in different applications and is known to be a cause of
their failure. Such systems are characterized by a time-varying parameter. Certain regimes of system parameters are known
to cause instability resulting in large growth of vibration amplitudes. It is imperative for the designer to know a priori the
parameter ranges for the stable and unstable operation to ensure the reliability of the system. In the present work, we
investigate the instability due to such parametric excitation in a system comprising an asymmetric shaft mounted on
symmetric bearings. Such asymmetric shafts frequently arise practical applications as keyed or splined shafts. The varying
moments of inertia along different directions result in continuously changing stiffness of the rotating shaft, which even-
tually leads to parametric excitation and instability thereof.
Lee [1] in his works has presented a formulation for determining the stability or instability of rotating asymmetric shaft.
The equations of motion for such shaft formulated in the inertial reference frame show a parametric effect due to changing
stiffness. On transformation of the equations of motion to a reference frame fixed to the rotating shaft, the equations of
motion do not bear any time-varying parameter and are simplified to a constant coefficient differential equation. Through
these equations the unstable speed range for a given shaft cross-section is identified. However, in this work the authors limit
their analysis to a lumped two degree of freedom model of the shaft system. Arnold and Haft [2] have extended this
approach for continuous flexural deformation of the shaft. However, only transverse inertia and bending deformation are

n
Corresponding author.
E-mail address: as_sekhar@iitm.ac.in (A.S. Sekhar).

http://dx.doi.org/10.1016/j.jsv.2015.10.008
0022-460X/& 2015 Elsevier Ltd. All rights reserved.
R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291 277

Nomenclature Kn non-dimensional mean stiffness of the


asymmetric shaft
γ a number between 0 and 1 Kb bearing stiffness
δK difference in stiffness of the asymm- K b non-dimensional bearing stiffness
etric shaft l length of shaft
δK  non-dimensional difference in stiffness of the m equivalent mass of the shaft
asym- mb bearing
  mass
y
metric shaft r
yz
ζ damping factor r z
ω angular velocity R radius
b minor axis t time
C 1 ; C 2 … arbitrary constants u displacement y transformed into rotating
C equivalent viscous damping coordinates
Cm monodromy matrix v displacement z transformed into rotating
E Young's modulus of shaft material coordinates
fn non-dimensional displacement given by fl y; y1 shaft displacement along y-axis
g n
non-dimensional displacement given by gl yn, zn y z
,
l l
G n
non-dimensional gyroscopic parameter y1 non-dimensional displacement of shaft along
h major axis, key way hight y-axis
I1, I2 area moment of inertia evaluated about prin- y2 non-dimensional displacement of bearing
cipal coordinates of an asymmetric shaft cross along y-axis
section z; z1 shaft displacement along z-axis
Iyy area moment of inertia of asymmetric shaft z1 non-dimensional displacement of shaft along
cross section about y-axis z-axis
Izz area moment of inertia of asymmetric shaft z2 non-dimensional displacement of bearing
cross section about z-axis along z-axis
n
K equivalent mean stiffness of the represents non-dimensional quantities
asymmetric shaft € represents acceleration
_ represents velocity

accounted for in their formulations. Both analytical and experimental results of an overhung asymmetric shaft with
asymmetric disc at the end are presented. System gyroscopy is neglected in the study. Frohrib [3] studied the stability of a
rotor–shaft–bearing system accounting for gyroscopic effects. They evaluated the influence of shaft stiffness on instability.
The coupling effect between the bearing and the asymmetric shaft has been studied by Rajalingham et al. [4]. They
investigated the influence of bearing damping on stability using a lumped model with four degrees of freedom. The work is
useful to suitably chose bearing parameter so that the instability can be avoided. Kang et al. [5–7] presented a more
complete analysis with the inclusion of bending deformation, lateral inertia, rotary inertia, and gyroscopic inertia for sta-
bility. However, both internal damping and shear deformation are ignored. In contrast, Genta [8] proposed a finite element
based solution for the equations of motion for a general rotating system. These equations of motion, likewise are formulated
in a rotating reference frame. Further, the rotating system can include shafts, rotors, bearings or any combinations of them.
The model is capable of handling viscous, hysteretic damping, rotating and non-rotating part asymmetry, unbalance
distribution.
Frulla [9] used the Floquet theory to investigate the stability characteristics of a lumped bearing–shaft–rotor model.
Ganesan [10] obtained analytical expressions of asymmetric shaft response and phase. The author emphasizes on near
resonance behaviour of asymmetric shaft. Stability criteria are obtained based on effect of shaft asymmetry, and bearing
damping. Pei [11] has studied an asymmetric shaft system with axial loading and gyroscopic effect. Results of two solution
techniques namely Bolotin's method and the Floquet theory are compared. Author's main emphasis is on the assumptions of
Bolotin's method which results in enlarged stability boundary that contradict with Floquet theory results. Raffa and Vatta
[12] have studied a symmetric Bernoulli beam with shear effect and the Rayleigh beam with parametrically varying axial
force. The assumed solution to the governing equation with simply supported shaft boundary conditions results in classical
SDOF Matheiu equation. The stability of shaft with respect to parametrically varying axial force is described over stability
boundary of Mathieu equation reported in the literature. Bartylla [13] has studied asymmetric shaft with parametrically
varying axial loading. The Floquet theory is used to obtain the stability plot with respect to the parameters of parametrically
varying axial force.
Parametric vibrations are also observed in rotating machinery with faults such as crack and geared rotor bearing system.
Significant work is done in this area in which the authors studied the equation of motion in inertial frame. Shudeifat [14]
used time varying moment of inertia to model cracked rotor in inertial frame. The formulated equation of motion in finite
elements contains time varying stiffness matrix. The Harmonic Balance method is used to solve the governing equations. A
278 R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291

similar equation of motion is treated with both Floquet theory and harmonic balance method in [15] for stability analysis. In
this work the author has followed the sign of semi-infinite coefficient matrix formed out of harmonic balance method for
stability analysis. Guo et al. [16] used damage functions to model cracked rotor. The resulting finite element equations are
parametric in nature. The Floquet theory is used to obtain the stability plot. The authors presented the effect of damping on
stability boundaries. Several other authors who studied parametric vibrations in cracked rotors are [14,17–20], study on
parametric vibrations in geared system with slant and transverse crack are reported by [21,22] and geometrically nonlinear
shaft [23], etc. are also reported in the literature.
In the present work, analytical conditions for stability are obtained for a general asymmetric shaft system. Towards this
end, equations of motion are formulated in an inertial reference frame, progressively including the effects of shaft asym-
metry, bearing mass and bearing stiffness. As elaborated above, most other works reported in the literature approached this
problem by transforming the equations of motion in the rotating reference frame. In the later approach, the mathematical
solution is considerably simplified as the transformed equations of motion do not have any parametric excitation term. It is
believed that the present approach is beneficial for direct comparison with experimental results and other practical
applications (such as condition monitoring) wherein measurements are possible only on the stationary reference frame. The
scope of this work, however, is limited to theoretical investigations.
Despite the voluminous earlier works, not much clarity has evolved in terms of design guidelines which can be
applicable for a general asymmetric rotor–shaft–bearing system. Most of the above works have focused on the specific
systems and have not brought about any unification in the results presented for the different variety of possible rotor
dynamic systems with parametric excitation. In the present approach, the equations of motion are formulated and solved in
a non-dimensional form. Thus, the results obtained are applicable to different shaft cross-sections as elaborated through
examples in the subsequent sections. The goal of the study is to derive generalized criteria for determining stability/
instability zones in terms of the operating system parameters. Such generalized results will be useful to obtain design
guidelines for a wide variety of rotor applications.

2. Asymmetric shaft on rigid bearings

In this section, the work pertaining to the formulation of equations of motion of asymmetric shaft mounted on rigid
bearing and the solution of the formulated equations are presented. The equations and the solution process are performed
through a non-dimensionalization scheme. Finally, the results obtained are discussed in the context of specific examples of
elliptic shafts and shafts with key way as shown in Fig. 1.

2.1. Equation of motion

The deformations in the plane of cross-section of an asymmetric shaft are studied through a two degree of freedom
system model. Starting from the equilibrium conditions, the equations of motion for the in-plane deformation of the shaft
cross-section (denoted by y and z) are given by [1] as
my€ þC y_ þ ðK þ δK cos ð2ωtÞÞy þ δK sin ð2ωtÞz ¼ 0; (1a)

mz€ þC z_ þ ðK  δK cos ð2ωtÞÞz þ δK sin ð2ωtÞy ¼ 0; (1b)


Eq. (1) and its stability plot are presented in [15]. It is also analogous to the equation of motion that appear in [14].
However in the present work, Eq. (1) is studied by defining certain non-dimensional numbers which leads to different form
of stability plot as will be discussed later.
Note, these equations are applicable to the stationary inertial reference frame. Here, K is the mean of equivalent stiff-
nesses evaluated about the principal axes, δK is the difference of the equivalent stiffnesses evaluated about the principal

y y
Bearing
Asymmetric Shaft
h
x h z h z

Fig. 1. (a) Asymmetric shaft bearing system, (b) elliptic shaft cross section, (c) circular shaft with key way.
R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291 279

Table 1
Relation between the dimensional quantities in Eq. (1) and the corresponding non-dimensional quantities in Eq. (3).

Physical description Dimensional quantity Non-dimensional quantity Transformation relation

Time t tn
t  ¼ ωt
Mean stiffness K Kn K  ¼ mω
K
2

Difference in stiffness δK δK  δK  ¼ mω
δK
2
pffiffiffiffiffiffi
Damping C Cn 
C ¼ 2ζ K 
Displacement y, z, r yn, zn, rn y ¼ yl, z ¼ zl, r  ¼ fy z gT

axes. The equivalent stiffness at the appropriate cross-section of the shaft can be determined using the results of Euler–
Bernoulli or higher order beam theory. In the present work, attention is confined to the mid-section of the shaft which
3
results in the equivalent stiffness of K eq ¼ 48EI=l as per Euler–Bernoulli beam theory. Due to asymmetric shaft cross-
sections, the principal moments of inertia Izz and Iyy are unequal. The mean stiffness (K) and the difference in stiffness (δK)
in Eq. (1) are therefore given by
   
48E I yy þ I zz 48E 48E I yy I zz 48E
K¼ 3 ¼ 3 ðI m Þ; δK ¼ 3 ¼ 3 ðI d Þ: (2)
l 2 l l 2 l
δK represents the asymmetry parameter of the shaft. The equations of motion in Eq. (1) represent a linear time-varying
system as both the direct and cross-stiffness show periodic variation. For further analysis, these equations are transformed
into non-dimensional form, as follows:
By taking the transformation tn ¼ ωt, Eq. (1) is non-dimensionalized and represented in matrix form as
 (  ) "  #( ) "  #
1 0 y€ C 0 y_ K þ δK  cos ð2t  Þ δK  sin ð2t  Þ y
þ þ ¼0 (3)
0 1 z€
 0 C 
z_ δK 
sin ð2t 
Þ K 
 δ K 
cos ð2t 
Þ z

The transformation between the dimensional quantities in Eq. (1) and non-dimensional quantities in Eq. (3) are as
elaborated in Table 1. Subsequent analysis and solution will be limited to Eq. (3)
Eq. (3) contains the stiffness matrix which is a function of time and is π-periodic. This is in the form of classical Mathieu
equation but with two degrees of freedom. The Floquet theory is a well-known method for determining the stable/unstable
parameter zones of system of equations with periodic coefficients. In the present study, both asymptotic series expansion
and Floquet theory [24] are used to obtain the stability boundary of Eq. (3).

2.2. Solution

The two degrees of freedom Mathieu equation formulated in Eq. (3) are studied using a perturbation technique. In
particular, the focus is on determining the stability boundary as a function of different parameters in the equation. Viscous
damping is not included here because it does not result in any new unstable zone. Internal damping and hysteretic damping
are not in the scope of this work which lead to instability.

2.2.1. Perturbation method


Here a straight forward asymptotic series solution [24] is assumed for Eq. (3) with zero damping in power series of δK  .
y ðt  ; δK  Þ ¼ y0 ðt  Þ þ δK  y1 ðt  Þ þ ðδK  Þ2 y2 ðt  Þ… (4a)

z ðt  ; δK  Þ ¼ z0 ðt  Þ þ δK  z1 ðt  Þ þ ðδK  Þ2 z2 ðt  Þ… (4b)


In the above asymptotic series, the asymmetry of the shaft quantified by δK is chosen as the asymptotic parameter such


that 0 o δK  ⪡1. Substituting Eq. (4), in Eq. (3) and collecting terms at various orders of δK  , the following are obtained At
Oð1Þ
y€0 þ K  y0 ¼ 0 (5a)

z€0 þ K  z0 ¼ 0 (5b)

At OðδK Þ 

y€1 þ K  y1 þ cos ð2t  Þy0 þ sin ð2t  Þz0 ¼ 0 (6a)

z€1 þ K  z1  cos ð2t  Þz0 þ sin ð2t  Þy0 ¼ 0 (6b)

The general solution of Eq. (5) can be expressed as


pffiffiffiffiffiffi pffiffiffiffiffiffi
y0 ¼ Q 1 sin ð K  t  Þ þ Q 2 cos ð K  t  Þ (7a)
280 R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291

pffiffiffiffiffiffi pffiffiffiffiffiffi
z0 ¼ Q 3 sin ð K  t  Þ þ Q 4 cos ð K  t  Þ (7b)

Substituting Eq. (7) into Eq. (6) and solving it result in the following expression:
!
pffiffiffiffiffiffi pffiffiffiffiffiffi 1 pffiffiffiffiffiffi pffiffiffiffiffiffi
   

y1 ¼ Q 5 sin K t þ Q 6 cos K t þ pffiffiffiffiffiffi 
 K  þ K  ðQ 2 þQ 3 Þ cos K   2 t
K ð8K  8Þ
pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffi
þ ðK  K ÞðQ 2  Q 3 Þ cos ðð K þ2Þt Þ  ðK þ K Þð Q 4 þ Q 1 Þ sin ðð K  2Þt  Þ þ sin ðð K  þ2Þt  ÞðK   K  ÞðQ 1 þQ 4 ÞÞ
  

(8a)
!
pffiffiffiffiffiffi pffiffiffiffiffiffi 1 pffiffiffiffiffiffi pffiffiffiffiffiffi
z1 ¼ Q 7 sin K  t  þ Q 8 cos K  t   pffiffiffiffiffiffi 
 K  þ K  ðQ 2 þ Q 3 Þ cos K   2 t
K ð8K  8Þ
pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffi
þ ðK  K ÞðQ 2  Q 3 Þ cos ðð K þ2Þt Þ  ðK þ K Þð Q 4 þ Q 1 Þ sin ðð K  2Þt  Þ þ sin ðð K  þ2Þt  ÞðK   K  ÞðQ 1 þQ 4 ÞÞ
  

(8b)
 n n
Eq. (8) contains ð8K  8Þ in the denominator. Eq. (8) becomes infinity if K ¼1. Hence the equation breaks when K ¼ 1.
If the analysis is carried out for a very small values of δK (close to zero) which means Eq. (3) represents a single spring–
mass system that has only one unstable operating frequency. It is natural frequency of the system and the corresponding
non-dimensional number Kn is 1. Hence the system break down occurs here. The same result is obtained by perturbation
analysis too.
Analysis using straight forward expansion is limited to Oð1Þ and it can be proved that further analysis does not lead to
any value of Kn that breaks the solution. In contrast, SDOF Mathieu equation results in many values of non-dimensional
number for which the solution breaks. Hence, there are infinite number of unstable zones (given by [24]), whereas a lumped
model of asymmetric shaft (Eq. (3)) results in only one unstable zone. This is because of the type of cross coupling that exists
in the equation of motion. This phenomenon is also known as co-existence which is reported by [25,26]. The cross coupling
term is also a parametric term. If the cross coupling term is absent in Eq. (3) then it results in two uncoupled Mathieu
equations. Stability plot of these two Mathieu equations contains infinite number of unstable and stable regions.
Since the straight forward expansion fails when K  ¼ 1 this suggests expanding Kn around the value 1 in powers of δK  in
addition to expanding y ðt; δK  Þ and z ðt; δK  Þ. The viscous damping does not lead to a new unstable region. It mitigates
vibration hence reduces the existing unstable region. So, the damping is included here and expanded in power series of δK  .
The expansion is given by

y ðt  ; δK  Þ ¼ y0 ðt  Þ þ δK  y1 ðt  Þ þ ðδK  Þ2 y2 ðt  Þ⋯ (9a)

z ðt  ; δK  Þ ¼ z0 ðt  Þ þ δK  z1 ðt  Þ þðδK  Þ2 z2 ðt  Þ⋯ (9b)

K  ¼ 1 þ δK  K 1 þ ðδK  Þ2 K 2 þ ⋯ (9c)

C  ¼ δK  C 1 þ ðδK  Þ2 C 2 þ⋯ (9d)

Substituting Eq. (9) into Eq. (3) and equating the coefficients of each power of δK , the following are obtained At Oð1Þ


y€0 þy0 ¼ 0 (10a)

z€0 þz0 ¼ 0 (10b)

At OðδK Þ 

y€1 þC 1 y_1 þy1 þ ð cos ð2t  Þ þK 1 Þy0 þ sin ð2t  Þz0 ¼ 0 (11a)

z€1 þ C 1 z_1 þz1 þð  cos ð2t  Þ þ K 1 Þz0 þ sin ð2t  Þy0 ¼ 0 (11b)

The general solution of Eq. (10) can be expressed as


y0 ¼ Q 1 sin ðt  Þ þ Q 2 cos ðt  Þ (12a)

z0 ¼ Q 3 sin ðt  Þ þ Q 4 cos ðt  Þ (12b)


 
Substituting Eq. (12) into Eq. (12) and simplifying give rise to terms that contain sin ðt Þ or cos ðt Þ which are called as
secular terms. These terms are nothing but forcing terms with frequency equal to undamped natural frequency. Hence such
terms in Eq. (12) blow up the solution. The terms are eliminated by collecting the coefficients of sin ðt  Þ and cos ðt  Þ from Eq.
(12) which results in four simultaneous algebraic equations. The equations are listed in the form of matrix in Eq. (13).
R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291 281

Eliminating the secular terms is possible by solving these equations for the constants Q1, Q2, Q3 and Q4.
2 38 9 8 9
C 1 K 1 þ 0:5 0:5 0 >
> Q1 >
> >0>
6 7>> > >
> > >
>
6 0:5 0 C 1 K 1  0:5 7< Q 2 = < 0 =
6  7 ¼ (13)
6 K 0:5  C1
0:5 7
4 1 0 5>>
>
Q3 >
> >
> >0>
> >
>
:Q >
> ; :0;
0 0:5 K 1 þ 0:5  C 1 4

Eq. (13) is a homogeneous algebraic system of equations. Solving it for non-trivial solution gives the relationship between
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
C 1 and K 1 that is K 1 ¼ 7
ðC 1 Þ2 1. Substituting the value of K 1 in Eqs. (9c) and (9d) the following relation is obtained as

ðδK  Þ2 ¼ ðC  Þ2 þðK   1Þ2 (14)


The perturbation analysis is stopped at this point. It can be proved that even if it is continued further with higher order of
δK  , all the coefficients of higher order δK  (K 2 , K 3 …) are zero.
When Cn ¼0 Eq. (14) reduces to K  ¼ 7 δK  þ 1 which represents equation of two straight lines. Intercept of the straight
lines is at value 1 on Kn-axis. It is also clear that the product of slopes of both the straight lines is  1 that means the angle
between these two straight lines is 90°. This inference is an important advantage of the perturbation analysis as compared
to the Floquet theory which will be discussed in the next section. The magnitude of slope of each straight line is 1. Hence the
acute angle made by thepstraight
ffiffiffiffiffiffi lines with δK  ¼ 0 is 45°.
By assuming C as 2ζ K , Eq. (14) becomes,
n

pffiffiffiffiffiffi
ðδK  Þ2 ¼ ð2ζ K  Þ2 þ ððK  Þ2  1Þ2 (15)
which is the stability boundary equation.
In general at each order of δK  avoiding secular terms results in a relationship between the non-dimensional parameters
corresponding to periodic solution of Eq. (3). Hence, Eq. (15) is the analytical condition between non-dimensional numbers
which results in a periodic solution. In fact it determines a critical state crossing which instability sets in. But this analysis
does not give any information about the stable and unstable volume in the non-dimensional parametric space on either side
of this surface.
The Floquet theory can be used to determine the stable and unstable zone which is discussed in the next section.

2.2.2. Floquet theory


Application of the Floquet theory to Eq. (1) is discussed in this section (Detailed treatment of the Floquet theory on
classical Mathieu equation can be found in [24].) The equation of motion given by Eq. (3) can be represented in compact
form as
     
½I 22 r€ 21 þ C  22 r_ 21 þ K  ðtÞ 22 fr  g21 ¼ 0 (16)

It can be written in state space form as

R_ þ ½BðtÞR ¼ 0 (17)
  T h i
where R ¼ r r_ , ½Bðt Þ ¼ ½K   ½C  
 0 ½I

Since ½K   is periodic with period π, the matrix ½B becomes periodic with period π. The solution is carried out by
assuming initial conditions as columns of a unit matrix of size 4  4 which is called as fundamental solution matrix. With
this, Eq. (17) is solved numerically from t  ¼ 0 to t  ¼ π . Each initial condition vector gives another solution vector at the end
of time π and a matrix of size (4  4) is formed using all the solution vectors evaluated at the end of time π. This solution
matrix is called Monodromy matrix and is represented by Cm. The stability of the response depends on the eigenvalues of Cm
as given below

 If magnitude of maximum eigenvalue of Cm is more than 1 then the system is unstable.


 If magnitude of maximum eigenvalue of Cm is less than 1 then the system is stable.
 If magnitude of maximum eigenvalue of Cm is equal to 1 then the response of the system is periodic.

The eigenvalues of matrix Cm are called Floquet exponents. The same computation is repeated for all possible values of Kn,
δK  and ζ.
The values of Kn, δK  , ζ for which the maximal Floquet exponent is unity, are collected. All these points are plotted in the
Kn, δK  and ζ space, thus forming the stability boundary surface. It is verified that these points lie on the surface given by Eq.
(15) found by perturbation analysis. In Fig. 2, the analytical surface given by Eq. (15) is presented. The stable and unstable
parameter regimes falling in either side of this surface. It is verified that the criteria of instability as listed above are satisfied
by the parameters in the region marked as unstable in Fig. 2. Similarly, in the parameter zones marked as stable in Fig. 2, it is
found that the stability condition holds. For different values of damping (ζ), the stability boundary given by Eq. (15) is
presented in Fig. 3 as a two dimensional plot in the Kn, δK  space. The stability regions (S1, S2) and unstable region (U) are
identified over the non-dimensional parameter space as shown in Fig. 3.
282 R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291

Unstable Zone
Stable Zone
0.04

1.6
ζ

0.02
1.4

0
1.2
0.6
0.5 1 K*
0.4 Stable Zone
0.8
0.3
δ K* 0.2 0.6
0.1
0 0.4

Fig. 2. Stability zone surface plot using Eq. (15). The Floquet theory is used to determine stable/unstable region.

0.5 Comparison of analytical solution and Floquet theory

0.45

0.4

0.35 U
0.3
S2
δ K*

0.25

0.2
S1
0.15 Analytical soln
0.1
zero damping
4% damping
0.05 6% damping
8% damping
0
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
K*

Fig. 3. Comparison of the Floquet theory with analytical formulation.

2.3. Examples – results

In thepprevious
ffiffiffiffiffiffi section, the Floquet theory was applied to Eq. (3) which consists of three parameters Kn, δK  and ζ
( ¼C =2ζ K ). The stability boundary found in the parametric space Kn–δK  is presented in Fig. 3. As observed from this


result, the stability boundary is in V-shape. The region interior to the V-shape is unstable zone and rest is stable. Due to the
nature of non-dimensional numbers used in the present work the stability boundaries are in straight line form as opposed
to a curved form shown in [15]. Corner of V-shape gets filleted with an increase in damping ratio. Similar stability plots are
presented by [16]. In this work also inclusion of damping results in filleting corner of the stability plot and the stability plot
is de-attached from horizontal axis which results in reduction of unstable zone. Thus, with an increase in damping the
unstable zone is reduced. This is in line with the physical understanding that the instability effect is to be mitigated by
damping effects.
Any line drawn for a constant δK  intersects the stability boundary at two points which represents the milts of an
unstable parameter range. With increase in δK  , the unstable range increases. Physically, this implies dominance of
instability effect with greater asymmetry of the shaft cross-section. On the other extreme, a shaft system with δK  ¼ 0
R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291 283

represents a symmetric shaft system. From Fig. 3, it can be noted that this corresponds to the Kn-axis in the non-dimensional
parametric space. It can also be observed that in this case, the parametric nature disappears in Eq. (3) with choice of δK  ¼ 0.
Herein, the unstable range gets shrinked to one operating speed which for a symmetric shaft system is the undamped
critical speed. Perturbation analysis is done with increments of perturbation parameter i.e. δK  . When δK  is zero Eq. (3)
pffiffiffiffiffiffiffiffiffi
represents a simple spring–mass-damper system. There exists only a critical speed. That is when Kn ¼1 implies ω ¼ k=m
which is critical speed (refer to (1)). As the order of perturbation parameter increased the critical speed gets spilled over into
a range of operating speed in which the system is unstable. The extreme values of this range of unstable speeds correspond
to the natural frequencies computed with minimum area moment of inertia and maximum area moment of inertia [1].
As mentioned in Section (2.2) the lines in V-shaped stability plot that correspond to undamped system make 45° with
δK  ¼ 0. Hence the area within by δK  ¼ 0, Kn ¼ δK  þ1 and δK  ¼ 1 in Fig. 3 (marked with S2) indicates stable. The system
when operated at near zero angular velocity, corresponds to very high value of non-dimensional parameters Kn and δK  (see
Table 1). But the value of δK  is always smaller than Kn because the value of Id is less than Im for any asymmetric shaft.
Hence, at a very low operating speed the non-dimensional parameters always lie in the area S2. So an asymmetric shaft
being operated at a very low speed is always stable. Eq. (1) also validates this because at ω ¼ 0 the parametric terms vanish.
Another extreme case is operating speed being large which represents origin (and S1 region) in Fig. 3. From Fig. 3 the origin
S1 is in stable region. Thus asymmetric shafts rotating at very high speeds also are stable as per asymmetric shaft equation of
motion analysis. The stability issues associated with an elliptic and a circular shaft with key way are explained in the
following sections.

2.3.1. Elliptic shaft


As an example of asymmetric shaft instability analysis in this section a study of instability zones for an elliptic shaft is
performed.
The elliptic shaft cross section is shown in Fig. 1b. From Eq. (2) and Table 1, the non-dimensional numbers can be related
as δK  ¼ ðI d =I m ÞK  . It is an equation of straight line with slope I d =I m , passing through the origin. Here, I d =I m is a measure of
asymmetry. Referring to the geometry in Fig. 1b, I d ¼ ð1=2Þðπ bh =64  π hb =64Þ and I m ¼ ð1=2Þðπ bh =64 þ π hb =64Þ )
3 3 3 3

2 2
I d =I m ¼ ððh=bÞ  1Þ=ððh=bÞ þ 1Þ for all br h. Thus, asymmetry increases with the aspect ratio h=b. Thus, in the non-
dimensional parameter space, for a given asymmetric cross-section (hence given K, δK and m) the locus of points corre-
sponding to different operating speed ω forms a straight line passing through the origin, the equation of which is
δK  ¼ ðId =I m ÞK  . The slope of this line is proportional to the asymmetry of the cross-section.
Intersection of the straight lines with the V-shaped stability boundary indicates the unstable speed zone. As shown in
Fig. 5, for higher aspect ratio, the portion of straight line within the unstable zone is higher. This is because as discussed
above when the aspect ratio is higher, the slope of the non-dimensional parameter line is higher. Thus, with a higher
asymmetry a larger unstable speed range is obtained.
Lee [1] showed that the limits of the unstable speed range of an asymmetric cross section can be determined by eval-
uating the critical speeds corresponding to the maximal and minimal stiffness values. These correspond to the area
moments of inertia of the cross-section being taken about the principal axes. This was proved by formulating the equations
of motion in a rotating reference frame wherein the equation simplifies to a non-parametric form. The present formulation
in inertial reference is in agreement with the above finding as follows. Taking intersection of the straight line δK  ¼ ðI d =I m ÞK 

Stability plot w.r.t gyroscopic parameter


0.6

0.5

0.4
δ K*

0.3

0.2

G=0
0.1 G=0.1
G=0.2
G=0.3
0
0.4 0.6 0.8 1 1.2 1.4

K*
Fig. 4. Stability plot with gyroscopic effect.
284 R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291

(which represents locus of points with varying speed) with the stability boundaries K  ¼ 7ðδK   1Þ, gives δK  ¼ I d =ðI m 8I d Þ,
K  ¼ I m =ðI m 8I d Þ. Using definition of δK  and Kn from Eq. (2), the value of operating speed can be obtained as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ω1;2 ¼ 48EI1;2 =ml3 .
As stated earlier, it is expected that with an increase in damping the unstable zone will be reduced. This is verified from
Fig. 5, since higher the damping higher the fillet radius of the V-shaped stability boundary which results in a reduced
unstable region. Further, there exists a particular amount of damping for which the non-dimensional line corresponding to
varying speed of a given asymmetric shaft becomes tangent to the filleted stability boundary. Under such conditions the
unstable speed range is completely eliminated. Hence, the result presented in Fig. 5 is useful to decide upon the amount of
damping to be incorporated to stabilize an asymmetric shaft. As per this analysis an elliptic shaft with h=b ¼ 1:175 and
ζ ¼0.08 is noted to be completely stable irrespective of the material properties of the shaft.
Fig. 5 is also useful to select material. For example, a steel shaft of 0.5 m span 16 mm minor diameter and h=b ¼ 1:3 is
unstable between 280 rad/s to 364 rad/s using [1]. It is unstable at 350 rad/s. The geometry specified above leads to a fixed
I d =I m ¼ 0:2565. Corresponding to a the specified speed K  ¼ 0:8649 and δK  ¼ ðI d =I m ÞK  ¼ 0:2218. This point lies in the
unstable range of the non-dimensional results presented in Fig. 5. The same operating angular velocity can be stabilized by
changing the material. If titanium shaft is used the corresponding non-dimensional numbers K  ¼ 8683 and
δK  ¼ ðI d =Im ÞK  ¼ 2227 are in the stable region. This point is not be shown in Fig. 5 because it is far beyond the axis limits
chosen.

2.3.2. Circular shaft with key way


Fig. 1 c shows circular shaft with key way. In this case also δK  ¼ ðI d =I m ÞK  is valid because it is derived for an arbitrary
shaft cross section. Hence locus of non-dimensional numbers follows straight line that passes through origin. It can be
proved using Eq. (2) that I d =I m increases with increase in ðh=RÞ value. So ðh=RÞ increases the asymmetry and higher the value
of ðh=RÞ higher the slope of non-dimensional stability line. The non-dimensional stability lines for the cross section are

0.5
zero damping Unstable Zone
0.45 2% damping
4% damping
0.4
6% damping h/b=1.3
0.35 8% damping

0.3
δ K*

0.25
h/b=1.175
350 rad/s
0.2 Stable Zone Steel shaft
h/b=1.13
0.15 Increase in
angular velocity
h/b=1.0834
0.1

0.05 h/b=1.04
Stable Zone
0
0 0.5 1 1.5
K*

Fig. 5. Non-dimensional lines for elliptic shafts with different h=b value. The intersection of these lines with the stability boundary is shown.

0.25

0.2 h/R=0.8
δ K*

0.15 h/R=0.58

0.1 h/R=0.45
zero damping
2% damping
0.05 4% damping
6% damping
8% damping
0
0.7 0.8 0.9 1 1.1 1.2 1.3
K*

Fig. 6. Stability plot of circular shaft with key way.


R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291 285

plotted on V-shaped stability boundary as shown in Fig. 6. As explained in the elliptic shaft case, higher slope of the straight
line results in higher unstable speed range and it is evident from Fig. 6.
Identical design guidelines such as material change and appropriate amount of damping can be drawn for this case also.
An interesting condition that is common to any cross section is δK  ¼ 0, which represent a symmetric shaft system and as
discussed earlier the unstable range shrinks to a single point corresponding to the critical speed.
It can be observed that Id is always greater than zero and I d oI m . Hence the ratio I d =I m takes values between zero and one.
It is slope of non-dimensional straight line drawn for any asymmetric shaft. So the slope of non-dimensional straight line
δK  ¼ ðId =I m ÞK  lies between zero and 45°.

2.4. Effect of gyroscopic couple

The effect of gyroscopic couple on asymmetric shaft dynamics is incorporated by adding a skew-symmetric gyroscopic
matrix to the equation of motion of asymmetric shaft. Thus, the reformulated equations are
 ( € ) " #( ) "  #
1 0 y 0 G y_ K þ δK  cos ð2t  Þ δK  sin ð2t  Þ y
þ þ ¼ 0; (18)
0 1 z€ G 
0 z_ δK sin ð2t Þ
 
K  δK cos ð2t Þ
  
z

where G is the non-dimensional gyroscopic parameter.


The Floquet theory is applied to obtain the stability plot which is given in Fig. 4. From these results, it is clear that
gyroscopic action induces a leftward movement of the stability boundary. This in turn will affect the unstable operating
speed range as discussed in Section 2.3.1 and Fig. 6.

3. Asymmetric shaft supported on isotropic bearings

3.1. Massless bearings

In the previous section, the bearings are idealized as rigid. However, in reality bearings offer flexibility. The analysis of a
non-circular shaft supported on isotropic bearings is presented here. The equations of motion are formulated by including a
massless bearing (see (7), but mb is neglected) that offers stiffness Kb alone (studied by [4] in rotating coordinates) are given
by
 
y€1 þ K  þ δK  cos ð2t  Þ ðy1  y2 Þ þ ðδK  sin ð2t  ÞÞðz1  z2 Þ ¼ 0 (19a)
 
z€1 þ K   δK  cos ð2t  Þ ðz1  z2 Þ þ ðδK  sin ð2t  ÞÞðy1  y2 Þ ¼ 0 (19b)
 
K  þ δK  cos ð2t  Þ ðy2  y1 Þ þ ðδK  sin ð2t  ÞÞðz2 z1 Þ þ K b y2 ¼ 0 (19c)
  
K  δK  cos ð2t  Þ ðz2  z1 Þ þðδK  sin ð2t  ÞÞðy2  y1 Þ þK b z2 ¼ 0 (19c)

Here K bis the non-dimensional bearing stiffness. It is taken as an integer times K . n

In this case, there are three non-dimensional parameters arising in the system of equations above – Kn (representing the
mean stiffness of the asymmetric shaft), δK  (representing the difference stiffness of the asymmetric shaft) and K b
(representing the bearing stiffness). The stability boundary is obtained using the Floquet theory for different values of K b =K  .

z*2
z*1

Asymmetric Bearing *
shaft
y* K K
3 b
Y

Y m s
X
Z

y*
K K
1 2
m b

K = K*+δ K* cos(2t*) *
1 K
b

K = δ K* sin(2t*)
2

K = K*−δ K* cos(2t*) X
3 Z

Fig. 7. Equivalent spring mass model of asymmetric shaft supported in flexible bearings.
286 R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291

Rigid bearing
0.6 *
Kb=100K*
Unstable zone
*
Kb=30K*
0.5 *
Kb=20K*
*
Kb=10K*
0.4
K* =5K*
b Steel shaft
operating
δ K*

at 1170 rad/s
0.3 Unstable
zone h/R=0.8

0.2 U1
h/R=0.58
Stable zone
h/R=0.45
0.1

Stable zone
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
K*

Fig. 8. Stability plot of circular shaft with key way supported on isotropic massless bearings.

The stability plot is given in Fig. 8. The stability boundary is in V-shape that of the rigid bearing case shown in Fig. 5.
However, there is an additional stability boundary that emanates from origin. This branch is present for any value of K b . It
can be observed from Fig. 8 that this line emanating from origin has slope of 45°. Theoretically, it creates an extra unstable
region defined by K  ¼ 0 and δK  ¼ K  (marked as U1 in Fig. 8) which is not present in the case of asymmetric shaft sup-
ported on rigid bearings. The extra unstable region created has no practical significance. As explained in Section 2.3, for any
asymmetric shaft δK  oK  and thus it is impossible for a system to have Kn, δK  in the U1 region of Fig. 8.
Here, as the bearing stiffness changes in comparison to the shaft stiffness, the V-shaped stability boundary translates
along Kn-axis. At a very high value of stiffness of bearing the stability boundary coincides with the stability boundary
obtained in the case of asymmetric shaft on rigid bearings with zero damping (Fig. 3). This result corresponding to
undamped asymmetric shaft supported on rigid bearing is also presented in Fig. 8 with label “Rigid bearing” for reference.
As, in the previous study of rigid bearing, the non-dimensional numbers corresponding to a given cross section of
circular shaft with keyway (see Fig. 1c) for varying rotational speed are plotted on the stability plot shown in Fig. 8. Here,
also the locus of non-dimensional parameters Kn and δK  with changing ω, for any h=R is a straight line through the origin.
All the other characteristics related to the non-dimensional line and the stability boundary are analogous to the cases
discussed earlier. In this case, the designer has one more parameter to vary – namely the bearing stiffness relative to the
shaft stiffness – so as to stabilize the system at any particular speed of operation. Since the stability boundary moves along
positive Kn direction with decrease in stiffness, the choice of an appropriate bearing stiffness can stabilize the system at the
required operating speed. For example a steel circular shaft of diameter 16 mm, 0.5 m length, h=R ¼ 0:78 operating at
1170 rad/s is unstable if supported in bearings of stiffness K b ¼ 10  K  . Same shaft operating at same speed can be made
stable if K b ¼ 5  K  as shown in (8). Thus, here again we find that the general non-dimensional analysis presented in this
work can be used effectively in designing rotor–bearing system for stability.
Damping in both bearings as well as shaft is neglected in the present study. However, as in the previous cases it can be
shown that, addition of damping reduces the unstable speed range. The stability boundary will be similar to the case of
asymmetric shaft on rigid bearings, wherein it was observed that the corner of the stability boundary gets filleted due to
damping.

3.2. Bearings with mass

Here also the equivalent spring–mass model for an asymmetric shaft supported in flexible bearing with mass is assumed
as shown in Fig. 7. The equations of motion in non-dimensional form are given by
 
y€1 þ K  þ δK  cos ð2t  Þ ðy1  y2 Þ þ δK  sin ð2t  Þðz1 z2 Þ ¼ 0 (20a)

 
z€1 þ K   δK  cos ð2t  Þ ðz1  z2 Þ þ δK  sin ð2t  Þðy1 y2 Þ ¼ 0 (20b)

  
K δK    δK    K
y€2 þ þ cos ð2t  Þ y2  y1 þ sin ð2t  Þ z2  z1 þ b z2 ¼ 0 (20c)
γ γ γ γ
  
K δK    δK    K
z€2 þ  cos ð2t  Þ z2  z1 þ sin ð2t  Þ y2 y1 þ b y2 ¼ 0 (20d)
γ γ γ γ
R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291 287

Kb*=60K* Kb*=100K*
0.5 0.5
mb=0.1m mb=0.1m
0.45 mb=0.2m 0.45 mb=0.2m
mb=0.5m mb=0.5m
0.4 0.4

mb=0.1m
0.35 0.35

mb=0.2m
0.3 0.3 mb=0.1m
δ K*

δ K*
0.25 0.25
mb=0.2m
0.2 0.2

mb=0.5m
0.15 0.15

0.1 0.1
mb=0.5m

0.05 0.05

0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02

K* K*
n
Fig. 9. Stability curves at low K corresponding to very high speeds.

Kb*=60K* Kb*=100K*
mb=0.1m mb=0.1m
mb=0.2m mb=0.2m
1 mb=0.5m 1
mb=0.5m
Unstable zone Unstable zone
0.8 0.8
δ K*

δ K*

Steel shaft
0.6
operating at Stable 0.6
Steel shaft Stable
1170 rad/s zone operating at
1170 rad/s zone

0.4 h/R=0.58 h/R=0.8 0.4 Stable h/R=0.8


zone h/R=0.58
Stable
0.2
zone 0.2

h/R=0.45 h/R=0.45
0 0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

K* K*
Fig. 10. Stability curves at high Kn corresponding to lower speeds.

In this case along with Kn, δK  and K b , the bearing mass mb ¼ γ m; ð0 o γ o 1Þ is accounted. The Floquet theory is used to
obtain the stability boundary. The results are presented in two different plots with different non-dimensional number
scales, one in Fig. 9 and the other in Fig. 10 to show the boundaries clearly. Fig. 11 shows a zoomed in plot for the boxed
region in Fig. 9. For each value of bearing mass there are three values of Kn from which the stability curves emanate. Unlike
linear stability boundaries obtained in case of asymmetric shaft supported on rigid bearings or massless bearings (refer
results shown in Figs. 5 and 8), the stability boundaries obtained in this case are curves. This is similar to the stability
boundaries of classical single degree of freedom Mathieu equation [25].
Fig. 9 shows stability boundary for K b ¼ 100K  and K b ¼ 60K  at low Kn corresponding to very high speeds. In each case
of K b value three cases of bearing mass are assumed that is γ ¼0.1, 0.2, 0.5. For a particular value of bearing stiffness, it is
observed that the stability boundary is translated along positive Kn-axis as the mass of the bearing is changed. The
288 R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291

x 10 Kb*=60K* x 10 Kb*=100K*
4 5
mb=0.1m h/R=0.8
3.5
mb=0.2m 4.5
mb=0.5m
h/R=0.8 4
3
mb=0.5m
3.5
h/R=0.58
2.5 mb=0.2m 3
h/R=0.58 h/R=0.45
K*

2 2.5

δ K*
δ

2
1.5 mb=0.1m
h/R=0.45 1.5
1 mb=0.2m
1
mb=0.1m
0.5 mb=0.1m
0.5 mb=0.2m
mb=0.5m
mb=0.5m
0 0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05

K* K*
Fig. 11. Stability curves at very high speeds (zoomed view of box in Fig. 9).

separation can only be seen clearly with proper zooming into the plot since the translation is very small with respect to
change in mass. The separation of stability line are not seen clearly in Fig. 10 because it is plotted in a broader scale. In this
case, there are three unstable zones emanating from δK  ¼ 0. As in the previous cases, the non-dimensional straight lines
corresponding to given asymmetric shaft cross-section but continuously varying speed intersect these stability curves. One
unstable portion is shown in Fig. 10 and the other two are shown in Fig. 11. The plots are redrawn to appropriate scale in
these figures. In these figures, the non-dimensional plots with varying speed are presented for three different keyed shafts.
The appropriate h=R ratios are as indicated in the figures. In total there are three unstable zones for an asymmetric shaft
mounted on flexible bearings with mass. Whereas there is only one unstable zone in case of asymmetric shaft supported on
rigid bearings (Fig. 5) and asymmetric shaft supported on flexible massless bearings (Fig. 8).
Carrying on with the previous example of a steel circular shaft of diameter 16 mm, 0.5 m length, h=R ¼ 0:78 operating at
1170 rad/s is unstable for both K b ¼ 100  K  and K b ¼ 60  K  for different values of bearing mass as shown in Fig. 10. From
the results shown in Figs. 9 and 10, it is noted that the mass of the bearing has minimal influence on the stability boundary.
As in the previous models of rigid and spring-like bearing, in the present case too the non-dimensional plots as shown in
Figs. 10 and 11 can be used to decide appropriate materials, system damping, bearing stiffness, etc. to ensure that the system
remains stable during its operation.

4. Asymmetric continuous shaft

The equation of motion of a rotating Euler–Bernoulli beam in a reference frame mounted on the beam are given by [6]

  ∂4 u
ρA u€ 2ωv_  ω2 u þ EIu 4 ¼ 0 (21a)
∂x

  ∂4 v
ρA v€ þ 2ωu_  ω2 v þ EIv ¼0 (21b)
∂x4
The above equations are transformed from the non-inertial rotation frame above to an inertial frame [1]. Let y þiz ¼ R
represent the displacements in the inertial frame and u þ iv ¼ W represent the displacement in the rotating frame. These
displacements can be related as R ¼ Weiωt . Using this transformation Eq. (21) can be written in stationary coordinate system
as
   
∂4 y ∂4 z
ρAy€ þ E ðIm þId cos ð2ωt ÞÞ 4 þE ðId sin ð2ωt ÞÞ 4 ¼ 0 (22a)
∂x ∂x
   
∂4 z ∂4 y
ρAz€ þ E ðI m  Id cos ð2ωt ÞÞ þE ð I d sin ð 2 ωt ÞÞ ¼0 (22b)
∂x4 ∂x4
R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291 289

For simply supported boundary conditions, the solution of Eq. (22) can be written as yðx; tÞ ¼ f ðtÞ sin ðnπ x=lÞ,
zðx; tÞ ¼ gðtÞ sin ðnπ x=lÞ which satisfy the boundary conditions. Substituting the assumed solutions yðx; tÞ and zðx; tÞ in Eq.
(22) and simplifying results in
nπ 4 nπ 4
ρAf€ þ E ðI m þI d cos ð2ωt Þ f þ E ððI d sin ð2ωt ÞÞg ¼ 0 (23a)
l l
nπ 4 nπ 4
ρAg€ þ E ðI m I d cos ð2ωt Þ g þ E ððI d sin ð2ωt ÞÞf ¼ 0 (23b)
l l
Non-dimensionalizing equation (23) results in
  4   4
E nπ E nπ
f€ þ
  
ðI þ I cos ð 2t Þ f þ ððI d sin ð2t  ÞÞg  ¼ 0 (24a)
ρAω2 l
m d
ρAω2 l
  4   4
E nπ E nπ  
g€ þ ðI m  I d cos ð2t  Þ g  þ ðI d sin ð2t  ÞÞf ¼ 0 (24b)
ρAω 2 l ρAω 2 l
Eq. (24) is analogous to Eq. (3) without damping. Hence, the same result can be used to obtain the stability plot with
  4
E nπ
K ¼ Im (25a)
ρAω 2 l
  4
E nπ
δK  ¼ Id (25b)
ρAω2 l
This is in contrast to the lumped parameter model obtained in Eq. (3). In the lumped parameter model, the mean
stiffness (Kn) and difference in stiffness (δK  ) are obtained for a specific point on the shaft. In this work, we assume that this
point is chosen as the mid-point of the simply supported shaft. Accordingly, the non-dimensional stiffnesses are given in
Table 1. Hence, it is noted that the non-dimensional stiffnesses of lumped parameter model and continuous system model
are scaled by a factor of 48=n4 π 4 .
Using the analogy, the non-dimensional numbers of continuous beam when substituted in Eq. (14) for the stability
boundary take the form
    
EI d nπ 4 EI m nπ 4
¼ 7 1 (26)
ρAω2 l ρAω2 l
Whereas the stability boundary expression for lumped parameter model is of the form
    
48EI d 48EI m
¼ 7  1 (27)
mω2 l mω 2 l
3 3

Thus, for the continuous model there will be one stability boundary for each value of n. As brought out by the above two
equations, the nature of the stability plot for each n is identical to that obtained earlier in Fig. 3. However, due to the
rescaling of the non-dimensional stiffnesses the stability boundary for each mode of the continuous beam is rescaled. In
Fig. 12, the stability boundaries for the different modes of the continuous are plotted along with the stability boundary for
the lumped parameter model. Each of these stability boundaries is V-shaped. The separation between the stable and
unstable regions is analogous. These regions are as indicated in the figure. However, they are rescaled. For the nth mode the

0.2

0.18 Unstable
Unstable
Unstable
0.16

0.14

0.12
δ K*

0.1 n=1
stable
n=3
0.08
stable
0.06
n=2
0.04
stable
0.02 Lumped parameter model
Continuous system
0
0 0.5 1 1.5
K*

Fig. 12. Stability plot of continuous system model.


290 R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291

boundary emanate from K  ¼ 48=n4 π 4 ¼ 0:49=n4 . This is illustrated in Fig. 12 for n ¼ 1; 2; 3. Likewise, for the nth mode the

intercept on the δk axis is 48=n4 π 4 ¼ 0:49=n4 . For clarity, these intercepts are not shown in Fig. 12.

5. Conclusions

Dynamics of a rotating asymmetric shaft is studied. This system is known to have parametric excitation and thus
instability prediction for such systems is important. Most researchers have approached this problem by formulating the
equations of motion in a rotating reference frame attached to the shaft. This simplifies the mathematical treatment by doing
away with the parametric excitation term in the formulated equation. Related works in the context cracked rotating shafts
have formulated the governing equation in the inertial reference frame leading to differential equations with time varying
coefficients. In the present work, we propose a similar framework for presenting the stability boundary of any asymmetric
rotating shaft system. This is achieved through analysis of the equations of motion formulated in an inertial reference frame.
We believe the present formulation in the inertial reference frame as opposed to the rotating frame is more useful in
practical applications such as condition monitoring. The results are presented in non-dimensionalized form, making it
suitable for wide applicability. Also, in contrast to related previous studies which analyze similar equations, the present
scheme of non-dimensionalization results in stability boundary curves in the form of straight lines. This is beneficial in
arriving at design guidelines.
Different models of increasing level of complexity are investigated, namely (i) a 2-DOF lumped parameter system for
asymmetric shaft mounted on rigid bearing, (ii) a 4-DOF lumped parameter system for asymmetric shaft mounted on
flexible massless bearing, (iii) a 4-DOF lumped parameter system for asymmetric shaft mounted on flexible bearing with
mass, and finally (iv) an elastic asymmetric shaft mounted on rigid bearings. We seek to determine the parameter zones for
stability and instability for each of the above systems. Rather than finding these parameters for a specific system, we
propose a generalized approach based on non-dimensionalization of the governing equations of motion.
For each of the above systems, the non-dimensional parameter space corresponding to stable and unstable operations
are identified. It is found that the stability boundary is in the form of a V-shape. With increase in damping, the V-shape gets
filleted out resulting in a shrinkage of the unstable parameter regime. It is shown that in contrast to rigid bearing, flexibility
of the bearings results in an extra unstable region which does not have any practical significance. The bearing stiffness also
induces translation of the stability boundaries in the parameter space. Inclusion of bearing mass results in two more
branches of the stability boundary. Finally, it is shown that for a continuous elastic shaft there are infinite stability
boundaries corresponding to the different modes of vibration. Qualitatively these boundaries are similar with the incor-
poration of a scaling factor.
The utility of these results presented in a non-dimensional fashion lies in its applicability to all cases of rotating
asymmetric shaft (viz. irrespective of material and geometrical properties). For a given shaft configuration operating within
a speed range, the locus of the non-dimensional parameters follows a straight line equation. The intersection of this line
with the stability boundary gives the unstable operating speed range. Thus, a quick procedure is proposed to determine the
unstable operating speed range of a given asymmetric shaft. Further, these results can be used to suitably redesign the
system such that stability during operation is ensured. The effect of each system parameter on the stability is also high-
lighted in the work.

References

[1] C.-W. Lee, Vibration Analysis of Rotors, Vol. 21, Springer Science & Business Media, Dordrecht, The Netherlands, 1993.
[2] R.C. Arnold, E.E. Haft, Stability of an unsymmetrical rotating cantilever shaft carrying an unsymmetrical rotor, Journal of Manufacturing Science and
Engineering 94 (1) (1972) 243–249.
[3] D. Ardayfio, D. Frohrib, Vibration of an asymmetrically mounted rotor with gyroscopic effects, Journal of Manufacturing Science and Engineering 98 (1)
(1976) 327–331.
[4] C. Rajalingham, R. Bhat, G. Xistris, Influence of support flexibility and damping characteristics on the stability of rotors with stiffness anisotropy about
shaft principal axes, International Journal of Mechanical Sciences 34 (9) (1992) 717–726.
[5] Y. Kang, W.-W. Hwang, Influence of bearing damping on instability of asymmetric shafts—part i. stabilizing and destabilizing effects, International
Journal of Mechanical Sciences 38 (12) (1996) 1349–1358.
[6] Y. Kang, Y.-G. Lee, Influence of bearing damping on instability of asymmetric shafts—part ii. mode veering, International Journal of Mechanical Sciences
38 (12) (1996) 1359–1365.
[7] Y. Kang, Y.-G. Lee, Influence of bearing damping on instability of asymmetric shafts—part iii. disk effects, International Journal of Mechanical Sciences 39
(9) (1997) 1055–1065.
[8] G. Genta, Whirling of unsymmetrical rotors: a finite element approach based on complex co-ordinates, Journal of Sound and Vibration 124 (1) (1988)
27–53.
[9] G. Frulla, Rigid rotor dynamic stability using Floquet theory, European Journal of Mechanics-A/Solids 19 (1) (2000) 139–150.
[10] R. Ganesan, Effects of bearing and shaft asymmetries on the instability of rotors operating at near-critical speeds, Mechanism and Machine Theory 35
(5) (2000) 737–752.
[11] Y.-C. Pei, Stability boundaries of a spinning rotor with parametrically excited gyroscopic system, European Journal of Mechanics-A/Solids 28 (4) (2009)
891–896.
[12] F.A. Raffa, F. Vatta, Dynamic instability of axially loaded shafts in the Mathieu map, Meccanica 42 (6) (2007) 547–553.
[13] D. Bartylla, Stability investigation of rotors with periodic axial force, Mechanism and Machine Theory 58 (2012) 13–19.
[14] M.A. AL-Shudeifat, On the finite element modeling of the asymmetric cracked rotor, Journal of Sound and Vibration 332 (11) (2013) 2795–2807.
R. Srinath et al. / Journal of Sound and Vibration 362 (2016) 276–291 291

[15] M.A. AL-Shudeifat, Stability analysis and backward whirl investigation of cracked rotors with time-varying stiffness, Journal of Sound and Vibration 348
(2015) 365–380.
[16] C. Guo, M.A. AL-Shudeifat, J. Yan, L.A. Bergman, D.M. McFarland, E.A. Butcher, Stability analysis for transverse breathing cracks in rotor systems,
European Journal of Mechanics-A/Solids 42 (2013) 27–34.
[17] Q. Han, F. Chu, The effect of transverse crack upon parametric instability of a rotor–bearing system with an asymmetric disk, Communications in
Nonlinear Science and Numerical Simulation 17 (12) (2012) 5189–5200.
[18] Q. Han, F. Chu, Parametric instability of a Jeffcott rotor with rotationally asymmetric inertia and transverse crack, Nonlinear Dynamics 73 (1–2) (2013)
827–842.
[19] G. Giannopoulos, S. Georgantzinos, N. Anifantis, Coupled vibration response of a shaft with a breathing crack, Journal of Sound and Vibration 336 (2015)
191–206.
[20] C. Guo, M. Al-Shudeifat, J. Yan, L. Bergman, D. McFarland, E. Butcher, Application of empirical mode decomposition to a Jeffcott rotor with a breathing
crack, Journal of Sound and Vibration 332 (16) (2013) 3881–3892.
[21] Q. Han, J. Zhao, F. Chu, Dynamic analysis of a geared rotor system considering a slant crack on the shaft, Journal of Sound and Vibration 331 (26) (2012)
5803–5823.
[22] Q. Han, F. Chu, Parametric instability of a rotor–bearing system with two breathing transverse cracks, European Journal of Mechanics-A/Solids 36 (2012)
180–190.
[23] M. Shahgholi, S.E. Khadem, Stability analysis of a nonlinear rotating asymmetrical shaft near the resonances, Nonlinear Dynamics 70 (2) (2012)
1311–1325.
[24] A.H. Nayfeh, Introduction to Perturbation Techniques, John Wiley & Sons, Ithaca NY 14853, 2011.
[25] R.H. Rand, Lecture notes on nonlinear vibrations. —version 45. Available at 〈http://www.tam.cornell.edu/randdocs/〉.
[26] W. Magnus, S. Winkler, Hill's equation, Courier Corporation, 2013.

You might also like