Berichte des Instituts für Mechanik

Bericht 2/2010
Fikre Endashaw Boru

Numerical and Experimental Response and

Stability Investigations of Anisotropic Rotor-Bearing Systems

kassel
university
press
This work has been accepted by the faculty of Mechanical Engineering of the University of Kassel as a
thesis for acquiring the academic degree of Doktor der Ingenieurwissenschaften (Dr.-Ing.).

Supervisor: Prof. Dr.-Ing.Horst Irretier

Co-Supervisor: Prof. Dr.-Ing. Bernhard Schweizer
Examiners Prof. Dr.-Ing. Anton Matzenmiller
Prof. Dr. rer. nat. Andreas Meister

Defense day: 02 June 2010

nd

The publication was funded by Deutscher Akademischer Austauschdienst

Gedruckt mit Unterstützung des Deutschen Akademischen Austauschdienstes

Bibliographic information published by Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data is available in the Internet at http://dnb.d-nb.de.

Zugl.: Kassel, Univ., Diss. 2010

ISBN: 978-3-89958-544-5

© 2010, kassel university press GmbH, Kassel

www.upress.uni-kassel.de

Printed in Germany
 ACKNOWLEDGMENT

This dissertation was completed during my DAAD PhD scholarship at the Institute of Mechanics
at the University of Kassel under the supervision of Prof. Dr.-Ing. Horst Irretier. The completion
of this dissertation wouldn’t have been possible without the support and love of several people and
institutes.

I would like to foremost say my best thanks to Prof. Dr.-Ing. Horst Irretier for incepting the
research topic and accepting me in his workgroup. I would also like to thank him for his unfailing
support throughout my research stay and for his ardour to teach and disseminate knowledge in the
area of machinery and rotor dynamics. During my stay at the Institute of Mechanics at the
University of Kassel, I have enjoyed an estimable time in his workgroup and have learned a great
deal of lessons.

I would also like to thank Prof. Dr.-Ing. Bernhard Schweizer for his interest in the research topic
and for being the referee of this dissertation. I want to thank him for extending his supporting hand
during the uncertain times of my research stay. My special thanks also go to Prof. Dr.-Ing. Anton
Matzenmiller for his interest in my research, for the recommendations he wrote to my scholarship
institute throughout my research stay at the University of Kassel and for being in the examination
committee of my dissertation. I also like to thank Prof. Dr. rer. nat. Andreas Meister for his interest
in my research and for being in the examination committee of my dissertation as well.

I also like to give my heart-felt gratitude to the Deutscher Akademischer Austauschdienst (DAAD)
for the family scholarship during the large portion of my research stay at the Institute of Mechanics
at the University of Kassel. I also thank the DAAD for financing my language course at the Goethe
Institute, Göttingen, on my arrival in Germany. Without such a transition language course my life
here in Germany wouldn’t have been as enjoyable as it has been. This scholarship has given me the
opportunity to learn and enjoy the German culture.

My special thanks go to my colleague Dipl.-Ing. Daniel Strohschein for without his helpfulness and
encouragement my stay would not have been half as eventful. His unfailing on- and off-campus
help will stay with me for times to come. My thanks also go to Dipl.-Ing. Gerd Schneider for his
undeterred support and for investing his vast experimental knowledge on the experimental test-
setup of the research. I am indebted to him for the notable practical knowledge I have gained from
working with him at the Institute of Mechanics. Mrs. Christa Ellrich was an institute member to
whom I have frequently turned to seek support, and so I would like to take this opportunity to
express my gratefulness for the generous help I got from her and for her organisation at the
Institute of Mechanics. This acknowledgment wouldn’t be complete without due gratitude to the
members of the Institute of Mechanics at the University of Kassel. Every member of the institute
has helped to create the harmonious and friendly environment that I will always cherish.

Above all, I would like to thank God for his guidance through the jungles of life to this day. My
gratitude goes to the anchor and love of my life Selam Gebreselassie, without whose support this
achievement was impossible. My sincere thanks, love and respect also go out to my Mother
Etagegnehu Tessema and my Father the late Endashaw Boru whose love and upbringing I will
always cherish. My gratitude goes also to my brothers and sister and my whole extended family that
in one way or the other have significant mark in my life.

Kassel, March 2010 Fikre E. Boru

 TABLE OF CONTENTS

Acknowledgment i
List of Symbols vii
1. Introduction 1
1.1 Motivation and Objectives 1
1.2 Contents and Overview 2
2. Literature Review 3
2.1 Stability of Flexible Rotors with Anisotropies 4
2.2 Finite-Element-Modelling and Numerical Studies 6
2.3 Experimental Studies 8
3. LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings 13
3.1 Equations of Motion (in Fixed and Rotating Coordinate System) 13
3.1.1 Circular Shaft and Rigid Bearings 13
3.1.2 Circular Shaft and Anisotropic Elastic Bearings 15
3.1.3 Non-Circular Shaft and Rigid Bearings 17
3.1.4 Non-Circular Shaft and Anisotropic Elastic Bearings 19
3.2 HILL’s Method for Eigenvalue and Stability Analysis 21
3.2.1 Fixed Coordinate System 21
3.2.2 Rotating Coordinate System 23
3.3 Inclusion of Damping 25
3.3.1 Equations of Motion 25
3.3.2 Eigenvalue and Stability Analysis 26
3.4 Inclusion of Bearing Mass 27
3.5 Unbalance Vibrations 29
3.6 Vibrations due to Gravitational Load 30
4. Numerical and Experimental Studies 33
4.1 LAVAL-Rotor with Non-Circular Shaft and Different Types of Bearings 33
4.1.1 Rigid Bearings 33
4.1.2 Anisotropic Elastic Bearings 43
4.2 Numerical Studies for a Test Rotor 50
4.2.1 Rigid Bearings 50
4.2.2 Anisotropic Elastic Bearings 54
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

4.3 Experimental Studies and Findings 57

4.3.1 Test Rotor Design 57
4.3.2 Bearing Design and Properties 59
4.3.3 Experimental Results in Rigid Bearings 61
4.3.4 Experimental Results in Anisotropic Elastic Bearings 64
5. Fundamentals of Beam Theories and the Finite Element Method 65
5.1 Beam Theories 65
5.1.1 EULER-BERNOULLI Beam Theory 65
5.1.2 RAYLEIGH Beam Theory 67
5.1.3 TIMOSHENKO Beam Theory 68
5.2 Fundamentals of the Finite Element Method 71
5.2.1 Principle of Virtual Work 72
5.2.2 LAGRANGE’s Equations 74
5.2.3 Discretization and Finite Element Model of Rotors 77
6. General Anisotropic Rotor-Bearing Systems 81
6.1 Element Matrices 81
6.1.1 Shaft Elements 81
6.1.1.1 Circular Shaft Element in the Fixed Coordinate System 81
6.1.1.2 Circular Shaft Element in the Rotating Coordinate System 87
6.1.1.3 Non-Circular Shaft Element in the Rotating Coordinate System 88
6.1.1.4 Non-Circular Shaft Element in the Fixed Coordinate System 89
6.1.2 Disk Elements 90
6.1.2.1 Circular Disk Element in the Fixed Coordinate System 90
6.1.2.2 Circular Disk Element in the Rotating Coordinate System 92
6.1.3 Bearing Elements 93
6.1.4 External Forces 95
6.2 HILL’s Method for Free and Forced Vibration Analysis 97
6.2.1 Eigenvalues, Eigenvectors and Stability Analysis 99
6.2.2 Unbalance Vibrations 103
6.2.3 Vibrations due to Gravitational Load 106
6.3 Description of the Finite Element Program 108
7. Numerical Results 109
7.1 LAVAL-Rotor with Non-Circular Shaft and Different Types of Bearings 109
7.1.1 Rigid Bearings 109
7.1.2 Anisotropic Elastic Bearings 115

iv
 Table of Contents

7.2 Rotor with Rectangular Shaft and Different Types of Bearings 120
7.2.1 Uniform Rectangular Cross-Section 120
7.2.1.1 Rigid Bearings 120
7.2.1.2 Anisotropic Visco-Elastic Bearings 123
7.2.2 Non-Uniform Rectangular Cross-Section 126
7.2.2.1 Rigid Bearings 127
7.2.2.2 Anisotropic Visco-Elastic Bearings 135
7.3 Two Disk Rotor with Non-Circular Shaft and Anisotropic Elastic Bearings 137
8. Conclusion and Future Outlook 141
8.1 Conclusion 141
8.2 Future Outlook 143
A. Transformation Matrices between the Fixed and the Rotating Coordinate System 145
B. Transformation of the Equations of Motion 147
C. Stiffness Matrix of a LAVAL-Rotor with a Non-Circular Shaft and Anisotropic
Elastic Bearings 149
D. Linearization of the Stiffness Matrix of a LAVAL-Rotor with a Non-Circular Shaft
and Anisotropic Elastic Bearings with Respect to the Deviatoric Stiffnesses 151
E. Derivation of the HILL’s Equations for the Eigenvalues, Eigenvectors and the
Forced Responses 155
E.1 Eigenvalues and Eigenvectors 155
E.2 Forced Responses 158
E.2.1 Harmonic Force 158
E.2.2 Constant Force 161
F. Geometric and Material Data of the Investigated Rotors 165
F.1 LAVAL-Rotor with Non-Circular Shaft and Different Types of Bearings 165
F.2 Rotor with Rectangular Shaft and Different Types of Bearings 167
F.3 Two Disk Rotor with Non-Circular Shaft and Anisotropic Elastic Bearings 169
G. HERMITIAN Polynomials and Shape Functions for TIMOSHENKO Beam
Elements 171
H. Mass, Damping, Gyroscopic and Stiffness Matrices for the Shaft, Disk and
Bearing Elements 177
H.1 Shaft Elements 177
H.1.1 Circular Shaft Element in the Fixed Coordinate System 177
H.1.1.1 TIMOSHENKO Beam Theory 178
H.1.1.2 RAYLEIGH Beam Theory 178

v
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

H.1.1.3 EULER-BERNOULLI Beam Theory 179

H.1.2 Circular Shaft Element in the Rotating Coordinate System 179
H.1.3 Non-Circular Shaft Element in the Rotating Coordinate System 180
H.1.3.1 TIMOSHENKO Beam Theory 181
H.1.3.2 RAYLEIGH Beam Theory 182
H.1.3.3 EULER-BERNOULLI Beam Theory 182
H.1.4 Non-Circular Shaft Element in the Fixed Coordinate System 183
H.2 Disk Elements 184
H.2.1 Circular Disk Element in the Fixed Coordinate System 184
H.2.2 Circular Disk Element in the Rotating Coordinate System 185
H.2.3 Non-Circular Disk Element in the Rotating Coordinate System 185
H.2.4 Non-Circular Disk Element in the Fixed Coordinate System 186
H.3 Bearing Elements 186
H.3.1 Bearing Element in the Fixed Coordinate System 186
H.3.2 Bearing Element in the Rotating Coordinate System 187
References 189

vi
 LIST OF SYMBOLS

scalar quantities α angle between fixed and rotating

A cross-sectional area coordinate system
C centre of mass β angular position of the mass centre in
the rotating coordinate system,
d damping coefficients and damping rotation due to shear force
matrix elements ∆ deviatoric term
D damping ratio
δ virtual values
e independent generalized coordinate
ε eccentricity of mass centre from
E Young's modulus of elasticity geometric centre, strain
F unbalance force in the fixed coordinate θ angular rotation about the z-axis
system κ shear coefficient
f unbalance force in the rotating
Λ, λ eigenvalue in the fixed and rotating
coordinate system
coordinate system
g gravitational acceleration
µ non-circularity
G modulus of rigidity (shear modulus)
N, ν natural circular frequency in the fixed
I inertia force and area moment of a and rotating coordinate system
cross section ξ,η,ζ rotating coordinate system and rotating
J mass moment of inertia axis
j complex number ρ mass density
k stiffness matrix elements and variables σ Stress
associated with stiffness in the rotating φ angular rotation of disk centre
coordinate system
Φ, ϕ eigenvector in the fixed and rotating
K stiffness matrix elements and variables
coordinate system
associated with stiffness in the fixed
ψ angular rotation about the y-axis
coordinate system
L shaft length, shaft element length, Ω angular velocity in x-direction
lagrangian ω0 circular natural frequency
M bending moment
m disk mass, element mass matrices and vectors
Q shear force, generalized force C, c velocity proportional matrix in the
q distributed force, generalized fixed and rotating coordinate system


displacement c vector of centre of mass
R resultant external force other than D, d damping matrix in the fixed and
unbalance force rotating coordinate system



r radius of gyration F , f unbalance force vector in the fixed and
T time period, kinetic energy rotating coordinate system
G, g gyroscopic matrix in the fixed and
t time
rotating coordinate system
U FOURIER series limit H matrix of hermetian polynomials
V potential energy K, k stiffness and displacement
v displacement in y-direction proportional matrix in the fixed and
W geometric centre of disk, work rotating coordinate system
w displacement in z-direction M, m mass matrix in the fixed and rotating
coordinate system
x,y,z fixed coordinate system

q generalized displacement vector
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

T transformation matrix w related to geometric centre

w,v displacement vector in the fixed and y,z in the y, z-direction
rotating coordinate system yb,zb associated to the bearing in the y,z-
direction
subscript ybs, associated to the bearing and shaft in
0bs average value associated to the bearing zbs the y,z-direction
and shaft ys,zs associated to the shaft in the y,z-
0s average value associated to the shaft direction
0η, 0ζ average value in the η,ζ-direction ηbs, associated to the bearing and shaft in
av average ζbs the η,ζ-direction
associated to the bearing Ηs,ζs associated to the shaft in the η,ζ-
b
direction
bs associated to the bearing and shaft
c related to mass centre superscript
cs cosine coefficient matrix c constraints
D disk con conservative
d associated to disk e external
ex external f referred to the fixed coordinate system
g gravitational G associated to gravitational force
G associated to gravitational force i inertial
GH associated to gravitational force in the in internal
hyper matrix Int intermediate
H hyper matrix and hyper vectors
n-con non conservative
I inertial
r referred to the rotating coordinate
im imaginary system
in internal T transpose
LAG derived using LAGRANGE'S eq. -, u- associated to negative complex
m associated to mass matrix exponential function
n nodal +, u+ associated to positive complex
pvw derived using principle of virtual work exponential function
re real special symbols
s associated to the shaft associated to the hyper matrix,
֒
ss sine coefficient matrix amplitude of a harmonic term
u index of summation in FOURIER ~ time variable term
series __
complex number
U associated to unbalance forces
UH associated to unbalance forces in the
hyper matrix

viii
 1 Chapter 1

INTRODUCTION

1.1 Motivation and Objectives

The vibration phenomenon is an inherent property of all structures with dynamic structures being
more prone to it than static structures. In general, the vibration of these structures is undesired for
it causes noise, discomfort, malfunction, wear, fatigue, and sometimes a complete destruction of
the structure unless and otherwise the vibration is “designed into” the structure to achieve certain
tasks as in concrete vibrators, gravel shakers, etc..

The vibration phenomenon of rotor-bearing systems is the focus in the present dissertation. Rotating
systems are used in a number of industrial sectors most of which are in the power generation and
absorption by the use of turbines, generators, internal combustion engines, pumps, compressors,
flywheels etc. These rotating shafts possess one or more disks to achieve the task for which they are
designed. In general, industrial rotors are very expensive and so their functionality and safety are of
great interest. Hence, understanding the complete dynamic behaviour of rotors helps in designing
better equipment with prolonged, trouble free operation life.

Rotor vibrations may arise from a number of practical causes. Residual unbalance and instability rank
first from the rest. Rotor unbalance causes a rotating force to act in synchronism with running
speed while rotor instability is a self-excited vibration that may arise from a number of reasons.
Axial asymmetry, machine foundation with nonlinear properties, fluid bearings and seals are some
of the known causes of instabilities in rotors.

Most rotors are analysed as though they are axisymmetric contrary to the reality. In reality, there is
always some sort of asymmetry in rotors. Certain rotating shafts are even intentionally designed to
possess a non-circular cross-section to achieve the task for which they are built like for example in
2-pole generators or motor shafts, basically circular shafts with keyways grooved in to allow
coupling with other machine elements, etc.. On the other hand, cracks and manufacturing failures
of shafts are responsible for unintentional axial asymmetries for which a pre-production analysis is
difficult. When a rotor looses its axisymmetry for any reason, be it intentionally or unintentionally,
there is always a certain risk of a presence of an unstable operation speed range.

The main motivation of the present dissertation is the unstable behaviour of rotors with non-circular
shafts. The research with the above stated motivation sets the objective of determining the
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

conditions affecting the unstable operation speed and accurately determining the unstable speed
ranges. The assertion of the accuracy of the analytical methods present in literature for a simple
rotor model with the help of experimental investigation was also high on the list of objectives of
this work. Then the complete numerical investigation of a general axially asymmetric rotor by
developing a finite element model implementing the TIMOSHENKO beam theory was intended.

1.2 Contents and Overview

This thesis is organized in such a way that the second chapter deals with the state of the research in
the area of rotor dynamics. It discusses the literature reviewed in the course of this work by dividing it
in three further sections. In this chapter, the literature dealing with the main topic i.e. anisotropic
rotors, the mathematical methods and experimental verifications are investigated. Chapter three is
dedicated to the analytical analysis of LAVAL-rotors with non-circular shafts. Here, both rigid and
anisotropic elastic bearings (with and without damping) are discussed. The formulations of the
equations of motion are given in both a fixed and rotating coordinate system. The HILL’s method of
solution in both coordinate systems is also discussed in this chapter. The fourth chapter discusses
the executed experimental investigations. A LAVAL-rotor was analytically, numerically and
experimentally investigated and compared with each other. Three beam theories and the fundamentals
of the Finite Element Method are discussed in chapter five while the sixth chapter implements the
discussions of the fifth chapter to develop the circular and non-circular shaft element matrices of the
rotor in the fixed and rotating coordinate system. The disk and bearing element matrices are also
developed in both coordinate systems in this chapter. The seventh chapter is dedicated to the
implementation of the finite element model to analyse particular rotors. Here, a rectangular shaft rotor and
a two disk rotor are analysed besides the LAVAL-rotor already discussed in the fourth chapter. The
final chapter gives the conclusion and future outlook of this dissertation.

Besides the eight chapters, this dissertation has eight appendices to assist the reader with some
mathematical details of some chapters. The first two appendices are about the transformation of
the axis between the fixed and rotating coordinate system. The third and fourth appendices discuss the
resultant stiffness matrix of a LAVAL-rotor with a non-circular shaft mounted in anisotropic elastic bearings
and the linearization of this matrix. The fifth appendix outlines the exponent comparison used in
the HILL’s method. The data of the rotors considered in this dissertation are outlined in the sixth
appendix. The seventh appendix describes the HERMITIAN polynomials and the eighth appendix
gives all the element matrices used in the Finite Element Method used in this thesis.

2
 2 Chapter 2

LITERATURE REVIEW

The field of rotor dynamics is a relatively old science on which a number of books and literatures are
abundantly present. From simplicity and numerical dominance of rotors with circular shafts, most
technical manuscripts deal with such systems intensively. Only a few percentage of the publications
treat structures with non-circular rotating parts in depth. Having been studied for more than a
hundred years, it is difficult to give the complete chronological developments. Some authors such
as Rieger [87] and Nelson [69] have dedicated significant energy to study and publish some of the
early developments in rotor dynamics.

In the early stages of rotor dynamics constructions, the design of the rotors were based on the
theory of Rankine [83], who defined the first critical speed of the rotor as the whirling speed for the
first time but mistook it as the stability limit of the rotor assembly. This misconception was adopted
by a number of other prominent researchers of the time.

De Laval [35] in 1889 and Dunkerley [22] in 1894 reported the possibility of running a rotor above
the Rankine whirling speed. The work of Dunkerley was based on the theoretical work of Föppl
[26], which was not familiar to rotor dynamists of the time due to his choice of journal. According
to Nelson [69] it was in the paper from Dunkerley that the term “critical speed” was first used.

When Kerr [48] published an experimental result of the presence of a second “critical speed” in
1916, it became clear that both De Laval and Kerr have somehow passed the first critical speed
which was defined to be the stability limit of the assembly according to Rankine on which most
analysis were still based. The Royal Society of London gave Jeffcott the assignment to resolve the
difference between the experimental and theoretical results.

Jeffcott [45] released his findings of a simplified rotor model which is now used as the spring board
of further complicated models. In the dynamic analysis of rotor-bearing systems, it is now common
to start with this isotropic rotor (now called LAVAL- and sometimes JEFFCOTT-rotor). Unlike its
simplicity, the model can show the basic phenomenon of a practical rotor. Stodola [93] included the
CORIOLIS-effect to Jeffcott’s model in 1924. Since then, the LAVAL-rotor has been analytically
and numerically analysed and experimentally compared by numerous researchers. [13, 28, 29, 42,
58, 85, 98, 100] are some of the books which have explained the dynamics of the LAVAL-rotor in
detail.
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

With the above theoretical capacity the research went then in the design and analysis of flexible1
rotors. In 1925 Kimball [49] used internal damping to explain the instability in the supercritical region
in some practical rotors. In the same year, instabilities due to journal bearings were explained by
Newkirk and Taylor [72]. Instabilities due to different properties of the rotor-bearing assembly have
since then been investigated by a number of researchers. Then in 1933 Smith [92] conducted a
complete dynamic analysis of an unbalanced rotor bearing assembly.

2.1 Stability of Flexible Rotors with Anisotropies

As can be seen in a number of literatures, different types of anisotropies are present in rotor-bearing
systems. The anisotropies can be either in the rotating parts of the rotor, in the non-rotating parts of
the rotor or in both the rotating and non-rotating parts. According to Lee et al. [52], a rotor is said
to be “isotropic” if both the rotating and non-rotating parts of the rotor-bearing system are
axisymmetric. “Anisotropic rotor” is used to define rotors with an axially asymmetric non-rotating part
but an axisymmetric rotating part. “Asymmetric2 rotor” defines a rotor with an axially asymmetric
rotating and an axisymmetric non-rotating part. “General rotor” is used to define a rotor with axially
asymmetric rotating and non-rotating parts.

Considering the historical development of the analysis of a general rotor, the unfortunate model
taken by Rankine was axially asymmetric according to Nelson [69]. But it was Prandtl [81] who
published the first work on a LAVAL-rotor with a non-circular shaft in 1918. Rodgers [88] studied
horizontally and vertically undamped rotors with axially asymmetric shafts in 1922. In 1943 Foote
et. al. [25] formulated a mathematically exact solution for the coupled equation of a LAVAL-rotor
with non-circular shaft and anisotropic elastic bearings and showed the stability behaviour of such
rotor assembly.

Brosens and Crandall [9, 20] wrote papers detailing the stability analysis of a rotor with a rigid circular
shaft and an axially asymmetric disk mounted on anisotropic bearings that may be set to rotate with
the shaft. Their analysis considered gyroscopic forces and the conditions in which stability can be
achieved based on rotor rotational speed. The theoretical analysis is compared with experimental
results in their publication. Yamamoto and Ōta [102] considered asymmetric shaft rotors and
considered the effect of damping as well. Their work presented results of the experiment
conducted and also approximations to determine the unstable regions. Gladwell and Stammers [31]

1 These are rotors that are running near or above the first (fundamental) critical speed of the rotor-bearing assembly.
2 In this dissertation, this term will not be used to define rotors with axial asymmetry. This definition is avoided for it
may mislead readers as to expect laterally asymmetric rotors (for example an isotropic rotor mounted on two bearings
with different properties).

4
 Literature Review

compared three methods to determine the stability of a rigid rotor supported in anisotropic bearings.
Forrai [27] investigated the stability of asymmetric shaft rotors with two disks. He used the EULER-
BERNOULLI beam theory to generate four coupled HILL equations that were then solved using
the perturbation variation method. An example rotor was investigated to show the presence of six
unstable operation speed ranges.

Noah and Hopkins [74] described the HILL’s3 method for the stability analysis of a second order
differential equation with periodic coefficients for a general dynamic system. Iwatsubo et al. [44]
discussed the stability region of a LAVAL-rotor with a non-circular shaft mounted in anisotropic bearings.
A point mass model was used in the determination of the equations of motion and the Routh-
Hurwitz criterion was applied for the determination of system stability.

Genta [30] used the HILL’s method on a finite element model of a non-circular shaft rotor mounted in
asymmetric bearings in determining its stability. His paper details the whirl orbit, the CAMPBELL-
diagram and the unbalance response of an example rotor. Jei and Lee [46] conducted a modal analysis
of a non-circular RAYLEIGH-shaft with a non-circular disk mounted in isotropic bearings and
discussed the stability of the assembly in forward and backward precession. Person [79] investigated
the dynamics and stability of wind turbines with different number of blades. The different number
of blades may result in asymmetric distribution of the mass affecting stability. One year later, he
reported on the sensitivity of the HILL’s and FLOQUET’s method for the stability determination
along with Wiedemann [78]. Ertz et al. [23] discussed a non-circular shaft rotor mounted in
anisotropic bearings. Their paper starts off with a point mass model (LAVAL-rotor) and deepens
to discuss a finite element model for such systems. A method is suggested as to how the basic
eigenvalues and eigenvectors can be filtered from the redundant values generated from the HILL’s
method. Teichmann [97] discussed a LAVAL-rotor with a non-circular shaft and both rigid and
anisotropic elastic bearings. The work outlines formula for approximating the natural frequencies of the
LAVAL-rotor in anisotropic elastic bearings. Besides the LAVAL-rotor a general rotor was
numerically investigated using the RAYLEIGH beam theory. Afolabi [1] presented a method of
transforming the equation of motion of a LAVAL-rotor with a non-circular shaft, which is an
ordinary differential equation with time dependant coefficients, to one with constant coefficients.
Chen and Peng [12] discussed the effect of an axial force on the limits of stability of a
TIMOSHENKO beam element with dissimilar stiffness. Takashi and Murakami [96] conducted a
study of a rotor mounted on a base with stiffness and damping in one direction. Their study

3 G. W. Hill used a FOURIER-series to solve the differential equation with periodic coefficients during his study of the
motion of the moon. He published his work in a paper titled: “On the Part of the Motion of the Lunar Perigee which
is a function of the mean Motion of the Sun and Moon”. The solution method is therefore named after him.

5
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

includes the unbalance and gravity response. Further experimental studies were used to compare
results concerning the stability of the assembly and base flexibility. Oncescu et al. [77] investigated a
non-circular RAYLEIGH beam rotor mounted in anisotropic bearings using the finite element
method. They used the FLOQUET’s theory for the determination of the system stability. Lee et al.
[52, 53, 56] and Suh et al. [95] discussed the modal analysis of a non-circular shaft rotor mounted in
different types of bearings. They discussed the advantages of the directional frequency response functions
in axially asymmetric rotor bearing systems. Rao and Sreenivas [84] used a solid element unlike the
commonly used line element in their finite element model to investigate axially asymmetric general
rotor-bearing systems. In general, investigating the books dedicated to the analysis of rotor-bearing
systems shows that the authors mostly include only a short analytical analysis of LAVAL-rotors
with non-circular shafts in rigid or isotropic bearings [28, 29, 42, 51, 58, 85, 98].

2.2 Finite-Element-Modelling and Numerical Studies

The problems encountered in the real world of structures are often complicated requiring complex
physical models. Such complex physical models lead to complex mathematical models which are
often difficult to analyse with no known closed form solutions. The desire to analyse these more
realistic physical models has called for approximate methods to solve the resulting complex
mathematical formulae.

According to Huebner et. al. [36], the inception of the Finite Element Method (FEM) is claimed by
applied mathematicians, physicists and engineers. The claim comes as all three areas have
contributed essential ideas implemented in the FEM. Applied mathematicians were concerned with
boundary value problems of continuum mechanics. Physicists were also interested in continuum
mechanics but used piecewise approximation functions to represent the continuous functions. On
the other hand, engineers in the aerospace industry were being faced with complex structures and
needed to determine the stiffness coefficients of the structures. Courant [18] from the applied
mathematicians and Prager and Synge [80] from the physicists can be considered as novices of the
FEM in their respective fields.

In the engineering field, intuition has brought the concept of finite elements for the analysis of trusses
since 1930s. Hrenikoff [33] can be seen as the fountain head of the FEM as he extended this idea
for the analysis of trusses to continuous systems in 1941. McHenry [62] and Newmark [73] further
developed the idea of discretization in 1943 and 1949, respectively. The above studies were limited
to structures fixed at two points in space.

6
 Literature Review

In 1954 and 1955 Argyris published a series of papers unifying many approximate methods for the
solution of continuous and one dimensional problems [3]. Turner et. al. presented an actual
solution of a plane stress problem by means of triangular elements whose properties were
determined from the equation of elasticity in 1956. According to Clough [15, 17], the paper by
Turner was presented on a conference in January 1954. Argyris and Clough made significant
contributions for the development of the FEM and both have written books about their extensive
work in the area of finite elements [2, 14].

The term “Finite Element Method” first appeared in a paper dealing with plane elasticity problems by
Clough [16] in 1960. The terms “Finite Element Method” and “Direct Stiffness Method” were
parallely used until 1965 when Zienkiewicz and Cheung [103] gave the FEM a wider application by
showing that it can be used as long as a problem can be written in a variational form.

Huebner et. al. reports that, as the FEM became popular by engineers and physicists,
mathematicians became interested to give the method a firm mathematical foundation. Hence, a lot of
papers have been published in the 1960s and early 1970s laying the mathematical foundations. The
method was well in use for design by civil and aeronautics engineers before it got its rigorous
mathematical foundation in 1973 by Strang and Fix [94]. According to Champion and Ensminger
[11], the FEM was initially applied in the structural analysis area. But in the late 1970s and early
1980s the method was intensively used also in the thermal and fluid area.

Since then the FEM has won wide range of acceptance in the field of engineering. Hence,
numerous books and papers are published outlining the method and its application with the first
book written by Zienkiewicz and Cheung in 1967. Today there will appear more than 1.5 million
search results for the word “Finite Element Method” on Google search engine on the internet.

The study of rotor dynamics was also influenced by the emergence of this new analysis method. In the
early times of rotor dynamics studies the investigations of rotating bodies were based on analytical
methods. A rotor-bearing system would be modelled into simple blocks of mechanical parts whose
properties are completely known. This reduced model is then mathematically modelled and a closed
form solution is sought for it so as to be able to predict its properties. As actual rotor-bearing
assemblies are significantly more complicated than the models for which closed form solution are
calculated, the quality of the predicted values was obviously not sufficient.

The effort to generate better quality results lead to more complicated physical models including
realistic boundary conditions. The mathematical model for such refined physical models is intricate

7
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

and in most cases it was not possible to find its closed form solution. This directed rotor dynamic
analysts to look for numerical or approximate solution for specified operation conditions.

In 1944 Myklestad [68] used an approximate method for the dynamic analysis of an aircraft wing,
and in 1945 Prohl [82] published an approximate technique by which the natural frequencies of a
rotor system can be determined graphically. A broader numerical method known as the Transfer
Matrix Method (TMM) evolved based on the above two works for the analysis of rotors. Lund and
Orcutt [59] and Lund [60] extended the TMM for the determination of the unbalance response and
stability analysis of a rotor mounted in hydrodynamic bearings.

The FEM was not quickly integrated in the rotor dynamics analysis. According to Childs [13], Ruhl
and Booker [90] were the researchers responsible for the introduction of this solution method to
the field of rotor dynamics in 1972. Then the work of Nelson and McVaugh [70] in 1976 and
Nelson [71] in 1980 made further contributions for its use in the field.

To the author’s knowledge, the FEM was first used by Genta [30] for the analysis of an axially
asymmetric rotor-bearing system. The paper outlines the finite element formulation of a general rotor
besides other classical rotor bearing assemblies. Ertz et. al. [23] suggested theoretical methods of
identifying the basic eigenvalues from the large number of solutions that come from using the
HILL’s method of solving general rotors. But the suggested method is not suitable for automated
identification and visualization of the results due to the round-off errors of the common routines
used in the FEM. Most of the literatures written after 1988 applied the FEM as solution procedures
[12, 23, 52, 53, 56, 77, 95, 96]. All the above indicated literatures used beam elements based on either
EULER-BERNOULLI, RAYLEIGH or TIMOSHENKO beam theories. Recently, some authors
(for example [84]) have also used solid elements to model general rotors.

2.3 Experimental Studies

In the theoretical analysis of vibrating systems one is actually dealing with physical models that are
believed to represent the actual system. Hence, the results from analytical and numerical analysis are
only as good as the physical model developed for the system under investigation. The accuracy of
the physical model can only be checked by conducting measurements on the actual model. This
leads to the importance of experimental methods in the field of mechanical system design and control.

According to Wowk [101], the research in the field of vibrations was mainly in the field of music
before the industrial revolution. During this time, most of the mathematical backgrounds for

8
 Literature Review

vibration analysis were developed. Irretier [41] places CHLADNI4, who successfully showed the
modes of sound vibrations, as pioneer in the field of experimental studies. Campbell [10] was also
stated as to have pioneered in the field of experimental studies in his effort to photographically
show the backward wave propagation of a rotating disk.

During the industrial revolution the pursuit for maximized profit has lead to large speed machines
and the replacement of manual workforce with machines. This increased dynamism lead to
vibration and vibration related failure of these more complicated machines. During the Second
World War the importance of mechanical vibrations of equipments increased for failure of some
important machine parts like radio in fighter jets came as a result of vibrations. Further, vibration
signals (acoustics) were used for identification and location of submarines. During this time, the
measurement capacity was developed to the extent that the submarine’s motor condition could be
identified [101].

In the early 20th century the complete analytical or numerical dynamic analysis of high performance
machinery operating at different operation conditions was not successful for lack of analytical and
numerical capacity at that time. Hence, experimental investigations were deemed necessary for
complete dynamic analysis. In the mid 20th century the numerical capacity was significantly
improved with the development of the FEM but this has not diminished the importance of
experimental investigations. During this time the development of the electronic equipments were
also making great advances [34]. The vibration phenomena of industrial machines are
experimentally investigated with the technical background germinated during the Second World
War and the highly developed electronic equipments presently available.

The early experimental investigations were basically conducted for the determination of the resonance
frequencies and damping properties of the machines. The measuring instruments indicate the
vibration amplitudes and these measured amplitudes are analysed to determine the system
properties. But this analysis failed to identify natural frequencies lying very close to each-other and
those frequencies that are significantly damped. Around the 1940s, the Experimental Modal Analysis
(EMA) came into light to curb these shortcomings. The EMA is now the main experimental
procedure in structures, machines, rotors and rotating structures to identify their vibration
properties.

4 Ernst Florens Friedrich Chladni is a german physicist and musician (1756-1827) who researched in the vibration of
sound. He published his research in a book titled ‘Entdeckungen über die Theorie des Klanges’ in 1787.

9
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

The initial implementation of the EMA in the early 1950s was in the vibration analysis of structures
and machines. In later years, this method transgressed to rotating systems. The aim of the EMA is
the determination of modal parameters (eigenvalues, eigenvectors, modal mass, modal damping,
modal stiffness) from vibration measurements. To achieve this task a number of experimental
procedures and numerical methods have been developed over the years. Irretier [41] has given a
well constructed chronological development of the EMA used in structures and machines.

Irretier [41] traced the EMA back to de Vries and Kennedy and Pancu in 1942 and 1945,
respectively, who used experimental methods to generate the NYQUIST-diagram of vibrating
structures. This single point excitation based method was then extended to multi point excitation
methods from which the classical method of Phase Resonance Testing was developed in the 1950s.
Irretier [41] has also indicated the works of Lewis and Wrisley, de Vries, de Veubeke, Traill-Nash
and Asher as pioneering in the field. The great advances in the measuring techniques with the
development of the spectrum analyser and advances in digital data analysis with the Fast
FOURIER-Transformation (FFT) led to the development of the presently widely used, multifaceted
and efficient Phase Separation Method by Stahle and Natke in the 1960s.

The first EMA on a rotor was done by Nordmann and Schöllhorn [76] in 1978 on a LAVAL-rotor
using an impulse hammer and capacity displacement measuring devices. The EMA of an
intentionally deformed three meter axial compressor rotor mounted on three fluid film bearings
was investigated by Schlegel [91] in 1979. In 1986, Bently et. al. [5] described a method called
“Perturbation Technique” for a rigid rotor on elastic bearings which was later extended by
Muszynska et. al. [66, 67] to investigate a two disk elastic rotor to determine the bearing dynamic
forces. A two disk test rotor with a thin shaft was investigated by Rogers and Ewins [89] and, for
the first time, they showed experimentally the asymmetry in the FRF matrices arising from the fluid
film bearings and the gyroscopic effect as well as the difference between the right and left
eigenvectors.

The experimental and theoretical/numerical modal analysis was based on the classical EMA until
Lee [57] formulated the modal expression of a rotor showing all the special features such as the
forward and backward eigenforms in 1991. With the help of an autoregressive moving average model by
Lee et. al. [54] this method forms the foundation of the time series analysis which leads to the
identification of the eigenfrequencies and modal damping parameters.

A completely new method of the contactless EMA was carried out by Nordmann et. al. [75] in 1994
by using the magnetic bearings of a rotor as a vibration exciter. A magnetic exciter with integrated

10
 Literature Review

piezoelectric force sensors and later extended with a laser displacement measurement unit specially
made for rotor EMA was developed by Irretier and Kreuzinger-Janik [39,50]. Irretier [40] has given
very good chronological development of rotor and rotating structure EMA.

When considering time varying rotor systems, Bienert [6] published a paper on the relationship between
the EMA of a rotating and non-rotating structure. Irretier [40, 43] gave the mathematical
description for the EMA of such an assembly using the HILL’s method while Lee et. al. [52] used
the FLOQUET-theory.

In the majority of the literature discussed above it can be seen that the analytical investigation of a
LAVAL-rotor with a non-circular shaft and rigid bearings is presented. But the very few sources
that discuss the effect of bearing anisotropy besides shaft asymmetry include the bearing mass while
generating the rotor model. The consideration of the bearing mass which normally is seen as a
further extension of the basic LAVAL-rotor model doubles the degree of freedom. There is no
literature with a model for a LAVAL-rotor with a non-circular shaft and anisotropic elastic bearings and no
additional extension (i.e. bearing mass). This missing link is addressed in this dissertation by linearizing
the resultant stiffness of the series arranged time-dependent shaft stiffness and constant bearing stiffness in the
fixed coordinate system or vice versa in the rotating coordinate system. The linearized stiffness matrix of
such a system is shown to be suitable for a LAVAL-rotor system with a non-circular shaft and
anisotropic elastic bearings where the ratio of the mean shaft stiffness to the mean bearing stiffness
approaches either infinity or zero.

The majority of the literature that analyse general rotors with a non-circular shaft by the FEM used
the EULER-BERNOULLI beam theory while the remaining few sources used the RAYLEIGH
beam theory. To the author’s knowledge, there is no FEM analysis of such rotors using the
TIMOSHENKO beam theory. In this dissertation, the TIMOSHENKO beam theory is used for
generating all elements in the FEM analysis. Here, the effect of shear deformation is considered for the
mass, gyroscopic and stiffness matrices of both circular and non-circular shaft elements in both fixed and rotating
coordinate system.

The above literature further discusses the stability of a uniform cross-section along the length of
the rotor shaft and so the effect of the longitudinal asymmetry and the distribution of the non-
circularity along the length of the rotor have not been studied. This dissertation investigates the
effect of longitudinal asymmetry besides axial asymmetry of the rotor. To analyse this effect, the shaft
cross-section is varied along the length of the rotor shaft. This variation of the model also gave way
to investigate the effect the rotor’s resultant asymmetry.

11
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

The experimental investigations presented in few of the above cited literature are of LAVAL-rotor
mounted in rigid bearings. Some of the literature that considered anisotropic elastic bearings
considered asymmetry of the lumped masses i.e. disks and not the shaft. Hence, besides the above
stated numerical investigations, this dissertation has for the first time presented an experimental
investigation of a rotor with a non-circular shaft and rigid as well as anisotropic elastic bearings. The
numerical results from the developed computer code using the Finite Element Method during the
course of the dissertation is compared with the experimental findings.

12
 3 Chapter 3

LAVAL-ROTOR WITH NON-CIRCULAR SHAFT AND ANISOTROPIC

BEARINGS

As stated in the second chapter, it is common to start a rotor dynamic analysis with a LAVAL-
rotor. This rotor is selected for its simplicity for analytic analysis but has enough level of intricacy to
depict the basic phenomena of practical rotors. It is particularly suited in showing the unbalance
response of a rotor-bearing system which in reality is practically unavoidable. After the basic
properties and the terms associated with rotor systems are presented using this simple model, the
assembly is then modified to investigate the effect of a shaft non-circularity. This modification
shows that the single mass model is still an effective tool in successfully showing the stability
problems associated with such non-circular shaft rotors.

3.1 Equations of Motion (in Fixed and Rotating Coordinate System)

3.1.1 Circular Shaft and Rigid Bearings

The LAVAL-rotor (with a circular shaft and rigid bearings) has been investigated in depth by a number
of researchers. The derivations presented in this dissertation follow from the work of Irretier [42]
and Gasch et. al. [28]. As shown in figure 3.1, this model is a simple rotor with a massive rigid disk
of mass m mounted in the middle of a massless elastic circular shaft of isotropic stiffness
k ys = k zs = k s which in turn is placed in rigid bearings at its two ends. This assembly is assumed to

be conservative i.e. no damping is present in the system.

Figure 3.1: LAVAL-rotor in deflected position

Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 3.1 further shows the x,y,z- and the ξ,η,ζ-coordinate system which are the fixed and rotating
coordinate system, respectively. The origins of both coordinate systems are set on the undeflected disk
geometric centre W of the rotor-bearing assembly with the x- and ξ-axis coinciding with the
longitudinal axis (the straight line drawn between the centres of the end bearings) of the shaft. The
ξ,η,ζ-coordinates rotate with the shaft’s angular velocity Ω (positive direction given in the figure).

To generate the equation of motion in the fixed coordinate system, consider the movement of the mass
which is assumed to be concentrated on the mass centre C located at a distance ε from the
geometric centre W of the disk. NEWTON’s 2nd law is used to develop the equation of motion as
follows

R = ma = mcɺɺ . (3.1)

The vector position of the mass centre C is given by

 yC   yW  ε y  ε y   sin ϕ 
c = w+ε ⇒   =  +  ;   =ε  . (3.2)
 zC   zW  ε z  ε z  cos ϕ 

Taking the time derivative of equation (3.2) gives

ɺɺ
ɺɺ


ɺɺ ϕɺ 2 −ϕɺɺ

cɺɺ = w +ε ; ε = − ε . (3.3)
 ϕɺɺ ϕɺ 2 

In general, a rotor is acted upon by a number of different forces depending on the purpose for
which it is built. In the fixed coordinate system, the resultant external force acting on the mass centre
can be resolved in two mutually perpendicular axis, namely Rye and Rze . The elastic strain forces

due to the shaft deflection in the y- and z-directions are − k ys yW and − k zs zW , respectively.

Substituting equation (3.3) into (3.1) gives5

y   k ys
 m 0   ɺɺ 0   y   Rye  ϕɺ 2 −ϕɺɺ

 0 m   ɺɺ +     
= + m  ε (3.4)
 z  0 k zs   z   Rze   ϕɺɺ ϕɺ 2 

48 EI
where k ys = k zs = ks = .
L3

5 The subscript W is neglected for presentation simplicity.

14
 LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings

Equation (3.4) can be written in compact form using matrix and vector notation as

ɺɺ
+ K w

e ϕɺ 2 −ϕɺɺ

Mw s = R + m  ε (3.5)
 ϕɺɺ ϕɺ 2 

Most industrial rotors are designed to operate at a preselected optimal rotational speed for most
part of their operational life. Hence, unless ones particular interest is to study the transient
properties of a rotor, analyses are usually conducted for constant rotational speed. For such a rotor
the angular rotation, velocity and acceleration are given by

ϕɺɺ = 0 ; ϕɺ = Ω ; ϕ = Ωt . (3.6)

Considering no other external force except the unbalance and substituting equations (3.6) into (3.5)
gives

ɺɺ
+ K w



Mw s = mΩ 2ε = F . (3.7)

Using the terms defined in figure 3.1, equation (3.7) can be expressed in the rotating coordinate system
using a coordinate transformation matrix detailed in Appendix A as

 y   cos α sin α  η 


 =   ; α = Ωt − β ⇒ w = Tv . (3.8)
 z   − sin α cos α  ζ 

Differentiating equation (3.8) twice and substituting it into equation (3.7) gives6

 m 0  ηɺɺ  0 2mΩ  ηɺ    −mΩ2 0   ks 0   η  2  sin β 

 0 m  ζɺɺ  −2mΩ
+  + +     = mΩ ε  
    0  ζɺ    0 − mΩ 2   0 ks   ζ  cos β 

⇒ mvɺɺ + gvɺ + ( k m + k s ) v = f (3.9)

Both equations of motion (3.7) and (3.9) are second order differential equations with constant
coefficients.

3.1.2 Circular Shaft and Anisotropic Elastic Bearings

Consider now the LAVAL-rotor with a circular shaft and anisotropic elastic bearings as shown in figure
3.2. In contrast to figure 3.1, the shaft is now mounted on two identical elastic supports which for

6 The detailed transformation of the equation of motion is given in Appendix B.

15
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

the sake of generality are assumed to have different stiffnesses k yb and k zb in the horizontal and

vertical direction, respectively. The bearings are fixed in space giving the assembly constant bearing
stiffnesses in the horizontal and vertical direction in the fixed coordinate system. The deflected disk
mass-centre is acted upon by the combined shaft and bearing elastic restoring forces. The bearings
and the shaft elasticity are arranged in series giving the resultant system stiffnesses as

Figure 3.2: LAVAL-Rotor in anisotropic elastic bearings

2k yb k ys 2k zb k zs
k ybs = ; k zbs = (3.10)
2k yb + k ys 2k zb + k zs

48 EI
where again k ys = k zs = k s = .
L3

To determine the equation of motion of the assembly in the fixed coordinate system, the coefficients in the
stiffness matrix in equation (3.7) are substituted by the equations (3.10). Hence, the equation of
motion in the fixed coordinate system becomes

y   k ybs
 m 0   ɺɺ 0   y
ɺɺ
+ K w


 0 m   ɺɺ +   =F ⇒ Mw =F. (3.11)
 z  0 k zbs   z  bs

Transforming equation (3.11) into the rotating coordinate system as outlined in Appendix B yields

16
 LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings

 m 0  ηɺɺ  0 2mΩ  ηɺ 

 0 m  ζɺɺ +  −2mΩ  
0  ζɺ 
   
  −mΩ 2 0   k0bs + ∆kbs cos 2α ∆kbs sin 2α   η 

+ +   = f

 0 − mΩ 2   ∆kbs sin 2α k0bs − ∆kbs cos 2α   ζ 

⇒ ( )
mvɺɺ + gvɺ + k m + kɶ bs v = f (3.12)

k ybs + k zbs k ybs − k zbs

where k0bs = ; ∆kbs = are the mean and deviatorc resultant bearing
2 2
and shaft stiffnesses, respectively.

From the equations (3.11) and (3.12) it can be seen that the equations of motion of the LAVAL-
rotor with circular shaft and anisotropic elastic bearings in the fixed and rotating coordinate
systems are mathematically completely different. In the fixed coordinate system, the equation of motion
(3.11) is a second order ordinary differential equation with constant coefficients. Since α = Ωt -β ,
the equation of motion (3.12) in the rotating coordinate system is a second order ordinary differential
equation with periodically time dependent coefficients as long as the bearing deviatoric stiffness
∆kbs is not zero. But for systems with isotropic elastic bearings, which sets the deviatoric stiffness
to zero, the equation of motion of the assembly becomes a second order ordinary differential
equation with constant coefficients as well.

3.1.3 Non-Circular Shaft and Rigid Bearings

In the analysis of a LAVAL-rotor with a non-circular shaft and rigid bearings as shown in figure 3.3, a
fixed coordinate system stages the difficulty of having to deal with varying stiffnesses along any fixed
direction in space. But considering a rotating coordinate system ensures that the stiffness variation
relative to the coordinate system disappears. The product area moment of the shaft cross section
becomes zero when orienting the rotating coordinate system such that the axes lie parallel to the
major and minor axis of the cross-section. Hence, such coordinate system decouples the equations of
motion in the two mutually perpendicular directions.

Now, the elastic restoring forces of the shaft are different in the η- and ζ-directions due to the loss
of shaft axisymmetry. The equation of motion in the rotating coordinate system can therefore be
determined from equation (3.9) by replacing the isotropic shaft stiffness ks by the corresponding

different shaft stiffnesses kη s and kζ s in the η- and ζ-directions, respectively, which yields

17
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 3.3: LAVAL-rotor with a non-circular shaft and

rigid bearings

 m 0  ηɺɺ  0 2mΩ  ηɺ    − mΩ 2 0   kη s 0   η 

+
 0 m  ζɺɺ  −2mΩ  + +   = f
    0  ζɺ    0 2
− mΩ   0 kζ s   ζ 

⇒ mvɺɺ + gvɺ + ( k m + k s ) v = f . (3.13)

The equation of motion of the system in the fixed coordinate system can then be determined by
transforming equation (3.13) from the rotating to the fixed coordinate system 7 . The resulting
equation follows as

y   k0 s + ∆k s cos 2α
 m 0   ɺɺ −∆ks sin 2α   y 
ɺɺ
+ K
ɶ w


 0 m   ɺɺ +   = F ⇒ Mw =F (3.14)
   z   −∆ks sin 2α k0 s − ∆k s cos 2α   z  s

kη s + kζ s kη s − kζ s
where k0 s = ; ∆k s = are the mean and deviatoric shaft stiffnesses,
2 2
respectively.

Unlike the previous case, for the LAVAL-rotor with a non-circular shaft and rigid bearings the
equation of motion (3.13) in the rotating coordinate system is a second order ordinary differential
equation with constant coefficients. On the other hand, the equation of motion (3.14) of the same
rotor in the fixed coordinate system is a second order ordinary differential equation with periodically
time dependent coefficients for non-zero deviatoric stiffness.

7 Refer to footnote 6 on page 15.

18
 LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings

3.1.4 Non-Circular Shaft and Anisotropic Elastic Bearings

In the LAVAL-rotor systems investigated above a proper selection of the coordinate system allows
the expression of the equation of motion using second order differential equation with constant
coefficients. In the present assembly, it can be seen that the stiffness of a LAVAL-rotor with a non-
circular shaft and anisotropic elastic bearings has time dependent properties in both the fixed and the
rotating coordinate system. In the fixed coordinate system, the bearing stiffnesses are constant in time
while the shaft stiffness varies between a maximum and a minimum value completing two cycles
for every shaft revolution. Considering the rotating coordinate system, the shaft stiffness appears to be
constant in time but now the bearing stiffnesses vary between the values of the bearing stiffnesses
in the two directions of the fixed coordinate system twice for every rotation.

Considering first the fixed coordinate system, the shaft stiffness is time dependent as in equation (3.14)
and the bearing stiffness is given by

k 0
K b =  yb . (3.15)
0 k zb 

As stated in section 3.1.2, the shaft and the bearing stiffnesses are in series with each other. Hence,
the total flexibility matrix of the system becomes

−1 −1
( Kɶ )
bs
−1
ɶ
= ( 2K b ) + K s( ) . (3.16)

The inverse of equation (3.16) yields the resulting stiffness matrix as8

ɶ = 1  K11
K
K12 
(3.17)
K  K 21 K 22 
bs

where K = k02s + 2 ( k yb + k zb ) k0 s − ∆k s2 + 4k yb k zb − 2 ( k yb − k zb ) ∆ks cos 2α

(
K11 = 2k yb 2k zb k0 s + k02s − ∆ks2 + 2k zb ∆ks cos 2α )
K12 = K 21 = −4k yb k zb ∆ks sin 2α

(
K 22 = 2k zb 2k yb k0 s + k02s − ∆ks2 − 2k yb ∆k s cos 2α )
The equation of motion in the fixed coordinate system will therefore be

8 The simplification of the resultant stiffness matrix for the particular cases treated in the sub-units 3.1.1 to 3.1.3 is
detailed in Appendix C.

19
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

y  1  K11
 m 0   ɺɺ K12   y 
ɺɺ
+ K
ɶ w


 +   =F ⇒ Mw =F. (3.18)
 0 m  ɺɺ
   z  K  K 21 K 22   z  bs

In the rotating coordinate system, the shaft stiffness is constant and the bearing stiffness is time
dependent. Hence, from the equation

−1 −1
( kɶ )
bs (
= 2kɶ b ) + (k ) s
−1
(3.19)

a relation similar to equation (3.17) can be generated by using the shaft stiffness matrix given in
equation (3.13) and the bearing stiffness matrix

 k + ∆kb cos 2α ∆kb sin 2α 

kɶ b =  0b (3.20)
 ∆kb sin 2α k0b − ∆kb cos 2α 

k yb + k zb k yb − k zb
where k0b = ; ∆kb = are the mean and deviatoric bearing stiffnesses,
2 2
respectively.

The resulting stiffness matrix is therefore

1 k k12 
kɶ bs =  11 (3.21)
k  k21 k22 

where k = k02b + 2 ( kη s + kζ s ) k0b − ∆kb2 + 4kη s kζ s − 2 ( kη s − kζ s ) ∆kb cos 2α

(
k11 = 2kη s 2kζ s k0b + k02b − ∆kb2 + 2kζ s ∆kb cos 2α )
k12 = k21 = −4kη s kζ s ∆kb sin 2α

(
k22 = 2kζ s 2kη s k0b + k02b − ∆kb2 − 2kη s ∆kb cos 2α . )
Hence, the equation of motion in the rotating coordinate system becomes

 m 0  ηɺɺ  0 2mΩ  ηɺ    − mΩ 2 0 1  k11 k12   η 

 0 m  ζɺɺ  −2mΩ
+  + 2
+    = f
    0  ζɺ    0 − mΩ  k  k21 k22   ζ 

⇒ ( )
mvɺɺ + gvɺ + k m + kɶ bs v = f . (3.22)

20
 LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings

One special case is when the bearing properties are isotropic. In this case the bearing stiffnesses in
the y- and z-directions will be the same making the bearing deviatoric stiffness in equation (3.20) to
be zero. Hence, for isotropic bearing stiffnesses the equation of motion (3.22) can be simplified to

mvɺɺ + gvɺ + ( k m + k bs ) v = f (3.23)

 kηbs 0  2kb kη s 2kb kζ s

where k bs =   ; kηbs = ; kζ bs = ; kb = k yb = k zb .
 0 kζ bs  2kb + kη s 2kb + kζ s

From the equations (3.18) and (3.22) it can be seen that the equations of motion of a general
LAVAL-rotor-bearing system with a non-circular shaft and anisotropic elastic bearings are second
order differential equations with time dependent coefficients irrespective of the coordinate system
selected. Furthermore, the stiffness matrices in the equations are non-linear with respect to the
deviatoric stiffnesses ∆k s and ∆kb of the shaft and the bearings, respectively. Such systems can not
be solved using the standard methods of solution for a rotor-bearing system with constant matrices.
Therefore, the next unit discusses a method for solving such equations.

3.2 HILL’s Method for Eigenvalue and Stability Analysis

As shown in the previous section, the equations of motion of a general LAVAL-rotor with a non-
circular shaft and anisotropic elastic bearings are second order differential equations with time dependent
coefficients. The HILL’s method is one known method to address such equations for eigenvalue and
stability analysis. In this section, the HILL’s equations will be developed in both the fixed and rotating
coordinate system.

3.2.1 Fixed Coordinate System

The equation of motion of a general LAVAL-rotor in the fixed coordinate system appears as given in
equation (3.18) (repeated here as 3.18r)

ɺɺ
+ K
ɶ w


Mw bs =F (3.18r)

ɶ is given in equation (3.17).

where the time dependent stiffness matrix K bs

21
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Linearizing this time dependent stiffness matrix with respect to the deviatoric shaft stiffness ∆k s as
discussed in Appendix D gives a harmonic stiffness matrix, which can then be written using
exponential functions9 as

ɶ = ∆K e − jΩɶ t + K + ∆K e jΩɶ t
K ; ɶ = 2Ω
Ω (3.24)
bs −1 0 +1

∆k s  1 − j  j 2β  k0 ybs 0  ∆k  1 j  − j 2 β
where ∆K −1 =  − j −1  e ; K =  0  ; ∆K +1 = s  e .
2  j −1
0
2    k0 zbs 

To solve the homogeneous equation (3.18r), the displacement w is assumed to be an exponential
function having periodic coefficients with the period Tɶ in the form

ɶ
ɶ
ɶ
ɶ
w = Φ e Λt ; Φ = Φ ( t ) = Φ t + Tɶ ( ) ; ɶ Tɶ = 2π .
Ω (3.25)

Deriving the first relation of equation (3.25) twice10 and inserting it into the homogeneous part of
equation (3.18r) yields

 ɺɺ
ɶ
ɺɶ
ɶ 
ɶ

ɶ Φ
M  Φ + 2ΛΦ + Λ 2 Φ  + K = 0. (3.26)
bs
 

ɶ
Periodic functions such as Φ can be written as the sum of simple sinusoidal functions using a
FOURIER-series as

ɶ ∞
ˆ
ɺɶ ∞
ɺɺ
ɶ ∞
ˆ juΩɶ t

ɶ ˆ e juΩɶ t
∑ ∑ = − ∑ u 2Ω
ɶ
Φ= Φ u e juΩt ; Φ = ju ΩΦ ; Φ ɶ 2Φ . (3.27)
u u e
u = −∞ u = −∞ u = −∞

Substituting the equations (3.24) and (3.27) into the equation of motion (3.26) gives

∞
ˆ juΩɶ t ∞


M ∑ −u 2 Ω
u =−∞
(
ɶ 2 + 2ΛjuΩ
ɶ + Λ2 Φ
ue
ɶ
) ɶ
+ ∆K −1e− jΩt + K 0 + ∆K +1e jΩt ( ) ∑ Φˆ e
u =−∞
u
ɶt
juΩ
= 0. (3.28)

For the above mathematical equation to be fulfilled, the terms having the same exponential factors
must be equal on both sides of the equation i.e. zero in this case. Hence, selecting the equal

9 e j 2α + e − j 2α e j 2α − e − j 2α
cos 2α = ; sin 2α = − j ; α = Ωt − β .
2 2


ɺɶ Λt
ɶ Λt ɺɺ
ɶ Λt
ɺɺ
= Φ

ɶɺ Λt
ɶ Λt
10
wɺ = Φ e + ΛΦ e ; w e + 2ΛΦ e + Λ 2Φ e .

22
 LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings

exponential terms and rearranging the resulting relations will give a HILL’s eigenvalue problem for the
eigenvalues Λ in the fixed coordinate system as11

( Λ Mˆ + ΛCˆ + Kˆ ) Φˆ = 0 .
2
(3.29)

The matrices and vectors in the preceding HILL’s eigenvalue problem are infinite in dimension
rendering it to be unsolvable. But it is a well known fact that components corresponding to higher
indices of summation of a FOURIER-series have less effect than those corresponding to lower
indices. Hence, truncating the index of summation from − ∞ to ∞ to –U to U will not affect the
accuracy of the eigenvalues nor the stability ranges calculated from equation (3.29) greatly. This
change of the bounds of summation limits the size of the matrices in equation (3.29) to
2(2U + 1) × 2(2U + 1) and of the vector to 2(2U + 1) × 1 for the LAVAL-rotor with one disk in the
middle of the shaft12.

3.2.2 Rotating Coordinate System

In the rotating coordinate system, the equation of motion of a general LAVAL-rotor is given in equation
(3.22) (repeated here as 3.22r)

( )
mvɺɺ + gvɺ + k m + kɶ bs v = f (3.22r)

where the time dependent stiffness matrix kɶ bs is given in equation (3.21).

Here again, the time dependent stiffness matrix is linearized as discussed in Appendix D and
written using exponential functions as

ɶ ɶ
kɶ bs = ∆k −1e − jΩt + k 0 + ∆k +1e jΩt ; ɶ = 2Ω
Ω (3.30)

1 j  j 2 β  k0ηbs 0   1 − j  − j2β
where ∆k −1 = ∆kb   e ; k0 =   ; ∆k +1 = ∆kb  e .
 j −1  0 k0ζ bs   − j −1 

Assuming again a homogeneous solution of the form

11 Details of this exponents comparison and the resulting matrices are given in Appendix E.1.
12 Note here that the given size corresponds to 2× 2 M and K matrices. In general, the size of the matrices M̂ , Ĉ and
ɶ in equation (3.18).
K̂ will be s (2U + 1) × s (2U + 1) where s is the size of the matrices M and K bs

23
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

ɶ
ɶ
ɶ
ɶ
v = φ eλ t ; φ = φ ( t ) = φ ( t + Tɶ ) ; ɶ Tɶ = 2π ,
Ω (3.31)

the derivation13 and substitution of this assumed solution in the homogeneous part of equation
(3.22r) will give

 ɺɺ
ɶ
ɺɶ
ɶ  
ɶɺ
ɶ 
ɶ

m  φ + 2λφ + λ 2φ  + g  φ + λφ  + k m + kɶ bs φ = 0 .
   
( ) (3.32)

ɶ
Again expressing φ using the FOURIER-series

ɶ ∞
ˆ
ɺɶ ∞
ɺɺ
ɶ ∞

ɶ φˆ e juΩɶ t ɶ 2φˆ e juΩɶ t .
∑φ e ∑ ; φ = − ∑ u 2Ω
ɶt
juΩ
φ= u ; φ= ju Ω u u (3.33)
u =−∞ u = −∞ u = −∞

and following the procedure outlined in the previous sub-unit gives

∞
∞
ˆ ɶ
m ∑ −u 2 Ω ɶ + λ 2 φˆ e juΩɶ t + g
(
ɶ 2 + 2λ juΩ
u =−∞
u ∑ juΩɶ + λ φu e juΩt
) u =−∞
( )
∞


( ɶ
+ k m + ∆k −1e − jΩt + k 0 + ∆k +1e jΩt
ɶ
) ∑ φˆ e
u =−∞
u
ɶt
ju Ω
=0 (3.34)

which leads to the HILL’s eigenvalue problem for the eigenvalues λ in the rotating coordinate system of the
form

(λ mˆ + λ cˆ + kˆ )φ
ˆ = 0
.
2
(3.35)

Here again, the size of the matrices and of the vector are 2(2U + 1) × 2(2U + 1) and 2(2U + 1) × 1 ,
respectively, for an index of FOURIER-series summation from − U to U .

Solving equation (3.29) or (3.35) for the eigenvalues Λ and λ , respectively, gives the free vibration
solution of the general LAVAL-rotor with a non-circular shaft and anisotropic elastic bearings from
which the assembly’s stability condition can be determined. But the stability of the assembly can be
affected by the presence of damping. Hence, before a stability analysis of the rotor system is carried
out damping is included in the next sub-unit.

ɺɶ
ɶ
ɺɺ
ɶ
ɶɺ
ɶ

13
vɺ = φ eλ t + λφ eλ t ; vɺɺ = φ eλ t + 2λφ eλ t + λ 2φ eλ t .

24
 LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings

3.3 Inclusion of Damping

In practical rotors there is always some sort of energy loss. This source of damping leads to a change
in the dynamic behaviour of the system requiring some way of integrating this phenomenon in the
model so as to get accurate results. Damping elements are mechanical elements which dissipate
energy and hence are usually integrated in rotor models to achieve this purpose. Unless otherwise
required, the commonly used dampers are velocity-proportional dampers. For the case of generality
a non-circular shaft is considered and for simplicity the damping is set on the central disk in the
derivation of the equations of motion.

3.3.1 Equations of Motion

The equation of motion of the LAVAL-rotor in figure 3.3 with velocity-proportional damping added
to the disk can be derived from equation (3.18) in the fixed coordinate system where the damper
provides a resistive force proportional to the velocities in the y- and z-direction. Hence, this
equation can be extended to incorporate the damping as

ɺ ɶ


ɺɺ
+ Dw
Mw + K bs w = F . (3.36)

For simplicity, the damping constants in the two directions are set equal to each other without
cross-coupling damper elements14.

The equation of motion in the rotating coordinate system can be derived using the coordinate
transformation outlined in the Appendices A and B. Hence, in extension of equation (3.22) one will
get

( )
mvɺɺ + ( g + d ) vɺ + k m + k d + kɶ bs v = f . (3.37)

As can be observed from the equations (3.36) and (3.37) such a LAVAL-rotor with the specified
type of damping has an equation of motion with time dependent displacement-proportional
matrices as in the previously considered undamped case and time independent mass and velocity-
proportional matrices in both the fixed and rotating coordinate systems. Therefore, here again the
HILL’s method is required to analyse the system.

14 The matrices are detailed in Appendix B.

25
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

3.3.2 Eigenvalue and Stability Analysis

For the homogeneous case, a solution similar to equation (3.25) and (3.31) can be assumed. For the
fixed coordinate system, substituting the assumed solution (3.25) and its derivatives into equation (3.36)
and dividing by the exponential term e Λt gives

 ɺɺ
ɶ
ɺɶ
ɶ  
ɶɺ
ɶ 
ɶ

ɶ Φ
M  Φ + 2ΛΦ + Λ 2 Φ  + D  Φ + ΛΦ  + K =0 (3.38)
bs
   

where again the stiffness matrix is written in the exponential form expressed in equation (3.24) by

ɶ = ∆K e − jΩɶ t + K + ∆K e jΩɶ t
K ; ɶ = 2Ω .
Ω (3.24r)
bs −1 0 +1

Writing the periodic function in equation (3.25) in terms of a FOURIER-series as in the equations
(3.27) and substituting them into the equation of motion (3.38) yields

∞
ˆ juΩɶ t ∞

M ∑ −u 2 Ω ( ) ∑ juΩɶ + Λ Φˆ u e juΩt ( )
ɶ
ɶ 2 + 2ΛjuΩ
ɶ +Λ2 Φ e + D
u
u =−∞ u =−∞
∞
ˆ

( ɶ ɶ
+ ∆K −1e − jΩt + K 0 + ∆K +1e jΩt )∑
u =−∞
ɶ
Φ u e juΩt = 0 . (3.39)

Here again, from an exponential comparison detailed in Appendix E.1 equation (3.39) will simplify
to a characteristic equation for the eigenvalues Λ in the fixed coordinate system similar to that given in
equation (3.29) (repeated here as 3.29r)

( Λ Mˆ + ΛCˆ + Kˆ ) Φˆ = 0 .
2
(3.29r)

Similarly, the eigenvalue problem for the eigenvalues λ in the rotating coordinate system corresponds to
equation (3.35) (repeated here as 3.35r)

( λ mˆ + λ cˆ + kˆ )φˆ = 0
2
(3.35r)

where the details of the matrices correspond to those given in Appendix E.1 for the preceding
eigenvalue problem in the fixed coordinate system.

The matrices and vectors in the equations (3.29r) and (3.35r) are again infinite in dimension. As
outlined in section 3.2.1 and 3.2.2, the dimensions of both can be reduced to 2(2U + 1) × 2(2U + 1)

26
 LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings

and 2(2U + 1) × 1 , respectively, without much loss of accuracy in the determination of the
eigenvalues and stability ranges.

3.4 Inclusion of Bearing Mass

In the LAVAL-rotor model all the rotor elements except the disk are taken to be massless. In this
unit the bearing mass is considered besides the disk mass. The disk is still mounted at the centre of
the non-circular shaft. In turn, the shaft is placed in two anisotropic visco-elastic bearings which now
possess both inertia and damped-elastic properties. Figure 3.4 shows a simple mass-damper-spring
model that can be used to represent such an assembly.

Figure 3.4: Simplified model of a damped LAVAL-rotor

with bearing mass and visco-elasticity

For the sake of generality, an anisotropic visco-elastic bearing is considered, but for simplicity
damper elements with equal resistances in the two directions and with no cross-coupling are
considered as in the previous sub-unit.

Using the subscripts ‘d’, ‘s’ and ‘b’ for ‘disk’, ‘shaft’ and ‘bearing’, respectively, the equation of motion of
this rotor-bearing assembly in the fixed coordinate system is given as

27
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

M d 0   w ɺɺ
  Dd

0   wɺ d   K ɶ ɶ
−K  wd   Fd 
d s s
 0  ɺɺ
 +    +  ɶ
=

ɶ + 2K   w
2M b   w 2Db   wɺ b   −K


 b 0 s K s b b  Fb 

ɺ ɶ


ɺɺ
+ Dw
⇒ Mw + K bs w = F (3.40)

ɶ , w
where M d , Dd , K s d and Fd are similar to the matrices and vectors in equation (3.36)

mb 0 d 0 k 0
 y 
F 
and M b =   ; Db =  b  ; K b =  yb  ; wb =  b  ; F =  yb  .
0 mb  0 db  0 k zb   zb   Fzb 

In the rotating coordinate system, the equation of motion of the above given rotor-bearing model can
be determined using again a coordinate transformation outlined in Appendix B. Hence, the
equation of motion becomes

m d


0  vɺɺd   g d + d d 0


 vɺd  k m + k s −k s  v
d   f d 

+  
+  
 = 

 0
 2mb   vɺɺb   0 2 ( g b + db )   vɺb   −k s ( )
k s + 2 k mb + k db + kɶ b   vb   f b 
 

⇒ mvɺɺ + cvɺ + kɶ v = f (3.41)

where m d , dd , g d , k m , k s , vd and f d are similar to the matrices and vectors in eq. (3.37)

mb 0  0 2mb Ω  − mb Ω 2 0 
and m b =  ; g = ; k =   ;
0 mb  b  −2 m Ω
 b 0  mb
 0 − mb Ω 2 

d 0  0 d b Ω
η 
 fηb 
db =  b  ; k db =   ; vb =  b  ; fb =   ;
0 db  − d b Ω 0  ζ b   fζ b 
 k + ∆kb cos 2α −∆kb sin 2α  k +k k −k
kɶ b =  0b  ; k0b = yb zb ; ∆kb = yb zb .
 −∆kb sin 2α k0b − ∆kb cos 2α  2 2

The equations (3.40) and (3.41) are the general equations of motion in the fixed and rotating coordinate
system, respectively, for a LAVAL-rotor with a non-circular shaft and anisotropic visco-elastic bearings with
bearing mass. The particular cases of undamped and/or isotropic elastic bearings can be determined
from the above two equations by setting the unnecessary matrices to zero. As an example,
considering isotropic bearing stiffnesses means the bearing deviatoric stiffness ∆kb is zero setting

kɶ b and hence kɶ to be time independent in the rotating coordinate system. If on the other hand, an

28
 LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings

undamped system is considered, then the matrices Dd , Db (hence D ), dd and d b in the fixed and
rotating coordinate system, respectively, are zero matrices15.

Being time dependent ordinary differential equations both equations (3.40) and (3.41) can be solved
using again the HILL’s method of solution. For the homogeneous equation, this leads to equations
similar to equation (3.29) and (3.35) in the fixed and rotating coordinate system, respectively, with
the main difference coming in the definition of the individual matrices. The size of the matrices in
the HILL’s equation also changes due to the increased degrees of freedom of the system. Hence,
the overall sizes of the matrices and the vectors after the implementation of the HILL’s method
become 4 ( 2U + 1) × 4 ( 2U + 1) and 4 ( 2U + 1) ×1 , respectively, for an index of FOURIER-series

summation from − U to U .

3.5 Unbalance Vibrations

In this section the unbalance response of a LAVAL-rotor with a non-circular shaft and anisotropic elastic
bearings is analysed. For the sake of generality, a damped rotor assembly is considered. For such an
assembly, the stiffness matrix is time dependent in both the fixed and rotating coordinate system.
But for ease of interpretation, the unbalance response is analysed in the fixed coordinate system in this
sub-unit. As seen from equation (3.36) the equation of motion is therefore

ɺ ɶ


ɺɺ
+ Dw


ˆ
ˆ
Mw + K bs w = FU ; FU = FU− e − jΩt + FU+ e jΩt (3.42)

ɶ are given in equation (3.18) and Appendix B

where M , D and K bs

ˆ − mε Ω 2  j
ˆ + mε Ω 2  − j 
and FU =   ; FU =  .
2 1  2 1

To solve equation (3.42) assume solutions of the form

∞
ˆ
∞
U−
∑ ∑ Φˆ
ɶ ɶ
wU+ = e jΩt ΦUu + e juΩt ; wU− = e− jΩt u e juΩt . (3.43)
u =−∞ u =−∞

Deriving the equations (3.43) once and twice with respect to time, substituting it into equation
(3.42) and again comparing the equal exponential terms gives the two sets of linear equations16

15 Cross coupled bearing damping and different damping coefficients in the two directions are considered in chapter 6
along with the Finite Element Method.
16 Details of this exponents comparison and the resulting matrices are given in Appendix E.2.1.

29
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

ˆ U +
ˆ +
ˆ U −
ˆ −
KU +Φ = FUH ; KU −Φ = FUH (3.44)

ˆ U +
ˆ U −
ˆ U +
ˆ U −
for the vectors Φ u and Φ u in the hypervectors Φ and Φ , respectively. Similar to the
analysis for the homogenous equation, the series in equation (3.43) can be broken after a finite

ˆ U +
ˆ U −
series of terms without incurring significant errors. Once the values of Φ and Φ are
determined from the above equations, their elements will be substituted into the equations (3.43) to
determine the unbalance response of the LAVAL-rotor.

3.6 Vibrations due to Gravitational Load

In this section the gravitational response of a LAVAL-rotor with a non-circular shaft and anisotropic elastic
bearings is analysed. Here again a damped rotor assembly is considered for the sake of generality.
Unlike a classical LAVAL-rotor, the response of the present assembly to the gravitational load is a
function of time in the fixed coordinate system since the stiffness matrix is time dependent in this
coordinate system. Note here that the stiffness matrix is time dependent in both the fixed and
rotating coordinate system and the gravitational force is constant in the fixed coordinate system.
Hence, the problem is simpler to tackle in the fixed coordinate system.

The equation of motion of a LAVAL-rotor with a non-circular shaft mounted in anisotropic

bearings with gravitational load in the fixed coordinate system follows from equation (3.36) as

ɺ ɶ


ɺɺ
+ Dw

0 
Mw + K bs w = FG ; FG = mg   (3.45)
1 

ɶ are given in equation (3.18) and Appendix B.

where M , D and K bs

To solve equation (3.45) assume a solution of the form

∞
G
∑ Φˆ
ɶ
wG = u e juΩt . (3.46)
u =−∞

Deriving equation (3.46) once and twice with respect to time, substituting it into equation (3.45)
and again comparing the equal exponential terms gives17

ˆ

K G Φ G = FGH (3.47)

17 Details of this exponents comparison and the resulting matrices are given in Appendix E.2.2.

30
 LAVAL-Rotor with Non-Circular Shaft and Anisotropic Bearings

ˆ G
ˆ G
ˆ G
where Φ u are the elements of Φ . Once the value of Φ is determined from the above equation,
its elements will be substituted into equation (3.46) to get the gravitational response of the rotor.

The next chapter uses the mathematical relations and solutions developed in this chapter to
investigate two particular LAVAL-rotor assemblies (one with rigid bearings and one with
anisotropic elastic bearings) which are also experimentally investigated so as to determine the
accuracy of the physical and mathematical models and the solution methods described in the
present chapter.

31
 4 Chapter 4

NUMERICAL AND EXPERIMENTAL STUDIES

In the previous chapter, a LAVAL-rotor with a circular or non-circular shaft mounted on rigid or
elastic bearings (isotropic and anisotropic) was modelled and the equations of motion were derived.
These equations of motion are solved in the present chapter so as to determine the stability and the
unbalance and gravitational response of the rotor, giving particular focus to a LAVAL-rotor with a non-
circular shaft.

This chapter also gives a general picture of the behaviour of a LAVAL-rotor with a non-circular
shaft based on the analytical results of a particular selected assembly. The different properties
influencing the dynamics of the assembly are varied in order to predict the behaviour of more
general rotor-bearing systems classified in this category.

Further, this chapter also explains the experimental set-up that is constructed and tested. The
constructed set-up is simple enough to be considered as a LAVAL-rotor. The response of this set-
up is compared against the theoretical findings. This way the validity of the solution path that is
being followed can be justified.

4.1 LAVAL-Rotor with Non-Circular Shaft and Different Types of Bearings

4.1.1 Rigid Bearings

Consider the LAVAL-rotor with a non-circular shaft and rigid bearings given in figure 3.3 whose equation
of motion is expressed in the equations (3.13) and (3.14) in the rotating and fixed coordinate
system, respectively. As can be observed from the two equations, the assembly is simpler to solve in
the rotating coordinate system, since all the coefficients of the differential equation

 m 0  ηɺɺ  0 2mΩ  ηɺ    − mΩ 2 0   kη s 0   η 

 0 m  ζɺɺ  −2mΩ
+  + +   = f
    0  ζɺ    0 − mΩ 2   0 kζ s   ζ 

⇒ mvɺɺ + gvɺ + ( k m + k s ) v = f (3.13r)

are constant in this reference system.

Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

• Free Vibrations and Stability

As in the analysis of a classical rotor-bearing assembly, the free vibrations are investigated first. For
that, the right hand side of equation (3.13r) is set to zero and the resulting equation is solved by the
methods known and discussed in a number of books and manuscripts [28, 29, 42, 58, 85, 98].
Dividing every element of the homogeneous equation (3.13r) by the mass of the disk gives

ηɺɺ  0 2Ω  ηɺ  ω02η − Ω 2 0  η  0 

 ɺɺ +    ɺ +  2 2 
=  (4.1)
ζ   −2Ω 0  ζ   0 ω0ζ − Ω  ζ  0 

kη s kζ s
where ω0η = ; ω0ζ = are the natural circular frequencies of the shaft-disk
m m
system in the η- and ζ-direction, respectively, for zero rotational speed.

Assume a solution of the form

η  ηˆ  λ t
v =   =   e where λ = λ re + jλ im . (4.2)
ζ  ζˆ 

Substituting equation (4.2) into the equation of motion (4.1) gives

λ 2 + ω02η − Ω 2
( ) 2Ωλ  ηˆ  0
    =   . (4.3)
 −2Ωλ λ 2 + ω02ζ − Ω 2  ζˆ  0 
( )


The non-trivial solution of equation (4.3) comes from setting the determinant of the matrix to zero.
This leads to a characteristic equation for the eigenvalues λ given by

λ 4 + (ω02η + ω02ζ + 2Ω 2 ) λ 2 + (ω02η − Ω2 )(ω02ζ − Ω 2 ) = 0 . (4.4)

As can be observed in this equation, it is a fourth order polynomial, which can be solved simply as
the coefficients of the odd exponents are zero. Hence, the four eigenvalues of the system are

λ = ± − (ωav2 + Ω 2 ) ± ωav 4Ω 2 + ωav2 µ 2 (4.5)

2
ω02η + ω02ζ ω02η − ω02ζ kη s − kζ s
where ω = av ; µ= = are the average natural circular
2 2ωav2 kη s + kζ s
frequency and the non-circularity of the shaft, respectively.

34
 Numerical and Experimental Studies

Considering the assumed solution (4.2), the real part of the exponential term has a significant role
for the system stability. If the value of λre is positive, then the solution in equation (4.2) increases
exponentially. This means that the system is unstable for the vibration amplitude tends to infinity
with time. If on the other hand λre is negative, the amplitude tends to zero with time, which is a

typical asymptotically stable behaviour of a damped system. If λre is zero, then the amplitude of
equation (4.2) remains constant with time and the system is stable.

Figure 4.1: Imaginary part of the eigenvalues in the

rotating coordinate system

Figure 4.2: Real part of the eigenvalues in the rotating

coordinate system

Figures 4.1 shows the imaginary parts of the four eigenvalues λ of a LAVAL-rotor with a non-
circular shaft in rigid bearings. From the figure presented on the left side, it can be seen that the
imaginary part of one eigenvalue pair decreases in magnitude to zero as the rotational speed ratio

35
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

increases. For rotor systems with a non-circularity greater than zero, it remains zero for some range
of rotational speed ratio before it starts increasing in magnitude. For an observer in a rotating
coordinate system rotating with the same rotational speed as the rotor, the rotor appears to be static.
Hence, the eigenvalues with zero imaginary part indicate a natural circular frequency which is equal
to the rotational speed of the rotor. From the classical LAVAL-rotor it is known when the
rotational speed equals the natural circular frequency, the rotor motion is accompanied by large
deflection i.e. a critical rotational speed. It is further known that this phenomenon occurs at one
particular rotational speed ratio (Ω ωav = 1) for the classical LAVAL-rotor. But for a LAVAL-
rotor with a non-circularity this phenomenon occurs over a range of operation speed ratio. To
make things worse, this range of speed is accompanied by an eigenvalue with positive real part as
shown in figure 4.2. This results in the already large deflection to increase exponentially with time
leading to instability.

In the fixed coordinate system, the eigenvalues Λ can be simply determined from the eigenvalues λ in
the rotating coordinate system. Since the rotational motion of the rotating reference frame is the
only difference, inclusion of this effect will suffice to determine the eigenvalues in the fixed
reference frame. Hence, the eigenvalues Λ become

Λ = λ + jΩ ⇒ Λ = Λ re + j Λ im = λre + j ( λim + Ω ) . (4.6)

The imaginary parts of the eigenvalues Λ are given in figures 4.3 while the real parts of the
eigenvalues Λ in the fixed coordinate system are equal to the real parts of the eigenvalues λ in the
rotating coordinate system. Hence, the plot is the same as that shown in figure 4.2.

Figure 4.3: Imaginary part of the eigenvalues in the fixed

coordinate system

36
 Numerical and Experimental Studies

It can be simply proved that there is one eigenvalue with positive real part (as in the left plot of
figure 4.2) leading to unstable rotor operation if the last term of the characteristic equation (4.4) is
less than zero. Hence, the unstable region of rotational speed is given by

(ω 2
0η )( )
− Ω 2 ω02ζ − Ω 2 < 0 ⇒ ωav 1 − µ < Ω < ωav 1 + µ . (4.7)

From this inequality a stability plot of the LAVAL-rotor can be plotted. The region of instability as
a function of rotational speed is determined from equation (4.7) and is plotted in figure 4.4.

unstable

Ω stable
= 1− µ
ωav

stable
Ω
= 1+ µ
ω av

Figure 4.4: Stability plot of LAVAL-rotor with a non-

circular shaft in rigid bearings

The stability plot shows that for any selected shaft non-circularity µ the rotor will have a stable
operation until it reaches the first natural frequency ω0ζ . Then the rotor has an unstable operation

as the rotational speed is increased further until the speed reaches the second natural frequency
ω0η . Once the second natural frequency is surpassed, the rotor retains its stability. Hence, the
region of zero imaginary part in figure 4.1 which corresponds to the non-zero real part in figure 4.2
is the same as the area marked as unstable in figure 4.4.

For one chosen non-circularity of µ = 0.4 , the imaginary and real parts of the eigenvalues appear as
shown in figures 4.5 and 4.6. These figures are the common presentations of the eigenvalues of a
LAVAL-rotor with a non-circular shaft in rigid bearings. These figures are actually a planar section
of the diagrams shown in figures 4.1-4.3. In the figures 4.5 and 4.6 it can be seen that over
operation speed ratio ranges where two eigenvalues have equal imaginary parts, they possess
opposing sign non-zero real parts resulting in unstable behaviour. In the rotating coordinate system, the

37
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

5 8

6
1− µ

Λim/ωav
λim/ωav

4
0 1− µ
2
1+ µ
0 1+ µ

-5 -2
0 1 2 3 0 1 2 3
Ω/ωav Ω/ωav
Figure 4.5: Imaginary part of the eigenvalues in the
rotating and fixed coordinate system for
µ = 0.4

0.2 0.2

0.1 1 − µ 0.1 1− µ
1+ µ 1+ µ
Λre/ωav
λre/ωav

0 0

-0.1 -0.1

-0.2 -0.2
0 1 2 3 0 1 2 3
Ω/ωav Ω/ωav

Figure 4.6: Real part of the eigenvalues in the rotating

and fixed coordinate system for µ = 0.4

imaginary parts of the eigenvalues over the unstable operation speed ratio are zero while they are
equal to the operation speed in the fixed coordinate system.

• Unbalance Vibrations
The next common step in the analysis of rotor bearing systems is the investigation of the unbalance
response. As previously shown in figure 3.1, it is assumed that the mass centre C of the disk is
somehow off its geometric centre W described by an unbalance radius ε. The equations of motion
of the system were derived in chapter 3.1.3 as shown in the equations (3.13) and (3.14) where the
unbalance forces are given in equations (3.9) and (3.7), respectively.

Again, the rotating coordinate system proves to have advantages over the fixed coordinate system in
expressing not only the stiffness matrix but also the unbalance force as time independent quantities.

38
 Numerical and Experimental Studies

Dividing equation (3.13) by the mass of the disk and introducing the natural circular frequencies
defined in equation (4.1) yields

ηɺɺ  0 2Ω  ηɺ  ω02η − Ω 2 0  η  2  sin β 

 ɺɺ +    ɺ +     = ε Ω  . (4.8)
ζ  − 2Ω 0  ζ   0 ω02ζ − Ω 2  ζ  cos β 

In the previously stated references, the solution method for such a non-homogeneous ordinary
differential equation is outlined extensively.

Being a constant load, a time independent solution for the unbalance response in the rotating
coordinate system is present. Hence, the particular solution of equation (4.8) reduces to

 sin β 
ω02η − Ω 2  η   sin β 
η   2 2 
0 2  ω0η − Ω 
 2 2  
= ε Ω2   ⇒ v = =εΩ 
cos β 
(4.9)
 0 ω0ζ − Ω  ζ  cos β  ζ   2 
2
ω0ζ − Ω 

which is equivalent to

 sin β 

 η 
ω (1 + µ ) − Ω 
2 2
v =   = ε Ω 2  av . (4.10)
ζ  cos β
 2 
 ωav (1 − µ ) − Ω 2 

Hence, the unbalance response at any operation speed is a constant deflection in the η,ζ-plane. Figure
4.7 shows the response in the rotating coordinate system graphically for a value of β= π 4 .

From figure 4.7 it can be seen that the magnitude of the disk deflection in the η- and ζ-direction is
dependent on the operation speed and the non-circularity. A closer look at the equations (4.9) and
(4.10) reveals that the deflection in the η-direction approaches infinity as the shaft rotational speed
nears the natural circular frequency in the same direction (Ω = ω0η ) . Similarly, the deflection in the

ζ-direction approaches infinity as the shaft rotational speed approaches the natural circular
frequency in the ζ-direction (Ω = ω0ζ ) . A comparison with the figures 4.1-4.4 shows that the η-

direction unbalance resonance is the upper stability limit while the ζ-direction unbalance resonance
is the lower stability limit.

39
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 4.7: Unbalance response magnitude in the rotating

coordinate system for β= π 4

It is worth noting here that the unbalance deflection in the fixed coordinate system can be determined
by transforming the responses given in the equations (4.9) and (4.10) using the procedure outlined
in Appendix A.

• Vibrations due to Gravitational Load

Unlike the classical rotor, the gravitational load has a significant effect on a rotor with a non-circular
shaft. That originates mainly due to the fact that under rotation the shaft stiffness varies with time
and, therefore, the gravitational force becomes time varying in the rotating coordinate system. Hence,
the equation of motion of the rotor in this reference system follows from equation (3.13) with the


gravitational force introduced in the vector f . Dividing the resulting equation by the disk mass
yields

ηɺɺ  0 2Ω  ηɺ  ω02η − Ω 2 0  η  − sin α 

 ɺɺ 
+   ɺ +  2 2 
= g . (4.11)
ζ   −2Ω 0  ζ   0 ω0ζ − Ω  ζ   cos α 

Being a harmonic excitation, the assumed response will also be harmonic of the form

η  ηˆ sin α 
v = = . (4.12)
ζ  ζˆ cos α 

Deriving this assumed solution with respect to time and substituting it into the equation of motion
(4.11) gives the matrix equation after eliminating the harmonic terms as

40
 Numerical and Experimental Studies

ω02η − 2Ω 2 −2Ω 2  ηˆ  −1

 2 2 2   ˆ
= g  . (4.13)
 −2Ω ω0ζ − 2Ω  ζ  1

Solving for the amplitude vector gives

ηˆ  g  4Ω 2 − ωav2 (1 − µ ) 
vˆ =   = 4  2 2
. (4.14)
ˆ ( 2
)
2
ζ  ωav 1 − µ − 4ωav Ω
i
2
ωav (1 + µ ) − 4Ω 

Figure 4.8 shows the non-dimensional gravitational response in the rotating coordinate system.

Figure 4.8: Gravitational response in the rotating

coordinate system

It is clear from figure 4.8 that the rotor assembly may experience large deflections due to the
gravitational load. This is called the gravitational resonance and occurs when the denominator in
equation (4.14) is equal to zero i.e. for

ωav
Ω= 1− µ 2 . (4.15)
2

For practical rotors where the non-circularity µ is very small, the gravitational resonance speed is
approximately equal to the half of the average natural circular frequency. In the figure it can be seen
that the resonance speed approaches an operation speed ratio of 0.5 as the non-circularity µ
decreases to 0. On the other hand, for a large non-circularity the gravitational resonance speed
becomes smaller than this upper limit.

41
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Here again, the gravitational response in the fixed coordinate system can be determined by transforming
the deflections given in equation (4.14) per procedure outlined in Appendix A.

• Damped Rotors
In the rotor system discussed above the effect of damping is neglected. But in real systems there is
always some sort of energy dissipation which can be modelled as damper elements in the assembly.
Considering such external damping elements at the disk location, the equations of motion were
given in the equations (3.36) and (3.37) in the fixed and rotating coordinate system, respectively.
Showing all the details of the matrices, the equation of motion in the rotating coordinate system appears
as

ηɺɺ  2 Dωav 2Ω  ηɺ  ωav2 (1 + µ ) − Ω 2 2 Dωav Ω  η  1

 ɺɺ +    +   = f (4.16)
ζ   −2Ω 2 Dωav  ζɺ   −2 Dωav Ω ωav2 (1 − µ ) − Ω 2  ζ  m

d
where D = is the damping ratio of the rotor.
2mωav

Considering the homogeneous equation and assuming a solution of the form stated in equation
(4.2) reduces equation (4.16) to

λ 2 + 2 Dωav λ + ωav2 (1 + µ ) − Ω 2 2Ωλ + 2 Dωav Ω  ηˆ  0 

    =  . (4.17)
 −2Ωλ − 2 Dωav Ω λ 2 + 2 Dωav λ + ωav2 (1 − µ ) − Ω2  ζˆ  0 
 

The non-trivial solution of this equation comes from setting the determinant of the matrix to zero.
This leads to a characteristic equation for the eigenvalues λ which again is a fourth order polynomial of
the form

λ 4 + 4 Dωav λ 3 + 2 (ωav2 + 2 D 2ωav2 + Ω 2 ) λ 2 + 4 Dωav (ωav2 + Ω 2 ) λ

i

(4.18)
+ ωav4 1 − µ 2 − 2ωav2 Ω 2 + 4 D 2ωav2 Ω 2 + Ω 4  = 0.
( )

In contrast to the polynomial (4.4) of the undamped case, all the coefficients of the above
polynomial are non-zero. The above characteristic equation can be used to show the stability of the
system as a function of the operation speed, the non-circularity and the damping ratio. From the
ROUTH-HURWITZ criterion [65], the system is unstable when the last term in equation (4.18) is
less than zero as given by

42
 Numerical and Experimental Studies

ωav4 (1 − µ 2 ) − 2ωav2 Ω 2 + 4 D 2ωav2 Ω 2 + Ω 4 < 0 . (4.19)

Hence, from equation (4.19) the stability plot in figure 4.9 can be developed.

unstable

stable stable

Figure 4.9: Stability plot of a general damped LAVAL-

rotor with a non-circular shaft

Figure 4.9 shows that with increasing damping ratio, the unstable operation speed ratio of the rotor
becomes smaller than in the undamped case shown in figure 4.4.

The free and forced vibration responses of damped rotors will be discussed in later chapters along
with the analysis of a general non-circular shaft rotor.

4.1.2 Anisotropic Elastic Bearings

The equations of motion of a LAVAL-rotor with a non-circular shaft and anisotropic elastic bearings were
given in the equations (3.18) and (3.22) in the fixed and rotating coordinate system, respectively, and are
repeated here in the following expressions

y  1  K11
 m 0   ɺɺ K12   y 
ɺɺ
+ K
ɶ w


 0 m   ɺɺ +   =F ⇒ Mw =F, (3.18r)
   z  K  K 21 K 22   z  bs

43
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

 m 0  ηɺɺ  0 2mΩ  ηɺ    − mΩ 2 0 1  k11 k12   η 

  ɺɺ 
+  + 2
+    = f

 0 m  ζ   −2mΩ 0  ζɺ    0 − mΩ  k  k21 k22   ζ 

⇒ ( )
mvɺɺ + gvɺ + k m + kɶ bs v = f . (3.22r)

In Appendix D the linearization of the periodic stiffness matrices given in the above two equations
is discussed. Further, the application of the HILL’s method to solve the linearized equation of
motion is discussed in the previous chapter in the sections 3.2.1 and 3.2.2. In this section, the
relations in the above stated sub-units are used to develop the free vibrations and stability of such a
LAVAL-rotor. The disadvantage of using small bearing masses to approximate such rotors as
considered by Teichmann [97] will also be discussed at the end of this section.

• Free Vibrations and Stability

As in the previous sub-unit, the rotor assembly with anisotropic bearings will be considered
systematically starting with free vibration analysis. The free vibration analysis i.e. the solution of
equation (3.29), suffices for the eigenvalue identification of the LAVAL-rotor. The terms in this
equation can be simply determined from the general presentation outlined in Appendix E.

As it will be shown in this section, the shaft average stiffness defined in equation (3.14), the shaft
non-circularity given in equation (4.5) (hereafter denoted as µs ), and, additionally, the bearing
average stiffness and the bearing anisotropy defined by

k yb − k zb
k0b = k yb + k zb 18 ; µb = , (4.20)
k yb + k zb

respectively, affect the assembly’s stability.

In Appendix D it was proven that the linearized equations of motion are suitable when the ratio of
the shaft non-circularity to bearing anisotropy is either significantly greater or less than one. Hence,
a rotor with a weak shaft in relatively stiff bearings, where the stability behaviour is similar to the case of
the rigid bearings, is considered first. As shown in figure 4.10 there will only be one unstable region
for the most bearing anisotropy values as in the case of a rotor mounted on rigid bearings, with the
closest approximation to rigid bearings being the plot for µ b = 0 . However, the unstable region
splits into three strips for low shaft non-circularity and large bearing anisotropies. One difference of

18 This average bearing stiffness is for the resultant stiffnesses of both bearings in the y- and z-direction (see figure 3.2).

44
 Numerical and Experimental Studies

the stability plot in figure 4.10 to that given by Gasch et. al. [28] is that in this plot the rotor
becomes unstable at lower shaft non-circularity for low operation speed ratio. The second
difference with the work from Gasch is seen for higher bearing anisotropy where there emerges
additional narrow unstable operation speed ranges at low operation speed range. Both differences
can be attributed to the linearization procedure as the linearized equation of motion is actually true
when the average shaft stiffness is insignificant in comparison to the average bearing stiffness.

Figure 4.10: Stability plot of a LAVAL-rotor with a non-

circular shaft and anisotropic elastic bearings
for k0b = 10k0 s

Next, figure 4.11 shows the variation of the stability zone of an assembly having equal average bearing
and shaft stiffness for different values of shaft non-circularity, bearing anisotropy and rotational speed
ratio. The graph shows that the unstable region for low operation speed ratio increases significantly.
It can further be seen that the main unstable operation region splits into three zones for low shaft
non-circularity as the bearing anisotropy increases. Here again, there are additional narrow unstable
operation speed ratio ranges besides the main unstable region for high bearing anisotropy. The
difference to the results from Gasch et. al. [28] is again attributed to the linearized equation of
motion used in generating the results in figure 4.11. For very high bearing anisotropy, the main and
45
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

narrow regions of instability overlap giving a highly unstable low operation speed ratio range with
only narrow stable operation speed ratios for low shaft non-circularity. It can also be noted from
figure 4.11 that the bearing anisotropy does not affect the upper stability limit. In general, as the
operation speed is increased from zero, a LAVAL-rotor with a non-circular shaft and anisotropic
elastic bearings may either pass through a number of stable and unstable operation speed ratios
before it becomes stable, it may pass through a single unstable operation speed ratio range after a
certain stable range, or it may stabilize after an initial unstable operation speed ratio range
depending on the shaft non circularity and the bearing anisotropy.

Figure 4.11: Stability plot of a LAVAL-rotor with a non-

circular shaft and anisotropic elastic bearings
for k0b = k0 s

As a third case, the stability plot of an assembly with a shaft considerably stiffer than the bearings is shown
in figure 4.12. The figure shows that the assembly is in general more stable than in the previous two
cases except for very high bearing anisotropy where the rotor will always be unstable. Here again,
the unstable operation speed range is seen to be wider for higher shaft non-circularity and the
unstable region splits into up to five regions as the bearing anisotropy increases. The figure further

46
 Numerical and Experimental Studies

Figure 4.12: Stability plot of a LAVAL-rotor with a non-

circular shaft and anisotropic elastic bearings
for k0b = 0.1k0 s

shows that the unstable region is generally over lower operation speed ratios as compared to the
previous cases.

• Damped Rotors
As in the case of the undamped system, the stability investigation of a LAVAL-rotor with damping is
now discussed. The stability plot of such a rotor in the fixed coordinate system can be determined from
the homogeneous part of the equation of motion (3.36). Applying the HILL’s method to solve the
equation for a number of damping ratios (as defined in equation 4.16) give the stability plots as
shown in figure 4.13.

As it can be seen from these plots, the case of zero damping shown on the upper left plot is similar
to that one shown in figure 4.10 corresponding to µ b = 0.75 . From the present figure, it can also
be seen that an increase of the damping ratio has a positive effect in stabilizing the rotor. It can
then be deduced that a further increase in damping ratio will completely stabilize the rotor.

47
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 4.13: Stability plot of a damped LAVAL-rotor

with a non-circular shaft and anisotropic
bearings for different damping ratios
( µb = 0.75 and k0b = 10k0 s )

• Bearing Mass
As a next extension, the inclusion of the bearing mass can be considered by the use of equation
(3.40). But the results of such analysis have been given in detail in the work of Teichmann [97]. In
the present dissertation, the results are given without further discussion of the equations. Figure
4.14 shows the stability plots of a LAVAL-rotor with a non-circular shaft mounted on damped
anisotropic elastic bearings considering the bearing mass. The figure also contains the results shown
in figure 4.13 to assist in comparison of the two methods.

As can be seen from the upper left plot in figure 4.14, the linearized equation of motion is more
conservative than the model including the bearing mass when considering the undamped case. But
as the damping ratio is increased, the results from the linearized equation of motion and the model
including the bearing mass becomes significantly different from each other. The result from the
linearized equation resembles the analytical results of the damped LAVAL-rotor mounted on rigid

48
 Numerical and Experimental Studies

Figure 4.14: Stability plot of a damped LAVAL-rotor

with a non-circular shaft and anisotropic
bearings for different damping ratios. o -
linearized stiffness matrix, + - with bearing
mass ( µb = 0.75 and k0b = 10k0 s )

bearings since the unstable region diminishes with increase in damping ratio. In the stability plot of
the model including the bearing mass, the lower stability limit is not greatly influenced by the
damping, but the upper stability limit is stretched to higher operating speed ratios for higher shaft
non-circularity. For low shaft non-circularity, the upper stability limit is also not significantly
influenced by the damping. This difference arises from the added bearing mass in the second case
due to which the assembly will have more than one mode shape. Further, an increase in damping
hinders the motion of the bearing mass reducing the model to approximate rotor mounted on rigid
bearings. Hence, the unstable region of the second model approaches that of a rigid bearing rotor
for low shaft non-circularity. The stretch of the upper stability limit for high shaft non-circularity
may arise from instabilities associated with the additional mode shapes of the model including the
bearing mass. It is therefore the author’s believe that the linearized equation of motion has

49
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

expressed the behaviour of such a LAVAL-rotor-bearing system better than the model with the
bearing mass.

4.2 Numerical Studies for a Test Rotor

In this sub-unit, a particular example for a test LAVAL-rotor with a non-circular shaft is selected and
its eigenvalues, unbalance response and gravitational response are discussed. The technical data of the
selected non-circular shaft LAVAL-rotor is given in Appendix F.1. Since the basic principles of
such a LAVAL-rotor are discussed in the previous sub-unit, the results will be given here without
further details of the equation used to generate the results. Here again, a rotor mounted in rigid and
anisotropic elastic bearings is investigated.

4.2.1 Rigid Bearings

The test rotor in rigid bearings shown in the sketch of figure 4.15 is addressed systematically as
discussed in the previous unit. First, the unstable operation speeds are determined and then the
unbalance and gravitational responses are discussed.

Figure 4.15: A sketch of the test rotor with a non-circular

shaft mounted in rigid bearings

For stability considerations, figure 4.16 shows the variation of the imaginary and real parts of the
eigenvalues in the rotating coordinate system calculated from equation (4.5) for different operation
speeds19.

19 Compare with the figures 4.5 and 4.6 left.

50
 Numerical and Experimental Studies

Figure 4.16: Imaginary and real part of the eigenvalues in

the rotating coordinate system

In the fixed coordinate system, the eigenvalues are determined from that in the rotating coordinate
system by equation (4.6). The results are shown in figure 4.1720.

Figure 4.17: Imaginary and real part of the eigenvalues in

the fixed coordinate system

Both figures show that the rotor is unstable in the speed range of 829-976 rpm for here the
eigenvalues have positive real parts. Since the system is undamped, the real parts of all the
eigenvalues are zero over the other operating speed ranges.

The unbalance response of the LAVAL-rotor in the rotating coordinate system is given in equation (4.9). In
the fixed coordinate system it can be determined from equation (4.9) by the coordinate transformation
outlined in Appendix A.

20 Compare with the figures 4.5 and 4.6 right.

51
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 4.18: Analytical unbalance response of the test

rotor in rigid bearings in the fixed and
rotating coordinate system for β= π 4

Figure 4.18 shows the unbalance response magnitude and radius of the rotor related to the unbalance
radius for the operation rotational speed shown. It can be seen that in the fixed and rotating coordinate
system the unbalance response have two unbalance resonance speeds which are equal to the circular
natural frequencies in the η- and ζ-directions.

The gravitational response in the rotating coordinate system is given in equation (4.14). Again, in the fixed
coordinate system it can be determined from equation (4.14) by the coordinate transformation outlined
in Appendix A.

Figure 4.19 shows the gravitational response of the above rotor for different operation rotational
speeds. From the figure it can be seen that the rotor assembly has one gravitational resonance
speed at 446 rpm when analysed in both the fixed and the rotating coordinate system. In the
previous sub-unit it was stated that the gravitational resonance circular frequency is approximately
half of the average natural circular frequency which is at 903 rpm for this test rotor. Hence, the
percentage error that will be incurred from approximating the gravitational resonance speed from
the average natural circular frequency is only -1.25%.

52
 Numerical and Experimental Studies

Figure 4.19: Analytical gravitational response of the test

rotor in rigid bearings in the fixed and
rotating coordinate system

Figure 4.20: Analytical total (unbalance and gravitational)

response of the test rotor in rigid bearings in
the fixed coordinate system for β= π 4

53
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Finally, the total response (unbalance and gravitational) of the test rotor is presented in figure 4.20. As
expected, it shows three resonance frequencies (i.e. one gravitational and two unbalance
resonances) which occur at 446 rpm, 829 rpm and 976 rpm. From the stability investigation it is
known that the test rotor is unstable between the two unbalance resonance speeds.

4.2.2 Anisotropic Elastic Bearings

The above rotor assembly is now analysed in anisotropic elastic bearings whose properties are also given
in Appendix F.1. Figure 4.21 gives the sketch of the investigated test rotor in the anisotropic elastic
bearings considered in this unit.

Figure 4.21: A sketch of the test rotor with a non-circular

shaft in anisotropic elastic bearings

Again, the rotor has been systematically investigated starting with the homogeneous equation of
motion. The HILL’s method has been applied to determine the eigenvalues of the assembly. Then
the non-homogeneous equation of motion is solved for the unbalance and gravitational response of
the rotor.

The eigenvalues and the stability are determined in the fixed coordinate system from equation (3.29) and
Appendix E.1. The imaginary and real parts of the eigenvalues are shown in figure 4.22. Figure 4.23
shows a magnified view of the imaginary parts of those eigenvalues which are associated with a
positive real part.

54
 Numerical and Experimental Studies

Figure 4.22: Imaginary and real part of the eigenvalues in

the fixed coordinate system

Figure 4.23: Magnified views of the imaginary part of the

eigenvalues in the fixed coordinate system
around unstable operation speed ranges

From the CAMPBELL-diagram in figure 4.22 it can be seen that there are six complex conjugate
pair eigenvalues differing the expected values of only two complex conjugate pair eigenvalues. This
comes from the application of the HILL’s method discussed in section 3.2 (refer to footnote 12 on
page 23). Figure 4.22 shows the variation of the basic as well as the redundant eigenvalues. The
other character of the assembly that can be concluded from the above two figures is that the real
part of the eigenvalues is non-zero when two eigenvalues possess equal imaginary parts over a
range of operation speed. In these operation speed ranges, one of the non-zero real parts is positive
resulting in the unstable behaviour of the rotor. The operation speed ranges with positive real parts
are found to be over 326-331 rpm, 387-391 rpm, 447-450 rpm, 608-711 rpm, 748-829 rpm and
875-946 rpm. Teichmann [97] has determined the three main unstable regions considering the

55
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

bearing mass. Teichmann has further identified that the average of the second unstable region is
approximately the average of the averages of the first and third unstable regions. Besides the three
main unstable regions, the linearized equation of motion has identified three minor unstable regions
whose averages are approximately equal to the half of the averages of the major unstable regions.

Next, the unbalance response of the assembly is determined also in the fixed coordinate system from
equation 3.44. Figure 4.24 shows the unbalance response for an assumed eccentricity of 1 mm. The
figure shows two local maxima at 219 rpm and 299 rpm besides the expected four unbalance
resonance speeds at 609 rpm, 711 rpm, 875 rpm and 948 rpm. The local maxima occur when the
operation speed is equal to the redundant eigenvalues and are equal to about one third of the
average of the pair of unbalance resonance speeds.

Figure 4.24: Analytical unbalance response of the test

rotor in anisotropic elastic bearings in the
fixed coordinate system

Finally, the gravitational response of the assembly follows from equation 3.47 and is shown in figure
4.25 giving two gravitational resonances at 331 rpm and 450 rpm. Here again, the gravitational
resonance speeds are close to half of the average unbalance resonance speeds.

56
 Numerical and Experimental Studies

Figure 4.25: Analytical gravitational response of the test

rotor in anisotropic elastic bearings in the
fixed coordinate system

In the next sub-unit, the preceding numerical results are compared with experimental results for the
considered test rotor.

4.3 Experimental Studies and Findings

The numerical investigation of a dynamic system is important in the design phase of structures that
will be subjected to external exciting forces. But before a certain numerical approach can be taken
as accurate enough to be used to model a design, it is advisable to compare the numerical results with
experimental results. This, however, can be best applied on a sufficiently simple model whose dynamic
responses can also be determined analytically but complicated enough to demonstrate as many
features of a real industrial rotor as possible.

4.3.1 Test Rotor Design

The previously analytically and numerically investigated LAVAL-rotor is constructed and its
properties and responses are observed and measured. As before, the technical data of the
investigated rotor is given in Appendix F.1.

57
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 4.26: Experimental set-up of a LAVAL-rotor with

a non-circular shaft in rigid bearings

Initially a rotor test set-up is constructed which simulates a simply supported LAVAL-rotor with a
non-circular shaft. The experimental set-up in figure 4.26 incorporates a relatively light foundation on
which two rigid bearing housings are mounted. Radial ball bearings are placed in each bearing
housing in which the non-circular shaft is placed. A relatively massive disk is fixed centrally on the
shaft. Besides the bearing housing, the foundation holds the driving motor that is connected to the
shaft through a flexible coupling, and a stopper which hinders excessive deflections of the shaft.
The whole assembly is relatively light (less than 20 kg) and portable. This assembly was good
enough to see the important phenomena (unbalance and gravitational resonances and instability
zones of operation) of a LAVAL-rotor with a non-circular shaft but too simple to make
quantitative analysis. Furthermore, the assembly can only simulate rigid support mounting
conditions.

In order to curb the above stated shortcomings, the assembly was extended as shown in figure 4.27.
The whole set-up was rigidly mounted on a massive foundation (weighing around 1 ton) which in
turn was placed on low stiffness springs. Such a foundation assures that the test set-up is well
isolated from surrounding disturbances. Further, there is enough space on the foundation to install
a vibration measurement equipment allowing the recording of the rotor motion relative to the
foundation.

58
 Numerical and Experimental Studies

Figure 4.27: Experimental set-up of a LAVAL-rotor with

a non-circular shaft in adjustable anisotropic
elastic bearings

The second modification is the reconstruction of the bearings to simulate both rigid and anisotropic
elastic bearings. The third inclusion to the previous design is a vibration measurement device. As shown in
figure 4.27, a rigid triangular structure is built on the foundation on which two compact optical
laser displacement measuring systems (COLDMS) are attached. One COLDMS measures the disk’s
vertical motion while the second measures the horizontal displacement relative to the foundation.
The analog data from the COLDMS are collected and digitalized by an A/D converter from
IOtech (Wavebook 512H). The digitized data are then sent by a serial data cable to a computer
where they are stored using a software called Waveview.

4.3.2 Bearing Design and Properties

The bearings of the reconstructed rotor test set-up are desired to have adjustable properties so as to
simulate rigid bearings (which may also be called isotropic bearings) and anisotropic elastic bearings. To
achieve this, the two properties are integrated on the same housing block. Ball bearings are press-
fitted in cylindrical aluminium housings. Rigidly fixing the aluminium cylinder on the foundation
gives a simple support. But fixing the aluminium cylinder to the end of a cantilever thin metallic
plate introduces anisotropy, as the metallic plate has different compliances in the vertical and
horizontal direction. Hence, introducing a mechanism that can integrate and exclude the effect of

59
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

the cantilever metallic strip will successfully give a bearing mounting that can simulate both an
isotropic and anisotropic rotor-bearing system.

Figure 4.28: The adjustable anisotropic bearing of the

experimental rotor set-up

The Institute of Mechanics at the University of Kassel has used the mechanism shown in figure
4.28 to achieve this purpose. The external frame, which is relatively rigid, is fixed to the foundation
with bolts. This rigid frame also provides the fixed end of the cantilever metallic plate. The two
threaded elements on either side of the frame can be tightened to give a rigid connection between
the frame and the aluminium cylinder resulting in a rigid bearing or loosened to give an anisotropic
bearing behaviour.

The total stiffness of the test rotor set-up is influenced by the stiffness of the cantilever metallic
strip. To determine the stiffness of the metallic strip consider figure 4.29. The material resistance to
external force can be found in many strength of material books. The approach of this work follows
that of Dubbel [21].

To determine the vertical stiffness of the metallic strip consider a vertically downward force Fz .
Sufficiently far from the boundaries this force will result in uniform distribution of the stress across
the cross-section. The total strain ε z at the end of the uniform cross-section is equal to the ratio of
the total change in length ∆L to the original length L . The stress and the strain are related to each

60
 Numerical and Experimental Studies

Figure 4.29: Sketch of the cantilever metallic strip

providing bearing stiffness anisotropy

other through the modulus of elasticity E . Using these information, the vertical spring stiffness can
be determined as21

Fz σ z A EA N
kz = = = = 9.33 ×107 . (4.21)
∆L ε z L L m

To determine the horizontal stiffness of the metallic strip consider now a concentrated horizontal
force Fy acting at the tip of the strip. From Dubbel [21] the end deflection yb can be taken.

Hence, the ratio of this force to the deflection gives the horizontal stiffness as

Fy Fy 3EI xx N
ky = = = = 1.152 ×10 4 . (4.22)
yb  Fy L
3
 L3
m
 3EI xx 
 

It can be seen that this metallic strip is much more compliance in the horizontal direction than in
the vertical direction giving the desired bearing anisotropic properties.

4.3.3 Experimental Results in Rigid Bearings

The test rotor in rigid bearings shown in figure 4.26 is operated and observed to see if the regions of
instability are indeed as found analytically. Hence, the rotor was photographed at different
rotational speeds corresponding to different regions of operation in figure 4.20. The considered
speeds were from 50 rpm up to 1350 rpm in steps of 50 rpm. It was indeed observed that the rotor
has gone through large amplitudes of deflection during the gravitational resonance frequency and
over the unstable rotational speed, including physical contact with the stopper.

21 For the data of the metallic strip see Appendix F.1.

61
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

In the two pictures in figure 4.30 it can be seen that the LAVAL-rotor is running smoothly in the
upper figure. This smooth operation leads to small deflections of the shaft and the disk giving a
clear picture. But in the lower figure, the rotor goes through large amplitude deflections during the
unbalance and gravitational resonances or in the unstable region of operation leading to the blurred
image.

Figure 4.30: The test rotor during low and high response
amplitudes

62
 Numerical and Experimental Studies

Although such qualitative results are important for basic understanding of the rotor-bearing system, it
is not assertive without quantitative measurement of the deflections. Hence, the deflection amplitudes
are measured using the experimental set-up shown in figure 4.27. The rotor was run at different
rotational speeds and the corresponding vertical and horizontal deflections of the disk were
recorded. Initially, the rotor was run from 50 rpm up to 1350 rpm in steps of 50 rpm. In regions
where there were recognizably large amplitudes, smaller rotational speed steps were taken to
generate a more accurate estimation of the resonance/unstable regions. The collected analog
measurements were digitized and saved in the binary text format (*.mat). These binary text files
were then imported in a MATLAB work environment where the software’s internal FFT-routine is
used to generate the amplitudes and the corresponding frequencies.

Figure 4.31 shows the filtered (Ω and 2Ω) amplitudes of the experimental unbalance and gravitational
responses which correspond well in the resonances and instability range to the theoretical results

Figure 4.31: Experimental unbalance and gravitational

response of the test rotor in rigid bearings

shown in the figures 4.16 to 4.19. The gravitational resonance is between 450 and 500 rpm and
agrees closely to the analytically determined value of 446 rpm. The low gravitational response in the
experimental findings is due to the fact that the results exclude the static deflection of the rotor
shaft which is about 1mm. The operation speed range of instability as determined experimentally lies
from 800 to 1040 rpm. This also agrees with the analytical values given in the figures 4.16 and 4.17.

63
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

4.3.4 Experimental Results in Anisotropic Elastic Bearings

The next experimental investigation considered is the test rotor of figure 4.27 in the anisotropic elastic
bearings. This is simply done by loosening the screws on the bearings shown in figure 4.28 in the
assembly shown in figure 4.21. Again, this assembly was run at different rotational speeds and the
corresponding vertical and horizontal deflections of the disk were recorded. Initially, the rotor was
run from 50 rpm up to 1350 rpm in steps of 25 rpm with smaller rotational speed steps in regions
of instability.

Figure 4.32 shows the filtered amplitudes of the experimental unbalance and gravitational responses which
correspond well to the theoretical results shown in the figures 4.24 and 4.25. The gravitational

Figure 4.32: Experimental unbalance and gravitational

response of the test rotor in anisotropic
elastic bearings

resonances are between 325-350 rpm and 475-505 rpm and agree closely to the analytically
determined values of 330 rpm and 450 rpm. Here again, the static deflection of the shaft of about 1
mm is not considered in the experimental results. The operation speed ranges of instability as
determined experimentally lie from 615-695 rpm, 760-820 rpm and 870-990 rpm. This also agrees
with the analytical values given in the figures 4.22 to 4.25.

64
 5 Chapter 5

FUNDAMENTALS OF BEAM THEORIES AND THE FINITE ELEMENT

METHOD

Before discussing the analysis of more complex rotor-bearing systems it is essential to discuss the
available beam theories and the mathematical background of the Finite Element Method (FEM). The
present chapter is therefore dedicated to discuss these background ideas.

5.1 Beam Theories

The beam theories that will be described here were actually all discovered before 1921. In this sub-
unit, three beam theories will be discussed with the first being the most basic and the other two
being extensions of the first to refine the solution. As will be shown at the end of this sub-unit the
TIMOSHENKO beam theory is the general form of presenting laterally vibrating beams from which
the other two theories can be derived. Hence, in the unit that follows, this theory will be exclusively
used in the derivation and a remark is included to indicate the appearance of the equations if the
other theories were used.

5.1.1 EULER-BERNOULLI Beam Theory

The EULER-BERNOULLI beam theory (which is sometimes called the classical beam theory, the
EULER theory, the BERNOULLI theory or the BERNOULLI-EULER theory) was formulated in
the 18th century. This theory is based on the knowledge that the bending effect is the most
significant property affecting the transverse vibrations of beams. This theory is discussed in the
majority of literatures dealing with dynamics of continua for its simplicity and demonstrative nature
before investigating more complex theories. The detailed derivation of this theory is found in the
books by Clough and Penzien [14], Meirovitch [63, 64], in the paper by Han et. al. [32] and many
other related monographs and publications.

To develop the theory consider the beam shown in figure 5.1 under the action of a variable
distributed load q( x, t ) and the lateral transverse bending deflection v( x, t ) in the y-direction. For
the sake of generality, the beam has a variable cross-section and general end conditions. In this
theory, the significant physical properties of the beam are the flexural stiffness EI ( x ) and the mass
per unit length ρA( x ) .
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 5.1: Vibrating EULER-BERNOULLI beam with

variable distributed load

Consider a differential element ‘ dx ’ acted upon by internal and external forces as show in figure 5.2
below.

Figure 5.2: Differential beam element including internal

and external forces

The equation of motion of the beam can be determined from NEWTON’s 2nd law of motion. The
sum of all forces in the vertical direction results in the acceleration of the beam element in the
lateral direction leading to the lateral equation as

∂Q ∂ 2v ∂Q ∂ 2v
q dx + dx = ρ A ( x ) dx 2 ⇒ q+ = ρ A( x) 2 . (5.1)
∂x ∂t ∂x ∂t

The equilibrium of moments about point ‘P’ gives

Q dx +
∂M
dx − q
( dx ) ≈ Q dx + ∂M dx = 0 ⇒ Q+
∂M
=0 (5.2)
∂x 2 ∂x ∂x

where the rotary inertia of the beam element is neglected.

66
 Fundamentals of Beam Theories and the Finite Element Method

Substituting the differential equation (5.2) into equation (5.1) and introducing the basic relation
between the bending moment and the curvature

∂ 2v
M = EI z ( x ) (5.3)
∂x 2

from the elementary beam theory which neglects the influence of the shear forces, gives the
EULER-BERNOULLI beam theory which is expressed by the partial differential equation as

∂2  ∂ 2v  ∂ 2v
q−  z( ) 2
EI x = ρ A ( ) 2.
x (5.4)
∂x 2  ∂x  ∂t

For constant cross section, equation (5.4) reduces to

∂ 4v ∂ 2v
EI z − q + ρ A =0. (5.5)
∂x 4 ∂t 2

As can be seen from the above equations, this theory expresses the beam vibrations as a function
of axial location and time. This is the most basic relation and will be present in all other beam
theories as well.

5.1.2 RAYLEIGH Beam Theory

Experience show that the natural frequencies of a free vibrating beam determined by the EULER-
BERNOULLI beam theory are higher than the values determined from experiments. The
RAYLEIGH beam theory therefore tries to reduce the discrepancy by considering the rotational
inertia besides the translational inertia. Considering the differential beam element given in figure 5.2
the dynamic equilibrium of the moments in the z-direction now becomes

Q dx +
∂M
dx − q
( dx )
≈ Q dx +
∂M ∂ 2θ
dx = ρ I z ( x ) dx 2
∂x 2 ∂x ∂t

∂M ∂ 2θ
⇒ Q+ = ρ Iz ( x) 2 . (5.6)
∂x ∂t

For the differential beam element the angular rotation can be related to the translational deflection
∂v
by θ = . Substituting this relation in equation (5.6) and equating for Q in the force balance
∂x
equation given in equation (5.1) yields

67
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

∂  ∂ 3v ∂M  ∂ 2v
q +  ρ Iz ( x) 2 −  = ρ A( x) 2 . (5.7)
∂x  ∂ t ∂x ∂x  ∂t

Substituting the moment-curvature relation from the elementary beam theory presented in equation
(5.3) and rearranging the previous equation gives the RAYLEIGH beam theory expressed by

∂2  ∂ 2v  ∂ 2v ∂  ∂ 3v 
 z
EI ( x )  − q + ρ A ( x ) − ρ
 z I ( x )  = 0. (5.8)
∂x 2  ∂x 2  ∂t 2 ∂x  ∂t 2 ∂x 

Here again, considering the particular case of a beam with constant cross section reduces equation
(5.8) to

∂ 4v ∂ 2v ∂ 4v
EI z 4 − q + ρ A 2 − ρ I z 2 2 = 0 . (5.9)
∂x  ∂t ∂t ∂x

EULER − BERNOULLI theory rotary inertia

As stated in the previous sub-unit, the RAYLEIGH beam theory is an extension of the EULER-
BERNOULLI beam theory. But considering the increased inertia in this theory does not lower the
over-estimated natural frequencies of a vibrating beam enough, calling for a more refined beam
theory including also the shear deflection as discussed in the next sub-unit.

5.1.3 TIMOSHENKO Beam Theory

It was determined by a number of authors that the inclusion of the rotary inertia only does not
improve the accuracy of the natural frequencies of a free vibrating beam sufficiently. Hence,
considering shear deformation (i.e. increasing the flexibility of the beam element) was the further
refinement taken in the TIMOSHENKO beam theory to bridge this result discrepancy. This reduces
the natural frequencies significantly to approach the experimentally determined results as
determined by Traill-Nash and Collar [99].

To determine the equation of motion of this beam theory consider the differential element in figure
5.3 with a magnified deformation22. For this differential TIMOSHENKO beam element, the total
∂v
angular rotation θ is equal to the sum of the slope v′ = of the elastic axis and the shear
∂x
deformation β . Applying again NEWTON’s 2nd law of motion for the translational and angular
dynamic equilibrium gives

22 The external and internal forces acting on this differential element are similar to those shown in figure 5.2.

68
 Fundamentals of Beam Theories and the Finite Element Method

Figure 5.3: Differential TIMOSHENKO beam element

showing the resulting deformations

∂Q ∂ 2v ∂Q ∂ 2v
qdx + dx = ρ A ( x ) dx 2 ⇒ q+ = ρ A( x) 2 , (5.10)
∂x ∂t ∂x ∂t

2
∂M ( dx ) ≈ Q dx + ∂M dx = ρ I x dx ∂ 2θ
Q dx + dx − q z ( )
∂x 2 ∂x ∂t 2

∂M ∂ 2θ
⇒ Q+ = ρ Iz ( x) 2 (5.11)
∂x ∂t

which are identical to the equations (5.1) and (5.6) in the previous sub-unit.

The TIMOSHENKO beam theory relates the deflection due to the shear force over a linearly
approached cross-section to be directly proportional to the shear force for materials with linear
property as

Q = κ GA ( x ) β . (5.12)

The proportionality constant κ depends on the material and the cross-sectional shape. κ A ( x )

represents the effective shear area of the cross-section. Literatures [19, 37, 47] give κ values for a
number of different cross-sections. A value of κ suggested by a number of books to give a good
result for rectangular cross-sections is 5/6. The other common value of κ found in the literature is
11/13 for circular cross-sections.

69
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Substituting equation (5.12) into equation (5.10) yields

∂ ∂ 2v
q+
∂x
( κ GA ( x ) β ) = ρ A ( x ) 2 .
∂t
(5.13)

The bending moment is related to the total angular rotation of the cross-section of the beam element
as

∂θ
M = EI z ( x ) . (5.14)
∂x

Substituting equation (5.12) and (5.14) into equation (5.11) gives

∂  ∂θ  ∂ 2θ
κ GA ( x ) β +  EI z ( x )  = ρ I z ( x ) 2 . (5.15)
∂x  ∂x  ∂t

From figure 5.3 it can be seen that the angular rotation due to the shear force and the total angular
rotation of the cross-section are related through

∂v
β= −θ (5.16)
∂x

and, therefore, substituting equation (5.16) into (5.13) and (5.15) gives

∂   ∂v  ∂ 2v
q +  κ GA ( x )  − θ   = ρ A ( x ) 2
∂x   ∂x  ∂t
(5.17)
 ∂v  ∂  
2
∂θ ∂θ
κ GA ( x )  − θ  +  EI z ( x )  = ρ Iz ( x) 2 .
 ∂x  ∂x  ∂x  ∂t

Assuming a constant cross section of the beam will reduce the first equation of (5.17) to

∂  ∂v  ∂ 2v ∂θ ∂ 2v 1  ∂ 2v 
q + κ GA  − θ  = ρ A ⇒ = 2−  ρ A − q. (5.18)
∂x  ∂x  ∂t 2 ∂x ∂x κ GA  ∂t 2


Because of the constant cross-section, differentiating the second equation of (5.17) with respect to
x will give

70
 Fundamentals of Beam Theories and the Finite Element Method

∂  ∂v  ∂ 3θ ∂ 3θ
κ GA − θ + EI = ρ I . (5.19)
∂x  ∂x  z z
 ∂x 3 ∂t 2 ∂x

Finally, substituting equation (5.18) into equation (5.19) and rearranging the different terms will give
the TIMOSHENKO beam theory described by

∂ 4v ∂ 2v ∂ 4v EI z ∂ 2  ∂ 2v  ρ I z ∂ 2  ∂ 2v 
EI z 4 − q + ρ A 2 − ρ I z 2 2 + 2 
q − ρ A 2 
− 2 
q − ρ A 2 
=0
∂x
 ∂ t ∂t ∂x κ GA ∂ x  ∂ t  κ GA ∂ t  ∂ t 
EULER − BERNOULLItheory
     
   shear deformation combined shear deformaion
and rotary inertia
RAYLEIGHtheory
 
TIMOSHENKOtheory

(5.20)

As can be seen from equation (5.20), the previous two beam theories are special cases of the
TIMOSHENKO beam theory. Assuming the beam to resist shear deformation, which implies
κ → ∞ , sets the last two terms in equation (5.20) to zero giving the RAYLEIGH beam theory. If
the area moment of the cross section is relatively small in comparison to the length of the beam,
then the third term in the equation can be neglected giving the EULER-BERNOULLI beam
theory.

If there is no external force acting on the beam, then equation (5.20) will reduce to

∂ 4v ∂ 2v ∂ 4v ρ EI z ∂ 4 v ρ 2 I z ∂ 4v
EI z + ρ A − ρ I z − + = 0. (5.21)
∂x 4 ∂t 2 ∂t 2 ∂x 2 κ G ∂t 2 ∂x 2 κ G ∂t 4

In chapter 6.1.1 it is discussed how the preceding continuous TIMOSHENKO beam whose
equation of motion is given by equation (5.21) can be discretized so as to develop a finite element
model.

5.2 Fundamentals of the Finite Element Method

The Finite Element Method (FEM) has been described in a number of books and manuscripts, some
of which have discussed the fundamentals of the method clearly [24, 61, 86, 104]. The process of
subdividing systems into finite elements, whose behaviour is readily understood, and rebuilding the
original system from the elements is the intuitive step taken because of our limitation to understand
and analyse complex systems in one step.

The fundamental idea of the FEM is the division of a given domain into a set of simple sub-
domains called finite elements, which have a proper geometric shape that allows the approximation of

71
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

the solution of the domain from the knowledge of the solution at selected points called nodes on the
sub-domain. The sub-domains can be either similar or different from each other.

Other features of the method include seeking a continuous function approximating the solution
over the sub-domain in terms of the nodal displacements and forces and assembling the elements by
imposing the inter-elemental continuity of the deformations and forces. In general, the assemblage
of the elements is subject to boundary conditions. The solutions for the sub-domain and the
complete domain can be determined only after these conditions have been imposed.

Being an approximate method, there are different errors associated with this technique. The sources of
errors are basically due to three reasons. The first source of error comes from the approximation of
the domain during the construction of the sub-domains. The second source of error is due to the
approximate solutions used in each sub-domain and the third one comes from the numerical
computation.

The next sub-unit discusses the application of the FEM as applied to mechanical structures. Two
principles used in the determination of the equation of motion are described. The discretization and
the finite element model are also discussed.

5.2.1 Principle of Virtual Work

The principle of virtual work states that “if a system of forces is in equilibrium, the work done by the
externally applied forces through virtual displacements compatible with the system constraints of
the system is zero” [63]. To illustrate the above statement an isolated particle from a system of
particles in equilibrium is considered as in figure 5.4 [8]

Figure 5.4: Forces F ji acting on the jth particle of a system

Being in equilibrium, the sum of all externally applied forces is equal to zero. The particle is then


subjected to a virtual displacement δ q j (small variation from the true position) compatible with the

72
 Fundamentals of Beam Theories and the Finite Element Method

constraints of the system. This is a virtual displacement and hence there is no change in time
associated with this displacement. The virtual work done by all forces over the virtual displacements
i.e. the dot product of the forces with the virtual displacements is zero. Hence, it holds

K




δ W j = ∑ Fji ⋅ δ q j = R j ⋅ δ q j = 0 (5.22)
i =1

where R j is the resultant force and K is the number of forces acting on the j th element.

The above relation is quite trivial as the sum of forces on a particle in equilibrium is zero. Consider
now a particle which is subjected to constraints. The total force is now the sum of external and
constraint forces as

R j = R ej + R cj (5.23)

where the superscripts ‘e’ and ‘c’ stand for ‘external’ and ‘constraints’, respectively.

Substituting equation (5.23) into (5.22) gives

δ W j = R ej ⋅ δ q j + R cj ⋅ δ q j = 0 . (5.24)

Hence, for a system of N particles the total virtual work will still be zero as

N N

N


δ W = ∑ δ W j = ∑ R ej ⋅ δ q j + ∑ R cj ⋅ δ q j = 0 . (5.25)
j =1 j =1 j =1

Since the induced virtual displacements are compatible with the constraints of the system, there is
no virtual displacement of the constraint points in the direction of the constraint forces. Hence, the
second term of equation (5.25) is automatically zero, which results in a basically different relation
from that one given in equation (5.22). In the present equation, the sum of the external forces is
not equal to zero and the virtual displacements may not be independent from each other. Therefore
the total virtual work becomes

N


δ W = ∑ R ej ⋅ δ q j = 0 . (5.26)
j =1

73
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Equation (5.26) is the principle of virtual work expressed mathematically. This equation can be used to
systems in both static and dynamic equilibrium. From D’ALEMBERT’s principle it is known that
any dynamic system with non-zero resultant forces can be considered as being in equilibrium with
its inertia forces. Hence, for a system with non-zero resultant forces equation (5.26) may be
rewritten as

N



δ W = ∑ ( R ej + I j ) ⋅ δ q j = 0 . (5.27)
j =1

In the above formulation the forces and the inertia are generalized and include both translational
and rotational motions. In general, the principle of virtual work can therefore be written as

δW = δW e + δW i = 0 (5.28)

where the superscript ‘i’ stands for ‘inertial’.

The principle of virtual work for a system of particles discussed above can be extended to a
continuum. Considering a finite element cut out from an accelerating (vibrating) continuum, there
are internal, external and inertial forces acting on it. From D’ALEMBERT’s principle, the element
is under dynamic equilibrium. Hence, the principle of virtual work including these forces is given by

δ W = δ W e + δ W i − δ W in = 0 (5.29)

where the superscripts ‘in’ stands for ‘internal’.

Before the discussion of the procedure of determining the equations of motion of a general rotor-
bearing system and the corresponding element matrices using equation (5.29) in the next chapter, the
next sub-unit will describe the LAGRANGE’s equations, which is another method to generate the
equations of motion and the element matrices of a finite element model.

5.2.2 LAGRANGE’s Equations

For the explanation of the basics of the LAGRANGE’s equations, the approach described by
Meirovitch [64] is followed in this work. Consider a system of N particles each having n degrees


of freedom. Hence, the position vector q j of the j th particle can be written in terms of n independent

coordinates ek as

q j = q j (e1 , e2 ,⋯, en ) ; j = 1, 2,⋯, N . (5.30)

74
 Fundamentals of Beam Theories and the Finite Element Method

The velocity of the j th particle will therefore become

ɺ ∂q j ∂q j ∂q j n ∂q
eɺn = ∑
j
qj = eɺ1 + eɺ2 + ⋯ + eɺk ; j = 1, 2,⋯ , N . (5.31)
∂e1 ∂e2 ∂en k =1 ∂e k

∂q j
Since the quantities do not depend on the generalized velocities eɺk explicitly, deriving
∂ek

equation (5.31) with respect to eɺk gives

∂qɺ j ∂q j
= ; j = 1, 2,⋯ , N ; k = 1, 2,⋯ , n . (5.32)
∂eɺk ∂ek

From equation (5.30) the virtual displacement can be determined as

∂q j ∂q j ∂q j n ∂q
δ qj = δ e1 + δ e2 + ⋯ + δ en = ∑ j δ ek ; j = 1, 2,⋯ , N . (5.33)
∂e1 ∂e2 ∂en k =1 ∂ek

The virtual work of the inertia forces of the system of particles can be written as

n  N


N

N


N

n ∂q j
∂q j 
δ W = ∑ I j ⋅ δ q j = −∑ m j q j ⋅ δ q j = −∑ m j q j ⋅ ∑
i ɺɺ ɺɺ δ ek = −∑  ∑ m j q j ⋅
ɺɺ  δ ek . (5.34)
j =1 j =1 j =1 k =1 ∂ek k =1  j =1 ∂ek 

For one particular value of j and k it can be seen that

∂q j d 
ɺ ∂q j 
ɺɺ
ɺ d  ∂q j  d 
ɺ ∂q j 
∂  dq j 
mjqj ⋅ =  mjq j ⋅  − mjqj ⋅   =  mjq j ⋅  − m j qɺ j ⋅   . (5.35)
∂ek dt  ∂ek  dt  ∂ek  dt  ∂ek  ∂ek  dt 

Substituting equation (5.32) into the first term on the right hand side of equation (5.35) gives

∂q j d 
ɺ ∂qɺ j
ɺɺ

ɺ ∂  dq j 
mjqj ⋅ =  mjq j ⋅  − m j q j ⋅  
∂ek dt  ∂eɺk  ∂ek  dt 

d  ∂  1
ɺ
ɺ   ∂  1
ɺ
ɺ   d  ∂  ∂   1
ɺ
ɺ 
=   mjqj ⋅ q j   −  mjqj ⋅ qj  =   ɺ −   m j q j ⋅ q j  . (5.36)
dt  ∂eɺk 2   ∂ek 2   dt  ∂ek  ∂ek   2 

75
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

From kinetics of particles and rigid bodies, the second term on the right hand side of equation
(5.36) is known to be the kinetic energy T j of the j th particle. Hence, replacing this term in equation
(5.36) and substituting the resulting relation into equation (5.34) gives

N


n  d  ∂  ∂  N  1
ɺ
ɺ   n 
∂T d  ∂T 
δ W i = −∑ m j qɺɺj ⋅ δ q j = −∑   −  ∑ ⋅  δ = ∑  −    δ ek .
 dt  ∂eɺk  m q q j  e
j j
 
k  dt  ∂eɺk
j =1 k =1  
  ∂ek  j =1  2 k =1  ∂ek 
(5.37)

Consider now the virtual work of the external forces on the above investigated system of particles due to
a virtual displacement. This external virtual work becomes

n  N




e

N N
n ∂q ∂q j  n
δW = ∑ Rj ⋅δ q j = ∑ Rj ⋅ ∑
e e j
δ ek = ∑  ∑ R j ⋅ e
 δ ek ⇒ δ W e = ∑ Qk δ ek . (5.38)
j =1 j =1 k =1 ∂ek k =1  j =1 ∂ek  k =1
  
Qk

This virtual work can be regarded as a product of generalized forces Qk acting over a virtual unit

displacement δ ek . These generalized forces are not necessarily forces but may also be moments.

The generalized forces acting on the system may be divided into conservative and non-conservative forces,
which may be derivable and non-derivable from a potential energy V , respectively. Writing the virtual
works due to these two types of forces separately and noting that the work of conservative forces
gives a negative potential energy yields

n
δ W e = δ W con + δ W n −con = −δ V + ∑ Qkn −conδ ek
k =1

 ∂V ∂V ∂V  n n
 ∂V 
= − δ e1 + δ e2 + ⋯ + δ en  + ∑ Qkn −conδ ek = −∑  − Qkn −con  δ ek (5.39)
 ∂e1 ∂e2 ∂en  k =1 k =1  ∂ek 

where the superscripts ‘con’ and ‘n-con’ stand for ‘conservative’ and ‘non-conservative’,
respectively.

Substituting the equations (5.37) and (5.39) into the principle of virtual work given in equation (5.28)
will give the relation

n  ∂T d  ∂T  ∂V 
∑  −   − + Qk
n − con
 δ ek = 0 . (5.40)
 ∂ek dt  ∂eɺk  ∂ek
k =1  

76
 Fundamentals of Beam Theories and the Finite Element Method

By definition, the virtual displacements δ ek are arbitrary and independent allowing any selection of

their values. Hence, n relations can be formulated by setting all but one of the values of δ ek to
zero. Hence, one gets the LAGRANGE’s equations expressed by

d  ∂T  ∂T ∂V
 − + = Qkn −con ; k = 1, 2,⋯ , n (5.41)
dt  ∂eɺk  ∂ek ∂ek

which may also be written in terms of the Lagrangian L = T − V (considering that the potential
energy does not depend on velocity) as

d  ∂L  ∂L
 − = Qkn− con ; k = 1, 2,⋯ , n . (5.42)
dt  ∂eɺk  ∂ek

Here again, the above relations may be extended for the analysis of a continuum. An element
section of a continuum under the influence of external and internal forces possesses potential
(strain) and kinetic energy as well as the effects of the non-conservative forces. Hence, the strain
and kinetic energy of the element is substituted in equation (5.41) or (5.42) to generate the equations
of motion and the element matrices.

As stated at the end of the previous sub-unit, the determination of the equations of motion of a
general rotor-bearing system and the element matrices using the LAGRANGE’s equations is also
discussed in the next chapter.

5.2.3 Discretization and Finite Element Model of Rotors

In the analysis of rotors, it is common to use hybrid models i.e. a combination of elements with
continuously distributed properties and lumped properties. The rotor shafts are usually treated as a
continuum with elasticity and inertia properties while the disks are treated as rigid masses and the
bearings as massless elastic bodies which may also possess dissipation properties. This sub-unit is
concerned with the discretization of the rotor to develop its finite element model.

As discussed in the introductory words for the Finite Element Method (FEM), the first step is the
subdivision of the given domain (the rotor shaft in this case) into a set of simple sub-domains.
Thus, the rotor shaft is divided into a number of one-dimensional sections that are called beam
elements (extended by the disk and bearing elements). The two ends of each section are the nodes of the
element. The knowledge of the deflection and rotation at the nodes gives a good picture of the
deflection of the complete rotor depending on the number of nodes considered. The deflection at

77
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

any point within the beam element is approximated by the linear superposition of the deflections at
the nodes using simple polynomials.

A typical hybrid rotor model with a single disk and two bearings is given in figure 5.5. As shown in the
figure, the number of the total possible deflections at a given node are its degrees of freedom
(generalized displacements). This implies that the sum of the degrees of freedom of all the nodes of
one element define the element’s degrees of freedom. The degrees of freedom of the complete
system is therefore the sum of the degrees of freedom of all nodes.

Figure 5.5: A hybrid rotor model with shaft, disk and

bearing elements, nodes and degrees of
freedom

For the bending vibrations of a rotor, the degrees of freedom of each node i are the displacements vi
and wi and the rotations ψi and θi in the y- and z-directions, respectively. They are related to the
corresponding generalized forces by element matrices which are derived using one of the mechanical
principles derived in the previous sub-units.

The element matrices are then assembled to form the global matrices of the rotor model in such a way
that the terms of different elements associated with the same node are added up. In the presence of
elements with lumped properties like the disks and the bearings, their properties are added to those
nodes where the elements are present23 (figure 5.6). It is therefore necessary to discretize the rotor

23 Note here that the shaft element matrix is twice as big as that of the disk and bearing since each shaft element has two
nodes.

78
 Fundamentals of Beam Theories and the Finite Element Method

shaft in such a way that these elements lie on a node. Finally, the global matrices are used to
determine the dynamic properties of the hybrid rotor.

Figure 5.6: Structure of the assembled global matrix of

the rotor model shown in figure 5.5

The next chapter now implements the background theories given in this chapter to develop the
element matrices of the shaft as well as the disks and the bearings in the fixed and rotating coordinate system of
a general hybrid rotor model.

79
 6 Chapter 6

GENERAL ANISOTROPIC ROTOR-BEARING SYSTEMS

As stated in the previous chapter, it is common to use a hybrid model in the numerical analysis of
rotor-bearing systems. In the present chapter, the different elements of the hybrid rotor will be
modelled using the foundations of the Finite Element Method (FEM) discussed in the previous
chapter.

The hybrid rotor will be modelled systematically starting from a circular shaft which will then be
extended to a non-circular shaft. Then the disk and the bearings will be modelled as inertia and
elastic/dissipative elements. The element matrices of the above elements will be developed in both
the fixed and rotating coordinate system.

6.1 Element Matrices

6.1.1 Shaft Elements

The equation of motion of the TIMOSHENKO beam element may be determined by using either the
principle of virtual work or the LAGRANGE’s equations. In this sub-unit, both methods are used to
determine the shaft element matrices and the equations of motion of a rotating TIMOSHENKO
beam element. The equations of the other two beam models can be determined from the
TIMOSHENKO beam by considering only the required terms of equation (5.20).

6.1.1.1 Circular Shaft Element in the Fixed Coordinate System

Consider figure 6.1 which shows a circular TIMOSHENKO beam element in the fixed x,y,z- and
rotating (body-fixed) ξ,η,ζ-coordinate system. Figure 6.2 shows the internal forces acting on the
beam element in the x,y- and x,z-plane and the deformation of the element due to bending and
shear.

The deflection of any point within the element can be given by the displacements v = v( x, t ) and
w = w( x, t ) and the rotations ψ = ψ ( x, t ) and θ = θ ( x, t ) in the y- and z-directions, respectively.



In the fixed coordinate system, the generalized deflections qxf = qxf ( x, t ) at any point in the element

will therefore be given by24

24 Since both fixed and rotating coordinate systems are used the generalized displacements in the two systems will be
distinguished by the superscript ‘f ’ and ‘r’, respectively.
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 6.1: Rotating circular TIMOSHENKO beam

element25

Figure 6.2: Forces and deformations of a

TIMOSHENKO beam element

qxf = [v w ψ θ ]T . (6.1)

As discussed in Appendix G, the deflection at any point in the element can be approximated by the



linear combination of the nodal displacement qnf = qnf (t ) using HERMITIAN polynomials included

in the shape function matrix H = H ( x ) . Hence, one gets

q xf ≈ H qnf (6.2)

T
where qnf = v1 w1 ψ 1 θ 1 v 2 w2 ψ 2 θ 2  is the generalized nodal displacement vector with the

superscripts 1 and 2 indicating the left and right node at the end of the finite element, respectively.

25 Note here that the shaft rotation ϕɺ = Ω is positive in the positive x-direction unlike the definition used in chapter 3.

82
 General Anisotropic Rotor-Bearing Systems

Next, the equation of motion for the element in terms of the nodal displacements is determined
using the two methods stated in the previous chapter. The shaft element matrices will also be given
during the derivation of the equation of motion.

• Principle of Virtual Work

The procedure to derive the shaft element matrices by the principle of virtual work is based on the work
of Belz [4]. In the absence of external loads the principle of virtual work relation (5.29) for the finite
element given in figure 6.2 is reduced to the virtual work of the internal and inertia forces.

The virtual work of the internal forces of the element is given by

L L
δ WSin = ∫
x =0
κ GA ( v′ − θ ) δ ( v′ − θ ) dx + ∫ κ GA ( w′ +ψ ) δ ( w′ +ψ ) dx
x =0
L L
+ ∫
x =0
EI yψ ′δψ ′dx + ∫ EI θ ′δθ ′dx.
x =0
z (6.3)

Equation (6.3) can be written in terms of the generalized displacements given in equation (6.1) as

L

fT



δ WSin = ∫ δq
x =0
∂ ∂x K int q∂f ∂x dx ; q∂f ∂x = T∂ ∂x q xf (6.4)

κ GA 0 0 0  ∂ ∂x 0 0 −1 
 0 κ GA 0 0    0 ∂ ∂x 1 0 
where K int = ; T∂ ∂x =  .
 0 0 EI y 0   0 0 ∂ ∂x 0 
   
 0 0 0 EI z   0 0 0 ∂ ∂x 

Writing the generalized displacements in terms of the nodal displacements given in equation (6.2)
will therefore give

L

T



δ WSin = δ qnf T ∫ (T ) ( )
H K int T∂ ∂x H dx qnf = δ qnf T K S qnf
∂ ∂x (6.5)

x =0
 
KS

where K S is the stiffness matrix of the shaft element.

Once the virtual work due to the internal forces is determined, the next step is the determination of
the virtual work due to the inertia forces of the finite beam element which moves generally in

83
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

space. Consider therefore the differential element shown in figure 6.1 which is assumed to move
laterally only in the y- and z-directions but rotates in all three axes x, y and z described by the
generalized displacements introduced with figure 6.1.

From D’ALEMBERT’s principle, the inertia force is the negative of the sum of the external forces.
Hence, considering all inertia forces and moments acting on the finite element shown in figure 6.1
and using the generalized deflections defined in equation (6.1) gives the virtual work of the inertia forces
as

L
δ WSi = − ∫ ( ρ Avɺɺδ v + ρ Aw ɺɺ − ρ ΩI ψδθ
ɺɺδ w + ρ I zθδθ x
ɺ + ρ I yψδψ ɺ ) dx
ɺɺ + ρ ΩI xθδψ (6.6)
x =0

which can be written as

L L



fT
ɺ f
δ WSi = − ∫ δ qxf T M int qɺɺxf dx − ∫ δq x G int
pvw q x dx (6.7)
x =0 x =0

ρ A 0 0 0  0 0 0 0 
 0 ρA 0 0  0 0 0 0 
where M int = ; G pvw = 
int
.
 0 0 ρIy 0  0 0 0 ρ ΩI x 
   
 0 0 0 ρIz  0 0 − ρ ΩI x 0 

Again, writing the generalized displacements in terms of the nodal displacements will give

L L


f
fT
f
fT
f
fT
f
∫x=0 H M Hdx qɺɺn − δ qn x∫=0 H G pvw Hdx qɺn = −δ qn M S qɺɺn − δ qn G S qɺn (6.8)
int int
δ WSi = −δ qnf T T T

  
MS GS

where M S and G S are the mass and gyroscopic matrices of the shaft element, respectively.

Substituting the equations (6.5) and (6.8) into the principle of virtual work in equation (5.29) for
δ W e = 0 gives

δ qnf T ( M S qɺɺnf + G S qɺnf + K S qnf ) = 0 . (6.9)

Since the virtual displacements are arbitrarily chosen displacements compatible with the system
constraints, they can be selected to be non-zero. This implies that in equation (6.9) the term in the

84
 General Anisotropic Rotor-Bearing Systems

brackets has to be equal to zero. Hence, the equation of motion of the TIMOSHENKO beam circular
shaft element in the fixed coordinate system becomes

M S qɺɺnf + G S qɺnf + K S qnf = 0 . (6.10)

The circular shaft element mass, gyroscopic and stiffness matrices in this equation are given in detail in
Appendix H.1.1. For a RAYLEIGH and EULER-BERNOULLI beam, the equation of motion is
similar to (6.10) with the only difference being in the terms of the element matrices as also shown
in Appendix H.1.1.

• LAGRANGE’s Equations

The equation of motion as well as the element matrices of the finite element shown in figure 6.1
can also be determined using the LAGRANGE’s equations. This sub-unit discusses the use of this
method to develop the element matrices per procedure discussed by Boru and Irretier [7]. To use
this method, one needs to determine the strain and kinetic energy of the deformed and vibrating
finite beam element.

Generally, the strain energy of the TIMOSHENKO beam element comes from the deformation
due to the shear forces in the y- and z-directions and from the moments in the x,y- and x,z-planes
as shown in figure 6.2. These internal loads are related to the material properties, the geometric
(cross-sectional) properties and the deformations by the equations (5.12) and (5.14) between the
shear forces and bending moments and the corresponding rotations of the cross-section.

The strain energy of the beam element becomes

L L L L
1 1 1 1
∫ κ GA ( v′ − θ ) dx + ∫ κ GA ( w′ + ψ ) dx + ∫ EI yψ ′2 dx + ∫ EI zθ ′2 dx . (6.11)
2 2
VS =
2 x =0 2 x =0 2 x =0 2 x =0

Using the vector and matrix definitions given in the equations (6.1), (6.2) and (6.4) reduces equation
(6.11) to

L
1
T
1


VS = qnf T
2 ∫ (T ∂ ∂x ) ( 2
)
H K int T∂ ∂x H dx qnf = qnf T K S qnf (6.12)

x=0
 
KS

where the matrices K int , T∂ ∂x and H are defined in equation (6.4) and Appendix G, respectively.

85
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

The kinetic energy of the rotating beam element is the sum of the translational and rotational kinetic
energy. Considering the beam to rotate with a constant angular velocity Ω about the x-axis and
neglecting small terms of 2nd order gives the kinetic energy to be

L L L
1 1
TS =
20∫ ( ) 20
( )
ρ A vɺ 2 + wɺ 2 dx + ∫ ρ I yψɺ 2 + I zθɺ 2 dx + Ω ∫ ρ I xψθ
ɺ dx (6.13)
0

which can be written as

L L
1



f
TS = ∫ qɺ xf T M int qɺ xf dx + ∫ qɺ xf T G int
LAG q x dx (6.14)
20 0

ρ A 0 0 0  0 0 0 0 
 0 ρA 0 0  0 0 0 0 
where M int = ; G LAG = 
int
.
 0 0 ρIy 0  0 0 0 ρ ΩI x 
   
 0 0 0 ρIz  0 0 0 0 

Again, the generalized displacements written in terms of the nodal displacements using equation
(6.2) will give

L L
1



f 1
ɺ f T



TS = qɺnf T ∫ HT M int Hdx qɺnf + qɺnf T ∫ HT G int
LAG Hdx qn = qn M S qɺnf + qɺnf T G ′S qnf . (6.15)
2 2

0
  
0
 
MS G ′S

Using the LAGRANGE’s equations (5.42) for qnf , substituting the equations (6.12) and (6.15) and
setting the non-conservative forces to zero gives

M S qɺɺnf + ( G ′S − G ′ST ) qɺnf + K S qnf = 0 (6.16)

from which again follows the equation of motion of the circular shaft element in the fixed coordinate system
as26

M S qɺɺnf + G S qɺnf + K S qnf = 0 . (6.17)

L L
26
(
Note that G ′S − G ′ST = ∫ H T G LAG
int int T
− G LAG )
Hdx = ∫ H T G pvw
int
Hdx = G S .
0 0

86
 General Anisotropic Rotor-Bearing Systems

It can be seen that the equations of motion (6.10) and (6.17) and the element matrices determined
by the two methods are identical as given in Appendix H.1.1.

6.1.1.2 Circular Shaft Element in the Rotating Coordinate System

The circular TIMOSHENKO beam element discussed in the previous section is readdressed using the
rotating coordinate system. Hence, the nodal displacement vector in equation (6.2) is reformulated using the
transformation matrix relating the two coordinate systems as27

qnf = Tn qnr (6.18)

T 0 0 0 
0 T 0 0
where Tn =   ; T = cos α − sin α 
; α = Ωt − β .
0 0 T 0  sin α cos α 

 
 0 0 0 T 

The first and second time derivative of the displacement vector in the two coordinate systems are
related through

ɺ q
r + T q
ɺ r
ɺɺ q
r + 2Tɺ q
ɺ r + T qɺ
ɺr .

qɺ nf = Tn n n n ; qɺɺnf = T n n n n n n (6.19)

Substituting the equations (6.18) and (6.19) into equation (6.10) or (6.17) gives

ɺɺ q
r + 2Tɺ q
ɺ r + T qɺ
ɺr + G T

MS T ( ) ɺ q
r + T q
ɺ r + K T q
r = 0 .
( ) (6.20)
n n n n n n S n n n n S n n

Pre-multiplying equation (6.20) by the transpose of the transformation matrix will give the equation
of motion of the circular shaft element in the rotating coordinate system as

ɺɺ + G Tɺ + K T ) q
r = 0
TnT M S Tn qɺɺnr + TnT ( G S Tn + 2M S Tɺ n ) qɺnr + TnT ( M S T
   
n S n S n n
mS cS kS

⇒ mS qɺɺnr + cS qɺnr + k S qnr = 0 . (6.21)

The circular shaft element mass, velocity-proportional and displacement-proportional matrices are computed using
a MATLAB program (symbolic math toolbox) and selectively checked with hand calculation. The
detailed matrices in equation (6.21) are given in Appendix H.1.2.

27 The different signs of the off-diagonal terms in the transformation matrix in comparison to equation (3.8) are due to
the opposite positive direction of Ω used in this chapter (see footnote 25 on page 82).

87
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

6.1.1.3 Non-Circular Shaft Element in the Rotating Coordinate System

If a non-circular TIMOSHENKO beam element is considered as shown in figure 6.3 then the element
matrices will have varying properties in the vertical and horizontal directions depending on the
orientation of the element. But using a coordinate system, which rotates with the same angular
velocity as the shaft element, will ensure constant properties in any direction in this rotating coordinate
system even though the values may be different in the two axes. In the analysis of such non-circular
systems it is common to set the rotating axis parallel to the principal axis of the cross-section so as
to simplify the output forms (figure 6.3).

Figure 6.3: Rotating non-circular TIMOSHENKO beam

element

Since using the principal axis sets the product moment of area to zero, the element matrices will be
similar to the ones developed in the previous section but with different values in the different axial
directions. Hence, the equation of motion of the non-circular shaft element in the rotating coordinate system can
be written as

mS qɺɺnr + cS qɺnr + k S qnr = 0 . (6.22)

As can be seen from equation (6.22), the matrices are constant in the rotating coordinate system. It
can further be seen that the equation of motion given in the equations (6.21) and (6.22) are similar
with the only difference coming from the elements in the matrices. The detailed values of the
elements in the matrices for this case are given in Appendix H.1.3.

88
 General Anisotropic Rotor-Bearing Systems

6.1.1.4 Non-Circular Shaft Element in the Fixed Coordinate System

Once the element matrices of the non-circular TIMOSHENKO beam element are determined in the
rotating coordinate system they can be transformed back into the fixed coordinate system similar as
discussed in the previous sections and Appendix B. But now, since the cross-sectional properties
vary in the fixed coordinate system as the shaft rotates the matrices are expected to become time
dependent.

The transformation matrix required is the inverse of the transformation matrix defined in equation
(6.18) which is equal to its conjugate. Hence, one gets

qnr = Tn−1qnf = TnT qnf (6.23)

Similar to equation (6.19), the first and second time derivative of the displacement vector in the two
coordinate systems are related by

ɺɺ T
f ɺ T
ɺ f
f
qɺnr = Tɺ nT qnf + TnT qɺnf ; qɺɺnr = T T ɺɺ
n qn + 2Tn qn + Tn qn . (6.24)

Substituting the equations (6.23) and (6.24) into the equation of motion (6.22) in the rotating
coordinate system gives

ɺɺ T q
f + 2T
mS T ( ɺ T q
ɺ f + TT qɺɺ
f + c Tɺ T q
f + TT q
ɺ f + k TT q
f = 0 .
) ( ) (6.25)
n n n n n n S n n n n S n n

Pre-multiplying by the transformation matrix will result in the equation of motion of the non-circular
shaft element in the fixed coordinate system as

ɺ T q
ɺ f + T m T T
f
  (
TnmS TnT qɺɺnf + Tn cS TnT + 2m S T

n

n )
n
ɺɺ T

(ɺT
S n + cS Tn + k S Tn qn = 0 )
ɶMS (t ) GS ɶ (t )
K S

⇒ ɶ qɺɺ
f + G q
ɺ f + K
M ɶ q
f = 0 (6.26)
S n S n S n

where now the non-circular shaft element mass and stiffness matrices are time dependent 28 while the
gyroscopic matrix remains time-independent.

28 These matrices are discussed further in chapter 6.2.

89
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

6.1.2 Disk Elements

In this sub-unit, the matrices of a disk element will be determined in the fixed and rotating coordinate
system. As stated in the previous chapter, the disk is considered as a lumped element with inertia
properties only. Hence, mass and gyroscopic matrices are the only element matrices expected. Here
again, the principle of virtual work and the LAGRANGE’s equations are used to determine the
equations of motion of the disk and the corresponding element matrices.

6.1.2.1 Circular Disk Element in the Fixed Coordinate System

• Principle of Virtual Work

In applying the principle of virtual work the procedure as for the shaft element is also followed in this
sub-unit. Unlike the shaft element, the disk element is assumed to be rigid and hence the virtual work
of the internal forces is zero. But since the rigid disk element has inertia forces, the virtual work due
to these forces has to be considered.

The virtual work of the inertia forces can be derived according to equation (6.6) as

δ WDi = − ( mvɺɺDδ vD + mw
ɺɺDδ wD + J zθɺɺDδθ D − Ω J xψɺ Dδθ D + J yψɺɺDδψ D + Ω J xθɺDδψ D ) (6.27)

which can be written as

δ WDi = −δ qDf T M D qɺɺDf − δ qDf T G D qɺ Df (6.28)

m 0 0 0 0 0 0 0 
0 m 0 0  0 0 0 0 
where M D =  ; GD = 
0 0 Jy 0 0 0 0 Ω Jx 
   
0 0 0 Jz  0 0 −Ω J x 0 

are the mass and gyroscopic matrices of the disk element, respectively, and q Df = [vD wD ψ D θ D ] is
T

the generalized nodal displacement vector of the disk element.

Substituting equation (6.28) into the principle of virtual work in equation (5.29) for δ W e = 0 and
δ W in = 0 gives

δ qDf T ( M D qɺɺDf + G D qɺ Df ) = 0 . (6.29)

90
 General Anisotropic Rotor-Bearing Systems

Here again, assuming non-zero virtual displacements gives the equation of motion of the circular disk
element in the fixed coordinate system as

M D qɺɺDf + G D qɺ Df = 0 . (6.30)

As can be seen from equation (6.30), the matrices defining the dynamic properties of the circular
disk are the circular disk element mass and gyroscopic matrices.

• LAGRANGE’s Equations

This sub-unit uses the LAGRANGE’s equations to determine the element matrices and the equation
of motion of the disk element as an alternative procedure to the use of the principle of virtual work
applied in the previous sub-unit.

The kinetic energy of the rotating disk is the sum of the translational and rotational kinetic energy.
Again, neglecting small terms of 2nd order gives the kinetic energy to be

1 1
TD =
2
( ) ( )
m vɺD2 + wɺ D2 + J yψɺ D2 + J zθɺD2 + Ω J xψɺ Dθ D
2
(6.31)

which can be written in terms of the generalized nodal displacements as

1
ɺ f T


f
TD = qD M D qɺ Df + qɺ Df T G int
D qD (6.32)
2

m 0 0 0 0 0 0 0 
0 m 0 0  0 0 0 0 
where M D =  ; G int = .
0 0 Jy 0 D
0 0 0 Ω Jx 
   
0 0 0 Jz  0 0 0 0 

Again, using the LAGRANGE’s equation (5.42), substituting equation (6.32) into it and setting the
potential energy and the non-conservative forces to zero gives the equation of motion of the circular
disk element in the fixed coordinate system as29

29 Compare footnote 26, page 86.

91
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

M D qɺɺDf + G D qɺ Df = 0 (6.33)

m 0 0 0 0 0 0 0 
0 m 0 0   0 0 0 0 
where M D =  ; GD = 
0 0 Jy 0 0 0 0 Ω Jx 
   
0 0 0 Jz  0 0 −Ω J x 0 

are the same matrices as previously defined in equation (6.28).

It can be seen that the equations of motion (6.30) and (6.33) and the element matrices determined
by the two methods are identical with each other.

Since the considered disk is circular, the mass moments of inertia J y and J z about the y- and z-

axis, respectively, are the same. They are commonly called the transverse mass moments of inertia J t in

the rotor dynamics literature. On the other hand, the mass moment of inertia J x about the x-axis is

the polar mass moment of inertia and is usually symbolized by J p .

6.1.2.2 Circular Disk Element in the Rotating Coordinate System

The fixed and rotating coordinate system are related to each other through the transformation
matrix described for the shaft element by equation (6.18) in section 6.1.1.2. Here, for the disk element
the transformation matrix is smaller in size as the disk is a lumped element on a single node. Hence,
the nodal displacement vector in the rotating coordinate system is given by

qDf = TD qDr (6.34)

T 0  cos α − sin α 
where TD =   ; T= ; α = Ωt − β .
 0 T  sin α cos α 

Deriving the displacement vector (6.34) once and twice and substituting it into equation (6.30) or
(6.33) gives

ɺɺ q
r + 2T
MD T ( ɺ q
ɺ r + T qɺɺ
r + G T
) ɺ q
r + T q
ɺ r = 0 .
( ) (6.35)
D D D D D D D D D D D

Pre-multiplying equation (6.35) by the transpose of the transformation matrix will give the equation
of motion of the circular disk element in the rotating coordinate system as

92
 General Anisotropic Rotor-Bearing Systems

ɺ ) q
ɺ r + TT ( M T ɺ
r
TDT M D TD qɺɺDr + TDT ( G D TD + 2M D T D D + G D TD ) qD = 0
ɺɺ
  
D D D

mD cD kD

⇒ m D qɺɺDr + c D qɺ Dr + k D qDr = 0 . (6.36)

The circular disk element mass, velocity-proportional and displacement-proportional matrices are given in detail in
Appendix H.2.2.

The disks considered in the above two sub-units are circular disks leading to time independent
element matrices in both fixed and rotating coordinate systems. A procedure similar to the one
described for the shaft can be followed to determine the element matrices of a non-circular disk. The
corresponding matrices are given in Appendix H.2.3 and H.2.4 without further explanation in this
chapter.

6.1.3 Bearing Elements

In this sub-unit, a general anisotropic visco-elastic bearing element is considered whose elastic and
damping properties in the fixed and rotating coordinate systems are discussed without investigating
the sources of the elasticity or damping. The bearing model used in this dissertation is based on a
linear relationship between forces and displacements and velocities as common in most rotor
dynamics literature.

Figure 6.4: Anisotropic bearing model

93
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 6.4 shows the bearing and shaft cross-section with the corresponding stiffness and damping
properties in the fixed coordinate system. Such bearing models develop radial support only and do not
provide any moment or longitudinal load on the shaft. For the sake of generality, a bearing with
cross-coupling is considered. For such types of bearings, the force-displacement/velocity relation of the
bearing element in the fixed coordinate system is given by Irretier [42], Lalanne [51] and others based on a
linearized solution of the REYNOLD’s equation for hydrodynamic journal bearings to be

FBf = DB qɺ Bf + K B qBf (6.37)

 d yy d yz 0 0  k yy k yz 0 0
d d zz 
0 0 k k zz 0 0 
where DB =  zy ; K B =  zy
 0 0 0 0 0 0 0 0
   
 0 0 0 0 0 0 0 0

are the bearing element damping and stiffness matrices of the bearing element, respectively, and

T
T
FBf =  FyB FzB M yB M zB  and qBf = [ vB wB ψ B θ B ] are the generalized nodal force and

displacement vectors of the bearing element, respectively.

The bearing properties need to be determined in the rotating coordinate system, too, as rotors with non-
circular shafts are better approached in this coordinate system. Hence, to determine the bearing
matrices in this coordinate system, one follows the transformation already described for the shaft
and the disk. Introducing the relation

qBf = TB qBr , (6.38)

where TB = TD is the transformation matrix as given in equation (6.34), one gets from equation
(6.37) the force-displacement relation of the bearing element in the rotating coordinate system as

ɺ ) q
r ⇒
TBT FB = TBT DB TB qɺ Br + TBT ( K B TB + DB T



f B ( t ) = dɶ B qɺ Br + kɶ B qBr . (6.39)

   
B

B
f B (t ) ɶ
dB ( t ) kɶ B ( t )

As can be seen from equation (6.39), the bearing element displacement proportional and damping matrices are
generally time dependent in the rotating coordinate system. Details of these matrices are discussed
in Appendix H.3.2.

94
 General Anisotropic Rotor-Bearing Systems

The element matrices of all the finite elements associated with a hybrid rotor model in both the fixed and
the rotating coordinate system are derived until this point. Since rotor systems are constructed to
achieve certain tasks there is always some sort of load acting on them. Besides these loads
associated with the tasks there are also some unavoidable undesired loads acting on the rotors.
Unbalance forces, which are always present, are one of the undesired loads which may greatly
influence the dynamics of rotors. In horizontally placed rotors, the gravitational load also affects the
rotor dynamics. Hence, it is essential to observe the appearance of these loads in both coordinate
systems which will be presented in the next sub-unit.

6.1.4 External Forces

As stated in the previous unit, there are always different types of external loads acting on a rotor
bearing system. Considering a nodal point in a finite element model, a general external force vector in
the fixed coordinate system is given by

T
F =  Fy Fz M y M z  . (6.40)

This force may be written in the rotating coordinate system by premultiplying equation (6.40) by the
transpose of the transformation matrix given in equation (6.34) or (6.38) as

f = TLT F (6.41)

where the general load transformation matrix is TL = TD = TB . Equation (6.41) is a general loading
and may be applied to all types of nodal loading.

One common loading expected on all rotors is the unbalance force. In the fixed coordinate system, the
nodal unbalance force vector (which is usually modelled to be located on the disks) of a rotor rotating
with a constant angular velocity Ω is given by

 FyU  − sin ( Ωt + δ ) 
   
2  cos ( Ωt + δ ) 

 FzU 
FU =   = mε Ω  . (6.42)
 M yU   0 
 M zU   0 

where m , ε and δ are the mass, unbalance radius and phase of the disk at a particular node,
respectively. Equation (6.42) shows that the unbalance force is a harmonic force acting at the node
where the unbalanced disk is located.

95
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

In the rotating coordinate system, the nodal unbalance force vector becomes

 fηU  − sin ( β + δ ) 


 fζ U    
2  cos ( β + δ ) 
fU =   = mε Ω  . (6.43)
 mηU   0 
 mζ U   0 

As opposed to the harmonic function of the unbalance force in the fixed coordinate system, the
force is constant in the rotating coordinate system. This also meets the intuitive expectation since
the unbalance is at a fixed point in this coordinate system.

The other common type of loading present in horizontally mounted rotors is the gravitational load.
This loading has a significant effect in rotor systems with non-circular shafts. The gravitational load
is a constant vertically downward force and is therefore constant in the fixed coordinate system. For the
selected orientation of the coordinate system the nodal gravitational load vector of a disk is therefore
given by

 FyG  0 


 FzG    
1 
FG =   = mg   . (6.44)
 M yG  0 
 M zG  0 

In the rotating coordinate system, the nodal gravitational load vector becomes

 fηG   sin α 


 fζ G    
cos α 
fG =   = mg   ; α = Ωt − β (6.45)
 mη G   0 
mζ G   0 

which unlike in the fixed coordinate system is harmonic in nature.

Once the element matrices of the shaft, the disk and the bearings and the nodal vectors of the
displacements, velocities, accelerations and forces are determined, then these terms are assembled
together to form the global matrices and vectors as outlined in chapter 5.2.3 which, in general, will
become time dependent. The next sub-unit uses the HILL’s method to solve the resulting second
order differential equation of motion with time dependent coefficients.

96
 General Anisotropic Rotor-Bearing Systems

6.2 HILL’s Method for Free and Forced Vibration Analysis

A hybrid rotor consists of a number of elements which were modelled in the previous sub-unit.
Further, for a rotor with a non-circular shaft and anisotropic bearings, it can be seen that there is always an
element whose properties are time dependent, requiring the use of the HILL’s method for the
analysis of the assembly in both the fixed and rotating coordinate system.

Considering the general rotor-bearing system in the fixed coordinate system, the global matrices are
composed of time dependent shaft mass and stiffness matrices, and time independent shaft
gyroscopic, disk mass and gyroscopic (for circular disks) and bearing stiffness and damping
matrices as seen from the element equations (6.26), (6.30) or (6.33) and (6.37) and the generalized
external force vector given in equation (6.40). Per procedure outlined in section 5.2.3 the global
equation of motion in the fixed coordinate system is therefore30

ɶ qɺɺ
f + Cq
ɺ f + K
M ɶ q
f = F (6.46)

where M ɶ =M +M ɶ , C = D +G +G , K ɶ =K +K ɶ denote the global matrices of the rotor in

D S B D S B S


the fixed coordinate system, and F is the global force vector including the unbalance and gravitational
forces determined from the nodal forces given in equations (6.42) and (6.44), respectively, besides
other external forces if any.

As can be observed from equation 6.26, the shaft element matrices and, therefore, also the global
mass and stiffness matrices are time-dependent and harmonic of the form

ɶ = M + ∆M cos 2α + ∆M sin 2α
M ; α = Ωt − β (6.47)
S 0S CS SS

which can be written to apply the HILL’s method outlined in chapter 3.2 using the trigonometric
identities31 as

ɶ = ∆M e − jΩɶ t + M + ∆M e jΩɶ t ; Ω
M ɶ = 2Ω (6.48)
S −1S 0S +1S

1 1
where ∆M −1S = ( ∆M CS + j∆M SS ) e j 2 β ; ∆M +1S = ( ∆M CS − j∆M SS ) e− j 2 β .
2 2

30 The global matrices for the EULER-BERNOULLI and RAYLEIGH beam theories can be determined from the
global matrices determined for the TIMOSHENKO beam theory.

31 e j 2α + e − j 2α e j 2α − e − j 2α .
cos 2α = ; sin 2α = − j
2 2

97
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Repeating the above procedure followed for the mass matrix the stiffness matrix becomes

ɶ = ∆K e − jΩɶ t + K + ∆K e jΩɶ t ; Ω
K ɶ = 2Ω . (6.49)
S −1S 0S +1S

1 1
where ∆K −1S = ( ∆K CS + j∆K SS ) e j 2 β ; ∆K +1S = ( ∆K CS − j∆K SS ) e− j 2 β .
2 2

The matrices in the preceding equations are given in detail in Appendix H.1.4. Hence, the global
equation of motion (6.46) may be written in exponential terms as

( ∆M −1
ɶ ɶ
) (
e − jΩt + M 0 + ∆M +1e jΩt qɺɺ f + Cqɺ f + ∆K −1e− jΩt + K 0 + ∆K +1e jΩt q f = F
ɶ ɶ
) (6.50)

where ∆M −1 = ∆M −1S , M 0 = M D + M 0S , ∆M +1 = ∆M +1S

and ∆K −1 = ∆K −1S , K 0 = K B + K 0S , ∆K +1 = ∆K +1S now represent the global

matrices of the rotor in the fixed coordinate system.

The same procedure can be followed to determine the global equation of motion of the rotor
bearing system in the rotating coordinate system. In this coordinate system, the global matrices are
composed of the time independent shaft and disk mass, velocity-proportional and displacement-
proportional matrices, and the time dependent bearing damping and displacement-proportional
matrices as seen from the element equations (6.22), (6.36) and (6.39) and the generalized external
force vector given in equation (6.41). Hence, the global equation of motion in the rotating coordinate system
is given by

mqɺɺr + cɶ qɺ r + kɶ q r = f (6.51)

where m = m D + m S , cɶ = dɶ B + c D + cS , kɶ = kɶ B + k D + k S denote the global matrices of the rotor in

the rotating coordinate system and f is the global force vector including the unbalance and
gravitational forces determined from the nodal forces given in equations (6.43) and (6.45) besides
other external forces if they exist.

Here again, the time dependent element and global matrices are harmonic and therefore equation
(6.51) may be written as

98
 General Anisotropic Rotor-Bearing Systems

( ɶ ɶ
)
ɶ
(
mqɺɺr + ∆c −1e − jΩt + c0 + ∆c+1e jΩt qɺ r + ∆k −1e − jΩt + k 0 + ∆k +1e jΩt q r = f
ɶ
) (6.52)

where ∆c −1 = ∆d −1B , c0 = d 0B + c D + cS , ∆c +1 = ∆d +1B

and ∆k −1 = ∆k −1B , k 0 = k 0B + k D + k S , ∆k +1 = ∆k +1B now represent the global

matrices of the rotor in the rotating coordinate system.

Hence, equation (6.52) for the rotating coordinate system is of a similar form as equation (6.50) for
the fixed coordinate system.

6.2.1 Eigenvalues, Eigenvectors and Stability Analysis

To determine the eigenvalues and eigenvectors for the stability analysis of the rotor consider the
homogeneous part of the equations (6.50) and (6.52) and follow the procedure already outlined in
chapter 3.2 for the eigenvalue and stability analysis of the LAVAL-rotor.

In the fixed coordinate system, assuming an exponential solution with periodic coefficients of the form

ɶ
ɶ
ɶ
ɶ
q f = Φ eΛt ; Φ = Φ ( t ) = Φ t + Tɶ ( ) ɶ Tɶ = 2π
; Ω (6.53)

and substituting the required derivatives into equation (6.50) yields

 ɺɺ
ɶ
ɶɺ
ɶ  
ɶɺ
ɶ 

( ∆M −1
ɶ ɶ


)
e − jΩt + M 0 + ∆M +1e jΩt  Φ + 2ΛΦ + Λ 2 Φ  + C  Φ + ΛΦ 
  

ɶ

ɶ
( ɶ
+ ∆K −1e− jΩt + K 0 + ∆K +1e jΩt Φ = 0 . ) (6.54)

ɶ
Similar to section 3.2.1, the periodic function Φ is written using a FOURIER-series as

ɶ ∞
ˆ
ɺɶ ∞
ɺɺ
ɶ ∞
ˆ juΩɶ t

ɶ ˆ e juΩɶ t
∑ ∑ = − ∑ u 2Ω
ɶ
Φ= Φ u e juΩt ; Φ = ju ΩΦ ; Φ ɶ 2Φ . (6.55)
u u e
u = −∞ u = −∞ u = −∞

Taking a finite index of summation U in equation (6.55) and substituting it into equation (6.54)
gives

U
ˆ juΩɶ t
( ɶ
∆M −1e− jΩt + M 0 + ∆M +1e jΩt
ɶ
) ∑ ( −u Ωɶ
u = −U
2 2 ɶ +Λ2 Φ
+ 2Λju Ω )
ue

U
ˆ juΩɶ t U


+ C ∑ juΩ
ɶ +Λ Φ
u = −U
(
ue ) ɶ
( ɶ
+ ∆K −1e − jΩt + K 0 + ∆K +1e jΩt ) ∑ Φˆ e
u = −U
u
ɶt
juΩ
=0 (6.56)

99
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

from which follows

U
ˆ j (u −1)Ωɶ t U
ˆ juΩɶ t
∆M −1 ∑ ( −u Ωɶ
u = −U
2 2 ɶ +Λ2 Φ
+ 2Λju Ω u e )
+ M0 ∑ ( −u Ωɶ
u = −U
2 2 ɶ +Λ2 Φ
+ 2Λju Ω u e )
U
ˆ j (u +1)Ωɶ t U
ˆ juΩɶ t
+ ∆M +1 ∑ ( −u Ωɶ
u = −U
2 2 ɶ +Λ2 Φ
+ 2Λju Ω u e ) + C ∑ ju ɶ +Λ Φ
Ω u e
u = −U
( )
U
ˆ j u −1 Ωɶ t U
ˆ U
ˆ j u +1 Ωɶ t

∑ ∑ ∑
ɶ
+ ∆K −1 Φu e ( ) + K 0 Φ u e juΩt + ∆K +1 Φ u e ( ) = 0. (6.57)
u = −U u = −U u = −U

Collecting equal exponential terms as outlined in Appendix E.1 and writing the resulting equations
in matrix form gives

 ⋱ ⋱ ⋱ ⋮ ⋮ ⋮ ⋰
 
 ⋯ ∆M +1 M0 ∆M −1 0 0 ⋯
 Λ 2 ⋯ 0 ∆M +1 M0 ∆M −1 0 ⋯
  
 ⋯ 0 0 ∆M +1 M0 ∆M −1 ⋯
 ⋰ ⋮ ⋮ ⋮ ⋱ ⋱ ⋱
 
⋱ ⋱ ⋱ ⋮ ⋮ ⋮ ⋰
⋯ − j 4 Ω
ɶ ∆M C − j 2ΩɶM 0 0 0 ⋯
 +1 0

+ Λ ⋯ 0 ɶ ∆M
− j2 Ω C j2 Ωɶ ∆M 0 ⋯
+1 −1
 ɶ ɶ ∆M 
⋯ 0 0 0 C + j 2ΩM 0 j4 Ω −1 ⋯
⋰ ⋮ ⋮ ⋮ ⋱ ⋱ ⋱
⋱ ⋱ ⋱ ⋮
⋯ ∆K − 4 Ωɶ 2 ∆M ɶ C−Ω
K 0 − jΩ ɶ 2M ∆K −1
 +1 +1 0

+ ⋯ 0 ɶ 2 ∆M
∆K +1 − Ω K0
+1

⋯ 0 0 ∆K +1
⋰ ⋮ ⋮ ⋮
 ⋮ 
⋮ ⋮ ⋰  
  ⋮ 
 ˆ 

0 0 ⋯  Φ −1  0 

ˆ  

∆K −1 − Ωɶ 2 ∆M
−1 0 ⋯   Φ  = 0  . (6.58)
ɶ ɶ 2 ɶ 2 ∆M   
0  

K + jΩC − Ω M ∆K −1 − 4 Ω ⋯  ˆ
0 0 −1
 Φ1  0 
⋱ ⋱  
⋱    ⋮ 
 ⋮   

In compact form equation (6.58) can be written as a HILL’s eigenvalue problem for the eigenvalues Λ in
the fixed coordinate system in the form

( Λ Mˆ + ΛCˆ + Kˆ ) Φˆ = 0 .
2
(6.59)

100
 General Anisotropic Rotor-Bearing Systems

Note here that the size of the matrices M̂ , Ĉ , and K̂ is N (2U + 1)× N (2U + 1) and of the vector

ˆ
Φ is N (2U + 1)×1 where N is the number of nodes times the number of degrees of freedom per
node used in the FEM to analyse the rotor-bearing assembly. The eigenvalues and eigenvectors are
determined by solving the HILL’s eigenvalue problem given in equation (6.59). In this work
MATLAB, which uses the QR and QZ factorization to address such problems, is implemented.

Next, to determine the eigenvalues and eigenvectors of the rotor in the rotating coordinate system,
consider the homogeneous part of equation (6.52). Assuming again an exponential solution with
periodic coefficients of the form

ɶ
ɶ
ɶ
ɶ
q r = φ eλ t ; φ = φ ( t ) = φ t + Tɶ ( ) ɶ Tɶ = 2π
; Ω (6.60)

and substituting the required derivatives into equation (6.52) gives

 ɺɺ
ɶ
ɺɶ
ɶ  
ɺɶ
ɶ 

ɶ
( ɶ
m  φ + 2λφ + λ 2φ  + ∆c−1e− jΩt + c0 + ∆c +1e jΩt  φ + λφ 
   
)

ɶ

( ɶ ɶ
+ ∆k −1e− jΩt + k 0 + ∆k +1e jΩt φ = 0 . ) (6.61)

ɶ
Again the periodic function φ is written using a FOURIER-series as

ɶ ∞
ˆ
ɺɶ ∞
ɺɺ
ɶ ∞

ɶ φˆ e juΩɶ t ɶ 2φˆ e juΩɶ t .
∑ φu e juΩt ; φ = ∑ ; φ = − ∑ u 2Ω
ɶ
φ= ju Ω u u (6.62)
u = −∞ u = −∞ u = −∞

Taking again a finite index of summation U in equation (6.62) and substituting it into equation
(6.61) gives

U

m ∑ −u 2 Ω ( ɶ + λ 2 φˆ e juΩɶ t
ɶ 2 + 2λ ju Ω )
u
u = −U
U

( ɶ
+ ∆c −1e − jΩt + c0 + ∆c +1e jΩt
ɶ
) ∑ ( juΩɶ + λ )φˆ e
u = −U
u
ɶt
ju Ω

U


( ɶ
+ ∆k −1e − jΩt + k 0 + ∆k +1e jΩt
ɶ
) ∑ φˆ e
u = −U
u
ɶt
ju Ω
=0 (6.63)

from which follows

101
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

U
U
ˆ
m ∑ −u 2 Ω ( ɶ + λ 2 φˆ e juΩɶ t + ∆c
ɶ 2 + 2λ ju Ω ) ∑ ( juΩɶ + λ )φ e
ɶt
j ( u −1) Ω
u −1 u
u = −U u = −U
U
U
ˆ
ɶ + λ φˆ e juΩɶ t + ∆c
∑( ∑ ( juΩɶ + λ ) φ
ɶt
+ c0
u = −U
juΩ u) +1
u = −U
u e
j ( u +1) Ω

U
ˆ U
ˆ U
ˆ

∑ φu e j(u −1)Ωt + k 0 ∑ φu e juΩt + ∆k +1 ∑φ
ɶ ɶ ɶt
j ( u +1) Ω
+ ∆k −1 u e =0. (6.64)
u = −U u = −U u = −U

Collecting equal exponential terms as already referred to Appendix E.1 and writing the resulting
equations in matrix form now gives

 ⋱ ⋱ ⋮ ⋮ ⋮ ⋮ ⋰
  
 ⋯ 0 m 0 0 0 ⋯
 λ 2 ⋯ 0 0 m 0 0 ⋯
  
 ⋯ 0 0 0 m 0 ⋯

 ⋰ ⋮ ⋮ ⋮ ⋮ ⋱ ⋱
⋱ ⋱ ⋱ ⋮ ⋮ ⋮ ⋰
⋯ ∆c c 0 − j 2Ωɶ m ∆c 0 0 ⋯
 +1 −1

+ λ ⋯ 0 ∆c +1 c0 ∆c−1 0 ⋯
 ɶ m ∆c 
⋯ 0 0 ∆c +1 c 0 + j 2Ω −1 ⋯
⋰ ⋮ ⋮ ⋮ ⋱ ⋱ ⋱
⋱ ⋱ ⋱ ⋮
⋯ ∆k − 2 jΩ
ɶ ∆c ɶ c −Ω
k 0 − jΩ ɶ m ∆k
2
 +1 +1 0 −1

+ ⋯ 0 ɶ
∆k +1 − jΩ∆c +1 k0

⋯ 0 0 ∆k +1
⋰ ⋮ ⋮ ⋮
⋮ 
⋮ ⋮ ⋰  
  ⋮ 
 ˆ 

0 0 ⋯  φ−1  0 

 

ɶ
∆k −1 + jΩ∆c−1 0 ⋯   φˆ0  = 0  . (6.65)


ɶ c −Ω
k 0 + jΩ ɶ m ∆k + 2 jΩ
2 ɶ ∆c ⋯  
ˆ  0 
0 −1 −1
 φ1   
⋱ ⋱ ⋱     ⋮ 
⋮ 

In compact form equation (6.65) leads to the HILL’s eigenvalue problem for the eigenvalues λ in the
rotating coordinate system as

( λ mˆ + λ cˆ + kˆ )φˆ = 0 .
2
(6.66)

102
 General Anisotropic Rotor-Bearing Systems

Note here that the size of the matrices m̂ , ĉ , and k̂ is again N (2U + 1)× N (2U + 1) and of the

ˆ
vector φ is N (2U + 1)×1 where, as before, N is the number of nodes times the number of
degrees of freedom per node used in the FEM to analyse the rotor bearing assembly.

As can be seen from the above derivation, the size of the mass, velocity-proportional and
displacement-proportional matrices in the equations (6.59) and (6.66) originally are infinite in
dimension. But as already stated in section 3.2, limiting the summation bounds to finite values from
-U to U does not affect the accuracy of the determined eigenvalues and eigenvectors greatly since higher
terms have less influence than lower terms in the FOURIER-series. Still after limiting the size of
the matrices to finite values, the size remains larger than the number of degrees of freedom of the
rotor model. This difference in size leads to a larger number 2 N (2U + 1) of eigenvalues and
eigenvectors than for a system with time independent matrices. But a closer investigation of the
solutions reveals that it is composed of 2 N basic eigensolutions, which describe the system properties,
and 2 N ⋅ 2U redundant eigensolutions, which give no additional information. As an example, the
displacement vector resulting from the assumed solution (6.53) (and (6.60) correspondingly) can be
written in terms of the basic and the redundant eigensolutions as

2N U
ˆ Λ t
qf =∑ ∑ ɶ
Φ i ,u e i ,u ; Λ i ,u = Λ i ,0 + juΩ (6.67)
i =1 u =−U

where the basic eigenvalues and eigenvectors are those determined for u = 0 while all other values
of u lead to the redundant eigenvalues and eigenvectors.

Besides the free vibration solution, the basic eigenvalues are used to identify the rotor stability. If there
are one or more basic eigenvalues with a positive real part, then the rotor is unstable. On the other
hand, if all basic eigenvalues have a non-positive real part then the assembly is stable for the
specified rotor speed of rotation.

6.2.2 Unbalance Vibrations

After the determination of the eigensolution of the rotor the response to commonly present
excitation forces is addressed. One such excitation, which can never be completely avoided, is the
unbalance force and therefore it is discussed in this section. As will be shown in this sub-unit, the
solution method for the unbalance response of a hybrid rotor depends on the bearing anisotropy.

If the bearing properties of the rotor-bearing assembly are isotropic then it is best analysed in the
rotating coordinate system. In this coordinate system, the unbalance force is constant as seen from

103
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

equation (6.43) giving a constant vector as solution as well. Hence, the unbalance response in the
rotating coordinate system is determined from equation (6.51) as

kɶ qUr = k qUr = fU (6.68)

where the global unbalance force vector consists of the nodal unbalance force vectors given in


equation (6.43). From equation (6.68) qUr can be determined using a number of commercially
available softwares. In this work MATLAB, which uses a GAUSS elimination to address such
problems, is implemented.

For a rotor with a non-circular shaft mounted in anisotropic bearings, there is at least one finite
element matrix which is time dependent in both the fixed and rotating coordinate system. Since the
investigation in the fixed coordinate system simplifies the interpretation of the results, the equations
are presented here in this coordinate system. A similar approach may be followed to determine the
unbalance response in the rotating coordinate system.

The unbalance response in the fixed coordinate system can be determined from the equation of motion
given in equation (6.46) and (6.50) with a global unbalance force vector composed by the nodal
unbalance force vectors as given in equation (6.42). To determine the rotor unbalance response this
nodal unbalance force vector is rewritten using exponential functions32 as

 FyU  − sin ( Ωt + δ )   − je− jδ   je jδ  

         
2  cos ( Ωt + δ ) 

 FzU  mε Ω 2   e − jδ  − jΩt  e jδ  jΩt 
FU =   = mε Ω  =  e + e 
 M yU   0  2  0   0 
 
 M zU   0  
 0 
   0 


ˆ
ˆ
⇒ FU = FU− e − jΩt + FU+ e jΩt (6.69)

which, for δ = 0 , corresponds to the unbalance force vector in equation (3.42) for the LAVAL-
rotor.

The global unbalance force vector is now composed by the nodal unbalance force vectors given in
equation (6.69) and introduced into equation (6.50). The response vectors due to this unbalance
global force vector are assumed to be of the form

32 See footnote 31 on page 97.

104
 General Anisotropic Rotor-Bearing Systems

∞
ˆ
∞
U−
∑ ∑ Φˆ
ɶ ɶ
qUf + = e jΩt ΦUu + e juΩt ; qUf − = e − jΩt u e juΩt . (6.70)
u =−∞ u =−∞

Substituting equation (6.70) into the equation of motion (6.50) and comparing again the
exponential terms as outlined in Appendix E.2.1 yields

⋱ ⋱ ⋱ ⋮
 2 2 2
⋯ ∆K +1 − 1.5Ωɶ ( ) ∆M +1 ( )
ɶ C − 0.5 Ω
K 0 − j 0.5 Ω ɶ( ) M0 ɶ
∆K −1 − 0.5Ω ( ) ∆M −1

⋯
2 2
0 ∆K +1 − ( 0.5 Ω
ɶ) ∆M +1 K0 + j ( 0.5 Ω
ɶ ) C − ( 0.5 Ω
ɶ) M0

 2

⋯ 0 0 ∆K − ( 0.5 Ω
+1
ɶ ) ∆M
+1
⋰ ⋮ ⋮ ⋮

⋮ ⋮ ⋰   ⋮   ⋮ 

 

0 0 ⋯  Φ ˆ
 U−1+   0 

 

⋯  Φ
2
ˆ U +
ˆ +
(
ɶ
∆K −1 − 1.5Ω ) ∆M −1 0 ˆ U +  =  Fˆ +  ⇒ K U + Φ
0 U
= FUH
2 2   
ˆ  

K0 + j (1.5Ω
ɶ ) C − (1.5Ω
ɶ) M0 (
ɶ
∆K −1 − 2.5Ω ) ∆M −1 ⋯  Φ U+
1
 0
    
⋱ ⋱ ⋱   ⋮   ⋮ 

⋱ ⋱ ⋱ ⋮
 2 2 2
(
ɶ
⋯ ∆K +1 − 2.5 Ω ) ∆M +1 ( ) (
ɶ C − 1.5 Ω
K 0 − j 1.5 Ω ɶ ) M0 ( ɶ
∆K −1 − 0.5 Ω ) ∆M −1
 2 2
⋯ 0 ∆K +1 − (1.5 Ω
ɶ) ∆M +1 K0 − j ( 0.5 Ω
ɶ ) C − ( 0.5 Ω
ɶ) M0

 2

⋯ 0 0 ∆K − ( 0.5 Ω
+1
ɶ ) ∆M
+1
⋰ ⋮ ⋮ ⋮

⋮ ⋮ ⋰   ⋮   ⋮ 

 

0 0 ⋯  Φ ˆ
 U−1−   0 

 

⋯  Φ
2
ˆ U −
ˆ −
ɶ
∆K −1 − 0.5 Ω ( ) ∆M −1 0 ˆ U −  =  ˆ −  ⇒ KU −Φ
0 FU = FUH
2 2 

ˆ   

K0 + j ( 0.5 Ω
ɶ ) C − ( 0.5 Ω
ɶ) M0 ɶ
∆K −1 − 1.5 Ω ( ) ∆M −1 ⋯  Φ U−
1
 0 
     
⋱ ⋱ ⋱   ⋮   ⋮ 
(6.71)

105
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

ˆ
ˆ
From equation (6.71) the values of ΦUu + and ΦUu − can be determined, which can then be
substituted into equation (6.70) to generate the required unbalance response in the fixed coordinate system
as

qUf = qUf + + qUf − (6.72)

where again a finite number of U terms is taken into consideration.

6.2.3 Vibrations due to Gravitational Load

The other commonly present force is the gravitational load on horizontal rotor assemblies. This force
is not interesting in the analysis of axisymmetric shaft rotors as it gives the rotor only a vertical
static sag. But in rotors with a non-circular shaft this load results in a gravitational response which has
an important effect as already discussed in chapters 3 and 4.

Here again, the rotor with a non-circular shaft and isotropic bearings is analysed in the rotating
coordinate system. In this coordinate system, the element and global matrices are constant and the
gravitational load is time dependent described by equation (6.45) here rewritten as

 fηG   sin α   1  0  


 fζ G          



cos α   0  1  
fG =   = mg   = mg    sin α +   cos α  ⇒ f G = fηG + fζ G (6.73)
 mηG   0  0
   0  
mζ G   0   0  0  

The gravitational response can be determined from the equation of motion (6.51) where now the
nodal gravitational force vectors given in equation (6.73) constitute the global gravitational force
vector. Being a harmonic type of excitation in the considered rotating coordinate system a solution
of the form

r
r
qGr = qGs sin α + qGc cos α ; α = Ωt − β (6.74)

is assumed. Substituting equation (6.74) into the equation of motion (6.51) and separating the sine
and cosine terms gives

ˆ
  fηG 

r
k − Ω 2 m −Ωc   qGs
  
r  = 
ˆ  . (6.75)
 Ωc k − Ω 2m  qGc   fζ G 

106
 General Anisotropic Rotor-Bearing Systems

From equation (6.75) the response vectors due to the gravitational force can be determined.
Substituting the results in equation (6.74) will then give the gravitational response in the rotating
coordinate system.

As presented in the previous sub-unit for a rotor with a non-circular shaft mounted in anisotropic
bearings there is at least one element matrix which is time dependent in both the fixed and rotating
coordinate system. To ease the interpretation of the results the rotor is investigated for this case in
the fixed coordinate system. Hence, the gravitational response can be determined from the equation of
motion given in equation (6.46) and (6.50) where in this case the global force vector is composed of
the nodal gravitational load vectors given in equation (6.44). Assuming a solution of the form

∞
G
∑ Φˆ
ɶ
qGf = u e juΩt , (6.76)
u =−∞

substituting it into the equation of motion (6.50) and comparing again the exponential terms as
outlined in Appendix E.2.2 gives an equation of the form

⋱ ⋱ ⋱ ⋮
⋯ ∆K − 4Ω ɶ 2 ∆M ɶ C−Ω
K 0 − jΩ ɶ 2M ∆K −1
 +1 +1 0

⋯ 0 ɶ 2
∆K +1 − Ω ∆M +1 K0

⋯ 0 0 ∆K +1
⋰ ⋮ ⋮ ⋮
 ⋮ 
⋮ ⋮ ⋰ 
  ⋮ 
ˆG 

0 0 ⋯ Φ −1   0 
ɶ 2∆M 
ˆ G  

ˆ G

∆K −1 − Ω −1 0 ⋯  Φ  =  FG  ⇒ KGΦ = FGH . (6.77)
ɶ ɶ 2 ɶ 2∆M 
0  

K + jΩC − Ω M ∆K −1 − 4Ω ⋯ ˆ G
0 0 −1
 Φ1   0 
⋱ ⋱ ⋱   ⋮ 
 ⋮   

ˆ
Once the values of Φ Gu are determined from the above equation, it will be substituted into
equation (6.76) to determine the gravitational response in the fixed coordinate system.

In this chapter all the element matrices required for the dynamic finite element model of a general rotor-
bearing system with a non-circular shaft and anisotropic bearings are developed in both the fixed and
rotating coordinate system. Further, the equations of motion of the complete system and the
corresponding global mass, velocity-proportional and displacement-proportional matrices are

107
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

developed and generally solved for the eigenvalues, eigenvectors and the unbalance and gravitational
responses. In the next chapter related results are given for particular rotor-bearing assemblies.

6.3 Description of the Finite Element Program

A finite element program called FEARS33 was developed to investigate the dynamic properties of
general anisotropic rotor-bearing-systems. This program, whose output GUI 34 is shown in the
figures 7.1 and 7.6 in the next chapter, is modular in construction. The rotor data can be supplied
either using the interactive input GUIs or through a separate input text file.

The first two modules of the program investigate LAVAL-rotors with circular and non-circular shafts
mounted in rigid bearings analytically. The third and fourth modules then investigate a LAVAL-rotor
with circular and non-circular shaft mounted in anisotropic elastic bearings. The investigation of LAVAL-
rotors with non-circular shafts in anisotropic elastic bearings implements the HILL’s method of
solution. The fifth and sixth modules are Finite Element Method (FEM) programs to analyse general
rotors. The fifth module is used for the analysis of rotors with circular shafts while the sixth module
analyses rotors with non-circular shafts. In the sixth module the rotor-bearing-system is investigated in
the rotating coordinate system if the rotor is mounted in either rigid, isotropic elastic or isotropic visco-elastic
bearings so as to avoid the use of the HILL’s method of solution. If on the other hand, the rotor is
mounted in anisotropic elastic, anisotropic visco-elastic or fluid film bearings then the rotor is analysed in the
fixed coordinate system using the HILL’s method of solution.

The FEM programs generate a 2-D sketch of the rotor-bearing assembly being analysed, the
vibration mode shapes (static plots and animations), the CAMPBELL-diagram of the natural
frequencies (in the fixed and rotating coordinate system and the identified basic natural frequencies
for the rotor analysed using the HILL’s method of solution), the damping ratios, the forced
responses (for unbalance and gravitational forces) and the frequency response functions.

Upon completion of the analysis, the outputs can be saved for later investigations. Besides the
outputs, the program can also generate a text file that can be modified and later used as an input file
for analysing a modified rotor-bearing-system.

33 FEARS : Finite Element Analysis of Rotor Systems.

34 GUI : Graphical User Interface.

108
 7 Chapter 7

NUMERICAL RESULTS

In the previous two chapters, the mathematical background of the Finite Element Method (FEM), the
element matrices and the equations of motion of a general hybrid rotor model are derived. Appendix
H further gives the element matrices for circular and non-circular shaft elements, circular and non-circular
disk elements and general bearing elements in both the fixed and rotating coordinate system. The equations of
motion are solved in the present chapter so as to determine the stability properties and the unbalance
and gravitational responses of selected rotor assemblies.

In this chapter, the LAVAL-rotor with a non-circular shaft (which was also experimentally investigated)
is reinvestigated using the FEM. Thus, the viability of the method and the effectiveness of the
developed program are investigated. This chapter further discusses two other selected rotor
assemblies, one of which is further varied to check the effect of the longitudinal symmetry and the
resultant non-circularity of the shaft. The third example is the two disk rotor analysed by Forrai L. [27].

7.1 LAVAL-Rotor with Non-Circular Shaft and Different Types of Bearings

7.1.1 Rigid Bearings

As stated in the introductory words, the rotor set-ups investigated in chapter 4 are reinvestigated in
this sub-unit. The LAVAL-rotor in rigid bearings was presented in figure 3.3 and the corresponding
experimental set-up was depicted in figure 4.26. In this sub-unit, this assembly, whose data are
given in Appendix F.1, is analysed using the FEM program written in MATLAB. This program
uses the rotating coordinate system to investigate all axially asymmetric rotors mounted in rigid bearings
so as to avoid the use of HILL’s method and the accompanying large sized matrices. Hence, the
equation of motion used for such rotor-bearing-systems comes from equation (6.51) by setting the
bearing properties to zero. Thus, the equation of motion becomes a second order differential
equation with constant coefficient which can be simply solved as described in chapter 3.

For the free vibration and stability analysis, the right hand side of equation (6.51) is set to zero so as to
investigate the homogeneous equation. From the resulting homogeneous equation, the eigenvalue
problem is solved. All the eigenvalues of the rotor-system are seen to be complex conjugate, with
(almost) zero real part over the stable operation region. But over the unstable operation speed
range there is at least one eigenvalue with positive real part as it was shown in chapter 3.
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Some of the outputs of the developed MATLAB-FEM program, named FEARS35, are displayed in
figure 7.1. Besides the static plots i.e. a 2D-sketch of the rotor system, the mode shapes, the natural
frequencies, the damping ratios and the frequency response functions (FRFs), the natural mode
animation with the corresponding increase or decrease of the amplitudes can be seen in the output
GUI36.

Figure 7.1: Outputs of FEARS for the LAVAL-rotor in

rigid bearings

In this sub-unit, the natural circular frequencies of the test rotor shown in figure 4.26 are presented
using the FEM with 6 shaft elements and compared with the analytical results from equation (4.5)
in table 7.1.

35 See footnote 33 on page 108.

36 See footnote 34 on page 108.

110
 Numerical Results

Beam Theory1
Operation Error Error Error
speed [rpm] Analytical E-B [%] Ray [%] Tim [%]
1st 41.21 39.58 3.96 39.58 3.96 39.55 4.04
500 2nd 147.59 145.92 1.13 145.92 1.13 145.89 1.15
st
1 31.66 28.94 8.60 28.94 8.60 28.91 8.70
600 2 nd
156.94 156.32 0.39 156.32 0.39 156.29 0.41
1st 19.68 17.97 8.69 17.97 8.69 17.94 8.86
700 2 nd
168.39 166.74 0.98 166.74 0.98 166.70 1.00
st
1 7.42 4.98 32.98 4.97 32.99 4.92 33.71
800 2nd 178.82 177.17 0.93 177.17 0.93 177.14 0.94
st
1 0 0 0 0 0 0 0
850 2 nd
184.04 182.39 0.90 182.39 0.90 182.36 0.92
1st 0 0 0 0 0 0 0
900 2 nd
189.27 187.61 0.87 187.61 0.87 187.58 0.89
st
1 0 0 0 0 0 0 0
950 2nd 194.49 192.84 0.85 192.84 0.85 192.80 0.87
1st 6.56 9.00 -37.26 9.00 -37.27 9.04 -37.93
1000 2 nd
199.72 198.06 0.83 198.06 0.83 198.03 0.84
st
1 20.20 20.89 -3.39 20.89 -3.39 20.92 -3.56
1100 2nd 211.21 208.52 1.28 208.52 1.28 208.48 1.29
1 ‘E-B’ : EULER-BERNOULLI, ‘Ray’ : RAYLEIGH, ‘Tim’ : TIMOSHENKO beam theory

Table 7.1: Natural circular frequencies [rad/s] of the test

rotor in rigid bearings (analytical and FEM-
results for different beam theories)

As can be seen from the table, the natural circular frequencies as determined by the three beam
theories are almost equal. The minimal changes in their values are due to the high shaft slenderness.
It has been proven by a number of authors that the beam theories affect the natural circular
frequencies as the slenderness ratio of the shaft decreases. The table also shows that the percentage
errors of the eigenvalues are relatively smaller when using the EULER-BERNOULLI beam theory,
even though the TIMOSHENKO beam theory gives a better description of the actual behaviour of
free vibrating beams. The reason is that the analytical method is based on the EULER-
BERNOULLI theory. The table further shows that the percentage difference is greater around the
boarder of the stable-unstable regions in comparison to the other operation regions. One final
point to see in the table is the comparatively lower percentage error of the second natural circular

111
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

frequency in comparison to the first. This is so because of the comparatively small absolute value of
the first natural circular frequency in comparison to the second one.

Figure 7.2 shows the CAMPBELL-diagram for the natural circular frequencies in the η- and ζ-
directions of the test rotor in figure 4.26 as determined analytically and using the FEM.

Figure 7.2: Natural circular frequencies of the test rotor

in rigid bearings in the rotating coordinate
system

From figure 7.2 it can be seen that the analytically determined natural circular frequencies are
slightly greater than those determined using the FEM. This means that the analytical model is
slightly stiffer and/or lighter than the FEM model. This is due to the additional flexibility incurred
due to the use of the TIMOSHENKO beam theory over the EULER-BERNOULLI beam theory.
Further, the discretization in the FEM procedure engenders more flexibility into the system. The
FEM also considers the shaft mass, all of which contribute to the lowering of the eigenvalues
leading to the differences seen in the figure.

Next, the unbalance response of the test rotor is determined using equation (6.68). Again, the FEM
results are compared with the analytical results from equation (4.10) as shown in figure 7.3. Here
again, the results in figure 7.3 were generated using 6 equal length shaft elements.

112
 Numerical Results

Figure 7.3: Unbalance response of the test rotor in rigid

bearings

It can be seen that the rotor has unbalance resonances at 814 rpm in the ζ-direction and at 960 rpm
in the η-direction in the rotating coordinate system for all three beam theories. As in the eigenvalues,
figure 7.3 shows the slight shift of the unbalance resonances to the left as determined by the FEM
as compared to that determined by using the analytical method coming from the higher flexibility
and considered shaft mass in the FEM.

Further, the gravitational response of the test rotor is determined from equations (6.74) and (6.75). The
FEM results are again compared with the analytical response from equation (4.14) in figure 7.4.

As stated in the fourth chapter, the gravitational resonance speed is approximately half of the
average natural circular frequency. Here, the gravitational resonance frequency as determined by the
FEM is 439 rpm which is slightly below half of the average natural circular frequency of 445.5 rpm.
Besides this shift of the gravitational resonance determined by the FEM to the left, one other
difference is the constant amplitude difference of 0.073 mm. This comes from the additional shaft
mass which is considered in the FEM in contrast to the analytical method. The average sag due to
the shaft gravitational load is found to be 0.0739 mm, leaving the insignificant difference of 0.9 µm.

113
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 7.4: Gravitational response of the test rotor in

rigid bearings

Figure 7.5: Total (unbalance and gravitational) response

of the test rotor in rigid bearings

114
 Numerical Results

Finally, figure 7.5 shows the total response (unbalance37 and gravitational) as determined using the
FEM and from the experimental test set-up shown in figure 4.26.

As can be seen from the figure, the response determined using the FEM is pretty accurate for the
operation speeds far away from the high amplitude operation areas, i.e. the gravitational resonance
and the unbalance resonance speeds. It can further be seen that there is no measurement between
the two unbalance resonances for this region is the unstable operation region as shown by the
stability plot in figure 4.4. The other fact to note here are the higher experimental gravitational and
unbalance resonance speeds in comparison to the values determined by the FEM. Hence, the actual
experimental test set-up is also stiffer than the FEM LAVAL-rotor model used in FEARS. That
may lie in the bearings which are normally modelled to provide only lateral forces in LAVAL-
rotors. This requirement means that the roller bearing thickness must be of infinitesimal dimension
which is impractical. The finite thickness of the bearing support will therefore give some moment
loading to the bearing making the assembly stiffer resulting in higher gravitational and unbalance
resonance speeds than calculated.

7.1.2 Anisotropic Elastic Bearings

In this sub-unit, the LAVAL-rotor mounted on anisotropic elastic bearings is reinvestigated. The
constructed experimental rotor set-up is shown in figure 4.27 and its responses as determined
analytically are given in the figures 4.22-4.25. The experimental results of the test set-up are
presented in figure 4.32. Similar to sub-unit 4.2.2, the free vibration analysis and stability of the
assembly is initially addressed, followed by the investigation of the unbalance and gravitational
response where in all cases the fixed coordinate system is used for the numerical representation.

From the free vibration and stability analysis, the eigenvalues of the system and the unstable operation
regions can be computed. Again, the program FEARS generates a number of outputs describing
the system properties and results some of which are presented in the outputs in figure 7.6 which
shows a sketch of the rotor system, the mode shapes, the natural frequencies, the damping ratios
and, if required, the FRFs on its output GUI.

Table 7.2 shows the first four natural circular frequencies for a rotational speed of n = 700 rpm as
determined analytically and with the FEM while figure 7.7 gives the comparison of the
CAMPBELL-diagram of the natural circular frequencies as determined from both methods.

37 Note here that the assumed eccentricity used in chapter 4 and in generating figure 7.3 has been modified to approach
the experimental results.

115
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 7.6: Outputs of FEARS for the test rotor in

anisotropic elastic bearings

Beam Theory1
Analytical Ray Error [%] Tim Error [%]
1st 52.82 54.47 3.12 54.51 3.18
nd
2 70.04 68.17 2.66 68.16 2.68
rd
3 76.57 78.42 2.41 78.43 2.43
4th 93.78 92.05 1.85 92.02 1.88
1 ‘Ray’: RAYLEIGH ,‘Tim’: TIMOSHENKO beam theory

Table 7.2: Natural circular frequencies [rad/s] of the test

rotor in anisotropic elastic bearings (analytical
and FEM-results for different beam theories)
for n = 700rpm

As it can be seen from the table, the percentage errors for the natural frequencies are less than
3.5%. Higher natural frequencies are not compared here for the analytical method gives the first
four eigenvalues only.

116
 Numerical Results

Figure 7.7: Natural circular frequencies of the test rotor

in anisotropic elastic bearings in the fixed
coordinate system

As it can be seen in figure 7.7, the CAMPBELL-diagram is much more complicated in some speed
ranges to identify the natural frequencies. It has been presented by Lee [55] that the observation of
the norm of the eigenvectors helps to identify the natural frequencies. In the work of Ertz et. al.
[23], a computer routine is proposed for the identification of the frequencies. As it is found in this
work, both methods are indicative and not absolute to identify the frequency values. It is found
better to present all the natural frequencies in the CAMPBELL-diagram and to use the natural
change of their values with operation speed besides the indications given by the other works.

As can be seen from figure 7.7, the FEM does not give large differences in the natural circular
frequencies as that determined from the analytical method at the majority of the operation speeds
for this simple assembly. The differences are seen to be slightly larger around the stability limits.
The figure further shows that the analytical method is more conservative in determining the
unstable operation region as compared to the FEM using the TIMOSHENKO beam theory.

The unbalance response of the test rotor in figure 4.27 is determined using the equations (6.69)-(6.72).
The FEM results are compared with the analytical unbalance response in figure 7.8.

117
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 7.8: Unbalance response of the test rotor in

anisotropic elastic bearings

As can be seen in figure 7.8, all the first four unbalance resonances have been identified by the
FEM in the operation speed shown. Here again, the conservative nature of the analytical method is
seen as the resonance speeds determined by this method are spaced further apart from each other
than those determined by the FEM. The unbalance response amplitudes determined by the
analytical method are also seen to be higher than that determined by the FEM.

Next, the gravitational response of the rotor is determined from the equations (6.76) and (6.77). Again,
the FEM results are compared with the analytical gravitational response and shown in figure 7.9.

Figure 7.9 shows that the gravitational responses as determined from the two methods are pretty
the same except around the gravitational resonance regions. Like in the previous assembly, it can
also be seen that the analytical model is slightly stiffer than the FE-model as the resonances
determined by the analytical method are higher than those determined by the FEM.

Finally, figure 7.10 shows the comparison of the total response (unbalance and gravitational) FEM
results and the experimental findings. Here again the results are generally in good agreement. As in
the previous case, it can be seen that the experimental set-up is slightly stiffer than the FE-model.
This can be seen from the slightly elevated gravitational and unbalance resonance speeds.

118
 Numerical Results

Figure 7.9: Gravitational response of the test rotor in

anisotropic elastic bearings

Figure 7.10: Total (unbalance and gravitational) response

of the test rotor in anisotropic elastic
bearings

119
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

7.2 Rotor with Rectangular Shaft and Different Types of Bearings

In the previous sub-units, a uniform shaft non-circularity is considered for the whole length of the
rotor. In this sub-unit, the effects of non-uniform shaft non-circularities associated with the terms
longitudinal asymmetries and resultant non-circularities are investigated. This is deemed necessary for in
practice it is common to modify the resultant non-circularity to improve the stability behaviour of
such rotors. For this analysis, the rotor set-up with a rectangular shaft as shown in figure 7.11, whose
technical data are given in Appendix F.2, is used. This sub-unit will also show the effect of
modifying the bearing properties on the rotor behaviour.

7.2.1 Uniform Rectangular Cross-Section

First, the rotor assembly shown in figure 7.11 is considered with rigid bearings. Then, the same rotor
is analysed considering anisotropic visco-elastic bearings. The eigenvalues and the unstable operation
speed ranges are determined and compared with each other to show the effect of the bearings on
the assembly’s properties.

Figure 7.11: Rotor with a uniform rectangular shaft and

rigid bearings

7.2.1.1 Rigid Bearings

The equations used in generating the figures in this sub-unit are already given in section 7.1. As in
the previous discussion, the investigation for rigid bearings is started with the free vibration and stability
analysis from which the CAMPBELL-diagram for the natural circular frequencies in figure 7.12 and
the damping ratios in figure 7.13 are generated. As it was indicated in chapter 4, the rotor has at
least one negative damping ratio when one eigenvalue is zero in the rotating coordinate system. Here,
the CAMPBELL and the stability diagrams are plotted over a larger operation speed range to show
120
 Numerical Results

the higher unstable operation speed ranges. As can be seen from the two figures the 1st, 2nd and 3rd
unstable speed ranges, which are at 1190-1540 rpm, 1.9x104-2.5x104 rpm and 2.2x104-2.8x104 rpm,
are associated with the 1st, 3rd and 5th natural circular frequencies. These natural frequencies
correspond to the 1st, 3rd and 5th mode shapes, respectively.

Figure 7.12: Natural circular frequencies of the rotor with

a uniform rectangular shaft in rigid bearings
in the rotating and fixed coordinate system

Figure 7.13: Damping ratios of the rotor with a uniform

rectangular shaft in rigid bearings

121
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Next, the forced responses of the rotor in figure 7.11 are investigated. The rotor was loaded by
unbalance and gravitational forces. The related responses from these two loadings are given in the
figures 7.14 and 7.15, respectively.

Figure 7.14 shows the unbalance response due to an eccentricity of 0.383 mm. As it can be seen in the
figure, the unbalance resonances in the η- and ζ-directions are at different operation speeds, i.e. at
1190 and 2.2x104 rpm in the ζ-direction and at 1540 and 2.8x104 rpm in the η-direction, due to the
difference in the stiffness in the two axes. It can further be seen that the unbalance resonance
speeds are either at the beginning or the end of the zero natural frequency speed ranges of the 1st
and 5th mode shapes in the rotating coordinate system. Hence, the unbalance resonance speeds give the
stability limits of these unstable ranges of the rotor assembly.

Figure 7.14: Unbalance response of the rotor with a

uniform rectangular shaft in rigid bearings

Figure 7.15 shows the gravitational response of the rotor in figure 7.11. As can be seen from the figure,
the gravitational resonance speeds are at 670 and 1.2x104 rpm. One other thing to note about the
occurrence of the gravitational resonance speeds is that they occur at speeds approximately half of
average unbalance resonance speeds.

122
 Numerical Results

Figure 7.15: Gravitational response of the rotor with a

uniform rectangular shaft in rigid bearings

7.2.1.2 Anisotropic Visco-Elastic Bearings

Next, the rotor with the uniform rectangular shaft shown in figure 7.11 is mounted in anisotropic
visco-elastic bearings whose properties are assumed to be operation speed independent. The stiffness
and damping properties of the bearings are given in Appendix F.2. From the free vibration and
stability analysis, the eigenvalues and damping ratios of the rotor-bearing system are determined first.
The results are then compared with the previous rotor mounted in rigid bearings.

Figure 7.16 shows the CAMPBELL-diagram for the natural circular frequencies in the fixed
coordinate system while figure 7.17 shows the damping ratios of the rotor considering (undamped)
anisotropic elastic bearings and (damped) anisotropic visco-elastic bearings as well as the previous
rotor with rigid bearings. Figure 7.16 shows the identified basic eigenvalues only since the
redundant eigenvalues do not give any additional information. The eigenvector norm and the
location of the dominant mode have been used to identify the basic eigenvalues. In the figure it can
be seen that there are operation speed ranges where there is a discontinuity in the identified basic
eigenvalues. These are the operation speeds where a redundant eigenvalue approaches one basic
eigenvalue. In such case there is no clear dominance in the eigenvector norm and so there appears
such discontinuity in the identified basic eigenvalue. The other discontinuity appears where the

123
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 7.16: Natural circular frequencies of the rotor with a

uniform rectangular shaft in anisotropic visco-
elastic bearings in the fixed coordinate system

Figure 7.17: Damping ratios of the rotor with a uniform

rectangular shaft in different types of bearings

124
 Numerical Results

basic eigenvalue approaches the operation speed given in red in the figure. Here, the eigenvalue is
equal to the rotor operation speed and has a positive real part i.e. unstable operation behaviour.

On the other hand, figure 7.17 shows that the rotor assembly mounted in anisotropic elastic
bearings has additional unstable regions. For rotors on rigid and on anisotropic elastic bearings, all
the damping ratios are zero except over the unstable regions where there is at least one damping
ratio less than zero. For the rotor mounted on anisotropic visco-elastic bearings the damping ratios
are greater or equal to zero over all the stable operation speeds. Over the unstable operation speed
ranges there is again at least one negative damping ratio. In comparison to the rotor mounted on
rigid bearings, the rotor on visco-elastic bearings possess some constant positive damping ratios for
the mode shapes which are associated with larger displacements at the bearings. It can therefore be
concluded that the rotor can be further stabilized by a proper placement of damping elements in
relation to the mode shapes associated with the instabilities.

Figure 7.18: Unbalance response of the rotor with a

uniform rectangular shaft in anisotropic
visco-elastic bearings

After the investigation of the free vibrations the rotor loaded by unbalance and gravitational forces
is considered. Figure 7.18 shows the unbalance response of the rotor with a uniform rectangular shaft
at the disk. The maximum deflection is affected by the introduced changes in the bearings. The

125
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

maximum deflection associated with the first unstable region about 1540 rpm which goes to infinity
for the system in rigid bearings reduces to 41.54 mm at 1380 rpm for the rotor with the visco-
elastic bearings.

Figure 7.19: Gravitational response of the rotor with a

uniform rectangular shaft in anisotropic
visco-elastic bearings

Finally, figure 7.19 shows the gravitational response of the rotor. It was found that the amplitudes at
the resonance rotational speeds are highly reduced due to the induced damping in the bearings. The
first vertical response has been reduced from infinity in the rigid bearing case to 18.65 mm for the
visco-elastic bearing case at 670 rpm. The higher gravitational resonances are also almost non-
existent due to the damping.

7.2.2 Non-Uniform Rectangular Cross-Section

Next, the effects of the longitudinal asymmetry and the resultant non-circularity of the rotor shaft are
considered. Hence, the rectangular shaft of the rotor shown in figure 7.11 is varied as shown in
Appendix F.2. Again, the rotor mounted in rigid bearings is considered first followed by the
analysis of the rotor mounted in anisotropic visco-elastic bearings.

126
 Numerical Results

7.2.2.1 Rigid Bearings

The first extension considered is the rotor with a 2 sections asymmetric rectangular shaft. This rotor is
(longitudinally) asymmetric unlike the previously investigated rotors, but it has a zero resultant non-
circularity. The analysis of this rotor is again started with the free vibration and stability analysis from
which the CAMPBELL-diagram for the natural circular frequencies shown in figure 7.20 and the
damping ratios shown in figure 7.21 are generated.

As can be seen from figure 7.20, the natural circular frequencies of the 2 sections symmetric and
asymmetric assembly are not significantly different from each other for the most operation speeds.
This is mainly due to the fact that the two assemblies possess the same mass and average stiffness.
But a closer inspection shows that the rotor with the 2 sections asymmetric rectangular shaft does
not possess operation speed ranges with zero natural circular frequencies in the rotating coordinate
system. Hence, there are no instabilities associated with zero natural circular frequencies in this
coordinate system. Instead of this, there are a number of regions where two natural circular
frequencies possess the same value and these are the operation speed ranges where at least one
eigenvalue has a positive real part.

Figure 7.20: Natural circular frequencies of the rotor with

a 2 sections rectangular shaft in rigid
bearings in the rotating coordinate system

127
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Considering the damping ratios in figure 7.21, it can be seen that the modification of the 2 sections
asymmetric shaft has completely removed the first unstable operation speed range from 1190 to
1550 rpm. But as a tradeoff a new unstable operation speed range from 8400 to 9470 rpm is
present which actually has a larger negative damping ratio. Considering the second and third
unstable regions from 1.9x104 to 2.8x104 rpm, the rotor with the 2 sections asymmetric rectangular
shaft achieves a narrower unstable operation speed range from 2.1x104 to 2.7x104 rpm in
comparison to the one with the 2 sections symmetric rectangular shaft. The other observed
behaviour of the rotor with the 2 sections asymmetric rectangular shaft is that the unstable
operation regions are not accompanied with zero natural circular frequencies in the rotating coordinate
system as already pointed out in the previous paragraph together with the description of figure 7.20.

Figure 7.21: Damping ratios of the rotor with a 2

sections rectangular shaft in rigid bearings

Next, the forced responses of the rotor loaded by unbalance and gravitational forces are
investigated.

Figure 7.22 shows the unbalance response of the rotor with the 2 sections symmetric and asymmetric
rectangular shaft. The rotor with the 2 sections asymmetric rectangular shaft has one less resonance
speed in the lower operation speed range. In the rotating coordinate system, the unbalance resonance
speeds are associated with zero natural circular frequencies, and there are three occurrences for this

128
 Numerical Results

rotor in the considered operation speed range. One other difference between the rotors with the 2
sections symmetric and asymmetric rectangular shaft is that the unbalance resonance speeds are the
stability operation speed limits in the 2 sections symmetric rotor but not in the 2 sections
asymmetric rotor. The first unbalance resonance speed of the rotor with the 2 sections asymmetric
shaft is not associated with instability and the second unstable operation speed range lies between
the second and the third unbalance resonance speed at 2.1x104 rpm and 2.7x104 rpm, respectively.

Figure 7.22: Unbalance response of the rotor with a 2

sections rectangular shaft in rigid bearings

Figure 7.23 shows the gravitational response of the two preceding rotor-bearing assemblies. For the
considered operation speed range, both rotors have two gravitational resonance speeds. The rotor
with the 2 sections symmetric rectangular shaft has gravitational resonances at 670 rpm and at
1.2x104 rpm. On the other hand, the rotor with the 2 sections asymmetric rectangular shaft has
gravitational resonances at 7590 rpm and 2.9x104 rpm.

Now, as a first further refinement, the number of sections of the non-uniform rectangular shaft was
increased considering both longitudinally symmetric as well as asymmetric rectangular shafts. Then,
a second refinement is discussed where two shafts with different resultant non-circularities are
considered.

129
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 7.23: Gravitational response of the rotor with a 2

sections rectangular shaft in rigid bearings

Unlike the longitudinal symmetry, which was simply qualitatively described, the resultant non-circularity is
defined by

n
µl ∑µ i
µR = ∑ i i ; µ= i =1
(7.1)
i =1 µ L n

where µi is the non-circularity of the i th section (equation 4.5), li is the length of the i th section, µ
is the average non-circularity and n is the total number of rotor shaft sections. The resultant non-
circularity µ R is the measure of the distribution of the non-circularities of all the shaft sections as
viewed in the η,ζ-plane and ranges from -1 and 1. When all the minor or major axes of the shaft
cross-sections are parallel to each other (a special case being a single shaft section as in the previous
sub-unit), then the resultant non-circularity is equal to 1 or -1. But if the shaft sections are oriented
in such a way that the sum of the second area moments of the cross-sections in the η- and ζ-axis
times the section lengths are equal then the resultant non-circularity is zero.

The increase of the number of sections gives the chance to study the effect of the longitudinal
symmetry and asymmetry of the rectangular rotor shaft. For this further refinement only the free

130
 Numerical Results

vibration and stability analysis is investigated from which the damping ratios in the figures 7.24 and
7.25 are generated.

The stability analysis shows that the rotors with (longitudinally) symmetric rectangular shafts possess
similar properties with each other and those which are asymmetric also have similar properties with
each other. But there is a difference in the properties between the symmetric and asymmetric
rectangular shaft rotors. Figure 7.24 shows rotors with symmetric rectangular shafts and a different
number of sections. In general, the stability of the rotor improves as the number of sections
increases. This phenomenon can be clearly seen in the first and third unstable operation speed
range around 1340 rpm and from 1.9x104 to 2.8x104 rpm. But considering the second unstable
operation speed range around 1.3x104 rpm shows that there are certain unstable operation speed
ranges which are differently affected by the number of sections. Here the rotor with the 2 sections
symmetric shaft is completely stable whereas the rotor with the 6 sections symmetric shaft has the
maximum instability.

Next, figure 7.25 shows rotors with asymmetric rectangular shafts and a different number of sections,
too. Again, a general improvement of the rotor stability is observed as the number of sections
increases. This phenomenon can be clearly seen in the first unstable operation speed range around
8850 rpm. But again considering the second unstable operation speed range around 1.3x104 shows
that there are certain unstable operation speed ranges which are less affected by the number of
sections. Here, the rotor with 4 sections is more stable than the rotor with 12 sections.

The direct comparison between the figures 7.24 and 7.25 shows that the unstable operation speed
ranges for the rotors with symmetric and asymmetric rectangular shafts are different from each
other. For the considered rotors and operation speed ranges the asymmetric rotors are relatively
more stable than the symmetric ones.

Finally, from figure 7.24 it can be seen that the rotors with 2 and 6 sections symmetric rectangular
shafts have the largest instabilities over the unstable operation speed ranges. This may be attributed
to the resultant non-circularities which for the 2 sections symmetric rotor is 1 while for the 6
sections symmetric rotor it is 1/3 but 0 for the other symmetric cases. Hence, a further refinement
is considered next where rotors with 2 sections asymmetric and 4 sections symmetric rectangular
shafts with different lengths of the sections A and B shaft segments and therefore different
resultant non-circularities are compared with each other. For both cases, the related shaft
dimensions used in this analysis are again given in Appendix F.2. For all variations of the

131
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 7.24: Damping ratios of the rotors with symmetric

rectangular shafts in rigid bearings

Figure 7.25: Damping ratios of the rotors with asymmetric

rectangular shafts in rigid bearings

132
 Numerical Results

rectangular shaft segments the disk remains in the central position of the shaft between the two
bearings.

First, figure 7.26 shows the damping ratios of the rotors with 2 sections asymmetric rectangular shafts and
different resultant non-circularities. The resultant non-circularities selected for the analysis are
chosen in such a way that the shaft cross-section discontinuities lie on the node of the vibration
mode shape. From the figure it is evident that the resultant non-circularity affects the stability of the
different vibration mode shapes. Considering the zero resultant non-circularity, one can see that the
first unstable operation region around 1350 rpm which is associated with the first mode shape
disappears. But an unstable operation speed range with a different mode shape appears around
8850 rpm. The effect of having two cross sections is completely removed by the presence of the
node at the point of cross-section discontinuity. The other difference that can be seen is that the
unstable operation speed is associated with two different mode shapes, i.e. during this operation
phase the mode shape along one axis is different from the mode shape in the other direction unlike
the rotor considered in section 7.2.1.1 in the rotating coordinate system. For the unstable region
around 8850 rpm the mode shapes during unstable operation speed are the first and the third while
for the unstable region around 1.3x104 rpm the associated mode shapes are the third and fifth
mode shapes in the η- and ζ-direction. The figure therefore shows how different mode shapes
become unstable over different operation speed ranges depending on the total non-circularity of
the shaft sections.

Considering now the rotors with 4 sections symmetric rectangular shafts, only three resultant non-
circularities are investigated because they suffice to explain the behaviour of the rotor in the
considered operation speed range. Figure 7.27 shows that a change in the cross-sections around the
location of the disk helps in stabilizing the rotor where the associated unstable vibration mode is
the first mode shape. This can be seen from the stability values around 1340 rpm. For the unstable
region around 1.3x104 rpm where the fifth mode shape is involved, it can be seen that the rotor
with a resultant non-circularity of 0.33 has the worst stability behaviour. Here again, the reason lies
in the fact that the effect of cross-section change has been removed by the location of the mode
shape nodes. Considering the unstable regions around 2.2x104 rpm and 2.5x104 rpm where the
third mode shape is involved, the rotor with zero resultant non-circularity has the best stability
behaviour since the sub-resultant non-circularity in every wavelength of this vibration mode shape
is also zero. Hence, in the stability analysis of a rotor with non-uniform sections it is difficult to
draw a single conclusion for the complete system.

133
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 7.26: Damping ratios of the rotors with 2 sections

asymmetric rectangular shafts in rigid bearings

Figure 7.27: Damping ratios of the rotors with 4 sections

symmetric rectangular shafts in rigid bearings

134
 Numerical Results

The resultant non-circularity defined in equation (7.1) works for a shaft where the principal axes of
the shaft cross-section segments are perpendicular to each other. In general, if this is not the case,
the resultant non-circularity of any non-uniform rotor shaft can be determined from

n
µi li χ i 1
µR = ∑ ; χi = − βi + 1 (7.2)
i =1 µL 45

where βi is the rotation of the minor axis of the i th section from the ζ-axis at x = 0 in degrees
which has values between -900 and 900.

7.2.2.2 Anisotropic Visco-Elastic Bearings

Before concluding this sub-unit, the rotor with a 2 sections asymmetric rectangular shaft is placed in
anisotropic visco-elastic bearings whose properties are given in Appendix F.2, to consider the combined
effect of damping, longitudinal asymmetry and resultant non-circularity. Again, only the free vibration and
stability analysis are investigated. From this analysis the eigenvalues and damping ratios of the rotor-
bearing system are determined. Figure 7.28 shows the CAMPBELL-diagram for the natural circular
frequencies in the fixed coordinate system of this rotor mounted in rigid and anisotropic visco-elastic

Figure 7.28: Natural circular frequencies of the rotor with a 2

sections asymmetric rectangular shaft in different
types of bearings in the fixed coordinate system

135
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

bearings. Here again, only the identified basic eigenvalues are presented in the figure. Similar to
figure 7.16 there are discontinuities in the identified basic eigenvalues at operation speeds where
these values lie close to a redundant eigenvalue.

Figure 7.29: Damping ratios of the rotor with a 2 sections

asymmetric rectangular shaft in different types of
bearings

On the other hand, figure 7.29 shows the damping ratios of the rotor with a 2 sections asymmetric
rectangular shaft in rigid and visco-elastic bearings. Both rotors have two unstable operation speed
ranges in the investigated operation speed. The rotor in the visco-elastic bearings has larger stable
operation speed range between the two unstable regions. The reduced first unstable operation
speed may be as a result of the reduced circular natural frequency due to the additional flexibility
from the bearings. From the figures 7.28 and 7.29 it may be seen that the second unstable region of
the rotor in rigid bearings is associated with circular natural frequency significantly different from
the rotor’s rotational speed, while the second unstable region of the rotor in visco-elastic bearings is
associated with a circular natural frequency equal to the rotor rotational speed.

136
 Numerical Results

7.3 Two Disk Rotor with Non-Circular Shaft and Anisotropic Elastic Bearings

In this sub-unit, the rotor assembly analysed by Forrai L. [27] is investigated. The rotor has a
massless non-circular shaft with two disks mounted in anisotropic elastic bearings. The rotor dimensions are
set such that the mean shaft and bearing stiffnesses agree with the data given in the paper. The
rotor is analysed using FEARS implementing the EULER-BERNOULLI beam theory. The data of
the analysed rotor shown in figure 7.30 are given in Appendix F.3.

Figure 7.30: Two disk rotor with a non-circular shaft and

anisotropic elastic bearings

Since the only output results in [27] are the analytical stability plots, here only the free vibration and
stability analysis are investigated. From the free vibration analysis the CAMPBELL-diagram for the
natural circular frequencies is shown in figure 7.31 and from the stability analysis the damping ratios
shown in figure 7.32 are generated.

As can be seen from the figures 7.31 and 7.32, the two disk rotor with the selected dimensions has
six unstable operation speed ranges which are at 339-343 rpm, 364-370 rpm, 389-396 rpm, 1124-
1134 rpm, 1269-1283 rpm and 1415-1434 rpm. As can be seen from figure 7.31, over the 1st, 3rd, 4th
and 6th unstable operation speeds there are two natural frequencies that are equal to the rotational
speed of the rotor. These two natural frequencies possess opposing sign real part which means at
least one positive real part over this speed ranges leading to instability. But over the 2nd and 5th
unstable operation speed there is no natural frequency equal to the operation speed. Instead, two
natural frequencies have the same imaginary part and opposite sign real parts leading to instabilities.
The unstable operation speed ranges are associated with non-zero damping ratio pair of opposing
sign as can be seen figure 7.32.

137
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Figure 7.31: Natural circular frequencies of the two disk

rotor in the fixed coordinate system

Figure 7.32: Damping ratios of the two disk rotor in the

fixed coordinate system

138
 Numerical Results

Additionally, figure 7.33 shows a general stability plot of the two disk rotor. Here the shaft non-
circularity and the bearing anisotropy are varied maintaining the mean shaft and bearing stiffness
constant as discussed in chapter 4.1.2.

Figure 7.33: Stability plot of the two disk rotor with a

non-circular shaft and anisotropic elastic
bearings

The results in figure 7.33 are in general agreement with the findings of Forrai [27]. There are two
unstable regions for the rotor mounted in isotropic bearings. As the bearing anisotropy increases,
each unstable operation region splits into three sections. These three unstable operation speed
ranges overlap with each other for high shaft non-circularities and low bearing anisotropies. As the
bearing anisotropy is further increased, each unstable operation speed range becomes narrower and,
hence, the three unstable regions separate completely from each other. The further increase in
bearing anisotropy to 1 results in the complete disappearance of the 1st, 2nd, 4th and 5th unstable
region. The results by Forrai are relatively conservative in comparison to results given in figure 7.33.

In this chapter, three rotor examples were investigated using the developed MATLAB-FEM
program using the element matrices discussed in chapter 6. The next and final chapter presents the
conclusion and the future outlook of this dissertation.

139
 8 Chapter 8

CONCLUSION AND FUTURE OUTLOOK

8.1 Conclusion

As stated in the first chapter, this dissertation has the motivation of investigating the unstable
behaviour of circularly asymmetric rotors. In the objectives, the development of a Finite Element
Method computer code was planned for the analysis.

A LAVAL-rotor with a non-circular shaft and rigid bearings was analytically investigated in the rotating
coordinate system. The model is simple enough to generate a closed form solution in this coordinate
system. Besides the analytical solution, the effectiveness of the HILL’s method was tested by
analysing the same rotor in the fixed coordinate system. Further, the rotor was constructed and the
responses were experimentally observed. The actual behaviour was seen to conform to the
mathematically determined behaviour with only minor differences which are expected due to the
simplicity of the mathematical model used.

The bearings of the above rotor were then complicated to be anisotropic elastic bearings. This rotor
does not have a closed form solution, and hence it was numerically solved using the HILL’s method.
Besides the numerical investigation, the bearing was constructed as well and the rotor was also
experimentally investigated. Here again, the results showed good agreement.

With this confidence in the HILL’s method as a good approximation method, an FEM program was
written using the MATLAB programming language. The developed program FEARS is an
extension of a previously developed FEM program called FEAR designed for the lateral vibration
investigation of circularly symmetric rotor-bearing systems. The present program has also the
capacity to operate in the fixed and rotating coordinate system besides the analysis of the unbalance
and gravitational force response.

During the course of the dissertation a user friendly computer code was generated which can
analyse the lateral vibration of a rotor using any of the three most important beam theories. The
generated FEM code can analyse both circularly symmetric and asymmetric shaft rotors with any
number of circularly symmetric and asymmetric disks mounted on stiff, fluid film and anisotropic
elastic bearings. Rotors with circularly asymmetric disks or shafts are analysed in the rotating
coordinate system by the program if the bearings are stiff. For all other cases, the fixed coordinate system is
used for the analysis.
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

The FEM computer code analyses the free and forced vibrations of the rotor. From the free vibrations,
the CAMPBELL-diagram for the eigenvalues (natural circular frequencies), the eigenvectors (mode
shapes), the damping ratios (the stability indicator), and the frequency response functions (FRFs) are
generated. From the information of the eigenvalues and eigenvectors, the rotor animation is also
displayed on the output GUI. Besides the free vibrations, the program can analyse the response due
to constant forces, and unbalance forces. For circularly asymmetric rotors, the shaft orientation, i.e.
vertical rotor or horizontal rotor, is analysed, for gravitational force leads to additional resonance
rotational speeds in such rotors.

Using the developed FEM computer code, the above stated LAVAL-rotor and two other rotors were
investigated. From the analysis of the LAVAL-rotor, it was seen that the FEM computer code
which implements the HILL’s method gives good accuracy. The gravitational and unbalance
resonances as well as the unstable operation speed ranges have been predicted with good level of
accuracy in comparison to the experimental results.

Besides the LAVAL-rotor, a rectangular shaft rotor and a two disk rotor were investigated. The basic
purpose of the rectangular shaft rotor was to investigate the effect of longitudinal asymmetry and resultant
non-circularity. From this investigation it was seen that distributing the asymmetry over the length of
the shaft has a positive effect on the stability by lowering the magnitude of negative damping of the
lower operation speed instabilities. But it has the opposite effect on higher operation speed
instabilities. In general, the unstable operation speed range gets slightly narrower as the asymmetry
is distributed over the length of the shaft. The analysis of this rotor further showed that a circularly
asymmetric and longitudinally symmetric rotor has different properties from a circularly and
longitudinally asymmetric rotor. It is difficult to talk generally of stability improvement, for they
possess some unstable regions over different operation speed ranges. On the other hand, lowering
the resultant asymmetry is seen to have a positive effect on the rotor stability, especially in the low
operating speed ranges. The two disk rotor was also used to show the accuracy of the method and the
program in determining the stability of the rotor by comparing the results with an analytically
investigated rotor.

All rotor data are given in detail in Appendix F so that other researchers can use these examples to
compare their solution methods or codes with the results presented here.

142
 Conclusion and Future Outlook

8.2 Future Outlook

Industrial rotors are usually more complicated than the numerically and experimentally analysed rotor
assemblies in this dissertation. The construction and investigation of a relatively massive rotor (with
one or more disks) is essential to show the viability of the solution method and the developed FEM
program. Such an experimental set-up can then be used to investigate the effect of shear deformation
on the stability, eigenvalues and forced responses. The construction of such a test rotor also assists
in generating better quality experimental results which can then be used in the study of experimental
modal analysis and model updating of general rotor-bearing systems.

The experimental and numerical investigation of practical general rotors, for example the shaft of an
industrial 2 pole generator or motor, can then be the next step. Such shafts have complicated
profiles and, hence, also have complicated dynamic properties. The complete dynamic investigation
and verification of such a complicated general rotor bearing system would then be the final aim.

143
 A Appendix A

TRANSFORMATION MATRICES BETWEEN THE FIXED AND THE

ROTATING COORDINATE SYSTEM

To determine the transformation matrix between the fixed and the rotating coordinate system, consider the
figure shown below.

Figure A.1: Fixed and rotating coordinate system

The position of any point on the Cartesian plane can be expressed in terms of either the y,z- or η,ζ-
coordinate system. For example, point W can be expressed in the two coordinate systems as being
located at (yW,zW) and (ηW,ζW) in the y,z- and η,ζ-coordinate system, respectively. Hence

yW = ηW cos α + ζ W sin α
; α =ϕ −β ; ϕ = Ωt . (a1)
zW = −ηW sin α + ζ W cos α

Neglecting the subscript W for simplification, this can be written in matrix form as

 y   cos α sin α  η 


 =   ⇒ w = Tv . (a2)
 z   − sin α cos α  ζ 

To transform from the η,ζ-coordinate system to the y,z-coordinate system, premultiply both sides
by the inverse (equal to the transpose) of the transformation matrix. Hence, it becomes

cos α − sin α 
v = T −1w ; T−1 = TT =  . (a3)
 sin α cos α 
 B Appendix B

TRANSFORMATION OF THE EQUATIONS OF MOTION

Consider the equation of motion of a LAVAL-Rotor with a circular shaft and (rigid or) anisotropic elastic
bearings including a viscous damping located on the disk. In this case, the equation of motion in the
fixed coordinate system is

ɺɺ
+ Dw

ɺ


Mw + Kw = F 38. (b1)

To express the equation of motion given above in the rotating coordinate system substitute equation
(a2) and its derivatives39 into equation (b1)

ɺɺ v
+ 2T
M T ( ɺ v
ɺ + Tvɺɺ
+ D Tɺ v
+ Tv
ɺ + KTv
= F .
) ( ) (b2)

Premultiplying equation (b2) with the transpose of the transformation matrix gives

ɺɺ v
+ 2TT MT
TT MT ɺ v
ɺ + TT MTvɺɺ
+ TT DTɺ v
+ TT DTv
ɺ + TT KTv
= TT F . (b3)

The first and second time derivatives of the transformation matrix are given by

ɺ = −Ω  sin α
T
− cos α  ɺɺ = −Ω 2  cos α
; T
sin α 
= −Ω 2 T (b4)
cos α sin α   − sin α 
cos α 
 

where α = ϕ − β and ϕ = Ωt .

Substituting the equations (b4) into equation (b3) gives

38 m 0 d 0 k 0 


0   sin ϕ  .
M=  , D= y  , K = K bs =  ybs  , F = FG + FU = mg   + mΩ 2ε  
0 m 0 dz   0 k zbs  1  cos ϕ 



ɺ ɺ


ɺɺ ɺɺ



39
w = Tv ; w = Tv + Tvɺ ; w = Tv + 2Tɺ vɺ + Tvɺɺ .
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

1 0 
 0 1 
ɺ 1 0  ɺɺ

− mΩ 2   v + 2mΩ   v +m v
0 1   −1 0  0 1 
 d y sin α cos α − d z sin α cos α − d y cos 2 α − d z sin 2 α 

−Ω v
 d y sin 2 α + d z cos 2 α − d y sin α cos α + d z sin α cos α 
 d y cos 2 α + d z sin 2 α d y sin α cos α − d z sin α cos α 
ɺ
+ v
 d y sin α cos α − d z sin α cos α d y sin 2 α + d z cos 2 α 
 k ybs cos 2 α + k zbs sin 2 α k ybs sin α cos α − k zbs sin α cos α 

+ v
 k ybs sin α cos α − k zbs sin α cos α k ybs sin 2 α + k zbs cos 2 α 
− sin α  2 sin ϕ cos α − cos ϕ sin α 
= mg   + mΩ ε   . (b5)
 cos α  sin ϕ sin α + cos ϕ cos α 

Using trigonometric identities40,41 and introducing the mean and deviatoric damping and stiffness

d y + dz dy − dz k ybs + k zbs k ybs − k zbs

d0 = ; ∆d = ; k0bs = ; ∆kbs = (b6)
2 2 2 2

equation (b5) can be reduced to

1 0 
 0 1 
ɺ 1 0  ɺɺ

− mΩ 2   v + 2mΩ   v +m v
0 1   −1 0  0 1 
 ∆d sin 2α − d 0 − ∆d cos 2α 
 d 0 + ∆d cos 2α ∆d sin 2α 
ɺ
−Ω  v+ v
 d 0 − ∆d cos 2α −∆d sin 2α   ∆d sin 2α d 0 − ∆d cos 2α 
 k + ∆kbs cos 2α ∆kbs sin 2α 
 − sin α  2  sin β 
+  0bs  v = mg   + mΩ ε  
 ∆kbs sin 2α k0bs − ∆kbs cos 2α   cos α  cos β 

⇒ k m v + gvɺ + mvɺɺ + kɶ d v + dɶ vɺ + kɶ bs v = f (b7)

Collecting the similar terms will lead to an equation of the form

( ) ( )
mvɺɺ + dɶ + g vɺ + k m + kɶ d + kɶ bs v = f 42 (b8)

1 1 1
40
sin α cos α = sin 2α ; cos 2 α = (1 + cos 2α ) ; sin 2 α = (1 − cos 2α ) .
2 2 2
41
sin (ϕ ± α ) = sin ϕ cos α ± cos ϕ sin α ; cos (ϕ ± α ) = cos ϕ cos α ∓ sin ϕ sin α .
42 Note that if the stiffness and damping properties of the system are symmetrical, then the deviatoric values will reduce
to zero leaving the mean values. In this case, all the matrices are time independent.

148
 C Appendix C

STIFFNESS MATRIX OF A LAVAL-ROTOR WITH A NON-CIRCULAR SHAFT

AND ANISOTROPIC ELASTIC BEARINGS

Considering a LAVAL-rotor with a non-circular shaft and anisotropic elastic bearings in the fixed coordinate
system, one can identify that the bearing stiffnesses are in series with the time dependent shaft
stiffness as seen in equation (3.16). Further, the shaft stiffness is time dependent due to the shaft
non-circularity.

From equation (3.17) the resultant time dependent stiffness matrix is given by

ɶ = 1  K11
K
K12 
(c1)
K  K 21 K 22 
bs

where K = k02s + 2 ( k yb + k zb ) k0 s − ∆k s2 + 4k yb k zb − 2 ( k yb − k zb ) ∆ks cos 2α

(
K11 = 2k yb 2k zb k0 s + k02s − ∆ks2 + 2k zb ∆ks cos 2α )
K12 = K 21 = −4k yb k zb ∆ks sin 2α

(
K 22 = 2k zb 2k yb k0 s + k02s − ∆ks2 − 2k yb ∆k s cos 2α . )
From this general stiffness matrix all particular cases considered in chapter 3 can be derived as
follows.

§ Circular Shaft and Rigid Bearings:

∆ks = 0 ; k0 s = ks ; k yb = k zb = ∞ (c2)

Substituting equation (c2) into equation (c1) reduces it to the stiffness matrix in equation (3.5)

ɶ = K =  ks
K
0
(c3)
bs s 0 ks 

Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

§ Circular Shaft and Anisotropic Elastic Bearings:

∆k s = 0 ; k 0 s = k s (c4)

In this case, the terms in equation (c1) reduce to

K = k s2 + 2 ( k yb + k zb ) ks + 4k yb k zb = ( 2k yb + ks ) ( 2k zb + ks )

( )
K11 = 2k yb 2k zb k s + ks2 = 2k yb k s ( 2k zb + k s )
(c5)
K12 = K 21 = 0
K 22 = 2k zb 2k yb k s + ks2 = 2k zb ks ( 2k yb + k s )
( )
Substituting the equations (c5) into equation (c1) gives the stiffness matrix in equation (3.11) with
the elements given in equation (3.10)

ɶ = K =  k ybs
K
0 
(c6)
bs bs  0 k zbs 


§ Non-Circular Shaft and Rigid Bearings:

k yb = k zb = ∞ (c7)

Substituting equation (c7) into the terms of equation (c1) gives the stiffness matrix in equation
(3.14)

ɶ =K
K ɶ =  k0 s + ∆k s cos 2α −∆k s sin 2α 
(c8)
bs s  −∆k sin 2α k0 s − ∆ks cos 2α 
 s

Following the same procedure, the stiffness matrices of the particular cases considered in chapter 3
in the rotating coordinate system can be determined from equation (3.21).

150
 D Appendix D

LINEARIZATION OF THE STIFFNESS MATRIX OF A LAVAL-ROTOR WITH A

NON-CIRCULAR SHAFT AND ANISOTROPIC ELASTIC BEARINGS WITH
RESPECT TO THE DEVIATORIC STIFFNESSES

From equation (3.16), the total flexibility matrix of a LAVAL-rotor with a non-circular shaft and
anisotropic elastic bearings in the fixed coordinate system is given by

−1 −1
( Kɶ )
bs
−1
ɶ
= ( 2K b ) + K( )
s (d1)

The shaft stiffness matrix as given in equation (3.14) can be written as

ɶ = k I + ∆k Iɶ = k  I + ∆ks
K

Iɶ  (d2)
s 0s s 0s
 k0 s 

1 0  cos 2α − sin 2α 
where I =   ; Iɶ =  .
0 1   − sin 2α − cos 2α 

∆k s
For << 1 , the inverse of the shaft stiffness matrix (d2) can be approximated in a linear form as
k0 s

−1
−1  ∆k   ∆k 
( Kɶ )
s
−1
= k I + s
0s
k0 s
Iɶ  ≈ k0−s1  I − s
k0 s
Iɶ  (d3)
   

Hence, from equation (d1) it follows

−1  ∆k  ∆k
( Kɶ )
bs
−1
≈ ( 2K b ) + k0−s1  I − s
k0 s
−1
Iɶ  = ( 2K b ) + k0−s1I − 2 s Iɶ
k0 s
(d4)
 

Introducing the matrices from equation (3.15) and (d2) yields

 2k yb + k0 s 
 0 
−1
 2k yb k0 s  − ∆k s  cos 2α − sin 2α 
( Kɶ )
bs =
 2k zb + k0 s  k0 s
2  − sin 2α − cos 2α 
= A−B (d5)

 0 
 2k zb k0 s 
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Further inverting and linearizing gives

−1 −1
ɶ = ( A − B )−1 =  A I − A −1B 
K ( ) (
= I − A −1B ) ( )
A −1 ≈ I + A −1B A −1 = A −1 + A −1BA −1 (d6)
bs  

Substituting the matrices A and B from equation (d5) into equation (d6) yields

 k02ybs cos 2α −k0 ybs k0 zbs sin 2α 

ɶ =  k0 ybs
K
0  ∆k s
+   (d7)
bs  0 k0 zbs  k02s − k02zbs cos 2α 
  − k0 ybs k0 zbs sin 2α

2k yb k0 s 2k zb k0 s
where k0 ybs = ; k0 zbs = .
2k yb + k0 s 2k zb + k0 s

For k0 s << 2k yb , 2k zb (i.e. the bearings are much stiffer than the shaft), the equivalent mean

stiffnesses are

2k yb k0 s 2k zb k0 s
k0 ybs = ≈ k0 s ; k0 zbs = ≈ k0 s (d8)
2k yb + k0 s 2k zb + k0 s

and therefore the second term in equation (d7) can be simplified to give

ɶ =  k0 ybs
K
0   cos 2α
+ ∆k s 
− sin 2α 
(d9)
bs  0  − cos 2α 
 k0 zbs   − sin 2α

which is now of linear form with respect to the deviatoric shaft stiffness.

In the rotating coordinate system, the linearization starts with equation (3.19) given as

−1 −1
( kɶ )
bs (
= 2kɶ b ) + (k ) s
−1
(d10)

The bearing stiffness matrix as given in equation (3.20) may be written as

 ∆k ɶi 
kɶ b = k0b i + ∆kb ɶi = k0b  i + b (d11)
 k0b 

1 0  ɶi = cos 2α sin 2α 
where i =   ;  sin 2α
0 1   − cos 2α 

152
Linearization of the Stiffness Matrix of a LAVAL-Rotor with a Non-Circular Shaft and Anisotropic Elastic …

∆k b
For << 1 , the inverse of the shaft stiffness matrix (d11) can be expressed in a linear form as
k0 b

−1
−1  ∆k ɶi  ≈ k −1  i − ∆kb ɶi 
( )
kɶ b −1
= k i + b
0b
k0b
0b
k0b
(d12)
   

Hence, from equation (d10) it follows

−1 k0−b1  ∆kb ɶi  + ( k ) −1 = k0b i − ∆kb ɶi + ( k ) −1

−1

( )
kɶ bs ≈
2 
i −
k0 b
s
2 2k02b
s (d13)


Introducing the matrices from equation (3.13) and (d11) yields

 2k0b + kη s 
 0 
−1
=
2k0b kη s  − ∆kb cos 2α sin 2α 
( kɶ )
bs  2k0b + kζ s  2k02b  sin 2α − cos 2α 
= a−b (d14)
 0 
 2k0b kζ s 

Further inverting and linearizing gives

−1 −1
kɶ bs = ( a − b ) = a i − a −1b 
( ) ( ) ( )
−1
= i − a −1b a−1 ≈ i + a −1b a −1 = a −1 + a −1ba −1
i (d15)

Substituting the matrices a and b from equation (d14) into equation (d15) yields

 k0ηbs 0  ∆k b  k02ηbs cos 2α k0ηbs k0ζ bs sin 2α 

kɶ bs =  +  
k0ζ bs  2k02b
(d16)
 0  k0ηbs k0ζ bs sin 2α − k02ζ bs cos 2α 

2k0b kη s 2k0b kζ s
where k0ηbs = ; k0ζ bs = .
2k0b + kη s 2k0b + kζ s

For 2k0b << kη s , kζ s (i.e. the shaft is much stiffer than the bearings), the equivalent mean

stiffnesses are

2k0b kη s 2k0b kζ s
k0ηbs = ≈ 2k 0 b ; k0ζ bs = ≈ 2 k0 b (d17)
2k0b + kη s 2k0b + kζ s

153
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

and therefore the second term in equation (d16) can be simplified to give

 k0ηbs 0  cos 2α sin 2α 

kɶ bs =   + 2∆kb  (d18)
 0 k0ζ bs   sin 2α − cos 2α 

which is now of linear form with respect to the deviatoric bearing stiffness.

154
 E Appendix E

DERIVATION OF THE HILL’S EQUATIONS FOR THE EIGENVALUES,

EIGENVECTORS AND THE FORCED RESPONSES

E.1 Eigenvalues and Eigenvectors

Consider a general homogeneous second order differential equation with time dependent
coefficients given as

ɶ qɺɺ
+ C
M ɶ q
ɺ + K
ɶ q
= 0 (e1)

where M ɶ and K
ɶ , C ɶ are the time dependent matrices of the mass, velocity- and displacement-
proportional coefficients in either the fixed or rotating coordinate system. The elements in the matrices
depend on the rotor physical properties and the selected analysis method. Hence, the
corresponding elements are defined in the text where the detailed equations are given in the
chapters 3 and 6 or in the related appendices.

Assume that all the time dependent matrices can be written as the sum of a constant mean matrix
and two deviatoric harmonic matrices as

ɶ = A ( t ) = ∆A e − jΩɶ t + A + ∆A e jΩɶ t ; A
A ɶ →M ɶ , or K
ɶ ,C ɶ ; Ω
ɶ = 2Ω (e2)
-1 0 +1

To solve the homogeneous equation (e1) for the eigenvalues and eigenvectors a periodic solution of the
form

ɶ
ɶ
ɶ
ɶ
q = Φ e Λt ; (
Φ = Φ ( t ) = Φ t + Tɶ ) ; ɶ Tɶ = 2π
Ω (e3)

is assumed. Deriving the first relation of equation (e3) once and twice with respect to time

ɶ Λt

ɺɶ Λt
ɶ Λt
ɺɺ
ɶ Λt
ɺɶ Λt
ɶ Λt

q =Φ e ; qɺ = Φ e + ΛΦ e ; qɺɺ = Φ e + 2ΛΦ e + Λ2Φ e , (e4)

inserting in equation (e1) and dividing by e Λt yields

 ɺɺ
ɶ
ɺɶ
ɶ 
( ∆M −1
ɶ ɶ


)
e − jΩt + M 0 + ∆M +1e jΩt  Φ + 2ΛΦ + Λ 2Φ 


ɺɶ
ɶ 
ɶ

( ɶ ɶ
+ ∆C−1e− jΩt + C0 + ∆C+1e jΩt  Φ

) + ΛΦ (
 + ∆K −1e

ɶt
− jΩ
)
ɶ
+ K 0 + ∆K +1e jΩt Φ = 0. (e5)
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

ɶ
Since the solution is assumed to be periodic the vector Φ and its derivatives can be written in terms
of a FOURIER-series as

ɶ ∞
ˆ
ɺɶ ∞
ɺɺ
ɶ ∞
ˆ juΩɶ t

Φ= ∑
ɶ
Φ u e juΩt ; Φ = ∑ ɶ ˆ e juΩɶ t
juΩΦ ; Φ = − ∑ u 2Ω
ɶ 2Φ . (e6)
u ue
u =−∞ u =−∞ u =−∞

Substituting the equations (e6) into equation (e5) yields

∞
ˆ juΩɶ t
( ɶ
∆M −1e− jΩt + M 0 + ∆M +1e jΩt
ɶ
) ∑ ( −u Ωɶ
u =−∞
2 2 ɶ + Λ2 Φ
+ 2ΛjuΩ ue ) ɶ
( ɶ
+ ∆C−1e − jΩt + C0 + ∆C+1e jΩt )
∞
ˆ juΩɶ t ∞


⋅ ∑(
u =−∞
ɶ +Λ Φ
juΩ ue ) ɶ
( ɶ
+ ∆K −1e− jΩt + K 0 + ∆K +1e jΩt ) ∑ Φˆ e
u =−∞
u
ɶt
juΩ
= 0.

(e7)

For the above mathematical relation to be fulfilled, the terms having the same exponential terms
must be equal on both sides of the equal sign i.e. zero in this case. From this condition the terms
for the first few exponents follow as

⋮ ⋮ ⋮


ˆ + M −4Ω

ˆ + ∆M −9Ω
ˆ +
ɶ
e−2 jΩt : ∆M −1 (
−Ω ɶ 2 − j 2ΛΩ ɶ + Λ2 Φ
−1 )
0
ɶ 2 − j 4ΛΩɶ + Λ2 Φ (−2 +1
ɶ 2 − j 6ΛΩ ɶ + Λ2 Φ ) −3 ( )

ˆ
ˆ
ˆ
ˆ
ˆ
ˆ
∆C−1 (− jΩ )
ɶ + Λ Φ + C − j 2Ω
−1 0 (
ɶ + Λ Φ + ∆C − j 3Ω
−2 +1 ) −3 −1 −1 (
ɶ + Λ Φ + ∆K Φ + K Φ + ∆K Φ = 0
0 −2 +1 −3 )

ˆ
ˆ 2 ˆ


ɶ
e− jΩt : ∆M −1 Λ 2 Φ 0 + M 0 −Ω ɶ2− ( j 2ΛΩ ɶ + Λ2 Φ ) ɶ2 ɶ
−1 + ∆M +1 −4Ω − j 4ΛΩ + Λ ( Φ −2 + )

ˆ
ˆ
ˆ
ˆ
ˆ
ˆ
∆C−1 ΛΦ ɶ
0 + C0 − j Ω + Λ ( ) Φ ɶ ( )
−1 + ∆C +1 − j 2Ω + Λ Φ −2 + ∆K −1 Φ 0 + K 0 Φ −1 + ∆K +1Φ −2 =0

ˆ

2 ˆ


2 ˆ
e j 0 : ∆M −1 −Ω (
ɶ 2 + j 2ΛΩ ɶ + Λ2 Φ ) ɶ2 ɶ
1 + M 0 Λ Φ 0 + ∆M +1 −Ω − j 2ΛΩ + Λ Φ −1 + ( )

ˆ
ˆ
ˆ
ˆ
ˆ
ˆ
∆C−1 jΩɶ +Λ Φ ( ) ɶ ( )
1 + C0 ΛΦ 0 + ∆C+1 − j Ω + Λ Φ −1 + ∆K −1 Φ1 + K 0 Φ 0 + ∆K +1Φ −1 =0

ˆ 2 ˆ


ˆ
e j Ωt :
ɶ
(
∆M −1 −4Ωɶ 2 + j 4ΛΩ ɶ + Λ2 Φ )
ɶ2 ɶ
2 + M 0 −Ω + j 2ΛΩ + Λ (
Φ1 + ∆M +1 Λ 2 Φ 0 + )

ˆ
ˆ
ˆ
ˆ
ˆ
ˆ
(ɶ + Λ Φ + C jΩ
∆C−1 j 2Ω 2 ) 0 ( 1 +1 0 )
ɶ + Λ Φ + ∆C ΛΦ + ∆K Φ + K Φ + ∆K Φ
−1 2 0 1 +1 0 =0

ˆ

2 ˆ

ˆ
ɶ
e 2 jΩt : ∆M −1 −9Ω (
ɶ 2 + j 6ΛΩ ɶ + Λ2 Φ ɶ2 ) ɶ
3 + M 0 −4Ω + j 4 ΛΩ + Λ ( ɶ 2 + j 2ΛΩ
Φ 2 + ∆M +1 −Ω ɶ + Λ2 Φ
1+ ) ( )

ˆ
ˆ
ˆ
ˆ
ˆ
ˆ
(
ɶ +Λ Φ
∆C−1 j 3Ω ) ɶ ( ɶ ) (
3 + C0 j 2Ω + Λ Φ 2 + ∆C+1 jΩ + Λ Φ1 + ∆K −1 Φ 3 + K 0 Φ 2 + ∆K +1 Φ1 ) =0
⋮ ⋮ ⋮
(e8)

Writing the above equations in matrix form will give

156
 Derivation of the HILL’s Equations for the Eigenvalues, Eigenvectors and the Forced Responses

 ⋱ ⋮ ⋮ ⋮ ⋮ ⋮ ⋰
 
 … M 0 ∆M −1 0 0 0 …
 … ∆M +1 M0 ∆M −1 0 0 …
 2 
 Λ … 0 ∆M +1 M0 ∆M −1 0 …
 … 0 0 ∆M +1 M0 ∆M −1 …
  
 … 0 0 0 ∆M +1 M0 …
 ⋰ ⋮ ⋮ ⋮ ⋮ ⋮ ⋱
 

⋱ ⋮ ⋮ ⋮ ⋮ ⋮ ⋰
… C0 − ɶM
j 4Ω ɶ ∆M
∆C −1 − j 2Ω 0 0 0 …
 0 −1

… ∆C+1 − ɶ ∆M
j 4Ω ɶ
C0 − j 2ΩM 0 ∆C−1 0 0 …
+1
 ɶ ∆M ɶ ∆M 
+Λ … 0 ∆C+1 − j 2Ω +1 C0 ∆C−1 + j 2Ω −1 0 …
… 0 0 ∆C+1 C0 + j 2Ω ɶM ɶ ∆M
∆C−1 + j 4Ω −1 …
 
0
ɶ ∆M ɶ
… 0 0 0 ∆C +1 + j 2Ω +1 C0 + j 4ΩM 0 …
⋰ ⋮ ⋮ ⋮ ⋮ ⋮ ⋱


⋱ ⋮ ⋮ ⋮
… K 0 − j 2Ω ɶ C − 4Ωɶ 2M ɶ ∆C − Ω
∆K −1 − jΩ ɶ 2 ∆M 0
 0 0 −1 −1
ɶ ∆C − 4Ω
… ∆K +1 − j 2Ω ɶ 2 ∆M K − jΩ ɶ C −Ω ɶ 2M ∆K −1
+1 +1 0 0 0
 ɶ ∆C − Ω
ɶ 2∆M
+ … 0 ∆K +1 − jΩ +1 +1 K0
… 0 0 ∆K +1

… 0 0 0
⋰ ⋮ ⋮ ⋮


 ⋮ 
⋮ ⋮ ⋰  
ˆ  
⋮ 
 Φ   
0 0 …  
−2  0 


…   Φ −1  0 
ˆ
0 0
ɶ ∆C − Ω ɶ 2∆M   
ˆ  

∆K −1 + jΩ −1 −1 0 …  ⋅  Φ 0
 = 0  .
ɶ ɶ 2
K 0 + jΩC0 − Ω M 0 ɶ
∆K −1 + j 2Ω∆C−1 − 4Ωɶ 2∆M …   
 

 ˆ  0 
−1
  Φ
ɶ ∆C − Ω ɶ 2 ∆M ɶ C − 4Ω
ɶ 2M …  
 0

1
∆K + jΩ
+1 +1 +1 K + j 2Ω
0 0 0
 ˆ   
⋮ ⋮ ⋱   Φ 2   ⋮ 
 
 ⋮ 
(e9)

The above equation is rewritten as HILL’s eigenvalue problem

( Λ Mˆ + ΛCˆ + Kˆ ) Φˆ = 0
2
(e10)

Note here that the matrices M̂ , Ĉ and K̂ can represent matrices either in the fixed or rotating
coordinate system and may be matrices developed using either the analytical methods for a LAVAL-
rotor without or with bearing masses or using the Finite Element Method for a general hybrid rotor
model.

157
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

E.2 Forced Responses

This sub-unit analyses the forced response to selected types of forcing functions. The two forcing
functions considered in this dissertation, i.e. the harmonic and the constant forcing functions, are
discussed in the following sub-units.

E.2.1 Harmonic Force

Consider a general presentation of a system with time-dependent matrices excited by a harmonic force
expressed by a complex exponential function whose equation of motion is

ɶ qɺɺ
+ C
M ɶ q
ɺ + K
ɶ q
= Fˆ + e jΩt (e11)

ˆ
where F + and Ω are the complex force amplitude vector and the excitation circular frequency,
respectively.

To determine the forced response assume a periodic solution of the form

ɶ
ɶ
ɶ
q = Φ + e jΩt ; Φ + ( t ) = Φ + t + Tɶ ( ) ɶ Tɶ = 2π .
; Ω (e12)

ɶ
Since the solution is assumed to be periodic again, the vector Φ + can be written using a
FOURIER-series as

ɶ ∞
ˆ
ɺɶ + ∞
ɺɺ
ɶ + ∞
ˆ + juΩɶ t

Φ+ = ∑
ɶ
Φ u+ e juΩt ; Φ = ∑ ɶ ˆ + e juΩɶ t
juΩΦ ; Φ = − ∑ u 2Ω
ɶ 2Φ . (e13)
u ue
u =−∞ u =−∞ u =−∞

Hence, the periodic forced response and its derivatives become

∞
ˆ
∞
ˆ + juΩɶ t
∞
ˆ
∑ ∑ ( ∑ ( Ω + uΩɶ )
2
q = e j Ωt
u =−∞
Φ u+ e juΩt ; qɺ = e jΩt
ɶ

u =−∞
ɶ Φ
j Ω + uΩ )
ue ; qɺɺ = −e jΩt
u =−∞
ɶ
Φ u+ e juΩt .

(e14)

Substituting the equations (e2) and (e14) into the equation of motion (e11) gives

 ∞
ˆ + juΩɶ t
( ) ( ɶ 2Φ ( )
∑ )
ɶt
− jΩ ɶt
jΩ ɶ ɶ
 − ∆ M −1 e + M 0 + ∆ M +1 e Ω + uΩ ue + ∆C−1e − jΩt + C0 + ∆C+1e jΩt
 u =−∞
∞
ˆ + juΩɶ t ∞

ˆ ɶ 
ˆ
⋅ ∑ j Ω + uΩ
u =−∞
( ɶ Φ
u e ) + ∆ K (
−1 e
ɶt
− jΩ
+ K 0 + ∆ K +1 e
ɶt
jΩ
∑
u =−∞
)
Φ u+ e juΩt  e jΩt = F + e jΩt . (e15)


158
 Derivation of the HILL’s Equations for the Eigenvalues, Eigenvectors and the Forced Responses

~
Eliminating the exponential factor e jΩt and, since Ω = 2Ω , equation (e15) may be simplified as

∞
ˆ + juΩɶ t
( ) ∑ ( u Ωɶ ) ( )
ɶ ɶ 2 ɶ ɶ
− ∆M −1e − jΩt + M 0 + ∆M +1e jΩt + Φ ue + ∆C−1e− jΩt + C0 + ∆C+1e jΩt
u =−∞
∞
ˆ + juΩɶ t ∞
+
ˆ
⋅ ∑ (
u =−∞
ɶ Φ
j u+ Ω )
ue
ɶ
( ɶ
+ ∆K −1e − jΩt + K 0 + ∆K +1e jΩt ) ∑ Φˆ e
u =−∞
u
ɶt
ju Ω
= F + (e16)

where u+ = u + 0.5 43.

The comparison of the exponential terms in the above equation will give

⋮ ⋮ ⋮

ˆ +
+
ˆ +

ɶ 2Φ ) Φˆ ɶ ˆ+
2
ɶ
e −2 jΩt : −∆M −1 −0.5Ω ( ) ɶ ( ɶ 2Φ ( ) ( )
−1 − M 0 −1.5Ω −2 − ∆M +1 −2.5Ω −3 + ∆C −1 j −0.5Ω Φ −1

ˆ +

ˆ +
ˆ +
ˆ +
ɶ Φ
+C0 j −1.5Ω ( )
−2 + ∆C +1 j −2.5Ω
ɶ ( ) Φˆ +
−3 + ∆K Φ
− 1 −1 + K Φ
0 −2 + ∆K Φ
+1 −3 =0

ˆ +
ˆ +
ˆ +
ˆ +
ɶ
e− jΩt : −∆M −1 0.5Ωɶ 2Φ ( − M − )
0.5Ωɶ 2Φ − ∆M ( − 1.5Ωɶ 2Φ ) + ∆C j 0.5Ωɶ Φ ( ) ( )
0 0 −1 +1 −2 −1 0

ˆ +
ˆ +
ˆ +
ˆ +
ˆ +
(
ɶ Φ + ∆C j −1.5Ω
+C0 j −0.5Ω −1 +1 ) −2 −1 0 (
ɶ Φ + ∆K Φ + K Φ + ∆K Φ
0 −1 ) +1 −2 =0

ˆ +
ˆ +
ˆ +

e j0 : − ∆M −1 1.5Ωɶ 2Φ
( ) ɶ 2Φ ( ɶ 2Φ
) ɶ ˆ+ ( ) ( )
1 − M 0 0.5Ω 0 − ∆M +1 −0.5Ω −1 + ∆C −1 j 1.5Ω Φ1

ˆ +

ˆ +
ˆ +
ˆ +
ˆ
+C0 j 0.5Ω(
ɶ Φ ) ɶ ˆ+ ( ) = F+
0 + ∆C +1 j −0.5Ω Φ −1 + ∆K −1 Φ1 + K 0 Φ 0 + ∆K +1Φ −1

ˆ +
ˆ +
ˆ +

ɶ
e jΩt : −∆M −1 2.5Ωɶ 2Φ
( ) ɶ 2Φ ( )ɶ 2Φ ɶ ˆ+ ( ) ( )
2 − M 0 1.5Ω 1 − ∆M +1 0.5Ω 0 + ∆C −1 j 2.5Ω Φ 2

ˆ +
ˆ +
ˆ +
ˆ +
ˆ +
(
ɶ Φ
+C0 j 1.5Ω 1 +) ∆C +1 j 0.5Ωɶ Φ
0 + ∆K ( Φ
−1 2 + )
K Φ
0 1 + ∆K Φ
+1 0 =0

ˆ +
+
+
+
ɶ 2Φ ) Φˆ − ∆M (1.5Ωɶ ) Φˆ + ∆C j (3.5Ωɶ ) Φˆ
2 2
ɶ
e 2 jΩt : −∆M −1 3.5Ω ( ) ɶ
3 − M 0 2.5Ω ( 2 +1 1 −1 3

ˆ +




ɶ Φ
+C0 j 2.5Ω ( )
2 + ∆C +1 j 1.5Ω
ɶ ( ) Φˆ + ∆K Φˆ + K Φˆ + ∆ K Φˆ
+
1 −1
+
3 0
+
2 1 +1
+
1 =0
⋮ ⋮ ⋮
(e17)

Compiling the above set of equations into a single matrix equation gives

43 For harmonic forces with negative exponents e− jΩ t , a term u-= u - 0.5 is used in place of u+= u + 0.5.

159
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

⋱ ⋮ ⋮ ⋮
 2 2
⋯ K 0 − j 1.5Ω (
ɶ C − 1.5Ω
0
ɶ M )0 ( )ɶ ∆C − 0.5Ω
∆K −1 − j 0.5Ω −1
ɶ ∆M
−1 0 ( ) ( )

⋯ ∆K +1 − j 1.5Ω
( ɶ ∆C − 1.5Ωɶ 2 ∆M
) (
K 0 − j 0.5Ω ) ɶ 2M
ɶ C − 0.5Ω (
ɶ ∆C − 0.5Ω
∆K −1 + j 0.5Ω ɶ 2 ∆M ) ( ) ( ) ( )
 +1 +1 0 0 −1 −1

 2 2

⋯ 0 ɶ ∆C − 0.5Ω
∆K +1 − j 0.5Ω +1
ɶ ∆M
+1 K0 + j 0.5Ω (
ɶ C − 0.5Ω
0
ɶ M
0 ) ( ) ( ) ( )
 2
⋯ 0 0 ɶ ∆C − 0.5Ω
∆K +1 + j 0.5Ω +1
ɶ ∆M
+1 ( ) ( )

⋯ 0 0 0
⋰ ⋮ ⋮ ⋮
⋮ ⋮ ⋰  ⋮ 

ˆ +  

⋮
0 0 ⋯ Φ   
−2
0
0 0 ⋯ 
ˆ +  

2  Φ−1   0 
(
∆K −1 + j 1.5Ω −1 )
ɶ ∆C − 1.5Ω
ɶ ∆M
−1 ( ) 0 ⋯ 
ˆ +  
ˆ + 
  Φ0  = F 
⋯ 
ˆ +   0 
2 2

K 0 + j 1.5Ω (
ɶ C − 1.5Ω
0 )
ɶ M
0 ( ) (
ɶ ∆C − 2.5Ω
∆K −1 + j 2.5Ω −1 )
ɶ ∆M
−1 ( )  Φ1
2 2  
ˆ   0

(
∆K +1 + j 1.5Ω +1 )
ɶ ∆C − 1.5Ω
ɶ ∆M
+1 ( ) K0 + j 2.5Ω 0(
ɶ C − 2.5Ω
ɶ M
0) ( ) ⋯  Φ +
2
  
    ⋮ 
⋮ ⋮ ⋱  ⋮ 

(e18)

which represents a set of linear equations of the form

ˆ +
ˆ +
K +Φ = FH . (e19)

ˆ
From the above equation the vectors Φ u+ can be determined and then substituted into the first
equation (e14) to determine the forced response for harmonic forcing functions with positive exponent.

A similar procedure can be followed for a harmonic force with a negative exponent of the form

ɶ qɺɺ
+ C
M ɶ q
ɺ + K
ɶ q
= Fˆ − e− jΩt . (e20)

Assuming now a solution of the form

ɶ
ɶ
ɶ
q = Φ − e − jΩt ; Φ − ( t ) = Φ − t + Tɶ ( ) ɶ Tɶ = 2π .
; Ω (e21)

ɶ
Since the solution is assumed to be periodic again, the vector Φ − can be written using a
FOURIER-series as

ɶ ∞
−
ɺɶ − ∞
ɺɺ
ɶ − ∞
ˆ − juΩɶ t

∑ Φˆ e ∑ ɶ ˆ − e juΩɶ t = − ∑ u 2Ω
ɶt
Φ− = juΩ
; Φ = juΩΦ ; Φ ɶ 2Φ . (e22)
u u ue
u =−∞ u =−∞ u =−∞

160
 Derivation of the HILL’s Equations for the Eigenvalues, Eigenvectors and the Forced Responses

Hence, the periodic forced response and its derivatives become

∞
−
∞
−
∞
ˆ
∑ Φˆ e ∑ j ( uΩɶ − Ω ) Φˆ e ∑ ( uΩɶ − Ω )
2
; qɺ = e − jΩt ; qɺɺ = −e − jΩt
ɶt ɶt ɶ
q = e − j Ωt u
ju Ω
u
juΩ
Φ u− e juΩt .
u =−∞ u =−∞ u =−∞
(e23)

Following the same procedure as given from equation (e15) to (e17) now leads to a final matrix
equation of the form

⋱ ⋮ ⋮ ⋮
 2 2
⋯ K0 − j 2.5Ω 0 (
ɶ C − 2.5Ω
ɶ M
0 ) ( )
ɶ ∆C − 1.5Ω
∆K −1 − j 1.5Ω −1
ɶ ∆M
−1 0 ( ) ( )

⋯ ∆K +1 − j 2.5Ω
2 2
(
ɶ ∆C − 2.5Ωɶ ∆M ) (
K 0 − j 1.5Ω )
ɶ C − 1.5Ω
ɶ M (
ɶ ∆C − 0.5Ω
∆K −1 − j 0.5Ω ɶ 2 ∆M ) ( ) ( ) ( )
 +1 +1 0 0 −1 −1

 2 2

⋯ 0 ɶ ∆C − 1.5Ω
∆K +1 − j 1.5Ω +1
ɶ ∆M
+1 K0 − j 0.5Ω (
ɶ C − 0.5Ω
0
ɶ M
0 ) ( ) ( ) ( )
 2
⋯ 0 0 ɶ ∆C − 0.5Ω
∆K +1 − j 0.5Ω +1
ɶ ∆M
+1 ( ) ( )

⋯ 0 0 0
⋰ ⋮ ⋮ ⋮
⋮ ⋮ ⋰  ⋮ 
⋮ 
0 0 ⋯ Φ

ˆ −  


−2
0
0 0 ⋯ 
ˆ −  

2  Φ−1   0 
(
∆K −1 + j 0.5Ω −1 )
ɶ ∆C − 0.5Ω
ɶ ∆M
−1 ( ) 0 ⋯ 
ˆ −  
ˆ − 
  Φ0  = F 
⋯ 
ˆ −   0 
ɶ 2M ɶ 2 ∆M

(
ɶ C − 0.5Ω
K0 + j 0.5Ω 0 ) 0 ( ) (
ɶ ∆C − 1.5Ω
∆K −1 + j 1.5Ω −1 ) −1 ( )  Φ1
2 2  
ˆ   0

(
∆K +1 + j 0.5Ω +1 )
ɶ ∆C − 0.5Ω
ɶ ∆M
+1 ( ) K0 + j 1.5Ω 0(
ɶ C − 1.5Ω
ɶ M)
0 ( ) ⋯  Φ −
2
  
    ⋮ 
⋮ ⋮ ⋱  ⋮ 
(e24)

which is equivalent to

ˆ −
ˆ −
K −Φ = FH . (e25)

ˆ
From the above equation the vectors Φ u− can be determined and then substituted into the first
relation of equation (e23) to determine the forced response for harmonic forcing functions with negative
exponent.

E.2.2 Constant Force

Consider a general representation of a system subjected to a constant force whose equation of motion
is

ɶ qɺɺ
+ C
M ɶ q
ɺ + K
ɶ q
= F (e26)

161
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

where F is the constant force vector.

To determine the forced response assume a periodic solution of the form

ɶ
ɶ
ɶ
q = Φ ; Φ ( t ) = Φ t + Tɶ ( ) ɶ Tɶ = 2π
; Ω (e27)

ɶ
Since the solution is assumed to be periodic, the vector Φ can again be written using a FOURIER-
series as

ɶ ∞
ˆ
ɺɶ ∞
ɺɺ
ɶ ∞
ˆ juΩɶ t

Φ= ∑
ɶ
Φ u e juΩt ; Φ = ∑ ɶ ˆ e juΩɶ t
juΩΦ ; Φ = − ∑ u 2Ω
ɶ 2Φ (e28)
u ue
u =−∞ u =−∞ u =−∞

Substituting equation (e28) in the equation of motion (e26) gives

∞

( ɶ
− ∆M −1e − jΩt + M 0 + ∆M +1e jΩt
ɶ
) ∑ u Ωɶ Φˆ e
u =−∞
2 2
u
ɶt
ju Ω
( ɶ
+ ∆C −1e− jΩt + C0 + ∆C+1e jΩt
ɶ
)
∞
∞


⋅ ∑ ɶ ˆ e juΩɶ t + ∆K e − jΩɶ t + K + ∆K e jΩɶ t
juΩΦ ( ) ∑ Φˆ e ɶt
ju Ω
= F . 44 (e29)
u −1 0 +1 u
u =−∞ u =−∞

The comparison of the exponential terms in the above equation yields

⋮ ⋮ ⋮


ˆ




ˆ
ɶ
ɶ 2∆M Φ
e−2 jΩt : −Ω ɶ2 ˆ ɶ2 ˆ ɶ ˆ ɶ ˆ ɶ
−1 −1 − 4Ω M0 Φ−2 − 9Ω ∆M+1Φ−3 − jΩ∆C−1Φ−1 − j 2ΩC0 Φ−2 − j3Ω∆C+1Φ−3

ˆ
ˆ
ˆ
+ ∆K −1Φ−1 + K0Φ−2 + ∆K +1Φ−3 = 0

ˆ
ˆ

ˆ
ˆ
ˆ
ˆ
ɶ
ɶ 2M Φ
e− jΩt : −Ω ɶ2 ɶ ˆ ɶ
0 −1 − 4Ω ∆M+1Φ−2 − jΩC0Φ−1 − j 2Ω∆C+1Φ−2 + ∆K −1Φ0 + K 0Φ−1 + ∆K +1Φ−2 =0

ɶ
ˆ
ˆ
ˆ
ˆ
ˆ
ˆ
ˆ

e0Ωt : ɶ 2∆M Φ
−Ω ɶ2 ɶ ɶ
−1 1 − Ω ∆M+1Φ−1 + jΩ∆C−1Φ1 − jΩ∆C+1Φ−1 + ∆K −1Φ1 + K 0Φ0 + ∆K +1Φ−1 =F

ˆ

ˆ

ˆ
ˆ
ˆ
ɶ
e jΩt : ɶ 2∆M Φ
−4Ω ɶ2 ˆ ɶ ɶ ˆ
−1 2 − Ω M0Φ1 + j 2Ω∆C−1Φ2 + jΩC0 Φ1 + ∆K −1Φ2 + K 0 Φ1 + ∆K +1Φ0 =0

ɶ
ˆ
ˆ
ˆ
ˆ
ˆ
ˆ
ɶ 2∆M Φ
e2 jΩt : −9Ω − 4Ωɶ 2M Φ − ɶ 2∆M Φ
Ω + j 3Ωɶ ∆C Φ + j 2ΩɶC Φ + jΩɶ ∆C Φ
−1 3 0 2 +1 1 −1 3 0 2 +1 1

ˆ
ˆ
ˆ
+ ∆K −1Φ3 + K0Φ2 + ∆K +1Φ1 = 0
⋮ ⋮ ⋮
(e30)

44 This equation is equivalent to equation (e15) for Ω =0.

162
 Derivation of the HILL’s Equations for the Eigenvalues, Eigenvectors and the Forced Responses

Compiling the above set of equations into a single matrix equation gives

⋱ ⋮ ⋮ ⋮ ⋮
⋯ ɶ
K 0 − j 2ΩC0 − 4Ω ɶ 2M ɶ
∆K −1 − jΩ∆C−1 − Ωɶ 2 ∆M 0 0
 0 −1
ɶ ɶ
⋯ ∆K +1 − j 2Ω∆C +1 − 4Ω ∆M +1
2 ɶ
K − j ΩC − Ω Mɶ 2
∆K −1 0
0 0 0
 ɶ ∆C − Ωɶ 2 ∆M ɶ ∆C − Ω ɶ 2 ∆M
⋯ 0 ∆K +1 − jΩ +1 +1 K0 ∆K −1 + jΩ −1 −1
⋯ 0 0 ∆K +1 K 0 + jΩ ɶ C −Ω ɶ 2M

0 0

⋯ 0 0 0 ɶ ∆C − Ω
∆K +1 + jΩ ɶ 2 ∆M
+1 +1
⋰ ⋮ ⋮ ⋮ ⋮

 ⋮ 
⋮ ⋰ 
ˆ  
⋮ 
Φ   
0 ⋯ 
−2   0 
ˆ

0 ⋯  Φ −1   0 
 
ˆ  

0 ⋯  Φ  = F 
0


ɶ
∆K −1 + j 2Ω∆C−1 − 4Ω ɶ 2 ∆M ⋯ ˆ  0
 

−1
  Φ  

ɶ ɶ 2
K 0 + j 2ΩC0 − 4Ω M 0
1
⋯ 
  0 
ˆ   
⋮ ⋱  Φ 2   ⋮ 
 
 ⋮ 

(e31)

which represents a set of linear equations of the form

ˆ
ˆ
KΦ = FH . (e32)

ˆ
From the above equation the vectors Φ u can be determined and then substituted into the
equations (e27) and (e28) to determine the forced response for constant forces.

163
 F Appendix F

GEOMETRIC AND MATERIAL DATA OF THE INVESTIGATED ROTORS

In the chapters 4.2 and 7 three different rotor assemblies are investigated. The detailed geometric
and material data are given in the following figures and tables.

F.1 LAVAL-Rotor with Non-Circular Shaft and Different Types of Bearings

Figure F.1: LAVAL-rotor with a non-circular shaft and

anisotropic elastic bearings

Shaft width b [m] 0.008

Shaft height h [m] 0.0065
Length L [m] 0.446
Modulus of elasticity E [N/m2 ] 2.10x1011
Density ρ [kg/m3] 7850.00
Second area moment Iη [m4] 1.4082 x10-10
Second area moment Iζ [m4] 1.9576 x10-10
Table F.1: Shaft data
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Density Radius Thickness Mass Eccentricity

ρ [kg/m3] R [m] H [m] m [kg] ε [mm]
7850.00 0.1 0.00861 2.125 1
Table F.2: Disk data

Circular frequency Circular frequency Average circular Shaft non-

for zero speed for zero speed frequency for zero circularity
ω0η [rad/s] ω0ζ [rad/s] speed ωav [rad/s] µs [-]
102.31 86.77 94.86 0.1632
Table F.3: Non-circular shaft parameters

kyb ∞
Rigid Bearing
kzb ∞

L 90 mm
kyb 11.52 kN/m
Anisotropic
W 20 mm
Elastic Bearing
kzb 93.3x103 kN/m
T 2 mm

Table F.4: Bearing data

166
 Geometric and Material Data of the Investigated Rotors

F.2 Rotor with Rectangular Shaft and Different Types of Bearings

Figure F.2: Rotor with different shafts composed by

rectangular sections

Section width b [m] 0.045

Section height h [m] 0.035
Length L [m] 1.20
Modulus of elasticity E [N/m2 ] 2.10x1011
Density ρ [kg/m3] 7850
Mass m [kg] 14.837
Second area moment Iη [m4] 1.6078 x10-7
Second area moment Iζ [m4] 2.6578 x10-7
Table F.5: Shaft data

167
 Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Shaft name (Number of sections; sym./asym.) Lengths of sections [m]

2 sections symmetric 0.6
2 sections asymmetric 0.6
4 sections symmetric 0.3
4 sections asymmetric 0.3
6 sections symmetric 0.2
6 sections asymmetric 0.2
12 sections symmetric 0.1
12 sections asymmetric 0.1
Table F.6: Shaft name and section data45

Density Inner radius Outer radius Thickness Mass Eccentricity

ρ [kg/m3] Ri [m] Ro [m] H [m] m [kg] ε [mm]
7850.00 0.0285 0.2 0.054 52.187 0.383
Table F.7: Disk data

Bearing kyy kyz kzy kzz dyy dyz dzy dzz

[N/m] [N/m] [N/m] [N/m] [Ns/m] [Ns/m] [Ns/m] [Ns/m]
1 2.49x107 0 0 1.76x107 0 0 0 0
2 1.10x108 0 0 2.17x108 0 0 0 0
Table F.8: Anisotropic elastic bearing data

Bearing kyy kyz kzy kzz dyy dyz dzy dzz

[N/m] [N/m] [N/m] [N/m] [Ns/m] [Ns/m] [Ns/m] [Ns/m]
1 2.49x107 -3.01x107 4.77x107 1.76x107 6.01x105 2.21x105 2.21x105 7.76x105
2 1.10x108 -1.34x107 2.21x108 2.17x108 1.01x106 9.95x105 9.95x105 3.12x106
Table F.9: Anisotropic visco-elastic bearing data

Shaft name Lengths of section A shaft Lengths of section B shaft

segments [m] segments [m]
µR = 0.00 0.60 0.60
µR = 0.33 0.80 0.40
2 sections 0.90 0.30
µR = 0.50
asymmetric
µR = 0.66 1.00 0.20
µR = 0.83 1.10 0.10
µR =0 0.30 0.30
4 sections 0.40 0.20
µR = 0.33
symmetric
µR = 0.66 0.50 0.10

Table F.10: The 2 sections and 4 sections variations46

45 The cylindrical sections of the shafts are locations of the bearings and the disk. Their lengths are not considered in the
FE-model for they refer to nodal points at the ends of the shaft and at the central shaft element, respectively.

168
 Geometric and Material Data of the Investigated Rotors

F.3 Two Disk Rotor with Non-Circular Shaft and Anisotropic Elastic Bearings

Figure F.3: Two disk rotor with a non-circular shaft

Shaft width b [m] 0.0062

Shaft height h [m] 0.0058
Length L [m] 0.25
Modulus of elasticity E [N/m2 ] 2.1x1011
Density ρ [kg/m3] 7850
Second area moment Iη [m4] 6.7308 x10-11
Second area moment Iζ [m4] 7.2391 x10-11
Table F.11: Shaft data

Mass m [kg]
Disks 0.5
Table F.12: Disk data

Circular frequency Circular frequency Average circular Shaft non-

for zero speed for zero speed frequency for zero circularity
ω1η , ω2η [rad/s] ω1ζ , ω2ζ [rad/s] speed ω1av , ω2 av [rad/s] µs [-]
41.59 35.42 38.22
0.0364
150.50 117.50 135.87
Table F.13: Non-circular shaft parameter

46 Note here that in the 4 sections symmetric shaft the length of section A is equally distributed to the two ends so as to
maintain the longitudinal symmetry.

169
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

kyy kyz kzy kzz dyy dyz dzy dzz

[N/m] [N/m] [N/m] [N/m] [Ns/m] [Ns/m] [Ns/m] [Ns/m]
Bearings 2000 0 0 6000 0 0 0 0
Table F.14: Bearing data

170
 G Appendix G

HERMITIAN POLYNOMIALS AND SHAPE FUNCTIONS FOR

TIMOSHENKO BEAM ELEMENTS

Consider a finite beam element cut out from the rotor shaft. This element will be subjected to internal
forces (and moments) resulting from the deflections in the x,y- and x,z-plane of the fixed coordinate
system. These internal forces deform the element as shown in Figure G.1.

Figure G.1: Beam element of the rotor shaft in the x,y-

and x,z-planes

From NEWTON’s 2nd law of motion and the TIMOSHENKO beam theory, the shear force and
bending moment constitutive equations can be written in the x,y-plane as

∂θ  ∂v 
M z = EI z = EI zθ ′ ; Qy = κ GAβ z = κ GA  − θ  = κ GA ( v′ − θ ) . (g1)
∂x  ∂x 

In the x,z-plane the bending moment and shear force and the deflections are related by

∂ψ  ∂w 
M y = EI y = EI yψ ′ ; Qz = κ GAβ y = κ GA  +ψ  = κ GA ( w′ +ψ ) . (g2)
∂x  ∂x 

From moment equilibrium in the x,y- and x,z-plane the relation between the forces and moments
can be determined as

M z′ + Qy = 0 ; M ′y − Qz = 0 . (g3)
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Substituting the equations (g1) and (g2) into equation (g3) and rearranging gives47

az L2 12 EI z
EI zθ ′′ + κ GA ( v′ − θ ) = 0 ⇒ θ ′′ + v′ − θ = 0 ; az =
12 κ GAL2
(g4)
a y L2 12 EI y
EI yψ ′′ − κ GA ( w′ + ψ ) = 0 ⇒ ψ ′′ − w′ −ψ = 0 ; a y = .
12 κ GAL2

Considering the beam element in a fixed coordinate system, it can be unequivocally located in space if
the displacements and rotations at both ends of the element are specified. Hence, a nodal
generalized displacement vector of an element of length L is defined as

T
qnf = v1 w1 ψ 1 θ 1 v 2 w2 ψ 2 θ 2  (g5)

where the superscripts 1 and 2 refer to the left and right node at the end of the element,
respectively.

It is assumed that the displacements and rotations of any point in the element can be approximated
by a linear combination of the nodal displacements and rotations at the ends of the element in the
corresponding directions. Hence, it is introduced48

v = H v1v1 + H v 2θ 1 + H v 3v 2 + H v 4θ 2 ; θ = Hθ 1v1 + Hθ 2θ 1 + Hθ 3v 2 + Hθ 4θ 2
(g6)
w = H w1w1 + H w 2ψ 1 + H w3 w2 + H w 4ψ 2 ; ψ = Hψ 1w1 + Hψ 2ψ 1 + Hψ 3 w2 + Hψ 4ψ 2 .

The “coefficients” in the above equations are shape functions of HERMITAN polynomials of the 3rd
order as49

H vi , wi = A3 x3 + A2 x 2 + A1 x + A0 ; Hθ i ,ψ i = B3 x 3 + B2 x 2 + B1 x + B0 ; i = 1,⋯ , 4 (g7)

Substituting the first two relations of equation (g6) into the first relation of equation (g4) gives

a z L2
H v′1v1 + H v′2θ 1 + H v′3v 2 + H v′4θ 2 = −
12
( )
Hθ′′1v1 + Hθ′′2θ 1 + Hθ′′3v 2 + Hθ′′4θ 2
(g8)
(
+ Hθ 1v1 + Hθ 2θ 1 + Hθ 3v 2 + Hθ 4θ 2 )
47 For a constant cross-section of the element.
48 The following derivations follow the procedures described in [2, 4, 8]
49 The subscripts vi, wi and θ i,ψ i are omitted on the polynomial coefficients Ai and Bi for presentation simplicity.

172
 HERMITIAN Polynomials and Shape Functions for TIMOSHENKO Beam Elements

Collecting equal terms of equation (g8) gives

az L2
H vi′ = − Hθ′′i + Hθ i ; i = 1,⋯ , 4 . (g9)
12

Using the equations (g9) and (g7) yields

az L2
3 A3 x 2 + 2 A2 x + A1 = − ( 6 B3 x + 2 B2 ) + B3 x3 + B2 x 2 + B1 x + B0 . (g10)
12

The comparison of equal ‘x’-coefficients will result in the relations between the coefficients of the
HERMITIAN polynomials as

a z L2
B3 = 0 ; B2 = 3 A3 ; B1 = 2 A2 ; B0 = A1 + A3 . (g11)
2

The coefficients in the preceding equations are then determined considering unit deflections of the
beam ends as discussed below.

• Unit displacement of the left end

v1 = v( x =0) = 1 ; θ 1 = θ( x = 0) = 0 ; v 2 = v( x = L ) = 0 ; θ 2 = θ( x = L ) = 0. (g12)

Substituting the preceding end conditions into the first two equations of equation (g6) gives

H v1 ( 0 ) = 1 ; H v1 ( L ) = 0 ⇒ A0 = 1 ; A3 L3 + A2 L2 + A1 L + A0 = 0
(g13)
Hθ 1 ( 0 ) = 0 ; H θ 1 ( L ) = 0 ⇒ B0 = 0 ; B3 L3 + B2 L2 + B1 L + B0 = 0 .

Using the equations (g11) and (g13) will give the values of the coefficients to be

2 3 az
A3 = ; A2 = − ; A1 = − ; A0 = 1
L (1 + a z )
3
L (1 + a z )
2
L (1 + a z )
(g14)
6 6
B3 = 0 ; B2 = 3 ; B1 = − 2 ; B0 = 0 .
L (1 + az ) L (1 + az )

173
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

Hence, substituting these coefficients into the equations (g7) gives

1 1 6 1 2
H v1 =  2 x3 − 3Lx 2 − az L2 x + (1 + a z ) L3  ; Hθ 1 =
3 
 x − Lx  . (g15)
1 + az L 1 + az L3 

• Unit rotation of the left end

v1 = v( x = 0) = 0 ; θ 1 = θ( x =0) = 1 ; v 2 = v( x = L ) = 0 ; θ 2 = θ( x = L ) = 0 . (g16)

Substituting the preceding end conditions into the first two equations of equation (g6) gives

H v2 ( 0) = 0 ; Hv2 ( L ) = 0 ⇒ A0 = 0 ; A3 L3 + A2 L2 + A1 L + A0 = 0
(g17)
Hθ 2 ( 0 ) = 1 ; Hθ 2 ( L ) = 0 ⇒ 3 2
B0 = 1 ; B3 L + B2 L + B1 L + B0 = 0 .

Using the equations (g11) and (g17) will give the values of the coefficients to be

1 4 + az 2 + az
A3 = ; A2 = − ; A1 = ; A0 = 0
L (1 + az )
2
2 L (1 + az ) 2 (1 + az )
(g18)
3 4 + az
B3 = 0 ; B2 = 2 ; B1 = − ; B0 = 1 .
L (1 + a z ) L (1 + az )

Hence, substituting these coefficients into the equations (g7) gives

1 1  3  a  2  az  2  1 1
H v2 = 2 
x −2+ z  Lx + 1 +  L x  ; Hθ 2 = 3 x 2 − ( 4 + az ) Lx + (1 + az ) L2 
2 
1 + az L   2   2   1 + a z L
(g19)

• Unit displacement of the right end

v1 = v( x = 0) = 0 ; θ 1 = θ( x =0) = 0 ; v 2 = v( x = L ) = 1 ; θ 2 = θ( x = L ) = 0 . (g20)

Substituting the preceding end conditions into the first two equations of equation (g6) gives

H v 3 ( 0 ) = 0 ; H v3 ( L ) = 1 ⇒ A0 = 0 ; A3 L3 + A2 L2 + A1 L + A0 = 1
(g21)
Hθ 3 ( 0 ) = 0 ; Hθ 3 ( L ) = 0 ⇒ B0 = 0 ; B3 L3 + B2 L2 + B1 L + B0 = 0 .

174
 HERMITIAN Polynomials and Shape Functions for TIMOSHENKO Beam Elements

Using the equations (g11) and (g21) will give the values of the coefficients to be

2 3 az
A3 = − ; A2 = ; A1 = ; A0 = 0
L (1 + az )
3
L (1 + a z )
2
L (1 + az )
(g22)
6 6
B3 = 0 ; B2 = − 3 ; B1 = 2 ; B0 = 0 .
L (1 + az ) L (1 + az )

Hence, substituting these coefficients into the equations (g7) gives

1 1 6 1
H v3 =  −2 x 3 + 3Lx 2 + az L2 x  ; Hθ 3 =
3 
 − x 2 + Lx 
3 
(g23)
1 + az L 1 + az L

• Unit rotation of the right end

v1 = v( x = 0) = 0 ; θ 1 = θ( x =0) = 0 ; v 2 = v( x = L ) = 0 ; θ 2 = θ( x = L ) = 1 . (g24)

Substituting the preceding end conditions into the first two equations of equation (g6) gives

H v4 ( 0) = 0 ; Hv4 ( L ) = 0 ⇒ A0 = 0 ; A3 L3 + A2 L2 + A1 L + A0 = 0
(g25)
Hθ 4 ( 0 ) = 0 ; Hθ 4 ( L ) = 1 ⇒ B0 = 0 ; B3 L3 + B2 L2 + B1 L + B0 = 1 .

Using the equations (g11) and (g25) will give the values of the coefficients to be

1 ( 2 − az ) az
A3 = ; A2 = − ; A1 = − ; A0 = 0
L (1 + az )
2
2 L (1 + a z ) 2 (1 + az )
(g26)
3 ( 2 − az )
B3 = 0 ; B2 = 2 ; B1 = − ; B0 = 0 .
L (1 + a z ) L (1 + az )

Hence, substituting these coefficients into the equation (g7) gives

1 1  3  az  2 az 2  1 1
H v4 = x − 1 −  Lx − L x  ; Hθ 4 = 3 x 2 − ( 2 − az ) Lx  (g27)
1 + a z L2   2  2  1 + az L2 

Using the second relation of equation (g4) and the same derivation procedure outlined from the
equations (g8) to (g27), the remaining HERMITIAN polynomials can be determined to be

175
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

1 1 6 1
H w1 =  2 x3 − 3Lx 2 − a y L2 x + (1 + a y ) L3  ; Hψ 1 =  − x 2 + Lx 
1 + a y L3   1 + a y L3 
1 1  3  ay   ay  2  1 1
H w2 = 2 
− x +  2 +  Lx 2 −  1 +  L x ; Hψ 2 = 3 x 2 − ( 4 + a y ) Lx + (1 + a y ) L2 
1 + ay L   2   2   1 + a y L2  
1 1 6 1 2
H w3 =  −2 x3 + 3Lx 2 + a y L2 x 
3 
; Hψ 3 =  x − Lx 
1 + ay L 1 + a y L3 
1 1  3  ay  2 ay 2  1 1
H w4 = − x + 1 −  Lx + L x  ; Hψ 4 = 3 x 2 − ( 2 − a y ) Lx 
2 
1 + a y L2  2 2 1 + ay L 
  
(g28)

If the shape function matrix H = H ( x ) of the HERMITIAN polynomials is defined as

 H v1 0 0 H v2 H v3 0 0 Hv4 
 0 H w1 H w2 0 0 H w3 H w4 0 
H= (g29)
 0 Hψ 1 Hψ 2 0 0 Hψ 3 Hψ 4 0 
 
 Hθ 1 0 0 Hθ 2 Hθ 3 0 0 Hθ 4 

and the generalized displacements of the element qxf = qxf ( x, t ) are given by

qxf = [ v w ψ θ]T , (g30)

then the four equations in equation (g6) can be written in matrix form as

qxf = Hqnf (g31)

where the generalized nodal displacement vector qnf is defined in equation (g5).

176
 H Appendix H

MASS, DAMPING, GYROSCOPIC AND STIFFNESS MATRICES FOR THE

SHAFT, DISK AND BEARING ELEMENTS

In this Appendix, the element matrices of the components of a general anisotropic rotor-bearing
system discussed in chapter 6 are given in both the fixed and rotating coordinate system.

H.1 Shaft Elements

H.1.1 Circular Shaft Element in the Fixed Coordinate System

M S qɺɺnf + G S qɺnf + K S qnf = 0 (6.10)

 m1 0 0 m2 m3 0 0 − m4 
 0 m1 −m2 0 0 m3 m4 0 

 0 −m2 m5 0 0 − m4 − m6 0 
 
m 0 0 m5 m4 0 0 −m6 
MS =  2
 m3 0 0 m4 m1 0 0 − m2 
 
 0 m3 −m4 0 0 m1 m2 0 
 0 m4 − m6 0 0 m2 m5 0 
 
 − m4 0 0 − m6 − m2 0 0 m5 

 0 g1 − g2 0 0 − g1 − g2 0 
− g 0 0 − g2 g1 0 0 − g 2 
 1
 g2 0 0 g4 − g2 0 0 − g3 
 
0 g2 − g4 0 0 − g2 g3 0 
GS = 
 0 − g1 g2 0 0 g1 g2 0 
 
 g1 0 0 g2 − g1 0 0 g2 
g 0 0 − g3 − g2 0 0 g4 
 2 
 0 g2 g3 0 0 − g2 − g4 0 

 k1 0 0 k2 − k1 0 0 k2 
 0 k1 − k2 0 0 − k1 − k2 0 

 0 − k2 k3 0 0 k2 k4 0 
 
k 0 0 k3 − k2 0 0 k4 
KS =  2
 − k1 0 0 − k2 k1 0 0 − k2 
 
 0 − k1 k2 0 0 k1 k2 0 
 0 − k2 k4 0 0 k2 k3 0 
 
 k2 0 0 k4 − k2 0 0 k3 
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

H.1.1.1 TIMOSHENKO Beam Theory

ρ AL  13 7 1 2 6 2
m1 = 2  35 + 10 a + 3 a + 5 µ 
(1 + a )
ρ AL2  11 11 1 2 2 1 1 
m2 = 2  210 + 120 a + 24 a + µ  10 − 2 a  
(1 + a )   
ρ AL  9 3 1 2 6 2
m3 = 2  70 + 10 a + 6 a − 5 µ 
(1 + a )  
ρ AL2  13 3 1 2 2 1 1 
m4 = 2  420 + 40 a + 24 a − µ  10 − 2 a  
(1 + a )   
ρ AL3  1 1 1 2 2 2 1 1 2 
m5 = 2 105 + 60 a + 120 a + µ  15 + 6 a + 3 a  
(1 + a )   
ρ AL3  1 1 1 2 2 1 1 1 2 
m6 = 2 140 + 60 a + 120 a + µ  30 + 6 a − 6 a  
(1 + a )   

12 EI I
where a = 2
; µ2 = 2 ; I = Iy = Iz
κ GAL AL

ρ Ω Ix  6  ρ Ω Ix  1 1 
g1 = 2   ; g2 = 2  − a
L (1 + a ) 5 (1 + a )  10 2 
ρ Ω L Ix  1 1 1 2 ρ Ω L Ix  2 1 1 
g3 = 2  + a − a  ; g4 = 2 
+ a + a2 
(1 + a )  30 6 6  (1 + a )  15 6 3 

12 EI 6 EI
k1 = ; k2 =
L (1 + a )
3
L (1 + a )
2

EI ( 4 + a ) EI ( 2 − a )
k3 = ; k4 =
L (1 + a ) L (1 + a )

H.1.1.2 RAYLEIGH Beam Theory

 13 6   11 1 2   9 6 
m1 = ρ AL  + µ 2  ; m2 = ρ AL2  + µ  ; m3 = ρ AL  − µ 2 
 35 5   210 10   70 5 
 13 1   1 2   1 1 
m4 = ρ AL2  − µ 2  ; m5 = ρ AL3  + µ 2  ; m6 = ρ AL3  + µ2 
 420 10   105 15   140 30 

6 ρ Ω Ix 1 1 2
g1 = ; g2 = ρ Ω I x ; g3 = ρ Ω L I x ; g 4 = ρ Ω L I x
5 L 10 30 15

12 EI 6 EI 4 EI 2 EI
k1 = ; k2 = ; k3 = ; k4 =
L3 L2 L L

178
 Mass, Damping, Gyroscopic and Stiffness Matrices for the Shaft, Disk and Bearing Elements

H.1.1.3 EULER-BERNOULLI Beam Theory50

13 11 9
m1 = ρ AL ; m2 = ρ AL2 ; m3 = ρ AL
35 210 70
13 1 1
m4 = ρ AL2 ; m5 = ρ AL3 ; m6 = ρ AL3
420 105 140

12 EI 6 EI 4 EI 2 EI
k1 = ; k2 = ; k3 = ; k4 =
L3 L2 L L

H.1.2 Circular Shaft Element in the Rotating Coordinate System

mS qɺɺnr + cS qɺnr + k S qnr = 0 (6.21)

 m1 0 0 m2 m3 0 0 − m4 
 0 m1 − m2 0 0 m3 m4 0 

 0 − m2 m5 0 0 − m4 − m6 0 
 
m 0 0 m5 m4 0 0 − m6 
mS =  2
 m3 0 0 m4 m1 0 0 − m2 
 
 0 m3 − m4 0 0 m1 m2 0 
 0 m4 − m6 0 0 m2 m5 0 
 
 − m4 0 0 − m6 − m2 0 0 m5 

 0 c1 −c2 0 0 c3 c4 0 
 −c 0 0 −c2 −c3 0 0 c4 
 1
 c2 0 0 c5 c4 0 0 −c6 
 
0 c2 −c5 0 0 c4 c6 0 
cS = 
 0 c3 −c4 0 0 c1 c2 0 
 
 −c3 0 0 −c4 −c1 0 0 c2 
 −c 0 0 −c6 −c2 0 0 c5 
 4 
 0 −c4 c6 0 0 −c2 −c5 0 

 k11 0 0 k22 k33 0 0 − k44 

 0 k11 − k22 0 0 k33 k44 0 

 0 − k22 k55 0 0 − k44 − k66 0 
 
k 0 0 k55 k44 0 0 − k66 
k S =  22
 k33 0 0 k44 k11 0 0 − k22 
 
 0 k33 − k44 0 0 k11 k22 0 
 0 k44 − k66 0 0 k22 k55 0 
 
 − k44 0 0 − k66 − k22 0 0 k55 

50 In the EULER-BERNOULI beam theory the gyroscopic forces are neglected ( g i = 0 ) .

179
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

where51
c1 = −2m1Ω + g1 ; c2 = −2m2 Ω + g 2 ; c3 = −2m3Ω − g1
c4 = −2m4 Ω − g 2 ; c5 = −2m5 Ω + g 4 ; c6 = −2m6 Ω + g3

k11 = −m1Ω 2 + g1Ω + k1 ; k22 = − m2 Ω 2 + g 2 Ω + k2 ; k33 = −m3Ω 2 − g1Ω − k1

k44 = −m4 Ω2 − g 2 Ω − k2 ; k55 = −m5 Ω2 + g 4 Ω + k3 ; k66 = −m6 Ω 2 + g3Ω − k4

H.1.3 Non-Circular Shaft Element in the Rotating Coordinate System

mS qɺɺnr + cS qɺnr + k S qnr = 0 (6.22)

 m1η 0 0 m2η m3η 0 0 − m4η 

 0 m1ζ − m2ζ 0 0 m3ζ m4ζ 0 

 0 − m2ζ m5ζ 0 0 − m4ζ − m6ζ 0 
 
m2η 0 0 m5η m4η 0 0 − m6η 
mS = 
 m3η 0 0 m4η m1η 0 0 − m2η 
 
 0 m3ζ − m4ζ 0 0 m1ζ m2ζ 0 
 0 m4ζ − m6ζ 0 0 m2ζ m5ζ 0 
 
 − m4η 0 0 − m6η − m2η 0 0 m5η 

 0 c1η −c2η 0 0 c3η c4η 0 

 −c 0 0 −c2ζ −c3ζ 0 0 c4ζ 
 1ζ
 c2ζ 0 0 c5ζ c4ζ 0 0 −c6ζ 
 
0 c2η −c5η 0 0 c4η c6η 0 
cS = 
 0 c3η −c4η 0 0 c1η c2η 0 
 
 −c3ζ 0 0 −c4ζ −c1ζ 0 0 c2ζ 
 −c 0 0 −c6ζ −c2ζ 0 0 c5ζ 
 4ζ 
 0 −c4η c6η 0 0 −c2η −c5η 0 

 k11η 0 0 k22η k33η 0 0 −k44η 

 0 k11ζ −k22ζ 0 0 k33ζ k44ζ 0 

 0 −k22ζ k55ζ 0 0 −k44ζ −k66ζ 0 
 
k22η 0 0 k55η k44η 0 0 −k66η 
kS = 
 k33η 0 0 k44η k11η 0 0 −k22η 
 
 0 k33ζ −k44ζ 0 0 k11ζ k22ζ 0 
 0 k44ζ −k66ζ 0 0 k22ζ k55ζ 0 
 
 −k44η 0 0 −k66η −k22η 0 0 k55η 

51
The expressions for mi , g i and ki are given in the units H.1.1.1 − H.1.1.3 .

180
 Mass, Damping, Gyroscopic and Stiffness Matrices for the Shaft, Disk and Bearing Elements

H.1.3.1 TIMOSHENKO Beam Theory

ρ AL  13 7 1 2 6 2 
m1η ,1ζ = 2  35 + 10 aη ,ζ + 3 aη ,ζ + 5 µζ ,η 
(1 + a ) η ,ζ
 

ρ AL2  11 11 1 2 2  1 1 
m2η ,2ζ = 2  210 + 120 aη ,ζ + 24 aη ,ζ + µζ ,η  10 − 2 aη ,ζ 
(1 + a ) η ,ζ   
ρ AL 9 3 1 2 6 2 
m3η ,3ζ = 2  70 + 10 aη ,ζ + 6 aη ,ζ − 5 µζ ,η 
(1 + a ) η ,ζ

ρ AL2  13 3 1 2 2  1 1 
m4η ,4ζ = 2  420 + 40 aη ,ζ + 24 aη ,ζ − µζ ,η  10 − 2 aη ,ζ 
(1 + a ) η ,ζ
  

ρ AL3  1 1 1 2 2  2 1 1 2 
m5η ,5ζ = 2 105 + 60 aη ,ζ + 120 aη ,ζ + µζ ,η  15 + 6 aη ,ζ + 3 aη ,ζ 
(1 + a ) η ,ζ   

ρ AL3  1 1 1 2 2  1 1 1 2 
m6η ,6ζ = 2 140 + 60 aη ,ζ + 120 aη ,ζ + µζ ,η  30 + 6 aη ,ζ − 6 aη ,ζ  
(1 + a )η ,ζ   

12 EIζ ,η Iζ ,η
where aη ,ζ = 2
; µζ2 ,η =
κ GAL AL2

c1η ,1ζ = −2m1η ,1ζ Ω + g1ηζ ; c2η ,2ζ = −2m2η ,2ζ Ω + g 2ηζ ; c3η ,3ζ = −2m3η ,3ζ Ω − g1ηζ
c4η ,4ζ = −2m4η ,4ζ Ω − g 2ηζ ; c5η ,5ζ = −2m5η ,5ζ Ω + g 4ηζ ; c6η ,6ζ = −2m6η ,6ζ Ω + g3ηζ

ρ ΩI x 6 ρ ΩI x 1 1 
g1ηζ =   ; g 2ηζ =  − ( aη + aζ ) 
L (1 + aη )(1 + aζ )  5  (1 + aη )(1 + aζ )  10 4 
ρ ΩLI x  1 1 1 2 
g3ηζ =  + ( aη + aζ ) − aη + aζ2  ( )
(1 + aη )(1 + aζ )  30 12 12 
ρ ΩLI x  2 1 1 2 2 
g 4ηζ =  + ( aη + aζ ) + aη + aζ  ( )
(1 + aη )(1 + aζ )  15 12 6 

k11η ,11ζ = − m1η ,1ζ Ω 2 + g1ηζ Ω + k1η ,1ζ ; k22η ,22ζ = −m2η ,2ζ Ω 2 + g 2ηζ Ω + k2η ,2ζ
k33η ,33ζ = − m3η ,3ζ Ω 2 − g1ηζ Ω − k1η ,1ζ ; k44η ,44ζ = −m4η ,4ζ Ω 2 − g 2ηζ Ω − k2η ,2ζ
k55η ,55ζ = − m5η ,5ζ Ω 2 + g 4ηζ Ω + k3η ,3ζ ; k66η ,66ζ = − m6η ,6ζ Ω 2 + g3ηζ Ω − k4η ,4ζ

12 EIζ ,η 6 EIζ ,η
k1η ,1ζ = ; k 2η ,2ζ =
L (1 + aη ,ζ )
3
L (1 + aη ,ζ
2
)
EIζ ,η ( 4 + aη ,ζ ) EIζ ,η ( 2 − aη ,ζ )
k3η ,3ζ = ; k4η ,4ζ =
L (1 + aη ,ζ ) L (1 + aη ,ζ )

181
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

H.1.3.2 RAYLEIGH Beam Theory

 13 6   11 1 
m1η ,1ζ = ρ AL  + µζ2 ,η  ; m2η ,2ζ = ρ AL2  + µζ2 ,η 
 35 5   210 10 
 9 6   13 1 
m3η ,3ζ = ρ AL  − µζ2 ,η  ; m4η ,4ζ = ρ AL2  − µζ2 ,η 
 70 5   420 10 
 1 2   1 1 
m5η ,5ζ = ρ AL3  + µζ2 ,η  ; m6η ,6ζ = ρ AL3  + µζ2 ,η 
 105 15   140 30 

c1η ,1ζ = −2m1η ,1ζ Ω + g1ηζ ; c2η ,2ζ = −2m2η ,2ζ Ω + g 2ηζ ; c3η ,3ζ = −2m3η ,3ζ Ω − g1ηζ
c4η ,4ζ = −2m4η ,4ζ Ω − g 2ηζ ; c5η ,5ζ = −2m5η ,5ζ Ω + g 4ηζ ; c6η ,6ζ = −2m6η ,6ζ Ω + g3ηζ

k11η ,11ζ = − m1η ,1ζ Ω 2 + g1ηζ Ω + k1η ,1ζ ; k22η ,22ζ = − m2η ,2ζ Ω 2 + g 2ηζ Ω + k2η ,2ζ
k33η ,33ζ = − m3η ,3ζ Ω 2 − g1ηζ Ω − k1η ,1ζ ; k44η ,44ζ = − m4η ,4ζ Ω 2 − g 2ηζ Ω − k2η ,2ζ
k55η ,55ζ = − m5η ,5ζ Ω 2 + g 4ηζ Ω + k3η ,3ζ ; k66η ,66ζ = − m6η ,6ζ Ω 2 + g3ηζ Ω − k4η ,4ζ

12 EIζ ,η 6 EIζ ,η 4 EIζ ,η 2 EIζ ,η

k1η ,1ζ = 3
; k2η ,2ζ = 2
; k3η ,3ζ = ; k4η ,4ζ =
L L L L

H.1.3.3 EULER-BERNOULLI Beam Theory

13 11 9
m1η ,1ζ = ρ AL ; m2η ,2ζ = ρ AL2 ; m3η ,3ζ = ρ AL
35 210 70
13 1 1
m4η ,4ζ = ρ AL2 ; m5η ,5ζ = ρ AL3 ; m6η ,6ζ = ρ AL3
420 105 140

c1η ,1ζ = −2m1η ,1ζ Ω ; c2η ,2ζ = −2m2η ,2ζ Ω ; c3η ,3ζ = −2m3η ,3ζ Ω
c4η ,4ζ = −2m4η ,4ζ Ω ; c5η ,5ζ = −2m5η ,5ζ Ω ; c6η ,6ζ = −2m6η ,6ζ Ω

k11η ,11ζ = − m1η ,1ζ Ω 2 + k1η ,1ζ ; k22η ,22ζ = − m2η ,2ζ Ω 2 + k 2η ,2ζ
k33η ,33ζ = − m3η ,3ζ Ω 2 − k1η ,1ζ ; k 44η ,44ζ = − m4η ,4ζ Ω 2 − k2η ,2ζ
k55η ,55ζ = − m5η ,5ζ Ω 2 + k3η ,3ζ ; k66η ,66ζ = − m6η ,6ζ Ω 2 − k4η ,4ζ

12 EIζ ,η 6 EIζ ,η 4 EIζ ,η 2 EIζ ,η

k1η ,1ζ = 3
; k2η ,2ζ = 2
; k3η ,3ζ = ; k 4η ,4ζ =
L L L L

182
 Mass, Damping, Gyroscopic and Stiffness Matrices for the Shaft, Disk and Bearing Elements

H.1.4 Non-Circular Shaft Element in the Fixed Coordinate System

ɶ qɺɺ
f + G q
ɺ f + K
M ɶ q
f = 0 (6.26)
S n S n S n

ɶ = ∆M e − jΩɶ t + M + ∆M e jΩɶ t
M (6.48)
S −1S 0S +1S

 m10 0 0 m20 m30 0 0 −m40 

 0 m10 −m20 0 0 m30 m40 0 

 0 −m20 m50 0 0 −m40 −m60 0 
 
m 0 0 m50 m40 0 0 −m60 
M 0S =  20
 m30 0 0 m40 m10 0 0 −m20 
 
 0 m30 −m40 0 0 m10 m20 0 
 0 m40 − m60 0 0 m20 m50 0 
 
 − m40 0 0 −m60 − m20 0 0 m50 

 ∆m1 − j ∆m1 j ∆m2 ∆m2 ∆m3 − j ∆m3 − j ∆m4 −∆m4 

 − j ∆m −∆m1 ∆m2 − j∆m2 − j ∆m3 −∆m3 −∆m4 j ∆m4 
 1

 j∆m2 ∆m2 −∆m5 j ∆m5 j ∆m4 ∆m4 ∆m6 − j ∆m6 

 
1 ∆m2 − j ∆m2 j∆m5 ∆m5 ∆m4 − j∆m4 − j ∆m6 −∆m6  − j 2 β
∆M +1S =  ⋅e
2  ∆m3 − j ∆m3 j ∆m4 ∆m4 ∆m1 − j∆m1 − j∆m2 −∆m2 
 
 − j ∆m3 −∆m3 ∆m4 − j∆m4 − j∆m1 −∆m1 −∆m2 j ∆m2 
 − j ∆m −∆m4 ∆m6 − j ∆m6 − j∆m2 −∆m2 −∆m5 j ∆m5 
 4

 −∆m4 j∆m4 − j ∆m6 −∆m6 −∆m2 j ∆m2 j ∆m5 ∆m5 

 0 g1ηζ − g 2ηζ 0 0 − g1ηζ − g 2ηζ 0 

− g 0 0 g 2ηζ g1ηζ 0 0 − g 2ηζ 
 1ηζ
 g 2ηζ 0 0 − g 4ηζ − g 2ηζ 0 0 − g3ηζ 
 
0 g 2ηζ − g 4ηζ 0 0 − g 2ηζ g 3ηζ 0 
GS = 
 0 − g1ηζ g 2ηζ 0 0 g1ηζ g 2ηζ 0 
 
 g1ηζ 0 0 g 2ηζ − g1ηζ 0 0 g 2ηζ 
g 0 0 − g3ηζ − g 2ηζ 0 0 g 4ηζ 
 2ηζ 
 0 g 2ηζ g3ηζ 0 0 − g 2ηζ − g 4ηζ 0 

183
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

ɶ = ∆K e − jΩɶ t + K + ∆K e jΩɶ t ; Ω
K ɶ = 2Ω (6.49)
S −1S 0S +1S

 k10 0 0 k20 −k10 0 0 k20 

 0 k10 − k20 0 0 − k10 − k20 0 
 
 0 − k20 k30 0 0 k20 k40 0 
 
k 0 0 k30 − k20 0 0 k40 
K 0S =  20
 − k10 0 0 −k20 k10 0 0 − k20 
 
 0 − k10 k20 0 0 k10 k20 0 
 0 − k20 k40 0 0 k20 k30 0 
 
 k20 0 0 k40 − k20 0 0 k30 

 ∆k1 − j ∆k1 j ∆k2 ∆k2 −∆k1 j ∆k1 j ∆k2 ∆k2 

 − j ∆k −∆k1 ∆k2 − j ∆k2 j ∆k1 ∆k1 ∆k2 − j ∆k2 
 1

 j ∆k2 ∆k2 −∆k3 j ∆k3 − j ∆k2 −∆k2 −∆k4 j ∆k4 

 
1  ∆k2 − j ∆k2 j ∆k3 ∆k3 −∆k2 j ∆k2 j ∆k4 ∆k4  − j 2 β
∆K +1S = ⋅e
2  −∆k1 j ∆k1 − j ∆k2 −∆k2 ∆k1 − j ∆k1 − j ∆k2 −∆k2 
 
 j ∆k1 ∆k1 −∆k2 j ∆k2 − j ∆k1 −∆k1 −∆k2 j ∆k2 
 j ∆k ∆k2 −∆k4 j ∆k4 − j ∆k2 −∆k2 −∆k3 j ∆k3 
 2

 ∆k2 − j ∆k2 j ∆k4 ∆k4 −∆k2 j ∆k2 j ∆k3 ∆k3 

miη + miζ miη − miζ

where52, 53 ∆M −1S = ∆M +*1S ; mi 0 = ; ∆mi = ; i = 1, 2,⋯ , 6
2 2
kiη + kiζ kiη − kiζ
∆K −1S = ∆K +*1S ; ki 0 = ; ∆ki = ; i = 1, 2,⋯ , 4
2 2

H.2 Disk Elements

H.2.1 Circular Disk Element in the Fixed Coordinate System

M D qɺɺDf + G D qɺ Df = 0 (6.30)

m 0 0 0 0 0 0 0 
0 m 0 0  0 0 0 0 
MD =  ; GD = 
0 0 Jy 0 0 0 0 Ω Jx 
   
0 0 0 Jz  0 0 −Ω J x 0 

52
The expressions for miη \ iζ , g iηζ and kiη \ iζ are given in unit H.1.3.1 − H.1.3.3.
53 The superscript * indicates complex conjugate.

184
 Mass, Damping, Gyroscopic and Stiffness Matrices for the Shaft, Disk and Bearing Elements

where J y = J z = J t ; J x = J p

H.2.2 Circular Disk Element in the Rotating Coordinate System

m D qɺɺDr + c D qɺ Dr + k D qDr = 0 (6.36)

m 0 0 0  0 −2m 0 0 
0 m 0  2m 0 
0  0 0
mD =  ; cD = Ω 
 0 0 Jt 0  0 0 0 J p − 2Jt 
   
0 0 0 Jt   0 0 2Jt − J p 0 

m 0 0 0 
0 m 0 0  54
k D = −Ω 2 
 0 0 Jt − J p 0 
 
 0 0 0 J t − J p 

H.2.3 Non-Circular Disk Element in the Rotating Coordinate System

m D qɺɺDr + c D qɺ Dr + k D qDr = 0 (6.36)

m 0 0 0  0 −2m 0 0 
0 m 0 0   2m 0 0 0 
mD =  ; cD = Ω 
 0 0 Jη 0  0 0 0 J ζ − Jη 
   
 0 0 0 J ζ   0 0 Jζ − Jη 0 

m 0 0 0 
0 m 0 0 
k D = −Ω 2 
 0 0 − Jζ 0 
 
 0 0 0 − Jη 

54 For a circular disk J p = 2 J t and therefore J p − 2 J t = 0 and J t − J p = − J t . But in order to generalize the matrices
cD and k D for a non-circular disk these expressions are left in this general form also for the circular disk.

185
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

H.2.4 Non-Circular Disk Element in the Fixed Coordinate System

ɶ qɺɺ
f + G q
ɺ f = 0
M D D D D

ɶ = ∆M e − jΩɶ t + M + ∆M e jΩɶ t ; Ω
M ɶ = 2Ω
D −1D 0D +1D

m 0 0 0 0 0 0 0 
0 m 0 0  
= ; ∆M +1D = 1 0 0 0 0  − j 2 β
where M 0D ⋅e
 0 0 J0 0 2 0 0 ∆J − j∆ J 
   
0 0 0 J0  0 0 − j∆ J −∆ J 

Jη + Jζ Jη − Jζ
with ∆M −1D = ∆M +*1D ; J 0 = ; ∆J = .
2 2

0 0 0 0 
0 0 0 0 
GD = 
0 0 0 Ω Jx 
 
0 0 −Ω J x 0 

H.3 Bearing Elements

H.3.1 Bearing Element in the Fixed Coordinate System

FBf = DB qɺ Bf + K B qBf (6.37)

 d yy d yz 0 0 k yy k yz 0 0
d d zz 0 0  k k zz 0 0
DB =  zy ; K B =  zy
 0 0 0 0 0 0 0 0
   
 0 0 0 0 0 0 0 0

186
 Mass, Damping, Gyroscopic and Stiffness Matrices for the Shaft, Disk and Bearing Elements

H.3.2 Bearing Element in the Rotating Coordinate System

f B ( t ) = dɶ B qɺ Br + kɶ B qBr . (6.39)

ɶ ɶ
dɶ B = ∆d −1B e− jΩt + d 0B + ∆d +1B e jΩt

 d 01 d 02 0 0  ∆d1 ∆d 2 0 0
 −d 0 0   0 0  − j 2 β
d 01 1  ∆d 2 −∆d1
where d 0B =  02 ; ∆d +1B = ⋅e
 0 0 0 0 2 0 0 0 0
   
 0 0 0 0  0 0 0 0

d yy + d zz d yz − d zy
with ∆d −1B = ∆d +*1B ; d 01 = ; d 02 = ;
2 2
d yy − d zz d yz + d zy d yz + d zy d yy − d zz
∆d1 = −j ; ∆d 2 = +j .
2 2 2 2

ɶ ɶ
kɶ B = ∆k −1B e − jΩt + k 0B + ∆k +1B e jΩt

 k01 k02 0 0  ∆k1 ∆k2 0 0

− k 0 0  0 0  − j 2 β
k01 1 ∆k −∆k1
where k 0B =  02 ; ∆k +1B =  2 ⋅e
 0 0 0 0 2 0 0 0 0
   
 0 0 0 0  0 0 0 0

k yy + k zz k yz − k zy
with ∆k −1B = ∆k +*1B ; k01 = + Ωd 02 ; k02 = − Ωd 01 ;
2 2
k yy − k zz k yz + k zy k yz + k zy k yy − k zz
∆k1 = −j + Ω∆d 2 ; ∆k2 = +j − Ω∆d1 .
2 2 2 2

If d yy = d zz and d yz = −d zy then the bearing damping matrix becomes constant in the rotating

coordinate system. If in addition to the above limitations k yy = k zz and k yz = − k zy then the

bearing stiffness matrix also becomes constant in the rotating coordinate system.

187
 REFERENCES

1. Afolabi D.: Elimination of Periodic Coefficients from the Equation of Motion of

Asymmetric Shafts in Anisotropic Bearings. Archive of Applied Mechanics, 65, 415-424,
1995.

2. Argyris J., Mlejnek H.P.: Die Methode der Finiten Elemente. Band I-III, Friedr. Vieweg
&Sohn, Braunschweig, 1986.

3. Argyris J.: Energy Theorems and Structural Analysis. Butterworth, London, 1954.

4. Belz J.: Modalanalysegestützte Unwuchtidentifikation elastischer Rotoren. Dissertation,

Institute of Mechanics, University of Kassel, 1997.

5. Bently D. E., Muszynska A., Olas A.: Identification of the Modal Parameters by
Perturbation Testing of a Rotor with Strong Gyroscopic Effect. Proceedings of the 2nd
International Conference on Rotordynamics, 189-194, Tokyo, Japan, 14-17 September
1986.

6. Bienert J.: Anwendung der Strukturmodifikation zur Vorhersage der Kreiselwirkung von
symmetrischen und unsymmetrischen Rotoren. Proceedings of the 4th ‘Schwingungen in
Rotierenden Maschinen’ (SIRM IV), Mar. 4-6 1997, 97-104, Kassel, Germany.

7. Boru F. E., Irretier H: Finite Element Investigation of an Asymmetric LAVAL-Rotor-

Bearing System. Institute of Mechanics, University of Kassel, 2nd edition, April 2008.

8. Boru F. E.: Lateral Vibration Analysis of Rotor-FEM. MSc Thesis, Department of

Mechanical Engineering at the Addis Ababa University, Addis Ababa, 2002.

9. Brosens P. J., Crandall S. H.: Whirling of Unsymmetrical Rotors. Transaction of the

ASME, Journal of Applied Mechanics, 28(3), 355-362, September 1961.

10. Campbell W. E.: The Protection of Steam Turbine Disk Wheels from Axial Vibration.
Transaction of the ASME, 31-83, 1924.

11. Champion E.R., Ensminger J. M.: Finite Element Analysis with Personal Computers. New
York, CRC press, 1988.
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

12. Chen L. W., Peng W. K.: Stability Analysis of a Timoshenko Shaft with Dissimilar Lateral
Moment of Inertia. Journal of Sound and Vibration, 207(1), 33-46, 1997.

13. Childs D.: Turbomachinery Rotordynamics-Phenomena Modelling & Analysis. New York,
John Wiley & Sons Inc., 1993.

14. Clough R. W., Penzien J.: Dynamics of Structures. Mcgraw-Hill professional, 1975.

15. Clough R. W., Wilson E. L.: Early Finite Element Research at Berkley. The 5th U.S.
National Conference on Computational Mechanics, August 4-6, 1999.

16. Clough R. W.: The Finite Element Method in Plane Stress Analysis. The 2nd Proceedings
of the ASCE Conference on Electronic Computation, Pittsburg, September 1960.

17. Clough R. W.: Thoughts about the Origin of the Finite Element Method. Computers and
Structures, 79, 2029-2030, 2001

18. Courant R.: Variational Methods for the Solution of Problems of Equilibrium and
Vibrations. Bulletin of the American Mathematical Society, 49, 1-23, 1943.

19. Cowper G. R.: The Shear Coefficient in Timoshenko’s Beam Theory. Journal of Applied
Mechanics, 33, 335-340, June 1966.

20. Crandall S. H., Brosens P. J.: On the Stability of Rotation of a Rotor with Rotationally
Unsymmetric Inertia and Stiffness Properties. Transaction of the ASME, Journal of
Applied Mechanics, 28(4), 567-570, December 1961.

21. Dubbel-Taschenbuch für den Maschinenbau. Springer-Verlag, Berlin/Heidelberg/New

York, 22nd edition, 2007.

22. Dunkerley S.: On the Whirling and Vibration of Shafts. Philosophical Transaction of the
Royal Society of London A, 185, 279-360, 1894.

23. Ertz M., Reister A., Nordmann R.: Zur Berechnung der Eigenschwingungen von
Strukturen mit periodisch zeitvarianten Bewegungsgleichungen. Referat der Tagung an der
Universität Kaiserslautern über Schwingungen in rotierenden Maschinen,
Braunschweig/Wiesbaden, 289-296, 1995.

190
 References

24. Felippa C. A.: Introduction to Finite Element Method. Manuscript at the Department of
Aerospace Engineering Science and Centre for Aerospace Structures, University of
Colorado, 2004.

25. Foote W. R., Poritsky H., Slade J. J.: Critical Speeds of Rotor with Unequal Shaft
Flexibilities Mounted in Bearings of Unequal Flexibility. Transaction of the ASME, Journal
of Applied Mechanics, 10, A77-A84, 1943.

26. Föppl A.: Das Problem der Lavalschen Turbinenwelle. Der Civilingenieur, 41, 335-342,
1895.

27. Forrai L.: Vibrations of a Rotating Asymmetric Shaft Carrying Two Disks, Supported in
Asymmetric Bearings. The IMechE International Conference on Vibrations in Rotating
Machinery, 43-48, University of Cambridge, Cambridge, 15-17 September 1976.

28. Gasch R., Nordmann H., Pfützner H.: Rotordynamik. 2nd edition, Berlin, Springer Verlag,
2002.

29. Genta G.: Vibration of Structures and Machines. 2nd edition, New York, Springer- Velag,
1995.

30. Genta G.: Whirling of Unsymmetrical Rotors: A Finite Element Approach Based on
Complex Coordinate. Journal of Sound and Vibration, 124(1), 27-53, 1988.

31. Gladwell G. M. L., Stammers C. W.: On the Stability of an Unsymmetrical Rigid Rotor
Supported in Unsymmetrical Bearings. Journal of Sound and Vibration, 3(3), 221-232,
1966.

32. Han S. M., Benaroya H., Wie T.: Dynamics of Transversely Vibrating Beams using Four
Engineering Theories. Journal of Sound and Vibration, 225(5), 935-988, 1999.

33. Hrenikoff A.: Solution of Problems in Elasticity by the Frame Work Method. Transaction
of the ASME, Journal of Applied Mechanics, 8, 169-175, 1941.

34. http://science.jrank.org/pages/2376/Electronics-History.html

35. http://www.delaval.com/About_DeLaval/TheCompany/History/Reflections.htm

191
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

36. Huebner K. H., Dewhirst D. L., Smith D. E., and Byrom T. G.: The Finite Element
Method for Engineers. 4th edition, Wiley-IEEE, 2001.

37. Hutchinson J. R.: On Timoshenko Beam of Rectangular Cross-Section. Transaction of the

ASME, Journal of Applied Mechanics, 71(2), 359-367, 2004.

38. Hutchinson J. R.: On Timoshenko Beam of Rectangular Cross-Section. Transaction of the

ASME, Journal of Applied Mechanics, 71, 359-367, May 2004.

39. Irretier H., Kreuzinger-Janik T.: Experimental Modal Analysis- A Tool for Unbalance
Identification of Rotating Machines. Proceedings of the 7th International Symposium on
Transport Phenomena and Dynamics of Rotating Machinery, Feb. 22-26, 1998, 286-295,
Honolulu, Hawaii, USA.

40. Irretier H.: Experimentelle Modalanalyse in der Rotordynamik. VDI-Report Nr 1550, 1-

27, 2000.

41. Irretier H.: Experimentelle Modalanalyse. 4th edition, Kassel, Universität Kassel, 2004.

42. Irretier H.: Maschinen- und Rotordynamik. Lecture Manuscript, 5th edition, Institute of
Mechanics, University of Kassel, Kassel, 2007.

43. Irretier H.: Mathematical Foundations of Experimental Modal Analysis in Rotordynamics.

Mechanical Systems and Signal Processing, 13(2), 183-191, 1999.

44. Iwatsubo T., Tsujiuchi N., Inoue T.: Vibration of Asymmetric Rotor Supported by Oil
Film Bearings. Ingenieur-Archiv, 56, 1-15, 1986.

45. Jeffcott H. H.: The Lateral Vibration of Loaded Shafts in the Neighbourhood of a
Whirling Speed- The Effect of Want of Balance. Philosophical Magazine, 37, 304-314,
1919.

46. Jei Y. G., Lee C. W.: Modal Analysis of Continuous Asymmetrical Rotor-Bearing System.
Journal of Sound and Vibration, 152(2), 245-262, 1992.

47. Kaneko T.: An Experimental Study of the Timoshenko’s Shear Coefficient for Flexural
Vibrating Beam. Journal of Physics D: Applied Physics, 11, 1979-1988, 1978.

48. Kerr W.: On the Whirling Speed of Loaded Shafts. Engineering, 150-, 1916.

192
 References

49. Kimball A. L.: Internal Friction as a Cause of Shaft Whirling. Philosophical Magazine, 49,
724-727, 1925.

50. Kreuzinger-Janik T., Irretier H.: New Excitation and Response Measurement Techniques
for Modal Testing of Flexible Rotors. Proceedings of the 5th IFToMM International
Conference on Rotordynamics, Sept. 7-10 1998, 695-708, Darmstadt, Germany.

51. Lalanne M., Ferraris G.: Rotordynamics Prediction in Engineering. Chichester, John Wiley
& Sons Inc., 1990.

52. Lee C. W., Han D. J. and Hong S. W.: Modal Analysis of Periodically Time-Varying Linear
Rotor System using Floquet Theory. The 7th IFToMM Conference on Rotor Dynamics,
Paper ID-78, Vienna, Austria, 25-28 September 2006.

53. Lee C. W., Han D. J., Suh J. H., Hong S. W.: Modal Analysis of Periodically Time-Varying
Linear Rotor System. Journal of Sound and Vibration, 303, 553-574, 2007.

54. Lee C. W., Park J. P., Yun J. S., Joh C. Y.: Complex Time Series Analysis for Rotor
Dynamics Identification. Mechanical System and Signal Processing, 11(6), 827-842, 1997.

55. Lee C. W., Seo Y. H.: Strong and Weak Whirling Modes in Rotating Machinery. The 9th
IMechE International Conference on Vibrations in Rotating Machinery, 277-288, Exeter,
United Kingdom, 8-11 September 2008.

56. Lee C. W., Suh J. H., Hong S. W.: Modal Analysis of Rotor System using Modulated
Coordinates for Asymmetric Rotor System with Anisotropic Stator. The 8th International
Conference on Vibrations in Rotating Machinery, University of Swansea, UK, 575-584,
2004.

57. Lee C. W.: A Complex Modal Testing Theory for Rotating Machinery. Mechanical System
and Signal Processing, 5(2), 119-137, 1991.

58. Lee C. W.: Vibration Analysis of Rotors. Dordrecht, Kluwer Academic Publishers, 1993.

59. Lund J. W., Orcutt F. K.: Calculations and Experiments on the Unbalance Response of a
Flexible Rotor. Transaction of the ASME, Journal of Engineering for Industry, 89, 785-
796, 1967.

193
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

60. Lund J. W.: Stability and Damped Critical Speed of a Flexible Rotor in Fluid Film
Bearings. Transaction of the ASME, Journal of Engineering for Industry, 96(2), 509-517,
1974.

61. Matzenmiller A.: Einführung in die Methode der Finiten Elemente. 2nd edition, Manuscript
at the Institute of Mechanics, University of Kassel, 2004.

62. McHenry D.: A Lattice Analogy for the Solution of Plane Stress Problem. Journal of
Institution of Civil Engineers, 21, 59-82, 1943.

63. Meirovitch L.: Analytical Methods in Vibration. The Macmillan Company, New York,
1967.

64. Meirovitch L.: Elements of Vibration Analysis. McGraw-Hill, Inc., New York, 1986.

65. Meirovitch L.: Methods of Analytical Dynamics. New York, Dover Publication Inc, 1998.

66. Muszynska, A., Bently, D. E., Franklin, W. D., and Hayashida, R. D.: 1989b. Identification
of Modal Parameters of Rotating Systems using Perturbation Techniques. Proceedings the
12th ASME Conference on Mechanical Vibration and Noise, 107-118, Montreal, Canada,
17-21 September 1989.

67. Muszynska, A.: Modal Testing of Rotor/Bearing Systems. International Journal of

Analytical and Experimental Modal Analysis, 1(3), 15-34, 1986.

68. Myklestad N. O.: A New Method for Calculating Natural Modes of Uncoupled Bending
Vibration of Airplane Wings and Other Types of Beams. Journal of Aeronautical Science,
11, 153-162, 1944.

69. Nelson F. C.: A Brief History of Early Rotor Dynamics. Sound and Vibration Magazine,
June 2003.

70. Nelson H. D., Mc Vaugh J. M.: The Dynamics of Rotor-Bearing System using Finite
Element. Transaction of the ASME, Journal of Engineering for Industry, 98(2), 593-600,
1976.

71. Nelson H. D.: A Finite Rotating Shaft Element using Timoshenko Beam Theory.
Transaction of the ASME, Journal of Mechanical Design, 102, 793-803, 1980.

194
 References

72. Newkirk B. L., Taylor H. D.: Shaft Whipping due to Oil Action in Journal Bearings.
General Electric Review, 28(7), 559-568, 1925.

73. Newmark N. M.: Numerical Methods of Analysis in Engineering. Macmillan, New York,
1949.

74. Noah S. T., Hopkins G. R.: A Generalized HILL’s Method for the Stability of
Parametrically Excited Dynamic System. Transaction of the ASME, Journal of Applied
Mechanics, 49(1), 217-223, March 1982.

75. Nordmann R., Matros M., Neumer T.: Parameter Identification in Rotating Machinery by
Means of Magnetic Bearings. Proceeding of the 4th IFToMM International Conference on
Rotordynamics, 7-16, Chicago, Illinois, USA, 7-9 September 1994.

76. Nordmann R., Schöllhorn K.: Experimentelle Ermittlung modaler Großen von Rotoren.
VDI-report Nr 320, 15-24, 1978.

77. Oncescu F., Lakis A. A., Ostiguy G.: Investigation of the Stability and Steady State
Response of Asymmetric Rotors using Finite Element Formulation. Journal of Sound and
Vibration, 245(2), 303-328, 2001.

78. Person M., Wiedemann M.: Stabilität und Empfindlichkeit der Stabilität bei periodisch
zeitvarianten Systemen - HILL versus FLOQUET. Report from Institute of Aerospace of
the University of Berlin, July 1989.

79. Person M.: Zur Dynamik von Windturbinen mit Gelenkflügeln- Stabilität und erzwungene
Schwingungen von Ein- und Mehrflüglern. VDI Progress Report, 1988.

80. Prager W., Synge J. L.: Approximation in Elasticity Based on the Concept of Function
Space. Quarterly of Applied Mathematics, 5, 241-269, 1947.

81. Prandtl L.: Beiträge zur Frage der kritischen Drehzahlen. Dinglers Polytechnisches Journal,
333, 179-182, 1918.

82. Prohl M. A.: A General Method for Calculating Critical Speeds of Flexible Rotors.
Transaction of the ASME, Journal of Applied Mechanics, 12(3), 142-148, 1945.

83. Rankine W. J.: On the Centrifugal Force of Rotating Shafts. Engineer, 27, 249-249, 1869.

195
Numerical and Experimental Response and Stability Investigations of Anisotropic Rotor-Bearing Systems

84. Rao J. S., Sreenivas R.: Dynamics of Asymmetric Rotors using Solid Models. Proceedings
of the International Gas Turbine Congress 2003 Tokyo, November 2-7, 2003.

85. Rao J. S.: Rotor Dynamics. 2nd edition, New Delhi, Wiley Eastern Limited, 1991.

86. Reddy J. N.: An Introduction to the Finite Element Method. 2nd edition, McGraw-Hill,
Inc., New Jersey, 1993.

87. Rieger N. F.: Vibration of Rotating Machinery. New York, Stress Technology
Incorporated, 1982.

88. Rodgers C.: On the Vibration and Critical Speed of Rotors. Philosophical Magazine, 44,
122-156, 1922.

89. Rogers P. J., Ewins D. J.: Modal Testing of an Operating Rotor System using a Structural
Dynamics Approach. Proceedings of the 7th International Modal Analysis Conference, Jan.
30 - Feb. 2, 1989, 466-473, Las Vegas, NV, USA.

90. Ruhl R. L., Booker J. F.: A Finite Element Model for Distributed Parameter Turborotor
System. Transaction of the ASME, Journal of Engineering for Industry, 94(1), 128-132,
1972

91. Schlegel V.: Zur Parameteridentifikation von Turboläufern aus dem Schwingungs-
verhalten. Dissertation, Faculty of Mechanical Engineering, University of Hannover, 1979.

92. Smith D. M.: The Motion of a Rotor Carried by a Flexible Shaft in Flexible Bearings.
Proceedings of the Royal Society of London A, 142, 92-118, 1933.

93. Stodola A.: Dampf- und Gas-Turbinen. Berlin, Springer Verlag, 1924.

94. Strang G., Fix G. J.: An Analysis of the Finite Element Method. New Jersey, Prentice-Hall,
1973.

95. Suh J. H., Hong S. W., Lee C. W.: Modal Analysis of Asymmetric Rotor System with
Isotropic Stator using Modulated Coordinate. Journal of Sound and Vibration, 284, 651-
671, 2005.

96. Takashi I., Murakami S.: Dynamic Response and Stability of a Rotating Asymmetric Shaft
Mounted on a Flexible Base. Nonlinear Dynamics, 20, 1-19, 1999.

196
 References

97. Teichmann H.: Biegeschwingungsverhalten von unrunden Rotoren in Gleitlagern.

Dissertation, Institute of Transport Systems and Applied Mechanics, Faculty of
Mechanical Engineering and Transport Systems, Technical University Berlin, 1995.

98. Tondl A.: Some Problems of Rotor Dynamics. Prague, Chapman & Hall, 1965.

99. Traill-Nash R. W., Collar A. R.: The Effects of Shear Flexibility and Rotatory Inertia on
the Bending Vibrations of Beams. Quarterly Journal of Mechanics and Applied
Mathematics, 6, 186-222, March 1953.

100. Vance J. M.: Rotordynamics of Turbomachinery. New York, John Wiley & Sons Inc.,
1988.

101. Wowk V: Machinery Vibration- Measurement and Analysis. New York, Book-mart Press,
1991.

102. Yamamoto T. and Öta H.: On the Unstable Vibrations of a Shaft Carrying an
Unsymmetrical Rotor. Transaction of the ASME, Journal of Applied Mechanics, 31(3),
515-522, September 1964.

103. Zienkiewicz O. C., Cheung Y. K.: Finite Element in the Solution of Field Problems.
Engineer, 220, 507-510, 1965.

104. Zienkiewicz O. C., Taylor R. L.: The Finite Element Method. 5th edition, Butterworth-
Heinemann, Oxford, 2000.

197